From 3c7c221ae6cfcdf6840da1d5262d94c75d56288b Mon Sep 17 00:00:00 2001 From: Oscar Wan Date: Wed, 26 Aug 2026 08:50:07 -0700 Subject: experiment: add completed A6000 crossover results --- .../r2/resnet-r2-bp-d20-lr0p1.json | 2537 ++ .../r2/resnet-r2-bp-d20-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-bp-d32-lr0p1.json | 2597 +++ .../r2/resnet-r2-bp-d32-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-bp-d56-lr0p1.json | 2717 +++ .../r2/resnet-r2-bp-d56-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-clean_kp-d20-lr0p1.json | 3869 ++++ ...resnet-r2-clean_kp-d20-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-clean_kp-d32-lr0p1.json | 4001 ++++ ...resnet-r2-clean_kp-d32-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-clean_kp-d56-lr0p1.json | 4265 ++++ ...resnet-r2-clean_kp-d56-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-dfa-d20-lr0p01.json | 2604 +++ .../r2/resnet-r2-dfa-d20-lr0p01.json.manifest.json | 132 + .../r2/resnet-r2-dfa-d32-lr0p01.json | 2700 +++ .../r2/resnet-r2-dfa-d32-lr0p01.json.manifest.json | 132 + .../r2/resnet-r2-dfa-d56-lr0p01.json | 2892 +++ .../r2/resnet-r2-dfa-d56-lr0p01.json.manifest.json | 132 + .../r2/resnet-r2-dualprop-d20-lr0p05.json | 2965 +++ ...esnet-r2-dualprop-d20-lr0p05.json.manifest.json | 132 + .../r2/resnet-r2-dualprop-d32-lr0p05.json | 2965 +++ ...esnet-r2-dualprop-d32-lr0p05.json.manifest.json | 132 + ...esnet-r2-dualprop-d56-lr0p05.json.manifest.json | 132 + .../r2/resnet-r2-ep-d20-lr0p0003.json | 1565 ++ .../resnet-r2-ep-d20-lr0p0003.json.manifest.json | 134 + .../r2/resnet-r2-ep-d32-lr0p0003.json | 1565 ++ .../resnet-r2-ep-d32-lr0p0003.json.manifest.json | 134 + .../r2/resnet-r2-ep-d56-lr0p0003.json | 1565 ++ .../resnet-r2-ep-d56-lr0p0003.json.manifest.json | 134 + .../r2/resnet-r2-fa-d20-lr0p1.json | 2669 +++ .../r2/resnet-r2-fa-d20-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-fa-d32-lr0p1.json | 2801 +++ .../r2/resnet-r2-fa-d32-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-fa-d56-lr0p1.json | 3065 +++ .../r2/resnet-r2-fa-d56-lr0p1.json.manifest.json | 130 + .../r2/resnet-r2-ff-d20-lr0p03.json | 8265 +++++++ .../r2/resnet-r2-ff-d20-lr0p03.json.manifest.json | 130 + .../r2/resnet-r2-ff-d32-lr0p03.json | 13125 +++++++++++ .../r2/resnet-r2-ff-d32-lr0p03.json.manifest.json | 130 + .../r2/resnet-r2-ff-d56-lr0p03.json | 22845 +++++++++++++++++++ .../r2/resnet-r2-ff-d56-lr0p03.json.manifest.json | 130 + .../r2/resnet-r2-pepita-d20-lr0p003.json | 1565 ++ ...resnet-r2-pepita-d20-lr0p003.json.manifest.json | 128 + .../r2/resnet-r2-pepita-d32-lr0p003.json | 1565 ++ ...resnet-r2-pepita-d32-lr0p003.json.manifest.json | 128 + .../r2/resnet-r2-pepita-d56-lr0p003.json | 1565 ++ ...resnet-r2-pepita-d56-lr0p003.json.manifest.json | 128 + .../r2/resnet-r2-sdil-d20-lr0p1.json | 7256 ++++++ .../r2/resnet-r2-sdil-d20-lr0p1.json.manifest.json | 150 + .../r2/resnet-r2-sdil-d32-lr0p1.json | 7472 ++++++ .../r2/resnet-r2-sdil-d32-lr0p1.json.manifest.json | 150 + .../r2/resnet-r2-sdil-d56-lr0p1.json | 7904 +++++++ .../r2/resnet-r2-sdil-d56-lr0p1.json.manifest.json | 150 + 53 files changed, 122492 insertions(+) create mode 100644 results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-dualprop-d56-lr0p05.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json create mode 100644 results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json create mode 100644 results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json create mode 100644 results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json create mode 100644 results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json.manifest.json create mode 100644 results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json create mode 100644 results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json.manifest.json (limited to 'results/resnet_crossover/r2') diff --git a/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json new file mode 100644 index 0000000..548e13e --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json @@ -0,0 +1,2537 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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, + "eval_accuracy": 0.4004, + "eval_loss": 1.7710138671875, + "eval_split": "validation", + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 1.6644643401463826 + }, + { + "epoch": 2, + "eval_accuracy": 0.5672, + "eval_loss": 1.221260009765625, + "eval_split": "validation", + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.2860143685658774 + }, + { + "epoch": 3, + "eval_accuracy": 0.6406, + "eval_loss": 1.0877959594726563, + "eval_split": "validation", + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.02516774397956 + }, + { + "epoch": 4, + "eval_accuracy": 0.6802, + "eval_loss": 0.9318223510742187, + "eval_split": "validation", + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 0.8597116688304477 + }, + { + "epoch": 5, + "eval_accuracy": 0.7032, + "eval_loss": 0.9318287963867188, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.7553892990854051 + }, + { + "epoch": 6, + "eval_accuracy": 0.737, + "eval_loss": 0.864343017578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.6687519911342197 + }, + { + "epoch": 7, + "eval_accuracy": 0.7616, + "eval_loss": 0.7147335205078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6076406918737624 + }, + { + "epoch": 8, + "eval_accuracy": 0.7384, + "eval_loss": 0.8417417602539062, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.5667279717922211 + }, + { + "epoch": 9, + "eval_accuracy": 0.7888, + "eval_loss": 0.6749005004882812, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5312818857087029 + }, + { + "epoch": 10, + "eval_accuracy": 0.8044, + "eval_loss": 0.602718798828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5028076828956604 + }, + { + "epoch": 11, + "eval_accuracy": 0.8276, + "eval_loss": 0.5217552856445312, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.4791169890032874 + }, + { + "epoch": 12, + "eval_accuracy": 0.7946, + "eval_loss": 0.63441904296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.4623024181683858 + }, + { + "epoch": 13, + "eval_accuracy": 0.6858, + "eval_loss": 1.0865167114257812, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.43909971352683175 + }, + { + "epoch": 14, + "eval_accuracy": 0.8122, + "eval_loss": 0.5628994506835937, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.42348679066763983 + }, + { + "epoch": 15, + "eval_accuracy": 0.8034, + "eval_loss": 0.6175327880859375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.41002807636260985 + }, + { + "epoch": 16, + "eval_accuracy": 0.8388, + "eval_loss": 0.50029931640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.39710193020502726 + }, + { + "epoch": 17, + "eval_accuracy": 0.8058, + "eval_loss": 0.578526904296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.3893488602956136 + }, + { + "epoch": 18, + "eval_accuracy": 0.8182, + "eval_loss": 0.56311103515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.38157559702661303 + }, + { + "epoch": 19, + "eval_accuracy": 0.8238, + "eval_loss": 0.5585554077148438, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.3700638519499037 + }, + { + "epoch": 20, + "eval_accuracy": 0.8182, + "eval_loss": 0.6097387573242188, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3558312208970388 + }, + { + "epoch": 21, + "eval_accuracy": 0.8246, + "eval_loss": 0.5422500793457031, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.3505221447202894 + }, + { + "epoch": 22, + "eval_accuracy": 0.8226, + "eval_loss": 0.5890034423828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3476858967781067 + }, + { + "epoch": 23, + "eval_accuracy": 0.8584, + "eval_loss": 0.45730882568359377, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.3428149689250522 + }, + { + "epoch": 24, + "eval_accuracy": 0.8518, + "eval_loss": 0.48339465942382814, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.3287880612267388 + }, + { + "epoch": 25, + "eval_accuracy": 0.8414, + "eval_loss": 0.4980908935546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.32596546165678236 + }, + { + "epoch": 26, + "eval_accuracy": 0.8368, + "eval_loss": 0.5045202209472657, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.31808142843776277 + }, + { + "epoch": 27, + "eval_accuracy": 0.8258, + "eval_loss": 0.5857073852539062, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.31575259125762517 + }, + { + "epoch": 28, + "eval_accuracy": 0.8596, + "eval_loss": 0.4352029235839844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.30214522661368054 + }, + { + "epoch": 29, + "eval_accuracy": 0.8388, + "eval_loss": 0.5140209655761718, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.3071780029588275 + }, + { + "epoch": 30, + "eval_accuracy": 0.844, + "eval_loss": 0.5278230224609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.30251549474928113 + }, + { + "epoch": 31, + "eval_accuracy": 0.8484, + "eval_loss": 0.4901044494628906, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.2975416218333774 + }, + { + "epoch": 32, + "eval_accuracy": 0.8412, + "eval_loss": 0.49973211669921874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.28743110587861803 + }, + { + "epoch": 33, + "eval_accuracy": 0.859, + "eval_loss": 0.45563653564453127, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.2848365594731437 + }, + { + "epoch": 34, + "eval_accuracy": 0.8556, + "eval_loss": 0.4716713928222656, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.2897914389504327 + }, + { + "epoch": 35, + "eval_accuracy": 0.8738, + "eval_loss": 0.4081214294433594, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.2804644786781735 + }, + { + "epoch": 36, + "eval_accuracy": 0.8702, + "eval_loss": 0.4287448791503906, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.27175004272990755 + }, + { + "epoch": 37, + "eval_accuracy": 0.8618, + "eval_loss": 0.4426470031738281, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.27865564083125854 + }, + { + "epoch": 38, + "eval_accuracy": 0.8396, + "eval_loss": 0.5234903503417969, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.26684371184772915 + }, + { + "epoch": 39, + "eval_accuracy": 0.855, + "eval_loss": 0.48744876708984375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.2674422241899702 + }, + { + "epoch": 40, + "eval_accuracy": 0.865, + "eval_loss": 0.4351408203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.2645341257254283 + }, + { + "epoch": 41, + "eval_accuracy": 0.8506, + "eval_loss": 0.4751714965820312, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.2622300498406092 + }, + { + "epoch": 42, + "eval_accuracy": 0.8486, + "eval_loss": 0.5085056274414063, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.25661233155727387 + }, + { + "epoch": 43, + "eval_accuracy": 0.8576, + "eval_loss": 0.4584455322265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.256074648004108 + }, + { + "epoch": 44, + "eval_accuracy": 0.805, + "eval_loss": 0.6773380615234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.25414007183975645 + }, + { + "epoch": 45, + "eval_accuracy": 0.867, + "eval_loss": 0.458577587890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.24826880225075615 + }, + { + "epoch": 46, + "eval_accuracy": 0.8544, + "eval_loss": 0.47776199951171877, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.2539279241243998 + }, + { + "epoch": 47, + "eval_accuracy": 0.864, + "eval_loss": 0.44519176635742186, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.2478049456278483 + }, + { + "epoch": 48, + "eval_accuracy": 0.8556, + "eval_loss": 0.466750341796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.24715948182211983 + }, + { + "epoch": 49, + "eval_accuracy": 0.8706, + "eval_loss": 0.45115664672851563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.24359825609525046 + }, + { + "epoch": 50, + "eval_accuracy": 0.8546, + "eval_loss": 0.4703095520019531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.24361659493313895 + }, + { + "epoch": 51, + "eval_accuracy": 0.8532, + "eval_loss": 0.5028222534179687, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.24154270447625054 + }, + { + "epoch": 52, + "eval_accuracy": 0.8768, + "eval_loss": 0.4012300231933594, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.23156300439834596 + }, + { + "epoch": 53, + "eval_accuracy": 0.8702, + "eval_loss": 0.43121511840820315, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.23431335463523864 + }, + { + "epoch": 54, + "eval_accuracy": 0.8506, + "eval_loss": 0.47547684326171874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.23226308568848503 + }, + { + "epoch": 55, + "eval_accuracy": 0.8574, + "eval_loss": 0.47655361328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.23324218198458355 + }, + { + "epoch": 56, + "eval_accuracy": 0.8666, + "eval_loss": 0.43212182006835936, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.23033563688066272 + }, + { + "epoch": 57, + "eval_accuracy": 0.8346, + "eval_loss": 0.5559804565429688, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.23134963205655415 + }, + { + "epoch": 58, + "eval_accuracy": 0.8638, + "eval_loss": 0.4514068420410156, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.2259135717921787 + }, + { + "epoch": 59, + "eval_accuracy": 0.8464, + "eval_loss": 0.4881473083496094, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.2296271209306187 + }, + { + "epoch": 60, + "eval_accuracy": 0.862, + "eval_loss": 0.48375531005859373, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.22572548664675818 + }, + { + "epoch": 61, + "eval_accuracy": 0.8684, + "eval_loss": 0.43897296142578124, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.22529616802533467 + }, + { + "epoch": 62, + "eval_accuracy": 0.8742, + "eval_loss": 0.39048966674804686, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.2210662999153137 + }, + { + "epoch": 63, + "eval_accuracy": 0.8618, + "eval_loss": 0.4357768005371094, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.2175708965169059 + }, + { + "epoch": 64, + "eval_accuracy": 0.8634, + "eval_loss": 0.46370264282226564, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.21798163765536413 + }, + { + "epoch": 65, + "eval_accuracy": 0.8574, + "eval_loss": 0.47452117309570313, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.22943376750946046 + }, + { + "epoch": 66, + "eval_accuracy": 0.8706, + "eval_loss": 0.42474490966796874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.22070531824032466 + }, + { + "epoch": 67, + "eval_accuracy": 0.8738, + "eval_loss": 0.43662334594726565, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.21509841103288863 + }, + { + "epoch": 68, + "eval_accuracy": 0.8518, + "eval_loss": 0.47741646728515624, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.21424531423250834 + }, + { + "epoch": 69, + "eval_accuracy": 0.8728, + "eval_loss": 0.445956689453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.20924902565744188 + }, + { + "epoch": 70, + "eval_accuracy": 0.8602, + "eval_loss": 0.4686448669433594, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.21512837672498492 + }, + { + "epoch": 71, + "eval_accuracy": 0.8564, + "eval_loss": 0.511124951171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.2124501904964447 + }, + { + "epoch": 72, + "eval_accuracy": 0.8566, + "eval_loss": 0.4767055786132812, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.2073291993273629 + }, + { + "epoch": 73, + "eval_accuracy": 0.8534, + "eval_loss": 0.5192368408203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.21053057844903733 + }, + { + "epoch": 74, + "eval_accuracy": 0.8766, + "eval_loss": 0.38260752563476563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.21595640751520792 + }, + { + "epoch": 75, + "eval_accuracy": 0.8756, + "eval_loss": 0.4221235717773438, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.21067103199164072 + }, + { + "epoch": 76, + "eval_accuracy": 0.876, + "eval_loss": 0.42229520263671877, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.20760168025758532 + }, + { + "epoch": 77, + "eval_accuracy": 0.8816, + "eval_loss": 0.41327566528320314, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.20539346924357943 + }, + { + "epoch": 78, + "eval_accuracy": 0.855, + "eval_loss": 0.4727892761230469, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.20733415001233418 + }, + { + "epoch": 79, + "eval_accuracy": 0.8538, + "eval_loss": 0.5247862976074219, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.20816828174591065 + }, + { + "epoch": 80, + "eval_accuracy": 0.8562, + "eval_loss": 0.5231691284179687, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.20593991249402363 + }, + { + "epoch": 81, + "eval_accuracy": 0.8644, + "eval_loss": 0.434323828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.20798121780554454 + }, + { + "epoch": 82, + "eval_accuracy": 0.8718, + "eval_loss": 0.43466107177734375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.2018360275056627 + }, + { + "epoch": 83, + "eval_accuracy": 0.873, + "eval_loss": 0.4336739562988281, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.20214488338894315 + }, + { + "epoch": 84, + "eval_accuracy": 0.8738, + "eval_loss": 0.4181335754394531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.20384543903138902 + }, + { + "epoch": 85, + "eval_accuracy": 0.8588, + "eval_loss": 0.5384344909667969, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.20373935463693407 + }, + { + "epoch": 86, + "eval_accuracy": 0.8714, + "eval_loss": 0.4508935974121094, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.20023504439459908 + }, + { + "epoch": 87, + "eval_accuracy": 0.8722, + "eval_loss": 0.4011263122558594, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.20357054704560174 + }, + { + "epoch": 88, + "eval_accuracy": 0.8724, + "eval_loss": 0.42971153564453124, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.2017909148454666 + }, + { + "epoch": 89, + "eval_accuracy": 0.8508, + "eval_loss": 0.5215956726074219, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.19414195964336395 + }, + { + "epoch": 90, + "eval_accuracy": 0.8584, + "eval_loss": 0.5140850219726563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.19675261680814954 + }, + { + "epoch": 91, + "eval_accuracy": 0.8542, + "eval_loss": 0.5065567810058593, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.20285941220124562 + }, + { + "epoch": 92, + "eval_accuracy": 0.8632, + "eval_loss": 0.4859201049804687, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.19762985503408645 + }, + { + "epoch": 93, + "eval_accuracy": 0.8656, + "eval_loss": 0.4395970275878906, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.18883612869050767 + }, + { + "epoch": 94, + "eval_accuracy": 0.8754, + "eval_loss": 0.4072813049316406, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.20028164596822526 + }, + { + "epoch": 95, + "eval_accuracy": 0.876, + "eval_loss": 0.4278625915527344, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.18935567673047385 + }, + { + "epoch": 96, + "eval_accuracy": 0.8754, + "eval_loss": 0.448179248046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.1924068394528495 + }, + { + "epoch": 97, + "eval_accuracy": 0.859, + "eval_loss": 0.4913785766601563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.19636243831846448 + }, + { + "epoch": 98, + "eval_accuracy": 0.8798, + "eval_loss": 0.39225457153320314, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.19402525822851394 + }, + { + "epoch": 99, + "eval_accuracy": 0.8718, + "eval_loss": 0.4399384094238281, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.19180790162351397 + }, + { + "epoch": 100, + "eval_accuracy": 0.8756, + "eval_loss": 0.44355946044921873, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.19071120822164747 + }, + { + "epoch": 101, + "eval_accuracy": 0.9076, + "eval_loss": 0.306420263671875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.12239563805394703 + }, + { + "epoch": 102, + "eval_accuracy": 0.9104, + "eval_loss": 0.30657166137695313, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.09218944064643649 + }, + { + "epoch": 103, + "eval_accuracy": 0.9124, + "eval_loss": 0.298024169921875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.08602881506813898 + }, + { + "epoch": 104, + "eval_accuracy": 0.91, + "eval_loss": 0.30647501831054685, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.07542413108613756 + }, + { + "epoch": 105, + "eval_accuracy": 0.9144, + "eval_loss": 0.3046997314453125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.07289705408944024 + }, + { + "epoch": 106, + "eval_accuracy": 0.9158, + "eval_loss": 0.30186732177734377, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.06816188242700365 + }, + { + "epoch": 107, + "eval_accuracy": 0.9152, + "eval_loss": 0.3045905822753906, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.06303669462733799 + }, + { + "epoch": 108, + "eval_accuracy": 0.9138, + "eval_loss": 0.31312174072265625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.06031152280039258 + }, + { + "epoch": 109, + "eval_accuracy": 0.913, + "eval_loss": 0.3162181823730469, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.05866686734623379 + }, + { + "epoch": 110, + "eval_accuracy": 0.916, + "eval_loss": 0.3138496337890625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.0575275326596366 + }, + { + "epoch": 111, + "eval_accuracy": 0.9168, + "eval_loss": 0.31661144409179687, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.05273066160678864 + }, + { + "epoch": 112, + "eval_accuracy": 0.9194, + "eval_loss": 0.3204001953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.05101389784415563 + }, + { + "epoch": 113, + "eval_accuracy": 0.9174, + "eval_loss": 0.33216910400390626, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.04935557192696465 + }, + { + "epoch": 114, + "eval_accuracy": 0.9158, + "eval_loss": 0.33527969970703125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.04651711027887132 + }, + { + "epoch": 115, + "eval_accuracy": 0.9148, + "eval_loss": 0.3381342224121094, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.046925569189919365 + }, + { + "epoch": 116, + "eval_accuracy": 0.9182, + "eval_loss": 0.33178223266601564, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.042672778570652006 + }, + { + "epoch": 117, + "eval_accuracy": 0.9158, + "eval_loss": 0.34461751098632815, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.044006208296616875 + }, + { + "epoch": 118, + "eval_accuracy": 0.917, + "eval_loss": 0.3422942260742188, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.04278108269870281 + }, + { + "epoch": 119, + "eval_accuracy": 0.9158, + "eval_loss": 0.34879398803710937, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.039726313829421996 + }, + { + "epoch": 120, + "eval_accuracy": 0.9192, + "eval_loss": 0.3475035583496094, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.0380223665383127 + }, + { + "epoch": 121, + "eval_accuracy": 0.9174, + "eval_loss": 0.35022559814453125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.037888774723476834 + }, + { + "epoch": 122, + "eval_accuracy": 0.9158, + "eval_loss": 0.34550585327148436, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.03741276666960783 + }, + { + "epoch": 123, + "eval_accuracy": 0.9174, + "eval_loss": 0.3567695190429688, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.03865679248174032 + }, + { + "epoch": 124, + "eval_accuracy": 0.9176, + "eval_loss": 0.35861453857421877, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.035508182628618344 + }, + { + "epoch": 125, + "eval_accuracy": 0.9162, + "eval_loss": 0.35652288208007815, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.036840546764598955 + }, + { + "epoch": 126, + "eval_accuracy": 0.918, + "eval_loss": 0.35606851806640627, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.03339896608259943 + }, + { + "epoch": 127, + "eval_accuracy": 0.9166, + "eval_loss": 0.36003599243164064, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.034402696437305876 + }, + { + "epoch": 128, + "eval_accuracy": 0.9166, + "eval_loss": 0.36662847900390627, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.03277938154472245 + }, + { + "epoch": 129, + "eval_accuracy": 0.9194, + "eval_loss": 0.36271561279296877, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.0318130090382364 + }, + { + "epoch": 130, + "eval_accuracy": 0.9186, + "eval_loss": 0.36240728149414064, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.0301810850335492 + }, + { + "epoch": 131, + "eval_accuracy": 0.9156, + "eval_loss": 0.36919815063476563, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.03182596413758066 + }, + { + "epoch": 132, + "eval_accuracy": 0.9184, + "eval_loss": 0.3730984680175781, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.03062839696539773 + }, + { + "epoch": 133, + "eval_accuracy": 0.9164, + "eval_loss": 0.37771748046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.028816458198097018 + }, + { + "epoch": 134, + "eval_accuracy": 0.9186, + "eval_loss": 0.3720545104980469, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.029397612818082175 + }, + { + "epoch": 135, + "eval_accuracy": 0.9176, + "eval_loss": 0.3824631408691406, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.028310401866502233 + }, + { + "epoch": 136, + "eval_accuracy": 0.9184, + "eval_loss": 0.3975944519042969, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.027445427227682537 + }, + { + "epoch": 137, + "eval_accuracy": 0.9154, + "eval_loss": 0.388535009765625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.028050902290145556 + }, + { + "epoch": 138, + "eval_accuracy": 0.9144, + "eval_loss": 0.39537998046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.027626405088106792 + }, + { + "epoch": 139, + "eval_accuracy": 0.9174, + "eval_loss": 0.3798079833984375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.02575553153918849 + }, + { + "epoch": 140, + "eval_accuracy": 0.9154, + "eval_loss": 0.3928522644042969, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.026475244792302448 + }, + { + "epoch": 141, + "eval_accuracy": 0.9176, + "eval_loss": 0.38615267944335935, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.026054457494947645 + }, + { + "epoch": 142, + "eval_accuracy": 0.9152, + "eval_loss": 0.3925371520996094, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.024948420998040174 + }, + { + "epoch": 143, + "eval_accuracy": 0.9142, + "eval_loss": 0.40529662475585937, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.025066455528471206 + }, + { + "epoch": 144, + "eval_accuracy": 0.9148, + "eval_loss": 0.3978242797851563, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.023501766060127154 + }, + { + "epoch": 145, + "eval_accuracy": 0.9146, + "eval_loss": 0.3887966552734375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.02534153993262185 + }, + { + "epoch": 146, + "eval_accuracy": 0.9182, + "eval_loss": 0.3958949401855469, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.025001446268293592 + }, + { + "epoch": 147, + "eval_accuracy": 0.9152, + "eval_loss": 0.3987830993652344, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.024140962322553 + }, + { + "epoch": 148, + "eval_accuracy": 0.9128, + "eval_loss": 0.3882256103515625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.02524253683288892 + }, + { + "epoch": 149, + "eval_accuracy": 0.9156, + "eval_loss": 0.39553275756835937, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.024576971726285087 + }, + { + "epoch": 150, + "eval_accuracy": 0.9168, + "eval_loss": 0.389462451171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.022645456993414297 + }, + { + "epoch": 151, + "eval_accuracy": 0.9182, + "eval_loss": 0.38770169067382815, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.01990334998567899 + }, + { + "epoch": 152, + "eval_accuracy": 0.9184, + "eval_loss": 0.38250919189453125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.018938107203278275 + }, + { + "epoch": 153, + "eval_accuracy": 0.9152, + "eval_loss": 0.38870892333984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.019380051201913093 + }, + { + "epoch": 154, + "eval_accuracy": 0.9174, + "eval_loss": 0.38606708984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.01811076421704557 + }, + { + "epoch": 155, + "eval_accuracy": 0.917, + "eval_loss": 0.38509130859375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.017462889037529628 + }, + { + "epoch": 156, + "eval_accuracy": 0.9178, + "eval_loss": 0.3827947204589844, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.017325110151370365 + }, + { + "epoch": 157, + "eval_accuracy": 0.9174, + "eval_loss": 0.3864782897949219, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.017640994083881378 + }, + { + "epoch": 158, + "eval_accuracy": 0.9168, + "eval_loss": 0.3881909729003906, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.017731406083785824 + }, + { + "epoch": 159, + "eval_accuracy": 0.9178, + "eval_loss": 0.38386309204101565, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.01699283284942309 + }, + { + "epoch": 160, + "eval_accuracy": 0.9176, + "eval_loss": 0.38543032836914065, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.017252897179871798 + }, + { + "epoch": 161, + "eval_accuracy": 0.917, + "eval_loss": 0.3836103637695312, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.016278477271108163 + }, + { + "epoch": 162, + "eval_accuracy": 0.9164, + "eval_loss": 0.388681640625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.01604987183109754 + }, + { + "epoch": 163, + "eval_accuracy": 0.9186, + "eval_loss": 0.38605330200195315, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.01634608430729972 + }, + { + "epoch": 164, + "eval_accuracy": 0.918, + "eval_loss": 0.38527861328125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.015447496734228398 + }, + { + "epoch": 165, + "eval_accuracy": 0.9162, + "eval_loss": 0.39043060913085936, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.016543969903389612 + }, + { + "epoch": 166, + "eval_accuracy": 0.9168, + "eval_loss": 0.3898049377441406, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.016445220247407754 + }, + { + "epoch": 167, + "eval_accuracy": 0.917, + "eval_loss": 0.3887971130371094, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.016121218882666693 + }, + { + "epoch": 168, + "eval_accuracy": 0.9172, + "eval_loss": 0.3925276428222656, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.015302956238720153 + }, + { + "epoch": 169, + "eval_accuracy": 0.9184, + "eval_loss": 0.3891949768066406, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.01579265182117621 + }, + { + "epoch": 170, + "eval_accuracy": 0.916, + "eval_loss": 0.3865777526855469, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.015928712855113877 + }, + { + "epoch": 171, + "eval_accuracy": 0.9184, + "eval_loss": 0.3891114501953125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.01610312660369608 + }, + { + "epoch": 172, + "eval_accuracy": 0.9178, + "eval_loss": 0.3897466430664063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.015435847022467189 + }, + { + "epoch": 173, + "eval_accuracy": 0.917, + "eval_loss": 0.39024139404296876, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.016074572068452836 + }, + { + "epoch": 174, + "eval_accuracy": 0.9178, + "eval_loss": 0.3874936279296875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.01542677370963825 + }, + { + "epoch": 175, + "eval_accuracy": 0.9168, + "eval_loss": 0.3924413757324219, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.014539433761768871 + }, + { + "epoch": 176, + "eval_accuracy": 0.9176, + "eval_loss": 0.38622310791015624, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.01533671252793736 + }, + { + "epoch": 177, + "eval_accuracy": 0.9186, + "eval_loss": 0.39207570190429686, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.01441938015388118 + }, + { + "epoch": 178, + "eval_accuracy": 0.9172, + "eval_loss": 0.394850341796875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.015913767455352678 + }, + { + "epoch": 179, + "eval_accuracy": 0.9184, + "eval_loss": 0.3906371459960937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.015869515031576156 + }, + { + "epoch": 180, + "eval_accuracy": 0.9184, + "eval_loss": 0.3919496215820312, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.014408293558988306 + }, + { + "epoch": 181, + "eval_accuracy": 0.9196, + "eval_loss": 0.3897723571777344, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.01504027492403984 + }, + { + "epoch": 182, + "eval_accuracy": 0.9188, + "eval_loss": 0.3978993347167969, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.015615745646092628 + }, + { + "epoch": 183, + "eval_accuracy": 0.919, + "eval_loss": 0.39478623046875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.015364365304840935 + }, + { + "epoch": 184, + "eval_accuracy": 0.9186, + "eval_loss": 0.39719932250976564, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.014254255270957947 + }, + { + "epoch": 185, + "eval_accuracy": 0.9188, + "eval_loss": 0.3978095458984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.01493217945512798 + }, + { + "epoch": 186, + "eval_accuracy": 0.9186, + "eval_loss": 0.39569725341796874, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.013478640712632073 + }, + { + "epoch": 187, + "eval_accuracy": 0.9166, + "eval_loss": 0.39535325317382813, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.013779820157090823 + }, + { + "epoch": 188, + "eval_accuracy": 0.9184, + "eval_loss": 0.3943298767089844, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.014935846232540078 + }, + { + "epoch": 189, + "eval_accuracy": 0.9178, + "eval_loss": 0.39513787841796877, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.014631074411628975 + }, + { + "epoch": 190, + "eval_accuracy": 0.917, + "eval_loss": 0.3951256713867187, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.01456148929099242 + }, + { + "epoch": 191, + "eval_accuracy": 0.9182, + "eval_loss": 0.3951748596191406, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.013960837727391885 + }, + { + "epoch": 192, + "eval_accuracy": 0.9178, + "eval_loss": 0.39580638427734377, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.014646381699293852 + }, + { + "epoch": 193, + "eval_accuracy": 0.9184, + "eval_loss": 0.3964273864746094, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.013860816383361816 + }, + { + "epoch": 194, + "eval_accuracy": 0.9186, + "eval_loss": 0.3953535705566406, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.013962972517146005 + }, + { + "epoch": 195, + "eval_accuracy": 0.9196, + "eval_loss": 0.3967518371582031, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.013807961952024037 + }, + { + "epoch": 196, + "eval_accuracy": 0.9182, + "eval_loss": 0.3972501220703125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.014517988362577226 + }, + { + "epoch": 197, + "eval_accuracy": 0.9184, + "eval_loss": 0.3951761962890625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.01375881134768327 + }, + { + "epoch": 198, + "eval_accuracy": 0.9164, + "eval_loss": 0.39577669677734373, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.013727891698976358 + }, + { + "epoch": 199, + "eval_accuracy": 0.918, + "eval_loss": 0.3989869873046875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.014122135368155108 + }, + { + "epoch": 200, + "eval_accuracy": 0.9176, + "eval_loss": 0.39629163818359375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.014614876673618953 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9176, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.39629163818359375, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2017119744, + "peak_memory_reserved_bytes": 2464153600, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 21.029086351394653, + "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": 1224.873178243637, + "train_wall_s": 1203.823739528656 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 729918720000000, + "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": 364959360000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 1094878080000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json.manifest.json new file mode 100644 index 0000000..5790984 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "bp", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "0", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787035719.2218466, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 1228.3254659175873, + "epochs": 200, + "experiment_name": "resnet-r2-bp-d20-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "bp", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d20-lr0p1.json", + "output_exists": true, + "output_sha256": "5cf3b2928b4d27a0968fb42d633d5c23d91dad233b24a25daffb0fa9cd2fe7ee", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json new file mode 100644 index 0000000..532bc80 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json @@ -0,0 +1,2597 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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, + "eval_accuracy": 0.4654, + "eval_loss": 1.4851486328125, + "eval_split": "validation", + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 1.7367673873901368 + }, + { + "epoch": 2, + "eval_accuracy": 0.4898, + "eval_loss": 1.725862939453125, + "eval_split": "validation", + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.3518918582280477 + }, + { + "epoch": 3, + "eval_accuracy": 0.5918, + "eval_loss": 1.3513257568359376, + "eval_split": "validation", + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.0766678378211127 + }, + { + "epoch": 4, + "eval_accuracy": 0.6896, + "eval_loss": 0.9191304931640625, + "eval_split": "validation", + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 0.8923857101016575 + }, + { + "epoch": 5, + "eval_accuracy": 0.7498, + "eval_loss": 0.7415965698242187, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.767474871105618 + }, + { + "epoch": 6, + "eval_accuracy": 0.7114, + "eval_loss": 0.9565532348632813, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.6632765820503235 + }, + { + "epoch": 7, + "eval_accuracy": 0.7736, + "eval_loss": 0.671961181640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6028720584339565 + }, + { + "epoch": 8, + "eval_accuracy": 0.7782, + "eval_loss": 0.70337177734375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.5587517645835877 + }, + { + "epoch": 9, + "eval_accuracy": 0.7972, + "eval_loss": 0.6166611938476563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.516674728096856 + }, + { + "epoch": 10, + "eval_accuracy": 0.7304, + "eval_loss": 0.9077167846679688, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.4872661090373993 + }, + { + "epoch": 11, + "eval_accuracy": 0.7792, + "eval_loss": 0.687501904296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.46315930485195583 + }, + { + "epoch": 12, + "eval_accuracy": 0.8068, + "eval_loss": 0.602014111328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.4376789878103468 + }, + { + "epoch": 13, + "eval_accuracy": 0.7708, + "eval_loss": 0.7402918701171874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.4173557227876451 + }, + { + "epoch": 14, + "eval_accuracy": 0.8168, + "eval_loss": 0.5883634521484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.4026338017993503 + }, + { + "epoch": 15, + "eval_accuracy": 0.836, + "eval_loss": 0.5093399963378906, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.38672829824553595 + }, + { + "epoch": 16, + "eval_accuracy": 0.826, + "eval_loss": 0.5640574462890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.37543256346384685 + }, + { + "epoch": 17, + "eval_accuracy": 0.8372, + "eval_loss": 0.48900656127929687, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.3632183216094971 + }, + { + "epoch": 18, + "eval_accuracy": 0.8556, + "eval_loss": 0.4457982177734375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.35278323027822706 + }, + { + "epoch": 19, + "eval_accuracy": 0.8474, + "eval_loss": 0.462450390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.344698157954216 + }, + { + "epoch": 20, + "eval_accuracy": 0.8326, + "eval_loss": 0.5396587097167969, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.33207386435932584 + }, + { + "epoch": 21, + "eval_accuracy": 0.8402, + "eval_loss": 0.5037933044433593, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.3211999874644809 + }, + { + "epoch": 22, + "eval_accuracy": 0.8164, + "eval_loss": 0.5877996826171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3155520805676778 + }, + { + "epoch": 23, + "eval_accuracy": 0.8612, + "eval_loss": 0.45332628173828127, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.30989496260748967 + }, + { + "epoch": 24, + "eval_accuracy": 0.8544, + "eval_loss": 0.455251806640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.29552004895475176 + }, + { + "epoch": 25, + "eval_accuracy": 0.8598, + "eval_loss": 0.4580159118652344, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.29108409735891555 + }, + { + "epoch": 26, + "eval_accuracy": 0.8516, + "eval_loss": 0.45268468017578123, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.290835695192549 + }, + { + "epoch": 27, + "eval_accuracy": 0.8604, + "eval_loss": 0.4285613464355469, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.2841388143777847 + }, + { + "epoch": 28, + "eval_accuracy": 0.8634, + "eval_loss": 0.4222153564453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.2743539887905121 + }, + { + "epoch": 29, + "eval_accuracy": 0.8402, + "eval_loss": 0.5224630859375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.27201610174179075 + }, + { + "epoch": 30, + "eval_accuracy": 0.8474, + "eval_loss": 0.4655813903808594, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.26682973040474783 + }, + { + "epoch": 31, + "eval_accuracy": 0.8522, + "eval_loss": 0.4705499938964844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.26481335138214956 + }, + { + "epoch": 32, + "eval_accuracy": 0.8432, + "eval_loss": 0.5508898071289062, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.26295674863391455 + }, + { + "epoch": 33, + "eval_accuracy": 0.868, + "eval_loss": 0.4072789001464844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.25482014421357047 + }, + { + "epoch": 34, + "eval_accuracy": 0.8502, + "eval_loss": 0.4737674072265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.25594374244478013 + }, + { + "epoch": 35, + "eval_accuracy": 0.8636, + "eval_loss": 0.43671715698242186, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.24550499177773794 + }, + { + "epoch": 36, + "eval_accuracy": 0.8504, + "eval_loss": 0.5042031433105468, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.24113377828068203 + }, + { + "epoch": 37, + "eval_accuracy": 0.8564, + "eval_loss": 0.456636279296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.24051628319422405 + }, + { + "epoch": 38, + "eval_accuracy": 0.8644, + "eval_loss": 0.455577294921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.23370062657992044 + }, + { + "epoch": 39, + "eval_accuracy": 0.8668, + "eval_loss": 0.4439416625976563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.2288697966337204 + }, + { + "epoch": 40, + "eval_accuracy": 0.876, + "eval_loss": 0.40666158447265627, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.2292720184326172 + }, + { + "epoch": 41, + "eval_accuracy": 0.874, + "eval_loss": 0.439620068359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.2260119872940911 + }, + { + "epoch": 42, + "eval_accuracy": 0.8744, + "eval_loss": 0.410631787109375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.22043352348274656 + }, + { + "epoch": 43, + "eval_accuracy": 0.8492, + "eval_loss": 0.5022360961914063, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.2215538299176428 + }, + { + "epoch": 44, + "eval_accuracy": 0.861, + "eval_loss": 0.4755216369628906, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.22306056750350528 + }, + { + "epoch": 45, + "eval_accuracy": 0.8722, + "eval_loss": 0.43333682250976563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.2135424311319987 + }, + { + "epoch": 46, + "eval_accuracy": 0.8742, + "eval_loss": 0.44612857666015626, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.21563978569772507 + }, + { + "epoch": 47, + "eval_accuracy": 0.8712, + "eval_loss": 0.4285671081542969, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.21078225373162163 + }, + { + "epoch": 48, + "eval_accuracy": 0.88, + "eval_loss": 0.3932395385742187, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.21454219284587436 + }, + { + "epoch": 49, + "eval_accuracy": 0.848, + "eval_loss": 0.5626197265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.21450545298788282 + }, + { + "epoch": 50, + "eval_accuracy": 0.8706, + "eval_loss": 0.4620752685546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.20769385964075723 + }, + { + "epoch": 51, + "eval_accuracy": 0.8528, + "eval_loss": 0.49712684326171874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.21213035280969408 + }, + { + "epoch": 52, + "eval_accuracy": 0.8806, + "eval_loss": 0.4304973815917969, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.2030811865568161 + }, + { + "epoch": 53, + "eval_accuracy": 0.8882, + "eval_loss": 0.36722509765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.2000371700392829 + }, + { + "epoch": 54, + "eval_accuracy": 0.8664, + "eval_loss": 0.4356230651855469, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.19917343689335718 + }, + { + "epoch": 55, + "eval_accuracy": 0.8716, + "eval_loss": 0.43104464721679686, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.19837952054871452 + }, + { + "epoch": 56, + "eval_accuracy": 0.8512, + "eval_loss": 0.49728991088867186, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.19955373254352146 + }, + { + "epoch": 57, + "eval_accuracy": 0.851, + "eval_loss": 0.5140265563964844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.19353483844863043 + }, + { + "epoch": 58, + "eval_accuracy": 0.8788, + "eval_loss": 0.41481446533203126, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.19488789674705928 + }, + { + "epoch": 59, + "eval_accuracy": 0.8672, + "eval_loss": 0.4865489929199219, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.1920552120367686 + }, + { + "epoch": 60, + "eval_accuracy": 0.8594, + "eval_loss": 0.5026045593261719, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.18388274129231771 + }, + { + "epoch": 61, + "eval_accuracy": 0.8774, + "eval_loss": 0.421068212890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.18769480295181273 + }, + { + "epoch": 62, + "eval_accuracy": 0.8728, + "eval_loss": 0.406421240234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.193258737399843 + }, + { + "epoch": 63, + "eval_accuracy": 0.858, + "eval_loss": 0.5191707092285156, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.1826606490029229 + }, + { + "epoch": 64, + "eval_accuracy": 0.8816, + "eval_loss": 0.41435826416015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.18747838848696816 + }, + { + "epoch": 65, + "eval_accuracy": 0.8784, + "eval_loss": 0.40977137451171874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.18679874833689797 + }, + { + "epoch": 66, + "eval_accuracy": 0.8744, + "eval_loss": 0.4389544189453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.18423253170251846 + }, + { + "epoch": 67, + "eval_accuracy": 0.878, + "eval_loss": 0.43681519775390626, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.18141219682428572 + }, + { + "epoch": 68, + "eval_accuracy": 0.8818, + "eval_loss": 0.3969242736816406, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.1803228695180681 + }, + { + "epoch": 69, + "eval_accuracy": 0.8846, + "eval_loss": 0.40659450073242187, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.1756576668156518 + }, + { + "epoch": 70, + "eval_accuracy": 0.8704, + "eval_loss": 0.4603804931640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.18106614275773367 + }, + { + "epoch": 71, + "eval_accuracy": 0.8642, + "eval_loss": 0.4872785766601562, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.17989884740511577 + }, + { + "epoch": 72, + "eval_accuracy": 0.872, + "eval_loss": 0.41939688110351564, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.17623830699788198 + }, + { + "epoch": 73, + "eval_accuracy": 0.8822, + "eval_loss": 0.4099404174804688, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.17245613755914901 + }, + { + "epoch": 74, + "eval_accuracy": 0.8872, + "eval_loss": 0.37339692993164064, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.17451896119647556 + }, + { + "epoch": 75, + "eval_accuracy": 0.878, + "eval_loss": 0.4249140869140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.17885927769872878 + }, + { + "epoch": 76, + "eval_accuracy": 0.8756, + "eval_loss": 0.41772691650390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.16925643913613425 + }, + { + "epoch": 77, + "eval_accuracy": 0.8868, + "eval_loss": 0.40187772216796874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.16989823605219523 + }, + { + "epoch": 78, + "eval_accuracy": 0.8786, + "eval_loss": 0.4158897705078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.17419387707710265 + }, + { + "epoch": 79, + "eval_accuracy": 0.8718, + "eval_loss": 0.4515111572265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.1669309623559316 + }, + { + "epoch": 80, + "eval_accuracy": 0.8772, + "eval_loss": 0.4646386657714844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.1685080126285553 + }, + { + "epoch": 81, + "eval_accuracy": 0.8792, + "eval_loss": 0.4025916625976563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.1665036998987198 + }, + { + "epoch": 82, + "eval_accuracy": 0.856, + "eval_loss": 0.49355731201171876, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.16681153111457825 + }, + { + "epoch": 83, + "eval_accuracy": 0.8674, + "eval_loss": 0.4592894470214844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.17225918037626478 + }, + { + "epoch": 84, + "eval_accuracy": 0.8668, + "eval_loss": 0.49534368896484376, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.1667035684797499 + }, + { + "epoch": 85, + "eval_accuracy": 0.8798, + "eval_loss": 0.44453221435546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.16662992577022975 + }, + { + "epoch": 86, + "eval_accuracy": 0.8926, + "eval_loss": 0.3840708984375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.1652555229081048 + }, + { + "epoch": 87, + "eval_accuracy": 0.891, + "eval_loss": 0.382837548828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.16847461541493733 + }, + { + "epoch": 88, + "eval_accuracy": 0.886, + "eval_loss": 0.4266284973144531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.16859921624395582 + }, + { + "epoch": 89, + "eval_accuracy": 0.8776, + "eval_loss": 0.4573221496582031, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.16235651333332063 + }, + { + "epoch": 90, + "eval_accuracy": 0.8702, + "eval_loss": 0.46487689819335937, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.16057908042801752 + }, + { + "epoch": 91, + "eval_accuracy": 0.8654, + "eval_loss": 0.4855129089355469, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.16279987126191456 + }, + { + "epoch": 92, + "eval_accuracy": 0.881, + "eval_loss": 0.43439296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.16621251893308428 + }, + { + "epoch": 93, + "eval_accuracy": 0.8926, + "eval_loss": 0.3748136779785156, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.1611718230115043 + }, + { + "epoch": 94, + "eval_accuracy": 0.871, + "eval_loss": 0.4397787536621094, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.16647623745997747 + }, + { + "epoch": 95, + "eval_accuracy": 0.8814, + "eval_loss": 0.4218956359863281, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.16139989402559068 + }, + { + "epoch": 96, + "eval_accuracy": 0.887, + "eval_loss": 0.4029042053222656, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.15817028217315673 + }, + { + "epoch": 97, + "eval_accuracy": 0.8846, + "eval_loss": 0.426631787109375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.15957610891660054 + }, + { + "epoch": 98, + "eval_accuracy": 0.8962, + "eval_loss": 0.35802305908203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.15509013244311015 + }, + { + "epoch": 99, + "eval_accuracy": 0.8836, + "eval_loss": 0.4078686767578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.16025661662419638 + }, + { + "epoch": 100, + "eval_accuracy": 0.8698, + "eval_loss": 0.4947103332519531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.15619702618320783 + }, + { + "epoch": 101, + "eval_accuracy": 0.9172, + "eval_loss": 0.283290478515625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.0919350504954656 + }, + { + "epoch": 102, + "eval_accuracy": 0.916, + "eval_loss": 0.2839746948242188, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.06293851057622168 + }, + { + "epoch": 103, + "eval_accuracy": 0.9204, + "eval_loss": 0.28171181640625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.053611354100538626 + }, + { + "epoch": 104, + "eval_accuracy": 0.9216, + "eval_loss": 0.2781740661621094, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.04668084240456422 + }, + { + "epoch": 105, + "eval_accuracy": 0.9226, + "eval_loss": 0.2879105224609375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.04359266593654951 + }, + { + "epoch": 106, + "eval_accuracy": 0.9234, + "eval_loss": 0.29206656494140626, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.038716788862148924 + }, + { + "epoch": 107, + "eval_accuracy": 0.9246, + "eval_loss": 0.29641119384765624, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.036085647984345756 + }, + { + "epoch": 108, + "eval_accuracy": 0.9232, + "eval_loss": 0.30074464111328125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.03377311499449942 + }, + { + "epoch": 109, + "eval_accuracy": 0.9242, + "eval_loss": 0.30086492919921876, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.03104732546210289 + }, + { + "epoch": 110, + "eval_accuracy": 0.924, + "eval_loss": 0.3048627014160156, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.030359695131911172 + }, + { + "epoch": 111, + "eval_accuracy": 0.924, + "eval_loss": 0.3087982421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.027889062374830246 + }, + { + "epoch": 112, + "eval_accuracy": 0.924, + "eval_loss": 0.3089269226074219, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.027552290483315787 + }, + { + "epoch": 113, + "eval_accuracy": 0.9266, + "eval_loss": 0.3071017822265625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.025298097997903824 + }, + { + "epoch": 114, + "eval_accuracy": 0.9212, + "eval_loss": 0.3117630615234375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.02453123506307602 + }, + { + "epoch": 115, + "eval_accuracy": 0.9254, + "eval_loss": 0.31803255615234377, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.022481246304677592 + }, + { + "epoch": 116, + "eval_accuracy": 0.923, + "eval_loss": 0.3159791198730469, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.022579287205139797 + }, + { + "epoch": 117, + "eval_accuracy": 0.9256, + "eval_loss": 0.3194157165527344, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.021970157868001196 + }, + { + "epoch": 118, + "eval_accuracy": 0.924, + "eval_loss": 0.3299554748535156, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.020222671439912583 + }, + { + "epoch": 119, + "eval_accuracy": 0.9256, + "eval_loss": 0.3334829528808594, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.019925816707147492 + }, + { + "epoch": 120, + "eval_accuracy": 0.9262, + "eval_loss": 0.3268382507324219, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.017402058001400697 + }, + { + "epoch": 121, + "eval_accuracy": 0.925, + "eval_loss": 0.3368515625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.01769845224486457 + }, + { + "epoch": 122, + "eval_accuracy": 0.9274, + "eval_loss": 0.3335461181640625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.017841399673289723 + }, + { + "epoch": 123, + "eval_accuracy": 0.927, + "eval_loss": 0.33780157470703126, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.016299792051646445 + }, + { + "epoch": 124, + "eval_accuracy": 0.9268, + "eval_loss": 0.34122034301757814, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.015568790648380916 + }, + { + "epoch": 125, + "eval_accuracy": 0.9256, + "eval_loss": 0.34111203002929685, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.016821574689199526 + }, + { + "epoch": 126, + "eval_accuracy": 0.9244, + "eval_loss": 0.34721368408203124, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.015542411519090334 + }, + { + "epoch": 127, + "eval_accuracy": 0.9256, + "eval_loss": 0.34593865356445314, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.015203730338811875 + }, + { + "epoch": 128, + "eval_accuracy": 0.9262, + "eval_loss": 0.3516455078125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.013415788523107767 + }, + { + "epoch": 129, + "eval_accuracy": 0.9264, + "eval_loss": 0.35247120971679685, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.013082009490993288 + }, + { + "epoch": 130, + "eval_accuracy": 0.9238, + "eval_loss": 0.3506563232421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.012456533164613776 + }, + { + "epoch": 131, + "eval_accuracy": 0.9248, + "eval_loss": 0.3509290588378906, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.013791085198024909 + }, + { + "epoch": 132, + "eval_accuracy": 0.9244, + "eval_loss": 0.3555397583007813, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.012909759015010463 + }, + { + "epoch": 133, + "eval_accuracy": 0.9258, + "eval_loss": 0.3627603454589844, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.012380503948198424 + }, + { + "epoch": 134, + "eval_accuracy": 0.9258, + "eval_loss": 0.36262833251953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.011786968642969926 + }, + { + "epoch": 135, + "eval_accuracy": 0.9268, + "eval_loss": 0.3640386047363281, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.012401803022126357 + }, + { + "epoch": 136, + "eval_accuracy": 0.925, + "eval_loss": 0.36981715087890626, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.01138345877594418 + }, + { + "epoch": 137, + "eval_accuracy": 0.9254, + "eval_loss": 0.36221641845703123, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.01095986443956693 + }, + { + "epoch": 138, + "eval_accuracy": 0.9254, + "eval_loss": 0.37304404296875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.011773464465969139 + }, + { + "epoch": 139, + "eval_accuracy": 0.9246, + "eval_loss": 0.37177071533203127, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.010575827517815762 + }, + { + "epoch": 140, + "eval_accuracy": 0.9246, + "eval_loss": 0.37671232299804686, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.01054492362919781 + }, + { + "epoch": 141, + "eval_accuracy": 0.9246, + "eval_loss": 0.3732326416015625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.011174578830599784 + }, + { + "epoch": 142, + "eval_accuracy": 0.925, + "eval_loss": 0.373262890625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.009631121079499523 + }, + { + "epoch": 143, + "eval_accuracy": 0.9266, + "eval_loss": 0.3794520751953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.010214703265577555 + }, + { + "epoch": 144, + "eval_accuracy": 0.925, + "eval_loss": 0.3776400390625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.009037613876826233 + }, + { + "epoch": 145, + "eval_accuracy": 0.9218, + "eval_loss": 0.3810528930664063, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.009570675100634496 + }, + { + "epoch": 146, + "eval_accuracy": 0.9242, + "eval_loss": 0.3810512023925781, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.009632527514712679 + }, + { + "epoch": 147, + "eval_accuracy": 0.924, + "eval_loss": 0.3873039978027344, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.009912042107847002 + }, + { + "epoch": 148, + "eval_accuracy": 0.9258, + "eval_loss": 0.3850289794921875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.009672875482009517 + }, + { + "epoch": 149, + "eval_accuracy": 0.9262, + "eval_loss": 0.3870732299804687, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.008430721307463116 + }, + { + "epoch": 150, + "eval_accuracy": 0.925, + "eval_loss": 0.3842845153808594, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.00836476296401686 + }, + { + "epoch": 151, + "eval_accuracy": 0.9248, + "eval_loss": 0.3796745849609375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.007752285371720791 + }, + { + "epoch": 152, + "eval_accuracy": 0.926, + "eval_loss": 0.3802146240234375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.008240172734318508 + }, + { + "epoch": 153, + "eval_accuracy": 0.9252, + "eval_loss": 0.38012276000976564, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.006998140365506212 + }, + { + "epoch": 154, + "eval_accuracy": 0.9266, + "eval_loss": 0.3791354736328125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.006879981412159072 + }, + { + "epoch": 155, + "eval_accuracy": 0.9258, + "eval_loss": 0.37881537475585936, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.006742739064494769 + }, + { + "epoch": 156, + "eval_accuracy": 0.9264, + "eval_loss": 0.38017742919921876, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.006501108894248804 + }, + { + "epoch": 157, + "eval_accuracy": 0.9254, + "eval_loss": 0.3797306640625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.006946949698196517 + }, + { + "epoch": 158, + "eval_accuracy": 0.9252, + "eval_loss": 0.3788091552734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0069799597129432695 + }, + { + "epoch": 159, + "eval_accuracy": 0.9256, + "eval_loss": 0.378277099609375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.006285468248588343 + }, + { + "epoch": 160, + "eval_accuracy": 0.927, + "eval_loss": 0.37642570190429686, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.006426012566747765 + }, + { + "epoch": 161, + "eval_accuracy": 0.925, + "eval_loss": 0.37553992919921875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.005865842514650689 + }, + { + "epoch": 162, + "eval_accuracy": 0.9256, + "eval_loss": 0.38009876708984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.006607880344159074 + }, + { + "epoch": 163, + "eval_accuracy": 0.9254, + "eval_loss": 0.3770572082519531, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.006660557381560405 + }, + { + "epoch": 164, + "eval_accuracy": 0.9266, + "eval_loss": 0.38073275146484375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.006553415514073438 + }, + { + "epoch": 165, + "eval_accuracy": 0.9262, + "eval_loss": 0.3793877258300781, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.005639047432028585 + }, + { + "epoch": 166, + "eval_accuracy": 0.928, + "eval_loss": 0.3806024658203125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.006318632692760891 + }, + { + "epoch": 167, + "eval_accuracy": 0.9256, + "eval_loss": 0.3804926208496094, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.005757771301083267 + }, + { + "epoch": 168, + "eval_accuracy": 0.9266, + "eval_loss": 0.3818713256835938, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.005887672925574912 + }, + { + "epoch": 169, + "eval_accuracy": 0.9256, + "eval_loss": 0.38314972534179687, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.0062801586909840505 + }, + { + "epoch": 170, + "eval_accuracy": 0.9266, + "eval_loss": 0.3803857055664063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.005388368136166699 + }, + { + "epoch": 171, + "eval_accuracy": 0.9274, + "eval_loss": 0.38005389404296874, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.005335355056242811 + }, + { + "epoch": 172, + "eval_accuracy": 0.926, + "eval_loss": 0.3798578369140625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.005680962507923444 + }, + { + "epoch": 173, + "eval_accuracy": 0.9268, + "eval_loss": 0.37983125610351565, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.005903353605791926 + }, + { + "epoch": 174, + "eval_accuracy": 0.927, + "eval_loss": 0.3794900756835938, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.005541315438598394 + }, + { + "epoch": 175, + "eval_accuracy": 0.9274, + "eval_loss": 0.37762247924804687, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.006261111334876882 + }, + { + "epoch": 176, + "eval_accuracy": 0.9266, + "eval_loss": 0.37954530029296873, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.0054626991358068255 + }, + { + "epoch": 177, + "eval_accuracy": 0.926, + "eval_loss": 0.37930660400390626, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.0054296861392756305 + }, + { + "epoch": 178, + "eval_accuracy": 0.9274, + "eval_loss": 0.3783561584472656, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.005808477004700237 + }, + { + "epoch": 179, + "eval_accuracy": 0.9254, + "eval_loss": 0.3786826538085937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.005417727620734109 + }, + { + "epoch": 180, + "eval_accuracy": 0.9268, + "eval_loss": 0.37825419921875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.005960534370400839 + }, + { + "epoch": 181, + "eval_accuracy": 0.9276, + "eval_loss": 0.37896533203125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.005777932195696566 + }, + { + "epoch": 182, + "eval_accuracy": 0.9258, + "eval_loss": 0.3841477783203125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.00565131838483115 + }, + { + "epoch": 183, + "eval_accuracy": 0.926, + "eval_loss": 0.37931566162109376, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.006461898795018593 + }, + { + "epoch": 184, + "eval_accuracy": 0.9268, + "eval_loss": 0.3812836486816406, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.005682292328650753 + }, + { + "epoch": 185, + "eval_accuracy": 0.9268, + "eval_loss": 0.3813713623046875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.005426007693178124 + }, + { + "epoch": 186, + "eval_accuracy": 0.9268, + "eval_loss": 0.3784933288574219, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.005445328628644347 + }, + { + "epoch": 187, + "eval_accuracy": 0.925, + "eval_loss": 0.3797018127441406, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.0051834056191146375 + }, + { + "epoch": 188, + "eval_accuracy": 0.9268, + "eval_loss": 0.3820699096679688, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.005792698964145448 + }, + { + "epoch": 189, + "eval_accuracy": 0.9248, + "eval_loss": 0.3816817443847656, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.00558649147922794 + }, + { + "epoch": 190, + "eval_accuracy": 0.9264, + "eval_loss": 0.3805133972167969, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.005222728777097331 + }, + { + "epoch": 191, + "eval_accuracy": 0.9264, + "eval_loss": 0.38058997192382815, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.004994279914928807 + }, + { + "epoch": 192, + "eval_accuracy": 0.9268, + "eval_loss": 0.3790363037109375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.005501532499823305 + }, + { + "epoch": 193, + "eval_accuracy": 0.927, + "eval_loss": 0.3807512573242188, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.005647506911224789 + }, + { + "epoch": 194, + "eval_accuracy": 0.9254, + "eval_loss": 0.37857630004882814, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.005686942210959064 + }, + { + "epoch": 195, + "eval_accuracy": 0.928, + "eval_loss": 0.3788825134277344, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.005199345403330193 + }, + { + "epoch": 196, + "eval_accuracy": 0.9272, + "eval_loss": 0.3805579833984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.005200389449753695 + }, + { + "epoch": 197, + "eval_accuracy": 0.9274, + "eval_loss": 0.38221766967773435, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.005538683751142687 + }, + { + "epoch": 198, + "eval_accuracy": 0.927, + "eval_loss": 0.38246532592773436, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.005221181215594212 + }, + { + "epoch": 199, + "eval_accuracy": 0.9274, + "eval_loss": 0.37890045166015623, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.005684674559947517 + }, + { + "epoch": 200, + "eval_accuracy": 0.928, + "eval_loss": 0.38430252685546873, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.005607284581330087 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.928, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.38430252685546873, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2330820608, + "peak_memory_reserved_bytes": 3049259008, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 33.98609757423401, + "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": 1926.7095592021942, + "train_wall_s": 1892.7023992538452 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 1239526656000000, + "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": 619763328000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 1859289984000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json.manifest.json new file mode 100644 index 0000000..c6886b4 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "bp", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "0", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787031303.6175716, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 1930.4826476573944, + "epochs": 200, + "experiment_name": "resnet-r2-bp-d32-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "bp", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d32-lr0p1.json", + "output_exists": true, + "output_sha256": "2450755a60aeebffe5376d58ac95a5a17b1b4af03202b8a09e71ca6672555cef", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json new file mode 100644 index 0000000..933fa96 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json @@ -0,0 +1,2717 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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, + "eval_accuracy": 0.4062, + "eval_loss": 1.6383858154296875, + "eval_split": "validation", + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 1.8120329790115357 + }, + { + "epoch": 2, + "eval_accuracy": 0.5926, + "eval_loss": 1.153287646484375, + "eval_split": "validation", + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.3592044618182713 + }, + { + "epoch": 3, + "eval_accuracy": 0.5084, + "eval_loss": 1.90244970703125, + "eval_split": "validation", + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.0838589077843561 + }, + { + "epoch": 4, + "eval_accuracy": 0.693, + "eval_loss": 0.906397900390625, + "eval_split": "validation", + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 0.8900402983453539 + }, + { + "epoch": 5, + "eval_accuracy": 0.7638, + "eval_loss": 0.6916922607421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.7705084403991699 + }, + { + "epoch": 6, + "eval_accuracy": 0.7054, + "eval_loss": 0.94351083984375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.6632061253759596 + }, + { + "epoch": 7, + "eval_accuracy": 0.7784, + "eval_loss": 0.6527540771484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.5979521300209893 + }, + { + "epoch": 8, + "eval_accuracy": 0.7632, + "eval_loss": 0.7341368408203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.55327621333864 + }, + { + "epoch": 9, + "eval_accuracy": 0.8116, + "eval_loss": 0.5726607421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5106070032013788 + }, + { + "epoch": 10, + "eval_accuracy": 0.7938, + "eval_loss": 0.6018488891601562, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.47956937680774264 + }, + { + "epoch": 11, + "eval_accuracy": 0.8024, + "eval_loss": 0.61379228515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.45889593749046326 + }, + { + "epoch": 12, + "eval_accuracy": 0.8038, + "eval_loss": 0.6077332397460937, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.438798218621148 + }, + { + "epoch": 13, + "eval_accuracy": 0.7766, + "eval_loss": 0.7508157470703125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.4087996634165446 + }, + { + "epoch": 14, + "eval_accuracy": 0.8144, + "eval_loss": 0.5816595092773438, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.39873176553514267 + }, + { + "epoch": 15, + "eval_accuracy": 0.84, + "eval_loss": 0.48526373901367187, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.3821409888108571 + }, + { + "epoch": 16, + "eval_accuracy": 0.832, + "eval_loss": 0.5142687622070312, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.36610251981947156 + }, + { + "epoch": 17, + "eval_accuracy": 0.8192, + "eval_loss": 0.5443266357421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.35214182429843477 + }, + { + "epoch": 18, + "eval_accuracy": 0.8386, + "eval_loss": 0.49019057006835937, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.341371633805169 + }, + { + "epoch": 19, + "eval_accuracy": 0.8502, + "eval_loss": 0.4660570251464844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.333972642771403 + }, + { + "epoch": 20, + "eval_accuracy": 0.8532, + "eval_loss": 0.43096478271484373, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.31695578266514673 + }, + { + "epoch": 21, + "eval_accuracy": 0.8168, + "eval_loss": 0.60798935546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.30790028344260323 + }, + { + "epoch": 22, + "eval_accuracy": 0.8586, + "eval_loss": 0.4445894714355469, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3030786698553297 + }, + { + "epoch": 23, + "eval_accuracy": 0.8612, + "eval_loss": 0.431785498046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.2980527818997701 + }, + { + "epoch": 24, + "eval_accuracy": 0.8438, + "eval_loss": 0.5205806335449219, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.2837291211393144 + }, + { + "epoch": 25, + "eval_accuracy": 0.863, + "eval_loss": 0.4326460021972656, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.27650729268391927 + }, + { + "epoch": 26, + "eval_accuracy": 0.8456, + "eval_loss": 0.5047505187988282, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.270019421143002 + }, + { + "epoch": 27, + "eval_accuracy": 0.8434, + "eval_loss": 0.5102612182617188, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.2668691424396303 + }, + { + "epoch": 28, + "eval_accuracy": 0.8578, + "eval_loss": 0.4414801574707031, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.25704341024292837 + }, + { + "epoch": 29, + "eval_accuracy": 0.8668, + "eval_loss": 0.4320094970703125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.2572188554790285 + }, + { + "epoch": 30, + "eval_accuracy": 0.8714, + "eval_loss": 0.408448828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.250815947622723 + }, + { + "epoch": 31, + "eval_accuracy": 0.837, + "eval_loss": 0.5562254150390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.248208172082901 + }, + { + "epoch": 32, + "eval_accuracy": 0.8584, + "eval_loss": 0.46939553833007813, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.23669359466499754 + }, + { + "epoch": 33, + "eval_accuracy": 0.864, + "eval_loss": 0.44012159423828123, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.2358193315267563 + }, + { + "epoch": 34, + "eval_accuracy": 0.8726, + "eval_loss": 0.41900897827148437, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.23047334933280944 + }, + { + "epoch": 35, + "eval_accuracy": 0.8484, + "eval_loss": 0.5120837097167968, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.22745651706324682 + }, + { + "epoch": 36, + "eval_accuracy": 0.8572, + "eval_loss": 0.48343660888671874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.2273322538084454 + }, + { + "epoch": 37, + "eval_accuracy": 0.8596, + "eval_loss": 0.4978451171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.22201022511985566 + }, + { + "epoch": 38, + "eval_accuracy": 0.8446, + "eval_loss": 0.5256515441894531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.2145939707385169 + }, + { + "epoch": 39, + "eval_accuracy": 0.8696, + "eval_loss": 0.43493318481445314, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.20739925054709116 + }, + { + "epoch": 40, + "eval_accuracy": 0.8666, + "eval_loss": 0.4600316223144531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.20801596263779534 + }, + { + "epoch": 41, + "eval_accuracy": 0.8632, + "eval_loss": 0.4669625122070313, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.20819860476387872 + }, + { + "epoch": 42, + "eval_accuracy": 0.8686, + "eval_loss": 0.4578409973144531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.19786155010196899 + }, + { + "epoch": 43, + "eval_accuracy": 0.8662, + "eval_loss": 0.46103836059570313, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.20556741065979003 + }, + { + "epoch": 44, + "eval_accuracy": 0.8394, + "eval_loss": 0.5724481323242188, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.1993922583738963 + }, + { + "epoch": 45, + "eval_accuracy": 0.8724, + "eval_loss": 0.43330186157226563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.19485141789913177 + }, + { + "epoch": 46, + "eval_accuracy": 0.8764, + "eval_loss": 0.42409732666015626, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.19267305657333797 + }, + { + "epoch": 47, + "eval_accuracy": 0.885, + "eval_loss": 0.39403966674804686, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.1875338579866621 + }, + { + "epoch": 48, + "eval_accuracy": 0.8796, + "eval_loss": 0.40434483642578123, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.18509351403978136 + }, + { + "epoch": 49, + "eval_accuracy": 0.88, + "eval_loss": 0.4242395751953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.18649417563014561 + }, + { + "epoch": 50, + "eval_accuracy": 0.8704, + "eval_loss": 0.4550154113769531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.18717047963407304 + }, + { + "epoch": 51, + "eval_accuracy": 0.8744, + "eval_loss": 0.44483524169921873, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.1796641605509652 + }, + { + "epoch": 52, + "eval_accuracy": 0.8818, + "eval_loss": 0.40557909545898435, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.17563802019225228 + }, + { + "epoch": 53, + "eval_accuracy": 0.8742, + "eval_loss": 0.42595355224609377, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.17193800271352133 + }, + { + "epoch": 54, + "eval_accuracy": 0.8312, + "eval_loss": 0.6097184692382812, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.17408619323306612 + }, + { + "epoch": 55, + "eval_accuracy": 0.863, + "eval_loss": 0.47159163208007815, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.17443495961295233 + }, + { + "epoch": 56, + "eval_accuracy": 0.864, + "eval_loss": 0.47021598510742185, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.17730448887348174 + }, + { + "epoch": 57, + "eval_accuracy": 0.845, + "eval_loss": 0.5302630187988281, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.17513225899802315 + }, + { + "epoch": 58, + "eval_accuracy": 0.8764, + "eval_loss": 0.45107080078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.16688513896730212 + }, + { + "epoch": 59, + "eval_accuracy": 0.8506, + "eval_loss": 0.5074487365722656, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.17318025396664938 + }, + { + "epoch": 60, + "eval_accuracy": 0.872, + "eval_loss": 0.43073905639648435, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.16225387883716158 + }, + { + "epoch": 61, + "eval_accuracy": 0.8646, + "eval_loss": 0.4888678771972656, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.1628989502257771 + }, + { + "epoch": 62, + "eval_accuracy": 0.89, + "eval_loss": 0.3716917114257812, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.16541327353583443 + }, + { + "epoch": 63, + "eval_accuracy": 0.867, + "eval_loss": 0.48051559448242187, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.15507885179254743 + }, + { + "epoch": 64, + "eval_accuracy": 0.8856, + "eval_loss": 0.4047019226074219, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.15761819744375016 + }, + { + "epoch": 65, + "eval_accuracy": 0.8712, + "eval_loss": 0.4679113525390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.16318066529962752 + }, + { + "epoch": 66, + "eval_accuracy": 0.869, + "eval_loss": 0.46758770141601563, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.15842568448649513 + }, + { + "epoch": 67, + "eval_accuracy": 0.8764, + "eval_loss": 0.47932765502929686, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.15360924773481158 + }, + { + "epoch": 68, + "eval_accuracy": 0.8536, + "eval_loss": 0.52894326171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.1567973585764567 + }, + { + "epoch": 69, + "eval_accuracy": 0.8874, + "eval_loss": 0.3929457092285156, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.15907586552831862 + }, + { + "epoch": 70, + "eval_accuracy": 0.8816, + "eval_loss": 0.42247525634765626, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.15716197686725192 + }, + { + "epoch": 71, + "eval_accuracy": 0.8732, + "eval_loss": 0.4963585998535156, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.14983130187855825 + }, + { + "epoch": 72, + "eval_accuracy": 0.8732, + "eval_loss": 0.46758525390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.15397654530207316 + }, + { + "epoch": 73, + "eval_accuracy": 0.8784, + "eval_loss": 0.43674370727539064, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.14723493348757427 + }, + { + "epoch": 74, + "eval_accuracy": 0.8886, + "eval_loss": 0.39118277587890626, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.14652091862890457 + }, + { + "epoch": 75, + "eval_accuracy": 0.8752, + "eval_loss": 0.4687583251953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.1545022050380707 + }, + { + "epoch": 76, + "eval_accuracy": 0.876, + "eval_loss": 0.4439466125488281, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.14616850409242843 + }, + { + "epoch": 77, + "eval_accuracy": 0.8882, + "eval_loss": 0.39720289306640627, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.1465005384021335 + }, + { + "epoch": 78, + "eval_accuracy": 0.887, + "eval_loss": 0.4100860900878906, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.15173963310188718 + }, + { + "epoch": 79, + "eval_accuracy": 0.8672, + "eval_loss": 0.4720408081054687, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.1440488050142924 + }, + { + "epoch": 80, + "eval_accuracy": 0.8848, + "eval_loss": 0.4128118469238281, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.14068703662686877 + }, + { + "epoch": 81, + "eval_accuracy": 0.8622, + "eval_loss": 0.4955517456054688, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.14217469872898525 + }, + { + "epoch": 82, + "eval_accuracy": 0.8926, + "eval_loss": 0.3864740173339844, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.14880184517701467 + }, + { + "epoch": 83, + "eval_accuracy": 0.8698, + "eval_loss": 0.5053119140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.1299323884540134 + }, + { + "epoch": 84, + "eval_accuracy": 0.8806, + "eval_loss": 0.4496087341308594, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.146550003431903 + }, + { + "epoch": 85, + "eval_accuracy": 0.8928, + "eval_loss": 0.40095016479492185, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.14196164631048838 + }, + { + "epoch": 86, + "eval_accuracy": 0.8936, + "eval_loss": 0.38955796508789065, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.13579268198278216 + }, + { + "epoch": 87, + "eval_accuracy": 0.8726, + "eval_loss": 0.4710050598144531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.14606884476078882 + }, + { + "epoch": 88, + "eval_accuracy": 0.8852, + "eval_loss": 0.40862398071289063, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.14596923381487528 + }, + { + "epoch": 89, + "eval_accuracy": 0.8938, + "eval_loss": 0.37538721313476564, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.13962470082706877 + }, + { + "epoch": 90, + "eval_accuracy": 0.8798, + "eval_loss": 0.45380982666015623, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.136911416721344 + }, + { + "epoch": 91, + "eval_accuracy": 0.8842, + "eval_loss": 0.4425386474609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.13629336178965037 + }, + { + "epoch": 92, + "eval_accuracy": 0.8908, + "eval_loss": 0.419219775390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.1369322536945343 + }, + { + "epoch": 93, + "eval_accuracy": 0.893, + "eval_loss": 0.3760003112792969, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.13725584357579548 + }, + { + "epoch": 94, + "eval_accuracy": 0.8704, + "eval_loss": 0.4654718566894531, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.13652064226733313 + }, + { + "epoch": 95, + "eval_accuracy": 0.8868, + "eval_loss": 0.44302418212890626, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.12910948644744025 + }, + { + "epoch": 96, + "eval_accuracy": 0.8864, + "eval_loss": 0.42267026977539063, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.13443772603140938 + }, + { + "epoch": 97, + "eval_accuracy": 0.8926, + "eval_loss": 0.3810096862792969, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.13513379742834303 + }, + { + "epoch": 98, + "eval_accuracy": 0.8934, + "eval_loss": 0.3915801513671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.12859395507971447 + }, + { + "epoch": 99, + "eval_accuracy": 0.8918, + "eval_loss": 0.4213328430175781, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.13369986132780712 + }, + { + "epoch": 100, + "eval_accuracy": 0.886, + "eval_loss": 0.45385445556640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.13087427303592364 + }, + { + "epoch": 101, + "eval_accuracy": 0.9204, + "eval_loss": 0.2958070129394531, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.07391229368348916 + }, + { + "epoch": 102, + "eval_accuracy": 0.9242, + "eval_loss": 0.28999569702148437, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.04571805165542497 + }, + { + "epoch": 103, + "eval_accuracy": 0.9242, + "eval_loss": 0.2971516357421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.036192750480605496 + }, + { + "epoch": 104, + "eval_accuracy": 0.9262, + "eval_loss": 0.3033703857421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.03154912928475274 + }, + { + "epoch": 105, + "eval_accuracy": 0.9256, + "eval_loss": 0.30697164306640623, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.02741032799548573 + }, + { + "epoch": 106, + "eval_accuracy": 0.9268, + "eval_loss": 0.30802741088867186, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.02428152202632692 + }, + { + "epoch": 107, + "eval_accuracy": 0.9262, + "eval_loss": 0.314338623046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.021879386997885174 + }, + { + "epoch": 108, + "eval_accuracy": 0.925, + "eval_loss": 0.3235135375976563, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.020878948243459067 + }, + { + "epoch": 109, + "eval_accuracy": 0.9252, + "eval_loss": 0.3302208251953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.01850255666706297 + }, + { + "epoch": 110, + "eval_accuracy": 0.9262, + "eval_loss": 0.32489039916992185, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.01726404068602456 + }, + { + "epoch": 111, + "eval_accuracy": 0.9268, + "eval_loss": 0.3281711608886719, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.016243692442443634 + }, + { + "epoch": 112, + "eval_accuracy": 0.9268, + "eval_loss": 0.3317916564941406, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.014255007772313224 + }, + { + "epoch": 113, + "eval_accuracy": 0.9268, + "eval_loss": 0.3346408752441406, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.014053582479059697 + }, + { + "epoch": 114, + "eval_accuracy": 0.9258, + "eval_loss": 0.3327147705078125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.013204236159059737 + }, + { + "epoch": 115, + "eval_accuracy": 0.9264, + "eval_loss": 0.3379136962890625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.012523622225059403 + }, + { + "epoch": 116, + "eval_accuracy": 0.925, + "eval_loss": 0.3409310119628906, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.011970065443052185 + }, + { + "epoch": 117, + "eval_accuracy": 0.9276, + "eval_loss": 0.341762646484375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.010813051444954343 + }, + { + "epoch": 118, + "eval_accuracy": 0.9268, + "eval_loss": 0.338385888671875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.011047906530731254 + }, + { + "epoch": 119, + "eval_accuracy": 0.924, + "eval_loss": 0.3458584655761719, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.01061292423490021 + }, + { + "epoch": 120, + "eval_accuracy": 0.9268, + "eval_loss": 0.34402819213867186, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.009872842174685663 + }, + { + "epoch": 121, + "eval_accuracy": 0.9262, + "eval_loss": 0.3493848388671875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.008324120312101311 + }, + { + "epoch": 122, + "eval_accuracy": 0.9276, + "eval_loss": 0.3444608703613281, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.00849333224710491 + }, + { + "epoch": 123, + "eval_accuracy": 0.9264, + "eval_loss": 0.3520544494628906, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.008254868710206614 + }, + { + "epoch": 124, + "eval_accuracy": 0.9262, + "eval_loss": 0.3435565612792969, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.007909329082051084 + }, + { + "epoch": 125, + "eval_accuracy": 0.9258, + "eval_loss": 0.35690042114257814, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.008038395318388939 + }, + { + "epoch": 126, + "eval_accuracy": 0.9262, + "eval_loss": 0.35696865234375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.007849043796294265 + }, + { + "epoch": 127, + "eval_accuracy": 0.9256, + "eval_loss": 0.369244384765625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.008171430269297626 + }, + { + "epoch": 128, + "eval_accuracy": 0.924, + "eval_loss": 0.3637481201171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.007164580019977358 + }, + { + "epoch": 129, + "eval_accuracy": 0.9276, + "eval_loss": 0.35849393920898437, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.007030607246607542 + }, + { + "epoch": 130, + "eval_accuracy": 0.9268, + "eval_loss": 0.3662152770996094, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.006088745883686675 + }, + { + "epoch": 131, + "eval_accuracy": 0.9274, + "eval_loss": 0.355767724609375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.006438479425261418 + }, + { + "epoch": 132, + "eval_accuracy": 0.9272, + "eval_loss": 0.3704664978027344, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.006546230017228259 + }, + { + "epoch": 133, + "eval_accuracy": 0.9276, + "eval_loss": 0.37374725952148435, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.005257100144318409 + }, + { + "epoch": 134, + "eval_accuracy": 0.9288, + "eval_loss": 0.3644460998535156, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.005803062219565941 + }, + { + "epoch": 135, + "eval_accuracy": 0.9268, + "eval_loss": 0.3769530700683594, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.0059276600245075925 + }, + { + "epoch": 136, + "eval_accuracy": 0.9252, + "eval_loss": 0.3732922607421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.005344935539116462 + }, + { + "epoch": 137, + "eval_accuracy": 0.9256, + "eval_loss": 0.3735669189453125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.006024956390758355 + }, + { + "epoch": 138, + "eval_accuracy": 0.9274, + "eval_loss": 0.37350067749023436, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.00587833823416796 + }, + { + "epoch": 139, + "eval_accuracy": 0.93, + "eval_loss": 0.3658798034667969, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.005202372343217333 + }, + { + "epoch": 140, + "eval_accuracy": 0.928, + "eval_loss": 0.37819749145507814, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.00522608685038156 + }, + { + "epoch": 141, + "eval_accuracy": 0.9308, + "eval_loss": 0.3692561706542969, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.005239222901273105 + }, + { + "epoch": 142, + "eval_accuracy": 0.9276, + "eval_loss": 0.37637430419921875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.004970019086450338 + }, + { + "epoch": 143, + "eval_accuracy": 0.9266, + "eval_loss": 0.3803944091796875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.004805412318226364 + }, + { + "epoch": 144, + "eval_accuracy": 0.9282, + "eval_loss": 0.3740898742675781, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.004170282630084289 + }, + { + "epoch": 145, + "eval_accuracy": 0.9276, + "eval_loss": 0.38094301147460935, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.004932723968476057 + }, + { + "epoch": 146, + "eval_accuracy": 0.9262, + "eval_loss": 0.37821016845703126, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.00465814622434684 + }, + { + "epoch": 147, + "eval_accuracy": 0.9282, + "eval_loss": 0.374492822265625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.004465697120130062 + }, + { + "epoch": 148, + "eval_accuracy": 0.9272, + "eval_loss": 0.37554210205078126, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.0049913340316878425 + }, + { + "epoch": 149, + "eval_accuracy": 0.9274, + "eval_loss": 0.3821321533203125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.004511870892842611 + }, + { + "epoch": 150, + "eval_accuracy": 0.9278, + "eval_loss": 0.383819873046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.0036155341097464165 + }, + { + "epoch": 151, + "eval_accuracy": 0.9266, + "eval_loss": 0.3865296875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.0034015125179870263 + }, + { + "epoch": 152, + "eval_accuracy": 0.93, + "eval_loss": 0.38060599365234377, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.0038181717752996417 + }, + { + "epoch": 153, + "eval_accuracy": 0.928, + "eval_loss": 0.38813572998046875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.003342558716175457 + }, + { + "epoch": 154, + "eval_accuracy": 0.9288, + "eval_loss": 0.37930936889648437, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.0037702425530387294 + }, + { + "epoch": 155, + "eval_accuracy": 0.9294, + "eval_loss": 0.37951343994140624, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.0035697890337970523 + }, + { + "epoch": 156, + "eval_accuracy": 0.929, + "eval_loss": 0.37994603881835937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.0036909519653353425 + }, + { + "epoch": 157, + "eval_accuracy": 0.9288, + "eval_loss": 0.38521229858398437, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.0035891185532013576 + }, + { + "epoch": 158, + "eval_accuracy": 0.9296, + "eval_loss": 0.38098255004882814, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0032051278701103808 + }, + { + "epoch": 159, + "eval_accuracy": 0.93, + "eval_loss": 0.3787800048828125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.003374717134113113 + }, + { + "epoch": 160, + "eval_accuracy": 0.9296, + "eval_loss": 0.38265086059570314, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.002994217939178149 + }, + { + "epoch": 161, + "eval_accuracy": 0.9286, + "eval_loss": 0.3790346862792969, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.003431437090711875 + }, + { + "epoch": 162, + "eval_accuracy": 0.929, + "eval_loss": 0.3793521911621094, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.0029321252649649977 + }, + { + "epoch": 163, + "eval_accuracy": 0.9296, + "eval_loss": 0.3795800048828125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.0036018194581071534 + }, + { + "epoch": 164, + "eval_accuracy": 0.929, + "eval_loss": 0.3825399169921875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.0027656607893606026 + }, + { + "epoch": 165, + "eval_accuracy": 0.928, + "eval_loss": 0.37968671264648435, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.0026229754688011277 + }, + { + "epoch": 166, + "eval_accuracy": 0.929, + "eval_loss": 0.38276746826171876, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.0029054207376307912 + }, + { + "epoch": 167, + "eval_accuracy": 0.9292, + "eval_loss": 0.3794607421875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.003060741257626149 + }, + { + "epoch": 168, + "eval_accuracy": 0.929, + "eval_loss": 0.38132125244140624, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.002886012737949689 + }, + { + "epoch": 169, + "eval_accuracy": 0.929, + "eval_loss": 0.3822179992675781, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.003108068531089359 + }, + { + "epoch": 170, + "eval_accuracy": 0.9292, + "eval_loss": 0.38042718505859374, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.0029554093821181193 + }, + { + "epoch": 171, + "eval_accuracy": 0.929, + "eval_loss": 0.383385693359375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.0030305214785453344 + }, + { + "epoch": 172, + "eval_accuracy": 0.9304, + "eval_loss": 0.38013319091796877, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.0031653479740954937 + }, + { + "epoch": 173, + "eval_accuracy": 0.9288, + "eval_loss": 0.38019481811523437, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.0032616001981827947 + }, + { + "epoch": 174, + "eval_accuracy": 0.9294, + "eval_loss": 0.3823200317382813, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.0026846566768984 + }, + { + "epoch": 175, + "eval_accuracy": 0.9294, + "eval_loss": 0.38054329833984374, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.0028761062923197945 + }, + { + "epoch": 176, + "eval_accuracy": 0.9294, + "eval_loss": 0.37557706298828125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.002718526501969124 + }, + { + "epoch": 177, + "eval_accuracy": 0.9296, + "eval_loss": 0.38403538208007815, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.002765856211798059 + }, + { + "epoch": 178, + "eval_accuracy": 0.9288, + "eval_loss": 0.3827006591796875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.002846009177507626 + }, + { + "epoch": 179, + "eval_accuracy": 0.9304, + "eval_loss": 0.37569268798828126, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.0028106547514183656 + }, + { + "epoch": 180, + "eval_accuracy": 0.9308, + "eval_loss": 0.37752535400390624, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.002670673717599776 + }, + { + "epoch": 181, + "eval_accuracy": 0.9294, + "eval_loss": 0.37972734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.0029900958141932887 + }, + { + "epoch": 182, + "eval_accuracy": 0.93, + "eval_loss": 0.3836199890136719, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.0028344500539824367 + }, + { + "epoch": 183, + "eval_accuracy": 0.9298, + "eval_loss": 0.382769482421875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.002919924992824801 + }, + { + "epoch": 184, + "eval_accuracy": 0.93, + "eval_loss": 0.38172957153320314, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.0029668405563053157 + }, + { + "epoch": 185, + "eval_accuracy": 0.9286, + "eval_loss": 0.37836466674804686, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.002790116118432747 + }, + { + "epoch": 186, + "eval_accuracy": 0.9302, + "eval_loss": 0.3797643493652344, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.002748481237143278 + }, + { + "epoch": 187, + "eval_accuracy": 0.9288, + "eval_loss": 0.3819689208984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.002598481794156962 + }, + { + "epoch": 188, + "eval_accuracy": 0.929, + "eval_loss": 0.38522958984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.0023670312017202376 + }, + { + "epoch": 189, + "eval_accuracy": 0.9284, + "eval_loss": 0.38395161743164063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.0026725557989130417 + }, + { + "epoch": 190, + "eval_accuracy": 0.9278, + "eval_loss": 0.3851235168457031, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.002891599158321818 + }, + { + "epoch": 191, + "eval_accuracy": 0.9292, + "eval_loss": 0.38445775756835937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.0026930961543590658 + }, + { + "epoch": 192, + "eval_accuracy": 0.9286, + "eval_loss": 0.38256485595703127, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.002824814874927203 + }, + { + "epoch": 193, + "eval_accuracy": 0.93, + "eval_loss": 0.38405804443359376, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.002977210904844105 + }, + { + "epoch": 194, + "eval_accuracy": 0.93, + "eval_loss": 0.3784375061035156, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.0028413962472023237 + }, + { + "epoch": 195, + "eval_accuracy": 0.93, + "eval_loss": 0.3832097351074219, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.0030580059261837355 + }, + { + "epoch": 196, + "eval_accuracy": 0.9294, + "eval_loss": 0.37996414184570315, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.0026476701118155485 + }, + { + "epoch": 197, + "eval_accuracy": 0.9292, + "eval_loss": 0.38070816650390626, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.0029351543212205997 + }, + { + "epoch": 198, + "eval_accuracy": 0.9298, + "eval_loss": 0.3816475341796875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.0026628320027556685 + }, + { + "epoch": 199, + "eval_accuracy": 0.9306, + "eval_loss": 0.3811450012207031, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.002589297404181626 + }, + { + "epoch": 200, + "eval_accuracy": 0.93, + "eval_loss": 0.3817935302734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.002858151467392842 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.93, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.3817935302734375, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 3257588736, + "peak_memory_reserved_bytes": 4322230272, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 60.09430003166199, + "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": 3323.5966560840607, + "train_wall_s": 3263.4807090759277 + }, + "work": { + "apical_macs_per_example": 0, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 2258742528000000, + "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": 1129371264000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 3388113792000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json.manifest.json new file mode 100644 index 0000000..f17443f --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "bp", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "0", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787024180.1494296, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 3326.933441400528, + "epochs": 200, + "experiment_name": "resnet-r2-bp-d56-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "bp", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-bp-d56-lr0p1.json", + "output_exists": true, + "output_sha256": "e4876181a6d1d8ddbda3b439319a758b9c9e42170add0a78f72df5378abd4088", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json new file mode 100644 index 0000000..b440202 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json @@ -0,0 +1,3869 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 267904, + "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": 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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.9992254873116811, + "feedback_forward_cosine": [ + 0.9996554255485535, + 0.9996746182441711, + 0.9996549487113953, + 0.9996075630187988, + 0.9996635913848877, + 0.9995270371437073, + 0.9995354413986206, + 0.9996858835220337, + 0.9996363520622253, + 0.9995448589324951, + 0.999627411365509, + 0.9995063543319702, + 0.9995788335800171, + 0.999750554561615, + 0.9997305870056152, + 0.999675989151001, + 0.9997004270553589, + 0.9992839097976685, + 0.9999722242355347 + ], + "feedback_forward_norm_ratio": [ + 0.9991434216499329, + 1.0004491806030273, + 0.9987601041793823, + 1.0013868808746338, + 0.999297022819519, + 1.0023059844970703, + 0.9987834692001343, + 1.0002436637878418, + 0.9993070363998413, + 1.0014671087265015, + 0.9994328618049622, + 1.0015302896499634, + 0.9993026256561279, + 1.0005580186843872, + 1.0002413988113403, + 1.0010871887207031, + 1.0002120733261108, + 1.0017412900924683, + 1.0046420097351074 + ], + "feedback_forward_relative_error": [ + 0.02625316195189953, + 0.025520561262965202, + 0.026284009218215942, + 0.028065064921975136, + 0.025940094143152237, + 0.03087770752608776, + 0.03048705868422985, + 0.02506886050105095, + 0.026966994628310204, + 0.030226128175854683, + 0.02730029635131359, + 0.03148436173796654, + 0.029020819813013077, + 0.022351281717419624, + 0.02321099117398262, + 0.02549651823937893, + 0.024476630613207817, + 0.03791650012135506, + 0.008795272558927536 + ], + "innovation_negative_gradient_cosine": [ + 0.9992389678955078, + 0.9991669058799744, + 0.999206006526947, + 0.9992231726646423, + 0.9992378950119019, + 0.9992799758911133, + 0.9992815256118774, + 0.9993095397949219, + 0.999367356300354, + 0.9992690086364746, + 0.9994315505027771, + 0.999388575553894, + 0.9994919300079346, + 0.9995896816253662, + 0.9996410608291626, + 0.999697208404541, + 0.9998255968093872, + 0.9999359846115112, + 0.8563728332519531 + ], + "mean_feedback_forward_cosine": 0.9996322111079567, + "mean_feedback_forward_relative_error": 0.026618016491595068, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9992389678955078, + 0.9991669058799744, + 0.999206006526947, + 0.9992231726646423, + 0.9992378950119019, + 0.9992799758911133, + 0.9992815256118774, + 0.9993095397949219, + 0.999367356300354, + 0.9992690086364746, + 0.9994315505027771, + 0.999388575553894, + 0.9994919300079346, + 0.9995896816253662, + 0.9996410608291626, + 0.999697208404541, + 0.9998255968093872, + 0.9999359846115112, + 0.8563728332519531 + ], + "teaching_negative_gradient_cosine": [ + 0.9992389678955078, + 0.9991669058799744, + 0.999206006526947, + 0.9992231726646423, + 0.9992378950119019, + 0.9992799758911133, + 0.9992815256118774, + 0.9993095397949219, + 0.999367356300354, + 0.9992690086364746, + 0.9994315505027771, + 0.999388575553894, + 0.9994919300079346, + 0.9995896816253662, + 0.9996410608291626, + 0.999697208404541, + 0.9998255968093872, + 0.9999359846115112, + 0.8563728332519531 + ], + "wall_s": 0.06803202629089355 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.344, + "eval_loss": 1.81885517578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.396217942237854, + "mean_feedback_forward_cosine": 0.12113288064536296, + "mean_feedback_forward_relative_error": 1.3444790902890658, + "min_feedback_forward_cosine": 0.037865087389945984 + }, + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 1.9119543294694687 + }, + { + "epoch": 2, + "eval_accuracy": 0.4864, + "eval_loss": 1.565498681640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3504620790481567, + "mean_feedback_forward_cosine": 0.27226419001817703, + "mean_feedback_forward_relative_error": 1.222043868742491, + "min_feedback_forward_cosine": 0.11271074414253235 + }, + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.4822090652677749 + }, + { + "epoch": 3, + "eval_accuracy": 0.523, + "eval_loss": 1.6564058837890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3010425567626953, + "mean_feedback_forward_cosine": 0.39278248579878555, + "mean_feedback_forward_relative_error": 1.11243883873287, + "min_feedback_forward_cosine": 0.18469145894050598 + }, + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.19166419444614 + }, + { + "epoch": 4, + "eval_accuracy": 0.6216, + "eval_loss": 1.09284404296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.246029257774353, + "mean_feedback_forward_cosine": 0.49751683755924825, + "mean_feedback_forward_relative_error": 1.0094205486147028, + "min_feedback_forward_cosine": 0.25789177417755127 + }, + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 1.0127930777443779 + }, + { + "epoch": 5, + "eval_accuracy": 0.7116, + "eval_loss": 0.8747228881835938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1837794780731201, + "mean_feedback_forward_cosine": 0.5858596798620725, + "mean_feedback_forward_relative_error": 0.9147494780389887, + "min_feedback_forward_cosine": 0.3337680697441101 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.884297638130188 + }, + { + "epoch": 6, + "eval_accuracy": 0.6672, + "eval_loss": 0.985279248046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1347962617874146, + "mean_feedback_forward_cosine": 0.6410800830314034, + "mean_feedback_forward_relative_error": 0.8509495085791537, + "min_feedback_forward_cosine": 0.38860657811164856 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7689341519885593 + }, + { + "epoch": 7, + "eval_accuracy": 0.745, + "eval_loss": 0.7586026489257812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0899189710617065, + "mean_feedback_forward_cosine": 0.6840777710864419, + "mean_feedback_forward_relative_error": 0.798096848161597, + "min_feedback_forward_cosine": 0.4376838803291321 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.7034475085046556 + }, + { + "epoch": 8, + "eval_accuracy": 0.702, + "eval_loss": 1.04657080078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0519425868988037, + "mean_feedback_forward_cosine": 0.7168333953932712, + "mean_feedback_forward_relative_error": 0.7551552151378832, + "min_feedback_forward_cosine": 0.4767589867115021 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.6440844105932447 + }, + { + "epoch": 9, + "eval_accuracy": 0.7628, + "eval_loss": 0.7188642578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0166592597961426, + "mean_feedback_forward_cosine": 0.744070269559559, + "mean_feedback_forward_relative_error": 0.7177636921405792, + "min_feedback_forward_cosine": 0.5122615694999695 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5978653068330553 + }, + { + "epoch": 10, + "eval_accuracy": 0.7684, + "eval_loss": 0.7278686767578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9829434752464294, + "mean_feedback_forward_cosine": 0.7671844739662973, + "mean_feedback_forward_relative_error": 0.6843707843830711, + "min_feedback_forward_cosine": 0.5441762208938599 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5641649359703064 + }, + { + "epoch": 11, + "eval_accuracy": 0.7844, + "eval_loss": 0.6429546142578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9501974582672119, + "mean_feedback_forward_cosine": 0.787681868201808, + "mean_feedback_forward_relative_error": 0.6532509734756068, + "min_feedback_forward_cosine": 0.574102520942688 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.5336678455458747 + }, + { + "epoch": 12, + "eval_accuracy": 0.7888, + "eval_loss": 0.6300875366210937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.918483316898346, + "mean_feedback_forward_cosine": 0.8054801445258292, + "mean_feedback_forward_relative_error": 0.6250106943281073, + "min_feedback_forward_cosine": 0.6022139191627502 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.5122270748244392 + }, + { + "epoch": 13, + "eval_accuracy": 0.6694, + "eval_loss": 1.1285370361328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8890088796615601, + "mean_feedback_forward_cosine": 0.8212189956715232, + "mean_feedback_forward_relative_error": 0.5988754441863612, + "min_feedback_forward_cosine": 0.6274921894073486 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.4805156542142232 + }, + { + "epoch": 14, + "eval_accuracy": 0.8094, + "eval_loss": 0.5873873657226563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.859317421913147, + "mean_feedback_forward_cosine": 0.8354297970470629, + "mean_feedback_forward_relative_error": 0.5742914959004051, + "min_feedback_forward_cosine": 0.651719868183136 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.4709179120169746 + }, + { + "epoch": 15, + "eval_accuracy": 0.8018, + "eval_loss": 0.631718359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8302932977676392, + "mean_feedback_forward_cosine": 0.8483938072857103, + "mean_feedback_forward_relative_error": 0.550973863978135, + "min_feedback_forward_cosine": 0.674503743648529 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.44848755186398825 + }, + { + "epoch": 16, + "eval_accuracy": 0.8008, + "eval_loss": 0.6399617309570312, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8032397627830505, + "mean_feedback_forward_cosine": 0.8598326538738451, + "mean_feedback_forward_relative_error": 0.529568424350337, + "min_feedback_forward_cosine": 0.6950962543487549 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.4364324880705939 + }, + { + "epoch": 17, + "eval_accuracy": 0.8222, + "eval_loss": 0.5439112243652344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7772430777549744, + "mean_feedback_forward_cosine": 0.8702934415716874, + "mean_feedback_forward_relative_error": 0.5092110178972545, + "min_feedback_forward_cosine": 0.7146843671798706 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.4180992658085293 + }, + { + "epoch": 18, + "eval_accuracy": 0.8316, + "eval_loss": 0.5281030395507812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7511368989944458, + "mean_feedback_forward_cosine": 0.8800302555686549, + "mean_feedback_forward_relative_error": 0.48952555499578776, + "min_feedback_forward_cosine": 0.7333462238311768 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.4113643139521281 + }, + { + "epoch": 19, + "eval_accuracy": 0.834, + "eval_loss": 0.5129178466796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7260403633117676, + "mean_feedback_forward_cosine": 0.8888662520207857, + "mean_feedback_forward_relative_error": 0.47090583176989304, + "min_feedback_forward_cosine": 0.7505145072937012 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.398461009144783 + }, + { + "epoch": 20, + "eval_accuracy": 0.8144, + "eval_loss": 0.5892133605957032, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7021915912628174, + "mean_feedback_forward_cosine": 0.8966989234874123, + "mean_feedback_forward_relative_error": 0.45369475923086466, + "min_feedback_forward_cosine": 0.7663941979408264 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3833774073653751 + }, + { + "epoch": 21, + "eval_accuracy": 0.8162, + "eval_loss": 0.6017692626953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6776108145713806, + "mean_feedback_forward_cosine": 0.9041856401845029, + "mean_feedback_forward_relative_error": 0.4366687630352221, + "min_feedback_forward_cosine": 0.7823076248168945 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.373022927708096 + }, + { + "epoch": 22, + "eval_accuracy": 0.7892, + "eval_loss": 0.7000044067382812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6551166772842407, + "mean_feedback_forward_cosine": 0.911015090189482, + "mean_feedback_forward_relative_error": 0.4206081620956722, + "min_feedback_forward_cosine": 0.796286940574646 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.368713469309277 + }, + { + "epoch": 23, + "eval_accuracy": 0.8424, + "eval_loss": 0.4894224792480469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6320359110832214, + "mean_feedback_forward_cosine": 0.9173811485892848, + "mean_feedback_forward_relative_error": 0.4050847472328889, + "min_feedback_forward_cosine": 0.8102049827575684 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.36200492996639677 + }, + { + "epoch": 24, + "eval_accuracy": 0.8508, + "eval_loss": 0.4847625244140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6104835271835327, + "mean_feedback_forward_cosine": 0.923206090927124, + "mean_feedback_forward_relative_error": 0.39038152600589554, + "min_feedback_forward_cosine": 0.8226820230484009 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.35031071446206835 + }, + { + "epoch": 25, + "eval_accuracy": 0.8472, + "eval_loss": 0.48265894165039064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5895683169364929, + "mean_feedback_forward_cosine": 0.9285419740174946, + "mean_feedback_forward_relative_error": 0.37639613763282176, + "min_feedback_forward_cosine": 0.8343791961669922 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.3469265750249227 + }, + { + "epoch": 26, + "eval_accuracy": 0.8544, + "eval_loss": 0.44391314697265627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5687019228935242, + "mean_feedback_forward_cosine": 0.9335442655964902, + "mean_feedback_forward_relative_error": 0.3628646017689454, + "min_feedback_forward_cosine": 0.8457885384559631 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.34203060767385696 + }, + { + "epoch": 27, + "eval_accuracy": 0.7932, + "eval_loss": 0.6590688598632812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.549074113368988, + "mean_feedback_forward_cosine": 0.9380777917410198, + "mean_feedback_forward_relative_error": 0.35010523309833125, + "min_feedback_forward_cosine": 0.8560115098953247 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.3337548480987549 + }, + { + "epoch": 28, + "eval_accuracy": 0.8348, + "eval_loss": 0.5370311096191406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5305789709091187, + "mean_feedback_forward_cosine": 0.9421965573963366, + "mean_feedback_forward_relative_error": 0.338118544534633, + "min_feedback_forward_cosine": 0.8653731942176819 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.3222589003774855 + }, + { + "epoch": 29, + "eval_accuracy": 0.8502, + "eval_loss": 0.4641535888671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5120760798454285, + "mean_feedback_forward_cosine": 0.9460670446094713, + "mean_feedback_forward_relative_error": 0.32643518400819677, + "min_feedback_forward_cosine": 0.8744317293167114 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.31773917949464586 + }, + { + "epoch": 30, + "eval_accuracy": 0.8528, + "eval_loss": 0.4488963806152344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.49475499987602234, + "mean_feedback_forward_cosine": 0.9497027867718747, + "mean_feedback_forward_relative_error": 0.3151285601289649, + "min_feedback_forward_cosine": 0.882604718208313 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.3195970536020067 + }, + { + "epoch": 31, + "eval_accuracy": 0.83, + "eval_loss": 0.5759987670898438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.47704029083251953, + "mean_feedback_forward_cosine": 0.9531683921813965, + "mean_feedback_forward_relative_error": 0.3039606629233611, + "min_feedback_forward_cosine": 0.8906716704368591 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.312859878677792 + }, + { + "epoch": 32, + "eval_accuracy": 0.8274, + "eval_loss": 0.5443523193359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.46091774106025696, + "mean_feedback_forward_cosine": 0.9562506518865886, + "mean_feedback_forward_relative_error": 0.2936018642626311, + "min_feedback_forward_cosine": 0.8978977203369141 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.3004345349841648 + }, + { + "epoch": 33, + "eval_accuracy": 0.8412, + "eval_loss": 0.5110523986816407, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.44484663009643555, + "mean_feedback_forward_cosine": 0.9592594598468981, + "mean_feedback_forward_relative_error": 0.28327554934903193, + "min_feedback_forward_cosine": 0.904775857925415 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.3016047252337138 + }, + { + "epoch": 34, + "eval_accuracy": 0.8304, + "eval_loss": 0.5683611450195313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42912110686302185, + "mean_feedback_forward_cosine": 0.9620242558027569, + "mean_feedback_forward_relative_error": 0.2734151631593704, + "min_feedback_forward_cosine": 0.9112752079963684 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.29844820080068374 + }, + { + "epoch": 35, + "eval_accuracy": 0.8534, + "eval_loss": 0.47788355712890623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4136193096637726, + "mean_feedback_forward_cosine": 0.9646216285856146, + "mean_feedback_forward_relative_error": 0.26376486295147944, + "min_feedback_forward_cosine": 0.9175251126289368 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.29182118734253776 + }, + { + "epoch": 36, + "eval_accuracy": 0.829, + "eval_loss": 0.5739090393066406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3991897404193878, + "mean_feedback_forward_cosine": 0.9669801373230783, + "mean_feedback_forward_relative_error": 0.2547991220888339, + "min_feedback_forward_cosine": 0.9230731725692749 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.2934884912861718 + }, + { + "epoch": 37, + "eval_accuracy": 0.8344, + "eval_loss": 0.5700015991210937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3849577009677887, + "mean_feedback_forward_cosine": 0.9692228216873972, + "mean_feedback_forward_relative_error": 0.24591049551963806, + "min_feedback_forward_cosine": 0.9283632636070251 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.28581399149894715 + }, + { + "epoch": 38, + "eval_accuracy": 0.8496, + "eval_loss": 0.46519459228515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3715720474720001, + "mean_feedback_forward_cosine": 0.9713154152819985, + "mean_feedback_forward_relative_error": 0.2373458535263413, + "min_feedback_forward_cosine": 0.9331613779067993 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.2865924285676744 + }, + { + "epoch": 39, + "eval_accuracy": 0.8558, + "eval_loss": 0.46593084716796873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3588463366031647, + "mean_feedback_forward_cosine": 0.9731747727645071, + "mean_feedback_forward_relative_error": 0.22942296886130384, + "min_feedback_forward_cosine": 0.937605619430542 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.2737637662622664 + }, + { + "epoch": 40, + "eval_accuracy": 0.8636, + "eval_loss": 0.44457278442382814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3455275595188141, + "mean_feedback_forward_cosine": 0.9750205466621801, + "mean_feedback_forward_relative_error": 0.2213271828858476, + "min_feedback_forward_cosine": 0.9420970678329468 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.27786194910473294 + }, + { + "epoch": 41, + "eval_accuracy": 0.846, + "eval_loss": 0.5155881774902343, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3331117630004883, + "mean_feedback_forward_cosine": 0.9767144510620519, + "mean_feedback_forward_relative_error": 0.21364634640906988, + "min_feedback_forward_cosine": 0.946147084236145 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.2707260728518168 + }, + { + "epoch": 42, + "eval_accuracy": 0.8526, + "eval_loss": 0.49671691284179686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3209340572357178, + "mean_feedback_forward_cosine": 0.9782712177226418, + "mean_feedback_forward_relative_error": 0.20631324695913414, + "min_feedback_forward_cosine": 0.9499820470809937 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.2681165763537089 + }, + { + "epoch": 43, + "eval_accuracy": 0.8602, + "eval_loss": 0.4529365417480469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30942007899284363, + "mean_feedback_forward_cosine": 0.9797280869985882, + "mean_feedback_forward_relative_error": 0.19923599417272367, + "min_feedback_forward_cosine": 0.9534589648246765 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.26589359577231936 + }, + { + "epoch": 44, + "eval_accuracy": 0.816, + "eval_loss": 0.6069492797851562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29825714230537415, + "mean_feedback_forward_cosine": 0.9811234944745114, + "mean_feedback_forward_relative_error": 0.1922379583120346, + "min_feedback_forward_cosine": 0.9567015171051025 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.2647001281181971 + }, + { + "epoch": 45, + "eval_accuracy": 0.8246, + "eval_loss": 0.6435483520507812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28758206963539124, + "mean_feedback_forward_cosine": 0.9823637510600843, + "mean_feedback_forward_relative_error": 0.18576494329854062, + "min_feedback_forward_cosine": 0.9596829414367676 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.25985294319258795 + }, + { + "epoch": 46, + "eval_accuracy": 0.845, + "eval_loss": 0.5288216491699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2769462466239929, + "mean_feedback_forward_cosine": 0.9835391013245833, + "mean_feedback_forward_relative_error": 0.17942392904507487, + "min_feedback_forward_cosine": 0.96257483959198 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.2615274556848738 + }, + { + "epoch": 47, + "eval_accuracy": 0.8418, + "eval_loss": 0.54138017578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2668825387954712, + "mean_feedback_forward_cosine": 0.984656035900116, + "mean_feedback_forward_relative_error": 0.17317641526460648, + "min_feedback_forward_cosine": 0.9652228355407715 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.2564132101482815 + }, + { + "epoch": 48, + "eval_accuracy": 0.8372, + "eval_loss": 0.5357763732910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25704485177993774, + "mean_feedback_forward_cosine": 0.9856965698693928, + "mean_feedback_forward_relative_error": 0.1671780539970649, + "min_feedback_forward_cosine": 0.9677011966705322 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.25630424637264676 + }, + { + "epoch": 49, + "eval_accuracy": 0.8638, + "eval_loss": 0.44562494506835937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24754087626934052, + "mean_feedback_forward_cosine": 0.9866711967869809, + "mean_feedback_forward_relative_error": 0.1613692139324389, + "min_feedback_forward_cosine": 0.9700177311897278 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.25752347220314875 + }, + { + "epoch": 50, + "eval_accuracy": 0.865, + "eval_loss": 0.4701431762695312, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23877203464508057, + "mean_feedback_forward_cosine": 0.9875494618164865, + "mean_feedback_forward_relative_error": 0.1559329197594994, + "min_feedback_forward_cosine": 0.9720914363861084 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.2516157629966736 + }, + { + "epoch": 51, + "eval_accuracy": 0.854, + "eval_loss": 0.4998943542480469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22992095351219177, + "mean_feedback_forward_cosine": 0.9883927734274613, + "mean_feedback_forward_relative_error": 0.1505319970218759, + "min_feedback_forward_cosine": 0.9741135239601135 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.24805901011890835 + }, + { + "epoch": 52, + "eval_accuracy": 0.8616, + "eval_loss": 0.47119757690429687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22150832414627075, + "mean_feedback_forward_cosine": 0.9891759470889443, + "mean_feedback_forward_relative_error": 0.1453473183669542, + "min_feedback_forward_cosine": 0.9759514927864075 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.24537688566843668 + }, + { + "epoch": 53, + "eval_accuracy": 0.8556, + "eval_loss": 0.48224800415039065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21376891434192657, + "mean_feedback_forward_cosine": 0.989898261271025, + "mean_feedback_forward_relative_error": 0.14038500679950966, + "min_feedback_forward_cosine": 0.9775823354721069 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.2405990086025662 + }, + { + "epoch": 54, + "eval_accuracy": 0.8436, + "eval_loss": 0.5470617431640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20610255002975464, + "mean_feedback_forward_cosine": 0.9905684119776675, + "mean_feedback_forward_relative_error": 0.1356346622893685, + "min_feedback_forward_cosine": 0.9791518449783325 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.2398143388748169 + }, + { + "epoch": 55, + "eval_accuracy": 0.8622, + "eval_loss": 0.45401356811523436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19825775921344757, + "mean_feedback_forward_cosine": 0.9912248034226266, + "mean_feedback_forward_relative_error": 0.13083300131716227, + "min_feedback_forward_cosine": 0.9806845784187317 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.2417209722412957 + }, + { + "epoch": 56, + "eval_accuracy": 0.8272, + "eval_loss": 0.6230890991210938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19106264412403107, + "mean_feedback_forward_cosine": 0.9918148360754314, + "mean_feedback_forward_relative_error": 0.12633993947192243, + "min_feedback_forward_cosine": 0.9820552468299866 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.23602942765553792 + }, + { + "epoch": 57, + "eval_accuracy": 0.8342, + "eval_loss": 0.5783703979492187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18371108174324036, + "mean_feedback_forward_cosine": 0.9923666746992814, + "mean_feedback_forward_relative_error": 0.12198558371318013, + "min_feedback_forward_cosine": 0.9833942651748657 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.2385865336894989 + }, + { + "epoch": 58, + "eval_accuracy": 0.8152, + "eval_loss": 0.6801926147460937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17712336778640747, + "mean_feedback_forward_cosine": 0.992893824451848, + "mean_feedback_forward_relative_error": 0.11769728735089302, + "min_feedback_forward_cosine": 0.9845579266548157 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.23616963873969185 + }, + { + "epoch": 59, + "eval_accuracy": 0.8444, + "eval_loss": 0.5382642761230468, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1707431674003601, + "mean_feedback_forward_cosine": 0.9933741876953527, + "mean_feedback_forward_relative_error": 0.113632299594189, + "min_feedback_forward_cosine": 0.9856466054916382 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.2340568670378791 + }, + { + "epoch": 60, + "eval_accuracy": 0.8684, + "eval_loss": 0.42916826171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1647421419620514, + "mean_feedback_forward_cosine": 0.9938058602182489, + "mean_feedback_forward_relative_error": 0.10984297096729279, + "min_feedback_forward_cosine": 0.9866265654563904 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.22967808141708373 + }, + { + "epoch": 61, + "eval_accuracy": 0.8644, + "eval_loss": 0.4622958190917969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1589381992816925, + "mean_feedback_forward_cosine": 0.9942190270674857, + "mean_feedback_forward_relative_error": 0.1061004608085281, + "min_feedback_forward_cosine": 0.9875530004501343 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.22695105500353707 + }, + { + "epoch": 62, + "eval_accuracy": 0.8694, + "eval_loss": 0.4372173278808594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15320304036140442, + "mean_feedback_forward_cosine": 0.9946047820542988, + "mean_feedback_forward_relative_error": 0.1024947821309692, + "min_feedback_forward_cosine": 0.9884278774261475 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.22514011753400168 + }, + { + "epoch": 63, + "eval_accuracy": 0.8506, + "eval_loss": 0.534511572265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14787736535072327, + "mean_feedback_forward_cosine": 0.9949603143491244, + "mean_feedback_forward_relative_error": 0.0990475590683912, + "min_feedback_forward_cosine": 0.9892134666442871 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.22366815836164686 + }, + { + "epoch": 64, + "eval_accuracy": 0.8492, + "eval_loss": 0.4916988586425781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14245589077472687, + "mean_feedback_forward_cosine": 0.9952998600508037, + "mean_feedback_forward_relative_error": 0.09564386621901863, + "min_feedback_forward_cosine": 0.9899836778640747 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.22651206641462115 + }, + { + "epoch": 65, + "eval_accuracy": 0.8582, + "eval_loss": 0.4878775207519531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1371895670890808, + "mean_feedback_forward_cosine": 0.9956255743378087, + "mean_feedback_forward_relative_error": 0.09226729152233977, + "min_feedback_forward_cosine": 0.9907054901123047 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.23046863102383083 + }, + { + "epoch": 66, + "eval_accuracy": 0.8532, + "eval_loss": 0.5311714599609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13216394186019897, + "mean_feedback_forward_cosine": 0.9959230548457095, + "mean_feedback_forward_relative_error": 0.08907792697611608, + "min_feedback_forward_cosine": 0.991371750831604 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.22570315517319572 + }, + { + "epoch": 67, + "eval_accuracy": 0.8616, + "eval_loss": 0.47444708251953127, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12750616669654846, + "mean_feedback_forward_cosine": 0.9961913918194018, + "mean_feedback_forward_relative_error": 0.08607606923109606, + "min_feedback_forward_cosine": 0.9919623136520386 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.2197730148103502 + }, + { + "epoch": 68, + "eval_accuracy": 0.8686, + "eval_loss": 0.4310839172363281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12284544110298157, + "mean_feedback_forward_cosine": 0.9964495489471837, + "mean_feedback_forward_relative_error": 0.08310531922861149, + "min_feedback_forward_cosine": 0.9925363063812256 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.22210470549265543 + }, + { + "epoch": 69, + "eval_accuracy": 0.875, + "eval_loss": 0.39954692993164065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11851232498884201, + "mean_feedback_forward_cosine": 0.9966876569547152, + "mean_feedback_forward_relative_error": 0.08026892731064245, + "min_feedback_forward_cosine": 0.9930498600006104 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.22326679899957444 + }, + { + "epoch": 70, + "eval_accuracy": 0.8728, + "eval_loss": 0.431579052734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1143125668168068, + "mean_feedback_forward_cosine": 0.9969074663363005, + "mean_feedback_forward_relative_error": 0.07755606699930995, + "min_feedback_forward_cosine": 0.9935322403907776 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.21717988074090747 + }, + { + "epoch": 71, + "eval_accuracy": 0.8372, + "eval_loss": 0.6019211547851563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11009687930345535, + "mean_feedback_forward_cosine": 0.9971224697012651, + "mean_feedback_forward_relative_error": 0.07480305041137494, + "min_feedback_forward_cosine": 0.9939974546432495 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.21987548260688783 + }, + { + "epoch": 72, + "eval_accuracy": 0.844, + "eval_loss": 0.5271776062011718, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1060536578297615, + "mean_feedback_forward_cosine": 0.9973184591845462, + "mean_feedback_forward_relative_error": 0.07221450186089466, + "min_feedback_forward_cosine": 0.9944281578063965 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.21815802181826696 + }, + { + "epoch": 73, + "eval_accuracy": 0.8644, + "eval_loss": 0.4716430725097656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10231104493141174, + "mean_feedback_forward_cosine": 0.9974972291996604, + "mean_feedback_forward_relative_error": 0.06975691196949858, + "min_feedback_forward_cosine": 0.994812548160553 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.21477413632074993 + }, + { + "epoch": 74, + "eval_accuracy": 0.8756, + "eval_loss": 0.39738220825195314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09869157522916794, + "mean_feedback_forward_cosine": 0.9976618415430972, + "mean_feedback_forward_relative_error": 0.06742280917732339, + "min_feedback_forward_cosine": 0.9951719045639038 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.2134038785510593 + }, + { + "epoch": 75, + "eval_accuracy": 0.8396, + "eval_loss": 0.5778799682617187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09520690888166428, + "mean_feedback_forward_cosine": 0.9978216541440863, + "mean_feedback_forward_relative_error": 0.0650710579203932, + "min_feedback_forward_cosine": 0.9955058097839355 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.21306624372800193 + }, + { + "epoch": 76, + "eval_accuracy": 0.8564, + "eval_loss": 0.50355625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09168129414319992, + "mean_feedback_forward_cosine": 0.9979705936030338, + "mean_feedback_forward_relative_error": 0.06280808003717347, + "min_feedback_forward_cosine": 0.9958318471908569 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.21620365244812437 + }, + { + "epoch": 77, + "eval_accuracy": 0.8696, + "eval_loss": 0.44153466796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08841861039400101, + "mean_feedback_forward_cosine": 0.9981047730696829, + "mean_feedback_forward_relative_error": 0.0606959780776187, + "min_feedback_forward_cosine": 0.9961221218109131 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.2079311021433936 + }, + { + "epoch": 78, + "eval_accuracy": 0.8704, + "eval_loss": 0.4515508117675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08516950905323029, + "mean_feedback_forward_cosine": 0.9982353982172514, + "mean_feedback_forward_relative_error": 0.058566737037740256, + "min_feedback_forward_cosine": 0.9964010715484619 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.21112659882969326 + }, + { + "epoch": 79, + "eval_accuracy": 0.8624, + "eval_loss": 0.47880274658203126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08209028840065002, + "mean_feedback_forward_cosine": 0.9983568599349574, + "mean_feedback_forward_relative_error": 0.05651370122244483, + "min_feedback_forward_cosine": 0.9966554045677185 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.21050622888141207 + }, + { + "epoch": 80, + "eval_accuracy": 0.8636, + "eval_loss": 0.4666690185546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07915076613426208, + "mean_feedback_forward_cosine": 0.9984677810417978, + "mean_feedback_forward_relative_error": 0.05457279988025364, + "min_feedback_forward_cosine": 0.9968898296356201 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.21067021816306644 + }, + { + "epoch": 81, + "eval_accuracy": 0.8598, + "eval_loss": 0.5046466125488281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07637186348438263, + "mean_feedback_forward_cosine": 0.9985681176185608, + "mean_feedback_forward_relative_error": 0.05274927233786959, + "min_feedback_forward_cosine": 0.9971036911010742 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.20502160912619696 + }, + { + "epoch": 82, + "eval_accuracy": 0.8576, + "eval_loss": 0.4760834167480469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0736217275261879, + "mean_feedback_forward_cosine": 0.9986653202458432, + "mean_feedback_forward_relative_error": 0.050925137965302715, + "min_feedback_forward_cosine": 0.997308611869812 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.20797349837091234 + }, + { + "epoch": 83, + "eval_accuracy": 0.8376, + "eval_loss": 0.611455859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07095649093389511, + "mean_feedback_forward_cosine": 0.9987575662763495, + "mean_feedback_forward_relative_error": 0.049138400703668594, + "min_feedback_forward_cosine": 0.997499406337738 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.20943366742134095 + }, + { + "epoch": 84, + "eval_accuracy": 0.8716, + "eval_loss": 0.4457356201171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06834620982408524, + "mean_feedback_forward_cosine": 0.9988417311718589, + "mean_feedback_forward_relative_error": 0.04744118040329531, + "min_feedback_forward_cosine": 0.9976792335510254 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.2061102368089888 + }, + { + "epoch": 85, + "eval_accuracy": 0.8524, + "eval_loss": 0.5395504333496094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06592147052288055, + "mean_feedback_forward_cosine": 0.9989195466041565, + "mean_feedback_forward_relative_error": 0.04582055912990319, + "min_feedback_forward_cosine": 0.9978403449058533 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.20714529931810166 + }, + { + "epoch": 86, + "eval_accuracy": 0.875, + "eval_loss": 0.4334523864746094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06359504163265228, + "mean_feedback_forward_cosine": 0.9989917152806332, + "mean_feedback_forward_relative_error": 0.04426226617866441, + "min_feedback_forward_cosine": 0.9979900121688843 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.20295997999509174 + }, + { + "epoch": 87, + "eval_accuracy": 0.873, + "eval_loss": 0.43320928955078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06130767986178398, + "mean_feedback_forward_cosine": 0.9990615468276175, + "mean_feedback_forward_relative_error": 0.04270692159862895, + "min_feedback_forward_cosine": 0.9981316328048706 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.20736424657503763 + }, + { + "epoch": 88, + "eval_accuracy": 0.8666, + "eval_loss": 0.45640607299804686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.059117674827575684, + "mean_feedback_forward_cosine": 0.9991248218636763, + "mean_feedback_forward_relative_error": 0.04123614924518686, + "min_feedback_forward_cosine": 0.9982626438140869 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.20548082500298817 + }, + { + "epoch": 89, + "eval_accuracy": 0.874, + "eval_loss": 0.41520469360351564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056990038603544235, + "mean_feedback_forward_cosine": 0.9991847339429354, + "mean_feedback_forward_relative_error": 0.039800861948414853, + "min_feedback_forward_cosine": 0.9983851909637451 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.20585563192102643 + }, + { + "epoch": 90, + "eval_accuracy": 0.839, + "eval_loss": 0.569453662109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.055011142045259476, + "mean_feedback_forward_cosine": 0.9992381836238661, + "mean_feedback_forward_relative_error": 0.038467583197512124, + "min_feedback_forward_cosine": 0.9984948635101318 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.20076531094180214 + }, + { + "epoch": 91, + "eval_accuracy": 0.8722, + "eval_loss": 0.4262968078613281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.052972324192523956, + "mean_feedback_forward_cosine": 0.9992925712936803, + "mean_feedback_forward_relative_error": 0.03706920274386281, + "min_feedback_forward_cosine": 0.9986041784286499 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.20624251328044468 + }, + { + "epoch": 92, + "eval_accuracy": 0.871, + "eval_loss": 0.4491199462890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05109149217605591, + "mean_feedback_forward_cosine": 0.9993400699213931, + "mean_feedback_forward_relative_error": 0.03580162468317308, + "min_feedback_forward_cosine": 0.9987013339996338 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.19799577930238513 + }, + { + "epoch": 93, + "eval_accuracy": 0.8736, + "eval_loss": 0.4416592041015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0493144765496254, + "mean_feedback_forward_cosine": 0.9993836095458583, + "mean_feedback_forward_relative_error": 0.03459886961469525, + "min_feedback_forward_cosine": 0.9987897276878357 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.19725176190800137 + }, + { + "epoch": 94, + "eval_accuracy": 0.8626, + "eval_loss": 0.47947452392578127, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.047581423074007034, + "mean_feedback_forward_cosine": 0.9994248214520907, + "mean_feedback_forward_relative_error": 0.03341968107576433, + "min_feedback_forward_cosine": 0.9988730549812317 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.19693422346115114 + }, + { + "epoch": 95, + "eval_accuracy": 0.8508, + "eval_loss": 0.5477066467285157, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04589957743883133, + "mean_feedback_forward_cosine": 0.9994630186181319, + "mean_feedback_forward_relative_error": 0.032289010129476846, + "min_feedback_forward_cosine": 0.9989511966705322 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.19458590910699633 + }, + { + "epoch": 96, + "eval_accuracy": 0.8694, + "eval_loss": 0.4502195617675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04428688809275627, + "mean_feedback_forward_cosine": 0.9994986276877554, + "mean_feedback_forward_relative_error": 0.031198943376933273, + "min_feedback_forward_cosine": 0.9990235567092896 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.19458247496816847 + }, + { + "epoch": 97, + "eval_accuracy": 0.8682, + "eval_loss": 0.46479819946289064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04265602305531502, + "mean_feedback_forward_cosine": 0.9995334713082564, + "mean_feedback_forward_relative_error": 0.030095937563792655, + "min_feedback_forward_cosine": 0.9990940690040588 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.19885163509315915 + }, + { + "epoch": 98, + "eval_accuracy": 0.8678, + "eval_loss": 0.4972033447265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.041172705590724945, + "mean_feedback_forward_cosine": 0.9995646257149545, + "mean_feedback_forward_relative_error": 0.029072351851745656, + "min_feedback_forward_cosine": 0.9991556406021118 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.19531077443758646 + }, + { + "epoch": 99, + "eval_accuracy": 0.8752, + "eval_loss": 0.40619945068359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03969261795282364, + "mean_feedback_forward_cosine": 0.9995948766407213, + "mean_feedback_forward_relative_error": 0.028044214913327443, + "min_feedback_forward_cosine": 0.9992151260375977 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.1985523914443122 + }, + { + "epoch": 100, + "eval_accuracy": 0.8546, + "eval_loss": 0.5547154479980468, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03826097771525383, + "mean_feedback_forward_cosine": 0.9996227025985718, + "mean_feedback_forward_relative_error": 0.027064508856519273, + "min_feedback_forward_cosine": 0.9992707967758179 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.19719866987334358 + }, + { + "epoch": 101, + "eval_accuracy": 0.9096, + "eval_loss": 0.31504560546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03824806213378906, + "mean_feedback_forward_cosine": 0.9996229880734494, + "mean_feedback_forward_relative_error": 0.02705035456701329, + "min_feedback_forward_cosine": 0.9992711544036865 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.12498004048400455 + }, + { + "epoch": 102, + "eval_accuracy": 0.9116, + "eval_loss": 0.31542617797851563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03824153542518616, + "mean_feedback_forward_cosine": 0.9996231637502971, + "mean_feedback_forward_relative_error": 0.02704000105395129, + "min_feedback_forward_cosine": 0.9992715716362 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.09414871436489954 + }, + { + "epoch": 103, + "eval_accuracy": 0.9148, + "eval_loss": 0.3152183624267578, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03823477402329445, + "mean_feedback_forward_cosine": 0.9996233425642315, + "mean_feedback_forward_relative_error": 0.02702999928672063, + "min_feedback_forward_cosine": 0.999271810054779 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.08886900408930248 + }, + { + "epoch": 104, + "eval_accuracy": 0.9156, + "eval_loss": 0.3212423614501953, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038228511810302734, + "mean_feedback_forward_cosine": 0.999623533926512, + "mean_feedback_forward_relative_error": 0.027020629740467195, + "min_feedback_forward_cosine": 0.9992719888687134 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.07943104676802953 + }, + { + "epoch": 105, + "eval_accuracy": 0.915, + "eval_loss": 0.3224820953369141, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03822202607989311, + "mean_feedback_forward_cosine": 0.9996236876437539, + "mean_feedback_forward_relative_error": 0.02701137523706022, + "min_feedback_forward_cosine": 0.9992722272872925 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.07692426940202712 + }, + { + "epoch": 106, + "eval_accuracy": 0.9132, + "eval_loss": 0.3243196716308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03821607679128647, + "mean_feedback_forward_cosine": 0.9996238539093419, + "mean_feedback_forward_relative_error": 0.027002569288015366, + "min_feedback_forward_cosine": 0.9992725253105164 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.0701384553776847 + }, + { + "epoch": 107, + "eval_accuracy": 0.9132, + "eval_loss": 0.3256701171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038209863007068634, + "mean_feedback_forward_cosine": 0.9996240139007568, + "mean_feedback_forward_relative_error": 0.02699356859451846, + "min_feedback_forward_cosine": 0.9992727041244507 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.06748027348253462 + }, + { + "epoch": 108, + "eval_accuracy": 0.9144, + "eval_loss": 0.333216845703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038203537464141846, + "mean_feedback_forward_cosine": 0.999624208400124, + "mean_feedback_forward_relative_error": 0.026984702501642078, + "min_feedback_forward_cosine": 0.9992730021476746 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.06455075730615192 + }, + { + "epoch": 109, + "eval_accuracy": 0.9104, + "eval_loss": 0.34080203857421876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03819728270173073, + "mean_feedback_forward_cosine": 0.9996243401577598, + "mean_feedback_forward_relative_error": 0.02697615119579591, + "min_feedback_forward_cosine": 0.9992731809616089 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.06166491029262543 + }, + { + "epoch": 110, + "eval_accuracy": 0.914, + "eval_loss": 0.3373171447753906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038191262632608414, + "mean_feedback_forward_cosine": 0.9996245283829538, + "mean_feedback_forward_relative_error": 0.02696770125705945, + "min_feedback_forward_cosine": 0.999273419380188 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.05793297163115607 + }, + { + "epoch": 111, + "eval_accuracy": 0.9152, + "eval_loss": 0.3412690979003906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0381854884326458, + "mean_feedback_forward_cosine": 0.9996246883743688, + "mean_feedback_forward_relative_error": 0.026959390379488468, + "min_feedback_forward_cosine": 0.9992736577987671 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.05538705672224363 + }, + { + "epoch": 112, + "eval_accuracy": 0.9112, + "eval_loss": 0.34866596069335937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03817934915423393, + "mean_feedback_forward_cosine": 0.9996248546399569, + "mean_feedback_forward_relative_error": 0.026950864709521596, + "min_feedback_forward_cosine": 0.9992738962173462 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.05465268326600393 + }, + { + "epoch": 113, + "eval_accuracy": 0.9114, + "eval_loss": 0.3563314697265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03817310184240341, + "mean_feedback_forward_cosine": 0.9996250209055448, + "mean_feedback_forward_relative_error": 0.026942566282262926, + "min_feedback_forward_cosine": 0.9992742538452148 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.05348444962501526 + }, + { + "epoch": 114, + "eval_accuracy": 0.9146, + "eval_loss": 0.3597590148925781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03816705942153931, + "mean_feedback_forward_cosine": 0.9996251683486136, + "mean_feedback_forward_relative_error": 0.026934282364029633, + "min_feedback_forward_cosine": 0.9992744326591492 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.05038708394269149 + }, + { + "epoch": 115, + "eval_accuracy": 0.915, + "eval_loss": 0.3545485961914063, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038161080330610275, + "mean_feedback_forward_cosine": 0.9996253597108942, + "mean_feedback_forward_relative_error": 0.02692613059556798, + "min_feedback_forward_cosine": 0.9992746710777283 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.047704472191300654 + }, + { + "epoch": 116, + "eval_accuracy": 0.9148, + "eval_loss": 0.3614710632324219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038155052810907364, + "mean_feedback_forward_cosine": 0.9996255008797896, + "mean_feedback_forward_relative_error": 0.026917944368171066, + "min_feedback_forward_cosine": 0.999274730682373 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.046668650693363616 + }, + { + "epoch": 117, + "eval_accuracy": 0.9118, + "eval_loss": 0.36147152099609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038148727267980576, + "mean_feedback_forward_cosine": 0.9996256859678971, + "mean_feedback_forward_relative_error": 0.026909909701268924, + "min_feedback_forward_cosine": 0.9992750883102417 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.0468660827504264 + }, + { + "epoch": 118, + "eval_accuracy": 0.9142, + "eval_loss": 0.36134048461914064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03814224898815155, + "mean_feedback_forward_cosine": 0.9996258585076583, + "mean_feedback_forward_relative_error": 0.026901894298038985, + "min_feedback_forward_cosine": 0.9992753863334656 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.04667721270521482 + }, + { + "epoch": 119, + "eval_accuracy": 0.9118, + "eval_loss": 0.3710879089355469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03813629969954491, + "mean_feedback_forward_cosine": 0.9996260247732464, + "mean_feedback_forward_relative_error": 0.026893928549007365, + "min_feedback_forward_cosine": 0.9992755651473999 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.04263128900461727 + }, + { + "epoch": 120, + "eval_accuracy": 0.9124, + "eval_loss": 0.37179076538085937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038130130618810654, + "mean_feedback_forward_cosine": 0.9996262004500941, + "mean_feedback_forward_relative_error": 0.026885935301451308, + "min_feedback_forward_cosine": 0.9992758631706238 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.04206890949805578 + }, + { + "epoch": 121, + "eval_accuracy": 0.9148, + "eval_loss": 0.3723926452636719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038124118000268936, + "mean_feedback_forward_cosine": 0.9996263635785956, + "mean_feedback_forward_relative_error": 0.02687794715166092, + "min_feedback_forward_cosine": 0.9992761015892029 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.04026431170701981 + }, + { + "epoch": 122, + "eval_accuracy": 0.9142, + "eval_loss": 0.3758395263671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03811798617243767, + "mean_feedback_forward_cosine": 0.9996265361183568, + "mean_feedback_forward_relative_error": 0.02686990219119348, + "min_feedback_forward_cosine": 0.9992762804031372 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.03920797461072604 + }, + { + "epoch": 123, + "eval_accuracy": 0.9152, + "eval_loss": 0.3701650634765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03811146691441536, + "mean_feedback_forward_cosine": 0.9996267274806374, + "mean_feedback_forward_relative_error": 0.02686190257143033, + "min_feedback_forward_cosine": 0.9992765188217163 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.040231052759620876 + }, + { + "epoch": 124, + "eval_accuracy": 0.9148, + "eval_loss": 0.3734594177246094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038105301558971405, + "mean_feedback_forward_cosine": 0.9996268906091389, + "mean_feedback_forward_relative_error": 0.026853734676383044, + "min_feedback_forward_cosine": 0.9992768168449402 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.038030407783057954 + }, + { + "epoch": 125, + "eval_accuracy": 0.9142, + "eval_loss": 0.37734255981445314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038098517805337906, + "mean_feedback_forward_cosine": 0.9996270788343329, + "mean_feedback_forward_relative_error": 0.02684540046673072, + "min_feedback_forward_cosine": 0.9992770552635193 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.03932609410782655 + }, + { + "epoch": 126, + "eval_accuracy": 0.9162, + "eval_loss": 0.37866737670898437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0380915105342865, + "mean_feedback_forward_cosine": 0.999627267059527, + "mean_feedback_forward_relative_error": 0.026837026700377464, + "min_feedback_forward_cosine": 0.9992772936820984 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.037950586869650416 + }, + { + "epoch": 127, + "eval_accuracy": 0.9132, + "eval_loss": 0.39278297119140626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038084693253040314, + "mean_feedback_forward_cosine": 0.9996274395992881, + "mean_feedback_forward_relative_error": 0.02682868611851805, + "min_feedback_forward_cosine": 0.9992775321006775 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.03678174229264259 + }, + { + "epoch": 128, + "eval_accuracy": 0.915, + "eval_loss": 0.38518326416015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03807844966650009, + "mean_feedback_forward_cosine": 0.9996276246873956, + "mean_feedback_forward_relative_error": 0.026820657823823, + "min_feedback_forward_cosine": 0.9992778301239014 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.03376755590836207 + }, + { + "epoch": 129, + "eval_accuracy": 0.9154, + "eval_loss": 0.3975419860839844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03807184472680092, + "mean_feedback_forward_cosine": 0.9996278191867628, + "mean_feedback_forward_relative_error": 0.026812335966449035, + "min_feedback_forward_cosine": 0.9992780685424805 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.03459894592629539 + }, + { + "epoch": 130, + "eval_accuracy": 0.9134, + "eval_loss": 0.3980976745605469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03806491941213608, + "mean_feedback_forward_cosine": 0.9996280074119568, + "mean_feedback_forward_relative_error": 0.02680385995068048, + "min_feedback_forward_cosine": 0.9992783665657043 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.034467040161291755 + }, + { + "epoch": 131, + "eval_accuracy": 0.9106, + "eval_loss": 0.401936376953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03805799409747124, + "mean_feedback_forward_cosine": 0.999628157992112, + "mean_feedback_forward_relative_error": 0.026795381287995138, + "min_feedback_forward_cosine": 0.9992786049842834 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.03290436211890645 + }, + { + "epoch": 132, + "eval_accuracy": 0.9158, + "eval_loss": 0.39828818969726565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03805088624358177, + "mean_feedback_forward_cosine": 0.9996284026848642, + "mean_feedback_forward_relative_error": 0.02678674944725476, + "min_feedback_forward_cosine": 0.9992789030075073 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.03265204352868928 + }, + { + "epoch": 133, + "eval_accuracy": 0.9134, + "eval_loss": 0.4009542297363281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0380437932908535, + "mean_feedback_forward_cosine": 0.999628600321318, + "mean_feedback_forward_relative_error": 0.02677821461111307, + "min_feedback_forward_cosine": 0.9992791414260864 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.03201477077437772 + }, + { + "epoch": 134, + "eval_accuracy": 0.9138, + "eval_loss": 0.40583228759765627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038036979734897614, + "mean_feedback_forward_cosine": 0.9996287540385598, + "mean_feedback_forward_relative_error": 0.026769806189756645, + "min_feedback_forward_cosine": 0.999279260635376 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.031194838628504012 + }, + { + "epoch": 135, + "eval_accuracy": 0.9122, + "eval_loss": 0.3971060607910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038029663264751434, + "mean_feedback_forward_cosine": 0.999628976771706, + "mean_feedback_forward_relative_error": 0.026760906618284553, + "min_feedback_forward_cosine": 0.9992796182632446 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.03187097754743364 + }, + { + "epoch": 136, + "eval_accuracy": 0.9136, + "eval_loss": 0.40547701416015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038022592663764954, + "mean_feedback_forward_cosine": 0.9996291806823329, + "mean_feedback_forward_relative_error": 0.026752426828208723, + "min_feedback_forward_cosine": 0.999279797077179 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.029618267251054445 + }, + { + "epoch": 137, + "eval_accuracy": 0.9144, + "eval_loss": 0.40041107177734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038015320897102356, + "mean_feedback_forward_cosine": 0.9996294034154791, + "mean_feedback_forward_relative_error": 0.02674361695780566, + "min_feedback_forward_cosine": 0.9992802143096924 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.03065378427306811 + }, + { + "epoch": 138, + "eval_accuracy": 0.9152, + "eval_loss": 0.4129467590332031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03800799697637558, + "mean_feedback_forward_cosine": 0.9996295885035866, + "mean_feedback_forward_relative_error": 0.026734913454244013, + "min_feedback_forward_cosine": 0.9992804527282715 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.02922247618569268 + }, + { + "epoch": 139, + "eval_accuracy": 0.916, + "eval_loss": 0.40856820678710937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03800143301486969, + "mean_feedback_forward_cosine": 0.999629773591694, + "mean_feedback_forward_relative_error": 0.02672667232783217, + "min_feedback_forward_cosine": 0.9992806911468506 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.02787504948510064 + }, + { + "epoch": 140, + "eval_accuracy": 0.912, + "eval_loss": 0.43249401245117186, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03799450770020485, + "mean_feedback_forward_cosine": 0.999629977502321, + "mean_feedback_forward_relative_error": 0.026718208664341977, + "min_feedback_forward_cosine": 0.9992810487747192 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.027838887613349492 + }, + { + "epoch": 141, + "eval_accuracy": 0.9126, + "eval_loss": 0.4213609069824219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037987120449543, + "mean_feedback_forward_cosine": 0.9996301782758612, + "mean_feedback_forward_relative_error": 0.02670954457043033, + "min_feedback_forward_cosine": 0.9992812275886536 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.02806180555969477 + }, + { + "epoch": 142, + "eval_accuracy": 0.9142, + "eval_loss": 0.42442941284179686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037980616092681885, + "mean_feedback_forward_cosine": 0.9996303570897955, + "mean_feedback_forward_relative_error": 0.02670134569665319, + "min_feedback_forward_cosine": 0.9992814064025879 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.025535029594269065 + }, + { + "epoch": 143, + "eval_accuracy": 0.914, + "eval_loss": 0.42216495971679685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03797249123454094, + "mean_feedback_forward_cosine": 0.9996305798229418, + "mean_feedback_forward_relative_error": 0.026691938505360956, + "min_feedback_forward_cosine": 0.9992818236351013 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.028624223914245763 + }, + { + "epoch": 144, + "eval_accuracy": 0.9144, + "eval_loss": 0.4278658020019531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037965077906847, + "mean_feedback_forward_cosine": 0.9996307805964821, + "mean_feedback_forward_relative_error": 0.026683067559803788, + "min_feedback_forward_cosine": 0.9992820620536804 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.026180000118662915 + }, + { + "epoch": 145, + "eval_accuracy": 0.9138, + "eval_loss": 0.427544677734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03795746713876724, + "mean_feedback_forward_cosine": 0.9996310190150612, + "mean_feedback_forward_relative_error": 0.026674046377210242, + "min_feedback_forward_cosine": 0.9992823600769043 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.026351813669668305 + }, + { + "epoch": 146, + "eval_accuracy": 0.9158, + "eval_loss": 0.42938318481445314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03794943168759346, + "mean_feedback_forward_cosine": 0.9996312386111209, + "mean_feedback_forward_relative_error": 0.02666486049757192, + "min_feedback_forward_cosine": 0.999282717704773 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.026147830514444245 + }, + { + "epoch": 147, + "eval_accuracy": 0.9154, + "eval_loss": 0.43108264770507815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037941813468933105, + "mean_feedback_forward_cosine": 0.9996314613442672, + "mean_feedback_forward_relative_error": 0.026655893233653746, + "min_feedback_forward_cosine": 0.999282956123352 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.025178877196378176 + }, + { + "epoch": 148, + "eval_accuracy": 0.912, + "eval_loss": 0.43701239624023436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03793426975607872, + "mean_feedback_forward_cosine": 0.9996316495694613, + "mean_feedback_forward_relative_error": 0.026646792251420647, + "min_feedback_forward_cosine": 0.9992832541465759 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.024480403775639006 + }, + { + "epoch": 149, + "eval_accuracy": 0.9118, + "eval_loss": 0.4395541748046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037926580756902695, + "mean_feedback_forward_cosine": 0.9996318691655209, + "mean_feedback_forward_relative_error": 0.02663764043858177, + "min_feedback_forward_cosine": 0.9992836117744446 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.024357059074110454 + }, + { + "epoch": 150, + "eval_accuracy": 0.914, + "eval_loss": 0.4387383911132812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791917487978935, + "mean_feedback_forward_cosine": 0.9996321075841, + "mean_feedback_forward_relative_error": 0.02662865788136658, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.023285648666487798 + }, + { + "epoch": 151, + "eval_accuracy": 0.9142, + "eval_loss": 0.43026748046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0379190668463707, + "mean_feedback_forward_cosine": 0.9996320981728403, + "mean_feedback_forward_relative_error": 0.026628399218775724, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.0211612816479471 + }, + { + "epoch": 152, + "eval_accuracy": 0.914, + "eval_loss": 0.42774334716796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918996065855026, + "mean_feedback_forward_cosine": 0.9996321013099269, + "mean_feedback_forward_relative_error": 0.026628176142510614, + "min_feedback_forward_cosine": 0.9992837905883789 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.02011557665401035 + }, + { + "epoch": 153, + "eval_accuracy": 0.9134, + "eval_loss": 0.4272701232910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791893273591995, + "mean_feedback_forward_cosine": 0.9996320981728403, + "mean_feedback_forward_relative_error": 0.0266279557131623, + "min_feedback_forward_cosine": 0.9992837905883789 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.019754009400142564 + }, + { + "epoch": 154, + "eval_accuracy": 0.9158, + "eval_loss": 0.423570166015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791886940598488, + "mean_feedback_forward_cosine": 0.9996320950357538, + "mean_feedback_forward_relative_error": 0.026627739205172186, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.019484071310361225 + }, + { + "epoch": 155, + "eval_accuracy": 0.9162, + "eval_loss": 0.42812459716796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918820977211, + "mean_feedback_forward_cosine": 0.9996320981728403, + "mean_feedback_forward_relative_error": 0.02662751343297331, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.01772839131222831 + }, + { + "epoch": 156, + "eval_accuracy": 0.9168, + "eval_loss": 0.4284971862792969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791876137256622, + "mean_feedback_forward_cosine": 0.9996321044470134, + "mean_feedback_forward_relative_error": 0.026627286141248124, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.01800470546417766 + }, + { + "epoch": 157, + "eval_accuracy": 0.9156, + "eval_loss": 0.425212060546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791870176792145, + "mean_feedback_forward_cosine": 0.9996321138582731, + "mean_feedback_forward_relative_error": 0.026627064094339545, + "min_feedback_forward_cosine": 0.9992837905883789 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.019227959014640913 + }, + { + "epoch": 158, + "eval_accuracy": 0.9154, + "eval_loss": 0.4282466613769531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791864588856697, + "mean_feedback_forward_cosine": 0.9996321326807925, + "mean_feedback_forward_relative_error": 0.026626851164588804, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.017664594351748625 + }, + { + "epoch": 159, + "eval_accuracy": 0.9174, + "eval_loss": 0.42084263916015624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791859373450279, + "mean_feedback_forward_cosine": 0.9996321138582731, + "mean_feedback_forward_relative_error": 0.026626633970361007, + "min_feedback_forward_cosine": 0.9992839694023132 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.017749026487767695 + }, + { + "epoch": 160, + "eval_accuracy": 0.9164, + "eval_loss": 0.42557537231445314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918537855148315, + "mean_feedback_forward_cosine": 0.9996321201324463, + "mean_feedback_forward_relative_error": 0.02662641520758993, + "min_feedback_forward_cosine": 0.9992837905883789 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.01739938084317578 + }, + { + "epoch": 161, + "eval_accuracy": 0.9164, + "eval_loss": 0.42110528564453126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918489426374435, + "mean_feedback_forward_cosine": 0.9996321389549657, + "mean_feedback_forward_relative_error": 0.026626198503531907, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.01759473463681837 + }, + { + "epoch": 162, + "eval_accuracy": 0.9154, + "eval_loss": 0.42479563598632814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918440997600555, + "mean_feedback_forward_cosine": 0.999632151503312, + "mean_feedback_forward_relative_error": 0.026625978564353364, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.01624706518434816 + }, + { + "epoch": 163, + "eval_accuracy": 0.9128, + "eval_loss": 0.4251840393066406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791838884353638, + "mean_feedback_forward_cosine": 0.9996321075841, + "mean_feedback_forward_relative_error": 0.02662576034076904, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.01809051398370001 + }, + { + "epoch": 164, + "eval_accuracy": 0.9162, + "eval_loss": 0.42689862060546874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0379183292388916, + "mean_feedback_forward_cosine": 0.9996321232695329, + "mean_feedback_forward_relative_error": 0.02662554500918639, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.017154754579729505 + }, + { + "epoch": 165, + "eval_accuracy": 0.9156, + "eval_loss": 0.42907059326171876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791828826069832, + "mean_feedback_forward_cosine": 0.9996321420920523, + "mean_feedback_forward_relative_error": 0.026625329530552813, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.015940935181909137 + }, + { + "epoch": 166, + "eval_accuracy": 0.9146, + "eval_loss": 0.42847799072265624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791823983192444, + "mean_feedback_forward_cosine": 0.9996321483662254, + "mean_feedback_forward_relative_error": 0.026625118463447218, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.01682934757007493 + }, + { + "epoch": 167, + "eval_accuracy": 0.9148, + "eval_loss": 0.42574083251953126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791818395256996, + "mean_feedback_forward_cosine": 0.9996321169953597, + "mean_feedback_forward_relative_error": 0.026624903916136214, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.016822115646633835 + }, + { + "epoch": 168, + "eval_accuracy": 0.9148, + "eval_loss": 0.4300515563964844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918128073215485, + "mean_feedback_forward_cosine": 0.9996321326807925, + "mean_feedback_forward_relative_error": 0.026624686182721666, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.016840671036309665 + }, + { + "epoch": 169, + "eval_accuracy": 0.9152, + "eval_loss": 0.42753214111328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918079644441605, + "mean_feedback_forward_cosine": 0.999632157777485, + "mean_feedback_forward_relative_error": 0.026624476488091443, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.01653002626688944 + }, + { + "epoch": 170, + "eval_accuracy": 0.9152, + "eval_loss": 0.4329725341796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791803494095802, + "mean_feedback_forward_cosine": 0.9996321452291388, + "mean_feedback_forward_relative_error": 0.026624266499359357, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.01631236160033279 + }, + { + "epoch": 171, + "eval_accuracy": 0.915, + "eval_loss": 0.43096278686523437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791797533631325, + "mean_feedback_forward_cosine": 0.9996321483662254, + "mean_feedback_forward_relative_error": 0.026624056167508427, + "min_feedback_forward_cosine": 0.9992839694023132 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.017680182982815637 + }, + { + "epoch": 172, + "eval_accuracy": 0.915, + "eval_loss": 0.4310341735839844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791791573166847, + "mean_feedback_forward_cosine": 0.9996321640516582, + "mean_feedback_forward_relative_error": 0.0266238445121991, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.01705649436513583 + }, + { + "epoch": 173, + "eval_accuracy": 0.9148, + "eval_loss": 0.4284998229980469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037917863577604294, + "mean_feedback_forward_cosine": 0.999632157777485, + "mean_feedback_forward_relative_error": 0.02662362472007149, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.016325897006193798 + }, + { + "epoch": 174, + "eval_accuracy": 0.9156, + "eval_loss": 0.42368416137695314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791780769824982, + "mean_feedback_forward_cosine": 0.9996321671887448, + "mean_feedback_forward_relative_error": 0.026623414927407316, + "min_feedback_forward_cosine": 0.9992839694023132 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.016306531962255635 + }, + { + "epoch": 175, + "eval_accuracy": 0.915, + "eval_loss": 0.42613728637695314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791775926947594, + "mean_feedback_forward_cosine": 0.9996321546403986, + "mean_feedback_forward_relative_error": 0.026623194743143886, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.016608378148078917 + }, + { + "epoch": 176, + "eval_accuracy": 0.915, + "eval_loss": 0.42469633178710936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791771084070206, + "mean_feedback_forward_cosine": 0.9996321483662254, + "mean_feedback_forward_relative_error": 0.026622989509058625, + "min_feedback_forward_cosine": 0.9992839694023132 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.01599796301147176 + }, + { + "epoch": 177, + "eval_accuracy": 0.9142, + "eval_loss": 0.42996865234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791766241192818, + "mean_feedback_forward_cosine": 0.9996321766000045, + "mean_feedback_forward_relative_error": 0.026622781431988665, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.0145107354823086 + }, + { + "epoch": 178, + "eval_accuracy": 0.9154, + "eval_loss": 0.42992142333984373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0379176065325737, + "mean_feedback_forward_cosine": 0.9996321452291388, + "mean_feedback_forward_relative_error": 0.026622566737626727, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.016593861082527372 + }, + { + "epoch": 179, + "eval_accuracy": 0.9154, + "eval_loss": 0.429712548828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791755437850952, + "mean_feedback_forward_cosine": 0.9996321671887448, + "mean_feedback_forward_relative_error": 0.026622350278653596, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.015633646452095775 + }, + { + "epoch": 180, + "eval_accuracy": 0.9142, + "eval_loss": 0.4295874938964844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037917498499155045, + "mean_feedback_forward_cosine": 0.9996321671887448, + "mean_feedback_forward_relative_error": 0.026622139946802667, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.01656670587816172 + }, + { + "epoch": 181, + "eval_accuracy": 0.9166, + "eval_loss": 0.4243903991699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037917450070381165, + "mean_feedback_forward_cosine": 0.9996321797370911, + "mean_feedback_forward_relative_error": 0.02662193775177002, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.01582161951859792 + }, + { + "epoch": 182, + "eval_accuracy": 0.9154, + "eval_loss": 0.43365278930664064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791740536689758, + "mean_feedback_forward_cosine": 0.999632157777485, + "mean_feedback_forward_relative_error": 0.02662172746893607, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.014813974743419223 + }, + { + "epoch": 183, + "eval_accuracy": 0.9148, + "eval_loss": 0.42709647216796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791734576225281, + "mean_feedback_forward_cosine": 0.9996321891483507, + "mean_feedback_forward_relative_error": 0.026621517872339802, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.01592662673095862 + }, + { + "epoch": 184, + "eval_accuracy": 0.9162, + "eval_loss": 0.4307603088378906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791729733347893, + "mean_feedback_forward_cosine": 0.9996321671887448, + "mean_feedback_forward_relative_error": 0.026621317441918348, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.016338840604159566 + }, + { + "epoch": 185, + "eval_accuracy": 0.9146, + "eval_loss": 0.43272603149414063, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037917252629995346, + "mean_feedback_forward_cosine": 0.9996321860112642, + "mean_feedback_forward_relative_error": 0.02662111627624223, + "min_feedback_forward_cosine": 0.999284029006958 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.015533369680245717 + }, + { + "epoch": 186, + "eval_accuracy": 0.9148, + "eval_loss": 0.43155269775390626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791719302535057, + "mean_feedback_forward_cosine": 0.9996321640516582, + "mean_feedback_forward_relative_error": 0.02662091192446257, + "min_feedback_forward_cosine": 0.9992837905883789 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.015310422006249427 + }, + { + "epoch": 187, + "eval_accuracy": 0.9148, + "eval_loss": 0.431560595703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791714087128639, + "mean_feedback_forward_cosine": 0.9996321954225239, + "mean_feedback_forward_relative_error": 0.02662071384685604, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.015171491112642817 + }, + { + "epoch": 188, + "eval_accuracy": 0.9154, + "eval_loss": 0.4299284484863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791709616780281, + "mean_feedback_forward_cosine": 0.9996321703258314, + "mean_feedback_forward_relative_error": 0.026620496799679177, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.015132849889993668 + }, + { + "epoch": 189, + "eval_accuracy": 0.9144, + "eval_loss": 0.43136605224609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037917036563158035, + "mean_feedback_forward_cosine": 0.9996321922854373, + "mean_feedback_forward_relative_error": 0.02662028823243944, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.01564434177271194 + }, + { + "epoch": 190, + "eval_accuracy": 0.9138, + "eval_loss": 0.42769647216796874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916988134384155, + "mean_feedback_forward_cosine": 0.9996322142450433, + "mean_feedback_forward_relative_error": 0.02662007956716575, + "min_feedback_forward_cosine": 0.9992838501930237 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.015331490024593141 + }, + { + "epoch": 191, + "eval_accuracy": 0.9144, + "eval_loss": 0.43386570434570315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791693598031998, + "mean_feedback_forward_cosine": 0.9996321891483507, + "mean_feedback_forward_relative_error": 0.026619868353009224, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.015315960738725133 + }, + { + "epoch": 192, + "eval_accuracy": 0.9144, + "eval_loss": 0.43133213500976564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916891276836395, + "mean_feedback_forward_cosine": 0.9996322111079567, + "mean_feedback_forward_relative_error": 0.026619663118923966, + "min_feedback_forward_cosine": 0.9992839694023132 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.015171922739346822 + }, + { + "epoch": 193, + "eval_accuracy": 0.9144, + "eval_loss": 0.431591796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916846573352814, + "mean_feedback_forward_cosine": 0.9996322205192164, + "mean_feedback_forward_relative_error": 0.02661946067880643, + "min_feedback_forward_cosine": 0.9992839694023132 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.014588061490986083 + }, + { + "epoch": 194, + "eval_accuracy": 0.9144, + "eval_loss": 0.4299183837890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916794419288635, + "mean_feedback_forward_cosine": 0.9996322079708702, + "mean_feedback_forward_relative_error": 0.0266192503469555, + "min_feedback_forward_cosine": 0.9992837905883789 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.015228129545847575 + }, + { + "epoch": 195, + "eval_accuracy": 0.916, + "eval_loss": 0.4288531494140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791675344109535, + "mean_feedback_forward_cosine": 0.9996321985596105, + "mean_feedback_forward_relative_error": 0.026619042858089272, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.014976719170146518 + }, + { + "epoch": 196, + "eval_accuracy": 0.9144, + "eval_loss": 0.4313197021484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916697561740875, + "mean_feedback_forward_cosine": 0.9996321891483507, + "mean_feedback_forward_relative_error": 0.026618830761627146, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.01579895223768221 + }, + { + "epoch": 197, + "eval_accuracy": 0.9152, + "eval_loss": 0.4294109436035156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916649132966995, + "mean_feedback_forward_cosine": 0.9996321891483507, + "mean_feedback_forward_relative_error": 0.026618630282188718, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.014463566927115123 + }, + { + "epoch": 198, + "eval_accuracy": 0.9168, + "eval_loss": 0.42698069458007815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916600704193115, + "mean_feedback_forward_cosine": 0.9996321985596105, + "mean_feedback_forward_relative_error": 0.02661842259725458, + "min_feedback_forward_cosine": 0.9992839694023132 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.015557571692268053 + }, + { + "epoch": 199, + "eval_accuracy": 0.916, + "eval_loss": 0.43350626220703126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037916556000709534, + "mean_feedback_forward_cosine": 0.9996322205192164, + "mean_feedback_forward_relative_error": 0.0266182226570029, + "min_feedback_forward_cosine": 0.999284029006958 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.014629167970021565 + }, + { + "epoch": 200, + "eval_accuracy": 0.9144, + "eval_loss": 0.4333634704589844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03791650012135506, + "mean_feedback_forward_cosine": 0.9996322111079567, + "mean_feedback_forward_relative_error": 0.026618016491595068, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.016335480781065094 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9144, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.4333634704589844, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2010747392, + "peak_memory_reserved_bytes": 2552233984, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 21.03406047821045, + "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": 1780.5374817848206, + "train_wall_s": 1758.6518173217773 + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 360978048000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 360978048000000, + "local_weight_correlation_macs": 364959360000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 364959360000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 1451874816000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json.manifest.json new file mode 100644 index 0000000..1d2c016 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "kp", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787033093.4963362, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 1784.3792207241058, + "epochs": 200, + "experiment_name": "resnet-r2-clean_kp-d20-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "clean_kp", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d20-lr0p1.json", + "output_exists": true, + "output_sha256": "87914e8c8674f9fc2d136f6dbf11ee9512c7e797d2d757b675ea7610ba6db795", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json new file mode 100644 index 0000000..7ca9d56 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json @@ -0,0 +1,4001 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.9990160942077637, + "feedback_forward_cosine": [ + 0.9995843172073364, + 0.9995485544204712, + 0.9993047714233398, + 0.9992356300354004, + 0.999225914478302, + 0.9990967512130737, + 0.9992426037788391, + 0.9990879893302917, + 0.9994840621948242, + 0.9992990493774414, + 0.999346911907196, + 0.9995205402374268, + 0.9992691278457642, + 0.9992083311080933, + 0.9994052648544312, + 0.9992618560791016, + 0.9994874000549316, + 0.9992926120758057, + 0.999449610710144, + 0.9992502331733704, + 0.999462366104126, + 0.9996726512908936, + 0.9996097087860107, + 0.9995791912078857, + 0.9996154308319092, + 0.9994989633560181, + 0.9995627403259277, + 0.999337375164032, + 0.9994921088218689, + 0.9988247156143188, + 0.9999656677246094 + ], + "feedback_forward_norm_ratio": [ + 0.9991628527641296, + 1.001270055770874, + 0.9982960224151611, + 1.0040854215621948, + 0.9988148212432861, + 1.0012292861938477, + 0.9982872605323792, + 1.0016695261001587, + 0.9984201192855835, + 1.0020995140075684, + 0.9984967708587646, + 1.000747799873352, + 0.9983559846878052, + 1.0024127960205078, + 0.9985647797584534, + 1.0024219751358032, + 0.9993768334388733, + 1.0025209188461304, + 0.9989345669746399, + 1.0021547079086304, + 0.9993236064910889, + 1.000596046447754, + 0.9998040199279785, + 1.0014662742614746, + 1.0001678466796875, + 1.0017902851104736, + 0.9997692108154297, + 1.0016337633132935, + 0.9996170401573181, + 1.0019867420196533, + 1.005363941192627 + ], + "feedback_forward_relative_error": [ + 0.028834059834480286, + 0.03009512461721897, + 0.037293966859579086, + 0.03939209133386612, + 0.03934045135974884, + 0.04254467412829399, + 0.038926754146814346, + 0.04277598485350609, + 0.03213443234562874, + 0.03754224628210068, + 0.036144670099020004, + 0.030987465754151344, + 0.038239117711782455, + 0.039911918342113495, + 0.03449464589357376, + 0.038545139133930206, + 0.03201395645737648, + 0.037745725363492966, + 0.03317772224545479, + 0.03882520645856857, + 0.03278729319572449, + 0.02560167759656906, + 0.027935020625591278, + 0.02906847931444645, + 0.027740592136979103, + 0.03173808380961418, + 0.0295693501830101, + 0.03646891936659813, + 0.03187037259340286, + 0.04857011139392853, + 0.009890343062579632 + ], + "innovation_negative_gradient_cosine": [ + 0.999089241027832, + 0.9990065097808838, + 0.9990898370742798, + 0.9988292455673218, + 0.999106764793396, + 0.9988035559654236, + 0.9991393089294434, + 0.9989110231399536, + 0.999163031578064, + 0.9990224242210388, + 0.9991826415061951, + 0.9991794228553772, + 0.9992508888244629, + 0.9989583492279053, + 0.9992785453796387, + 0.9990521669387817, + 0.9993072748184204, + 0.9991351366043091, + 0.9993723630905151, + 0.9991949796676636, + 0.9994121789932251, + 0.999466598033905, + 0.9995043873786926, + 0.9994280338287354, + 0.999567985534668, + 0.9995042085647583, + 0.9996387362480164, + 0.9997044801712036, + 0.999808669090271, + 0.9998710751533508, + 0.7805278897285461 + ], + "mean_feedback_forward_cosine": 0.9993942726042963, + "mean_feedback_forward_relative_error": 0.034200180532230486, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.999089241027832, + 0.9990065097808838, + 0.9990898370742798, + 0.9988292455673218, + 0.999106764793396, + 0.9988035559654236, + 0.9991393089294434, + 0.9989110231399536, + 0.999163031578064, + 0.9990224242210388, + 0.9991826415061951, + 0.9991794228553772, + 0.9992508888244629, + 0.9989583492279053, + 0.9992785453796387, + 0.9990521669387817, + 0.9993072748184204, + 0.9991351366043091, + 0.9993723630905151, + 0.9991949796676636, + 0.9994121789932251, + 0.999466598033905, + 0.9995043873786926, + 0.9994280338287354, + 0.999567985534668, + 0.9995042085647583, + 0.9996387362480164, + 0.9997044801712036, + 0.999808669090271, + 0.9998710751533508, + 0.7805278897285461 + ], + "teaching_negative_gradient_cosine": [ + 0.999089241027832, + 0.9990065097808838, + 0.9990898370742798, + 0.9988292455673218, + 0.999106764793396, + 0.9988035559654236, + 0.9991393089294434, + 0.9989110231399536, + 0.999163031578064, + 0.9990224242210388, + 0.9991826415061951, + 0.9991794228553772, + 0.9992508888244629, + 0.9989583492279053, + 0.9992785453796387, + 0.9990521669387817, + 0.9993072748184204, + 0.9991351366043091, + 0.9993723630905151, + 0.9991949796676636, + 0.9994121789932251, + 0.999466598033905, + 0.9995043873786926, + 0.9994280338287354, + 0.999567985534668, + 0.9995042085647583, + 0.9996387362480164, + 0.9997044801712036, + 0.999808669090271, + 0.9998710751533508, + 0.7805278897285461 + ], + "wall_s": 0.08741545677185059 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.3352, + "eval_loss": 1.85539453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3970344066619873, + "mean_feedback_forward_cosine": 0.09891654959609432, + "mean_feedback_forward_relative_error": 1.3545093651740783, + "min_feedback_forward_cosine": 0.030816495418548584 + }, + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.0165474932140772 + }, + { + "epoch": 2, + "eval_accuracy": 0.474, + "eval_loss": 1.5613578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3619446754455566, + "mean_feedback_forward_cosine": 0.21640455626672314, + "mean_feedback_forward_relative_error": 1.2635457823353429, + "min_feedback_forward_cosine": 0.08645570278167725 + }, + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.574120481385125 + }, + { + "epoch": 3, + "eval_accuracy": 0.5534, + "eval_loss": 1.43058212890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.31855046749115, + "mean_feedback_forward_cosine": 0.33052211519210567, + "mean_feedback_forward_relative_error": 1.1678868647544616, + "min_feedback_forward_cosine": 0.15054291486740112 + }, + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.2379753240161473 + }, + { + "epoch": 4, + "eval_accuracy": 0.6548, + "eval_loss": 1.0312754150390624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2712821960449219, + "mean_feedback_forward_cosine": 0.42535690578722185, + "mean_feedback_forward_relative_error": 1.0820181158281141, + "min_feedback_forward_cosine": 0.21556095778942108 + }, + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 0.9983540509329902 + }, + { + "epoch": 5, + "eval_accuracy": 0.6956, + "eval_loss": 0.9139252319335938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2214312553405762, + "mean_feedback_forward_cosine": 0.5065497131116928, + "mean_feedback_forward_relative_error": 1.0019216018338357, + "min_feedback_forward_cosine": 0.2794457972049713 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.8451392686208089 + }, + { + "epoch": 6, + "eval_accuracy": 0.7304, + "eval_loss": 0.777873779296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1813551187515259, + "mean_feedback_forward_cosine": 0.559566403588941, + "mean_feedback_forward_relative_error": 0.9465686207817446, + "min_feedback_forward_cosine": 0.32730311155319214 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7229798585573832 + }, + { + "epoch": 7, + "eval_accuracy": 0.7702, + "eval_loss": 0.6828182861328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1461174488067627, + "mean_feedback_forward_cosine": 0.6001128660094354, + "mean_feedback_forward_relative_error": 0.9015945369197477, + "min_feedback_forward_cosine": 0.36858493089675903 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6540972160975138 + }, + { + "epoch": 8, + "eval_accuracy": 0.747, + "eval_loss": 0.7888611328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1137171983718872, + "mean_feedback_forward_cosine": 0.634972870349884, + "mean_feedback_forward_relative_error": 0.8613815144185097, + "min_feedback_forward_cosine": 0.4044458270072937 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.6051028917206658 + }, + { + "epoch": 9, + "eval_accuracy": 0.755, + "eval_loss": 0.8275123046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0851657390594482, + "mean_feedback_forward_cosine": 0.6635892391204834, + "mean_feedback_forward_relative_error": 0.8266697622114613, + "min_feedback_forward_cosine": 0.4357264041900635 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5544356402185228 + }, + { + "epoch": 10, + "eval_accuracy": 0.7566, + "eval_loss": 0.7798774658203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0581560134887695, + "mean_feedback_forward_cosine": 0.689135649511891, + "mean_feedback_forward_relative_error": 0.7944988915997167, + "min_feedback_forward_cosine": 0.4646669626235962 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5214713230662875 + }, + { + "epoch": 11, + "eval_accuracy": 0.8146, + "eval_loss": 0.5679540466308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0313493013381958, + "mean_feedback_forward_cosine": 0.7127979820774447, + "mean_feedback_forward_relative_error": 0.76344284222972, + "min_feedback_forward_cosine": 0.4922131299972534 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.4940733931541443 + }, + { + "epoch": 12, + "eval_accuracy": 0.8034, + "eval_loss": 0.5874515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0037760734558105, + "mean_feedback_forward_cosine": 0.7339891156842632, + "mean_feedback_forward_relative_error": 0.7344710971078565, + "min_feedback_forward_cosine": 0.5191805362701416 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.4745852570427789 + }, + { + "epoch": 13, + "eval_accuracy": 0.772, + "eval_loss": 0.727918994140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9788193106651306, + "mean_feedback_forward_cosine": 0.7526846374234846, + "mean_feedback_forward_relative_error": 0.7078268095370261, + "min_feedback_forward_cosine": 0.5433098077774048 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.447769571240743 + }, + { + "epoch": 14, + "eval_accuracy": 0.8288, + "eval_loss": 0.51325146484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.953144371509552, + "mean_feedback_forward_cosine": 0.7700060529093589, + "mean_feedback_forward_relative_error": 0.6821638220740903, + "min_feedback_forward_cosine": 0.5672977566719055 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.43158293512132434 + }, + { + "epoch": 15, + "eval_accuracy": 0.817, + "eval_loss": 0.5514772705078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9272802472114563, + "mean_feedback_forward_cosine": 0.7858748801292912, + "mean_feedback_forward_relative_error": 0.6579482872639933, + "min_feedback_forward_cosine": 0.590796709060669 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.41741799857351514 + }, + { + "epoch": 16, + "eval_accuracy": 0.8046, + "eval_loss": 0.6012989379882813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9033938050270081, + "mean_feedback_forward_cosine": 0.8003424771370427, + "mean_feedback_forward_relative_error": 0.6350040791496154, + "min_feedback_forward_cosine": 0.6120290756225586 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.39520628892580667 + }, + { + "epoch": 17, + "eval_accuracy": 0.8054, + "eval_loss": 0.6249626220703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8786085247993469, + "mean_feedback_forward_cosine": 0.8139687603519808, + "mean_feedback_forward_relative_error": 0.6127420500401528, + "min_feedback_forward_cosine": 0.6332476139068604 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.38424887357287935 + }, + { + "epoch": 18, + "eval_accuracy": 0.826, + "eval_loss": 0.5176603759765624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8530976176261902, + "mean_feedback_forward_cosine": 0.8264068634279312, + "mean_feedback_forward_relative_error": 0.5914711466720027, + "min_feedback_forward_cosine": 0.654157280921936 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.3791499035358429 + }, + { + "epoch": 19, + "eval_accuracy": 0.844, + "eval_loss": 0.4981465942382812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8279763460159302, + "mean_feedback_forward_cosine": 0.8380863012806061, + "mean_feedback_forward_relative_error": 0.5709285841834161, + "min_feedback_forward_cosine": 0.674217939376831 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.3649848249647352 + }, + { + "epoch": 20, + "eval_accuracy": 0.841, + "eval_loss": 0.4900862609863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8037641644477844, + "mean_feedback_forward_cosine": 0.8487852657994917, + "mean_feedback_forward_relative_error": 0.5513923168182373, + "min_feedback_forward_cosine": 0.6925684809684753 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3498300515757667 + }, + { + "epoch": 21, + "eval_accuracy": 0.8254, + "eval_loss": 0.5741294128417969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7798997163772583, + "mean_feedback_forward_cosine": 0.8588206902627022, + "mean_feedback_forward_relative_error": 0.5324669249596135, + "min_feedback_forward_cosine": 0.7107298374176025 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.3389130860964457 + }, + { + "epoch": 22, + "eval_accuracy": 0.8452, + "eval_loss": 0.48783974609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7567204236984253, + "mean_feedback_forward_cosine": 0.8682854021749189, + "mean_feedback_forward_relative_error": 0.5141304625618842, + "min_feedback_forward_cosine": 0.7275015115737915 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3313982831213209 + }, + { + "epoch": 23, + "eval_accuracy": 0.8432, + "eval_loss": 0.503389306640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7334077954292297, + "mean_feedback_forward_cosine": 0.8772400752190621, + "mean_feedback_forward_relative_error": 0.4961547774653281, + "min_feedback_forward_cosine": 0.7437876462936401 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.32460021476745604 + }, + { + "epoch": 24, + "eval_accuracy": 0.8662, + "eval_loss": 0.4224889404296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.710907518863678, + "mean_feedback_forward_cosine": 0.8852154189540494, + "mean_feedback_forward_relative_error": 0.4794098816571697, + "min_feedback_forward_cosine": 0.7589513063430786 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.3124096696138382 + }, + { + "epoch": 25, + "eval_accuracy": 0.8464, + "eval_loss": 0.5034165222167969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6883212327957153, + "mean_feedback_forward_cosine": 0.8927982334167727, + "mean_feedback_forward_relative_error": 0.46311643960014465, + "min_feedback_forward_cosine": 0.7737209796905518 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.30748651559087964 + }, + { + "epoch": 26, + "eval_accuracy": 0.8578, + "eval_loss": 0.4695193481445312, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6669498682022095, + "mean_feedback_forward_cosine": 0.8997289615292703, + "mean_feedback_forward_relative_error": 0.4476435371944981, + "min_feedback_forward_cosine": 0.7872694730758667 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.29803875002331204 + }, + { + "epoch": 27, + "eval_accuracy": 0.8472, + "eval_loss": 0.49544317626953127, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6458977460861206, + "mean_feedback_forward_cosine": 0.9063345193862915, + "mean_feedback_forward_relative_error": 0.43247731029987335, + "min_feedback_forward_cosine": 0.80039381980896 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.29557929752137924 + }, + { + "epoch": 28, + "eval_accuracy": 0.8522, + "eval_loss": 0.4892141296386719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.626682698726654, + "mean_feedback_forward_cosine": 0.9123510795254861, + "mean_feedback_forward_relative_error": 0.4181007968802606, + "min_feedback_forward_cosine": 0.8119982481002808 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.2796997773885727 + }, + { + "epoch": 29, + "eval_accuracy": 0.8618, + "eval_loss": 0.4524398193359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6067200303077698, + "mean_feedback_forward_cosine": 0.9181783026264559, + "mean_feedback_forward_relative_error": 0.4038001053756283, + "min_feedback_forward_cosine": 0.823555588722229 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.28144939953486126 + }, + { + "epoch": 30, + "eval_accuracy": 0.8616, + "eval_loss": 0.431435400390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5873559713363647, + "mean_feedback_forward_cosine": 0.9236067668084176, + "mean_feedback_forward_relative_error": 0.3899670858536997, + "min_feedback_forward_cosine": 0.8343836069107056 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.27674690407911934 + }, + { + "epoch": 31, + "eval_accuracy": 0.8578, + "eval_loss": 0.4860141845703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5684483051300049, + "mean_feedback_forward_cosine": 0.9285307911134535, + "mean_feedback_forward_relative_error": 0.3769292422840672, + "min_feedback_forward_cosine": 0.8449437618255615 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.2692386422687107 + }, + { + "epoch": 32, + "eval_accuracy": 0.84, + "eval_loss": 0.5316449401855469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5495582222938538, + "mean_feedback_forward_cosine": 0.9332671357739356, + "mean_feedback_forward_relative_error": 0.3640456214066475, + "min_feedback_forward_cosine": 0.8548435568809509 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.26496250779893665 + }, + { + "epoch": 33, + "eval_accuracy": 0.8594, + "eval_loss": 0.46718753662109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5314036011695862, + "mean_feedback_forward_cosine": 0.9376626629983226, + "mean_feedback_forward_relative_error": 0.351678614174166, + "min_feedback_forward_cosine": 0.864128053188324 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.2612344945536719 + }, + { + "epoch": 34, + "eval_accuracy": 0.8486, + "eval_loss": 0.5073669921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5140032768249512, + "mean_feedback_forward_cosine": 0.9416962304422932, + "mean_feedback_forward_relative_error": 0.3399790626379751, + "min_feedback_forward_cosine": 0.8728002309799194 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.25332926568190256 + }, + { + "epoch": 35, + "eval_accuracy": 0.8634, + "eval_loss": 0.4601386779785156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.49668073654174805, + "mean_feedback_forward_cosine": 0.9455604130221952, + "mean_feedback_forward_relative_error": 0.3283394209800228, + "min_feedback_forward_cosine": 0.8810791969299316 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.24765972827275595 + }, + { + "epoch": 36, + "eval_accuracy": 0.8416, + "eval_loss": 0.49482080688476565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48018109798431396, + "mean_feedback_forward_cosine": 0.949139166262842, + "mean_feedback_forward_relative_error": 0.3172376600484694, + "min_feedback_forward_cosine": 0.8887827396392822 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.24774828658103942 + }, + { + "epoch": 37, + "eval_accuracy": 0.858, + "eval_loss": 0.45331826782226564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4640062749385834, + "mean_feedback_forward_cosine": 0.9525774159739094, + "mean_feedback_forward_relative_error": 0.30619221156643284, + "min_feedback_forward_cosine": 0.8959776163101196 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.24434037954939736 + }, + { + "epoch": 38, + "eval_accuracy": 0.858, + "eval_loss": 0.45240281982421876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4489477574825287, + "mean_feedback_forward_cosine": 0.9556983863153765, + "mean_feedback_forward_relative_error": 0.29581394238818076, + "min_feedback_forward_cosine": 0.9025242328643799 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.23976409662034776 + }, + { + "epoch": 39, + "eval_accuracy": 0.8412, + "eval_loss": 0.5876475219726562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4343760013580322, + "mean_feedback_forward_cosine": 0.9584913234556874, + "mean_feedback_forward_relative_error": 0.2862166019697343, + "min_feedback_forward_cosine": 0.9086928963661194 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.23123540512985655 + }, + { + "epoch": 40, + "eval_accuracy": 0.8564, + "eval_loss": 0.4765188659667969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41979628801345825, + "mean_feedback_forward_cosine": 0.9612213392411509, + "mean_feedback_forward_relative_error": 0.2765878001528402, + "min_feedback_forward_cosine": 0.9146870970726013 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.2318395030286577 + }, + { + "epoch": 41, + "eval_accuracy": 0.8616, + "eval_loss": 0.44323438720703123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40531429648399353, + "mean_feedback_forward_cosine": 0.9637255207184823, + "mean_feedback_forward_relative_error": 0.26743238106850653, + "min_feedback_forward_cosine": 0.9204114675521851 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.22619207657972973 + }, + { + "epoch": 42, + "eval_accuracy": 0.8578, + "eval_loss": 0.4867693237304688, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.391878217458725, + "mean_feedback_forward_cosine": 0.966075841457613, + "mean_feedback_forward_relative_error": 0.2584934775387087, + "min_feedback_forward_cosine": 0.9255408644676208 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.22141475805441538 + }, + { + "epoch": 43, + "eval_accuracy": 0.8746, + "eval_loss": 0.41213989868164064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3782900869846344, + "mean_feedback_forward_cosine": 0.9682946012866113, + "mean_feedback_forward_relative_error": 0.2497686491377892, + "min_feedback_forward_cosine": 0.9305530786514282 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.2203050850868225 + }, + { + "epoch": 44, + "eval_accuracy": 0.8488, + "eval_loss": 0.5225678588867188, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36502012610435486, + "mean_feedback_forward_cosine": 0.9704237772572425, + "mean_feedback_forward_relative_error": 0.2411658537964667, + "min_feedback_forward_cosine": 0.9352737665176392 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.21957209897041322 + }, + { + "epoch": 45, + "eval_accuracy": 0.849, + "eval_loss": 0.5496326416015626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35209617018699646, + "mean_feedback_forward_cosine": 0.9723464500519537, + "mean_feedback_forward_relative_error": 0.23312005520828308, + "min_feedback_forward_cosine": 0.9396935701370239 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.21895495686531066 + }, + { + "epoch": 46, + "eval_accuracy": 0.86, + "eval_loss": 0.4849064758300781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33995580673217773, + "mean_feedback_forward_cosine": 0.9741548203652904, + "mean_feedback_forward_relative_error": 0.2253057812490771, + "min_feedback_forward_cosine": 0.9437055587768555 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.21140181116792892 + }, + { + "epoch": 47, + "eval_accuracy": 0.8596, + "eval_loss": 0.48711629028320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32750967144966125, + "mean_feedback_forward_cosine": 0.9759039148207633, + "mean_feedback_forward_relative_error": 0.21750406536363787, + "min_feedback_forward_cosine": 0.9477157592773438 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.21739240936703152 + }, + { + "epoch": 48, + "eval_accuracy": 0.8674, + "eval_loss": 0.49024164428710937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3152298629283905, + "mean_feedback_forward_cosine": 0.9775446115001556, + "mean_feedback_forward_relative_error": 0.20989031272549782, + "min_feedback_forward_cosine": 0.9515104293823242 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.21653300149705676 + }, + { + "epoch": 49, + "eval_accuracy": 0.8672, + "eval_loss": 0.4632137329101563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30373454093933105, + "mean_feedback_forward_cosine": 0.9790571889569682, + "mean_feedback_forward_relative_error": 0.20261672836157582, + "min_feedback_forward_cosine": 0.954938530921936 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.20910456528663635 + }, + { + "epoch": 50, + "eval_accuracy": 0.8686, + "eval_loss": 0.44734768676757813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2930586338043213, + "mean_feedback_forward_cosine": 0.980450362928452, + "mean_feedback_forward_relative_error": 0.19571300307589193, + "min_feedback_forward_cosine": 0.958027184009552 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.20682951056427426 + }, + { + "epoch": 51, + "eval_accuracy": 0.865, + "eval_loss": 0.4583529357910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28304439783096313, + "mean_feedback_forward_cosine": 0.9817609517805038, + "mean_feedback_forward_relative_error": 0.18900507039600803, + "min_feedback_forward_cosine": 0.9608170986175537 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.20303286356396144 + }, + { + "epoch": 52, + "eval_accuracy": 0.8758, + "eval_loss": 0.42540660400390623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27335530519485474, + "mean_feedback_forward_cosine": 0.9829263110314647, + "mean_feedback_forward_relative_error": 0.18280399518628274, + "min_feedback_forward_cosine": 0.9634319543838501 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.1975233178668552 + }, + { + "epoch": 53, + "eval_accuracy": 0.8726, + "eval_loss": 0.45839552612304685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2639857232570648, + "mean_feedback_forward_cosine": 0.9840461400247389, + "mean_feedback_forward_relative_error": 0.1766628747986209, + "min_feedback_forward_cosine": 0.9658682346343994 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.19599860397444832 + }, + { + "epoch": 54, + "eval_accuracy": 0.854, + "eval_loss": 0.5225734069824218, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25491610169410706, + "mean_feedback_forward_cosine": 0.9850916497169002, + "mean_feedback_forward_relative_error": 0.17073152082100992, + "min_feedback_forward_cosine": 0.9681482911109924 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.19324108230802747 + }, + { + "epoch": 55, + "eval_accuracy": 0.8656, + "eval_loss": 0.4879412841796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24540641903877258, + "mean_feedback_forward_cosine": 0.9861299530152352, + "mean_feedback_forward_relative_error": 0.1646571411721168, + "min_feedback_forward_cosine": 0.9704469442367554 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.20186779841846889 + }, + { + "epoch": 56, + "eval_accuracy": 0.8788, + "eval_loss": 0.4124189697265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23672057688236237, + "mean_feedback_forward_cosine": 0.9870482760090982, + "mean_feedback_forward_relative_error": 0.15906442100963286, + "min_feedback_forward_cosine": 0.9724932312965393 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.1911707753446367 + }, + { + "epoch": 57, + "eval_accuracy": 0.8544, + "eval_loss": 0.4946718139648438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22838672995567322, + "mean_feedback_forward_cosine": 0.9879102610772655, + "mean_feedback_forward_relative_error": 0.1536721505465046, + "min_feedback_forward_cosine": 0.9743702411651611 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.19003599887424044 + }, + { + "epoch": 58, + "eval_accuracy": 0.8834, + "eval_loss": 0.4138064086914062, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22036190330982208, + "mean_feedback_forward_cosine": 0.9887177309682292, + "mean_feedback_forward_relative_error": 0.14841678906832972, + "min_feedback_forward_cosine": 0.9761249423027039 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.19302229064305623 + }, + { + "epoch": 59, + "eval_accuracy": 0.862, + "eval_loss": 0.49976304931640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21271266043186188, + "mean_feedback_forward_cosine": 0.9894644348852096, + "mean_feedback_forward_relative_error": 0.14339836915173837, + "min_feedback_forward_cosine": 0.9777436852455139 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.1886476122485267 + }, + { + "epoch": 60, + "eval_accuracy": 0.886, + "eval_loss": 0.3878099365234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20517779886722565, + "mean_feedback_forward_cosine": 0.9901608478638434, + "mean_feedback_forward_relative_error": 0.1385545569562143, + "min_feedback_forward_cosine": 0.9792742133140564 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.184096456334326 + }, + { + "epoch": 61, + "eval_accuracy": 0.87, + "eval_loss": 0.45004430541992185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1976340413093567, + "mean_feedback_forward_cosine": 0.9908320807641552, + "mean_feedback_forward_relative_error": 0.13373789431587343, + "min_feedback_forward_cosine": 0.9807661771774292 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.1952857186794281 + }, + { + "epoch": 62, + "eval_accuracy": 0.8708, + "eval_loss": 0.42280288696289064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19097411632537842, + "mean_feedback_forward_cosine": 0.9914130280094762, + "mean_feedback_forward_relative_error": 0.12940076714561832, + "min_feedback_forward_cosine": 0.9820346832275391 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.1781928380449613 + }, + { + "epoch": 63, + "eval_accuracy": 0.8804, + "eval_loss": 0.40265328369140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18434743583202362, + "mean_feedback_forward_cosine": 0.9919711351394653, + "mean_feedback_forward_relative_error": 0.125099069769344, + "min_feedback_forward_cosine": 0.9832502007484436 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.18024166654480828 + }, + { + "epoch": 64, + "eval_accuracy": 0.8768, + "eval_loss": 0.4114610107421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17806272208690643, + "mean_feedback_forward_cosine": 0.9924912568061582, + "mean_feedback_forward_relative_error": 0.12095758090576818, + "min_feedback_forward_cosine": 0.9843647480010986 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.17777254914310242 + }, + { + "epoch": 65, + "eval_accuracy": 0.8712, + "eval_loss": 0.4435559143066406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1717073768377304, + "mean_feedback_forward_cosine": 0.9929938431709043, + "mean_feedback_forward_relative_error": 0.1168181098276569, + "min_feedback_forward_cosine": 0.9854501485824585 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.18547025267018213 + }, + { + "epoch": 66, + "eval_accuracy": 0.859, + "eval_loss": 0.5121481506347656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16578994691371918, + "mean_feedback_forward_cosine": 0.9934470845806983, + "mean_feedback_forward_relative_error": 0.11296291853631696, + "min_feedback_forward_cosine": 0.986427903175354 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.1755339354938931 + }, + { + "epoch": 67, + "eval_accuracy": 0.855, + "eval_loss": 0.5351258056640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15990713238716125, + "mean_feedback_forward_cosine": 0.993884357713884, + "mean_feedback_forward_relative_error": 0.10910652541825848, + "min_feedback_forward_cosine": 0.9873668551445007 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.1767852502849367 + }, + { + "epoch": 68, + "eval_accuracy": 0.8736, + "eval_loss": 0.44542548828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15426933765411377, + "mean_feedback_forward_cosine": 0.9942979985667814, + "mean_feedback_forward_relative_error": 0.10534040029010465, + "min_feedback_forward_cosine": 0.9882389307022095 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.17708395219379 + }, + { + "epoch": 69, + "eval_accuracy": 0.8866, + "eval_loss": 0.37270454711914064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14849302172660828, + "mean_feedback_forward_cosine": 0.9946968151677039, + "mean_feedback_forward_relative_error": 0.10158244051760243, + "min_feedback_forward_cosine": 0.9890974164009094 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.18407729478412205 + }, + { + "epoch": 70, + "eval_accuracy": 0.8848, + "eval_loss": 0.3816723022460938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14344274997711182, + "mean_feedback_forward_cosine": 0.9950423971299203, + "mean_feedback_forward_relative_error": 0.09819682723572178, + "min_feedback_forward_cosine": 0.989820122718811 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.17256281439993115 + }, + { + "epoch": 71, + "eval_accuracy": 0.871, + "eval_loss": 0.47614798583984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13826242089271545, + "mean_feedback_forward_cosine": 0.9953846220047243, + "mean_feedback_forward_relative_error": 0.09473993376858773, + "min_feedback_forward_cosine": 0.9905393719673157 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.17987764293617672 + }, + { + "epoch": 72, + "eval_accuracy": 0.8598, + "eval_loss": 0.46955314331054687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13329215347766876, + "mean_feedback_forward_cosine": 0.9956990288149926, + "mean_feedback_forward_relative_error": 0.09145044511364352, + "min_feedback_forward_cosine": 0.991203784942627 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.17530702959431543 + }, + { + "epoch": 73, + "eval_accuracy": 0.8896, + "eval_loss": 0.37162421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12890847027301788, + "mean_feedback_forward_cosine": 0.99597418115985, + "mean_feedback_forward_relative_error": 0.08846384130658642, + "min_feedback_forward_cosine": 0.9917711019515991 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.16974435313542685 + }, + { + "epoch": 74, + "eval_accuracy": 0.8748, + "eval_loss": 0.4273910888671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12445683777332306, + "mean_feedback_forward_cosine": 0.9962359916779303, + "mean_feedback_forward_relative_error": 0.0855239714465795, + "min_feedback_forward_cosine": 0.9923257827758789 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.16758987526098887 + }, + { + "epoch": 75, + "eval_accuracy": 0.8744, + "eval_loss": 0.44162723388671876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1202189177274704, + "mean_feedback_forward_cosine": 0.996487186801049, + "mean_feedback_forward_relative_error": 0.08261592692184833, + "min_feedback_forward_cosine": 0.9928371906280518 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.16726809644434187 + }, + { + "epoch": 76, + "eval_accuracy": 0.8748, + "eval_loss": 0.4214301574707031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11572326719760895, + "mean_feedback_forward_cosine": 0.9967352132643422, + "mean_feedback_forward_relative_error": 0.07963655528522307, + "min_feedback_forward_cosine": 0.993360161781311 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.1751393345170551 + }, + { + "epoch": 77, + "eval_accuracy": 0.8956, + "eval_loss": 0.34109723510742185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11186333745718002, + "mean_feedback_forward_cosine": 0.9969496880808184, + "mean_feedback_forward_relative_error": 0.07697323192992518, + "min_feedback_forward_cosine": 0.9937939643859863 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.16733670739597745 + }, + { + "epoch": 78, + "eval_accuracy": 0.872, + "eval_loss": 0.44417415771484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10790038108825684, + "mean_feedback_forward_cosine": 0.9971546011586343, + "mean_feedback_forward_relative_error": 0.07434488015790139, + "min_feedback_forward_cosine": 0.9942234754562378 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.17388162036736807 + }, + { + "epoch": 79, + "eval_accuracy": 0.8782, + "eval_loss": 0.438529736328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10394710302352905, + "mean_feedback_forward_cosine": 0.997349654474566, + "mean_feedback_forward_relative_error": 0.07174537811548479, + "min_feedback_forward_cosine": 0.9946376085281372 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.1652422828303443 + }, + { + "epoch": 80, + "eval_accuracy": 0.8816, + "eval_loss": 0.44652198486328126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10049350559711456, + "mean_feedback_forward_cosine": 0.9975251978443515, + "mean_feedback_forward_relative_error": 0.06931953246314679, + "min_feedback_forward_cosine": 0.9949864745140076 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.16283351455264622 + }, + { + "epoch": 81, + "eval_accuracy": 0.8752, + "eval_loss": 0.4451551513671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09698614478111267, + "mean_feedback_forward_cosine": 0.9976857150754621, + "mean_feedback_forward_relative_error": 0.06702608723313577, + "min_feedback_forward_cosine": 0.9953293800354004 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.1627018078883489 + }, + { + "epoch": 82, + "eval_accuracy": 0.8712, + "eval_loss": 0.4183915771484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09360911697149277, + "mean_feedback_forward_cosine": 0.9978386798212605, + "mean_feedback_forward_relative_error": 0.06476938940825001, + "min_feedback_forward_cosine": 0.9956478476524353 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.1625231981145011 + }, + { + "epoch": 83, + "eval_accuracy": 0.864, + "eval_loss": 0.479559423828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09036646038293839, + "mean_feedback_forward_cosine": 0.9979804735029897, + "mean_feedback_forward_relative_error": 0.06260339933778009, + "min_feedback_forward_cosine": 0.9959428310394287 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.15952994990878636 + }, + { + "epoch": 84, + "eval_accuracy": 0.8664, + "eval_loss": 0.5053081115722656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08696158975362778, + "mean_feedback_forward_cosine": 0.9981216576791578, + "mean_feedback_forward_relative_error": 0.06037566810846329, + "min_feedback_forward_cosine": 0.9962424039840698 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.16951513698101045 + }, + { + "epoch": 85, + "eval_accuracy": 0.878, + "eval_loss": 0.46674307861328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08394859731197357, + "mean_feedback_forward_cosine": 0.9982455045946182, + "mean_feedback_forward_relative_error": 0.05834392362063931, + "min_feedback_forward_cosine": 0.996497392654419 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.16335849794281854 + }, + { + "epoch": 86, + "eval_accuracy": 0.8914, + "eval_loss": 0.3704887084960938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08105872571468353, + "mean_feedback_forward_cosine": 0.9983624296803628, + "mean_feedback_forward_relative_error": 0.05636267388059247, + "min_feedback_forward_cosine": 0.9967339634895325 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.15780939542982314 + }, + { + "epoch": 87, + "eval_accuracy": 0.8826, + "eval_loss": 0.3811180114746094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0780579075217247, + "mean_feedback_forward_cosine": 0.9984782568870052, + "mean_feedback_forward_relative_error": 0.054333017117554144, + "min_feedback_forward_cosine": 0.9969696998596191 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.16681171383327908 + }, + { + "epoch": 88, + "eval_accuracy": 0.8872, + "eval_loss": 0.4038883117675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07522793114185333, + "mean_feedback_forward_cosine": 0.998582543865327, + "mean_feedback_forward_relative_error": 0.05243379346305324, + "min_feedback_forward_cosine": 0.9971845746040344 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.16635543229844835 + }, + { + "epoch": 89, + "eval_accuracy": 0.8768, + "eval_loss": 0.4656008239746094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07262639701366425, + "mean_feedback_forward_cosine": 0.9986755482612117, + "mean_feedback_forward_relative_error": 0.05068007844590371, + "min_feedback_forward_cosine": 0.9973758459091187 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.15748041656017303 + }, + { + "epoch": 90, + "eval_accuracy": 0.8846, + "eval_loss": 0.4177832824707031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06999342143535614, + "mean_feedback_forward_cosine": 0.9987672355867201, + "mean_feedback_forward_relative_error": 0.048894857086481586, + "min_feedback_forward_cosine": 0.9975628852844238 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.16211600266562567 + }, + { + "epoch": 91, + "eval_accuracy": 0.8716, + "eval_loss": 0.4431572265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06744895130395889, + "mean_feedback_forward_cosine": 0.9988512204539391, + "mean_feedback_forward_relative_error": 0.04720152908515546, + "min_feedback_forward_cosine": 0.997735857963562 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.16203772835466596 + }, + { + "epoch": 92, + "eval_accuracy": 0.8826, + "eval_loss": 0.40832467041015624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06500688195228577, + "mean_feedback_forward_cosine": 0.9989292698521768, + "mean_feedback_forward_relative_error": 0.04557065719798688, + "min_feedback_forward_cosine": 0.9978965520858765 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.15830527657402887 + }, + { + "epoch": 93, + "eval_accuracy": 0.8902, + "eval_loss": 0.37456282958984377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06275323033332825, + "mean_feedback_forward_cosine": 0.9990006685256958, + "mean_feedback_forward_relative_error": 0.04402178369702831, + "min_feedback_forward_cosine": 0.9980393648147583 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.1551544521861606 + }, + { + "epoch": 94, + "eval_accuracy": 0.8852, + "eval_loss": 0.43639714965820314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0605853907763958, + "mean_feedback_forward_cosine": 0.9990669758089127, + "mean_feedback_forward_relative_error": 0.04253085158885487, + "min_feedback_forward_cosine": 0.998172402381897 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.15768385762373607 + }, + { + "epoch": 95, + "eval_accuracy": 0.8828, + "eval_loss": 0.42318448486328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05844554677605629, + "mean_feedback_forward_cosine": 0.9991313084479301, + "mean_feedback_forward_relative_error": 0.04103709913549885, + "min_feedback_forward_cosine": 0.9982991218566895 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.1587442197839419 + }, + { + "epoch": 96, + "eval_accuracy": 0.8878, + "eval_loss": 0.4036041809082031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05644100531935692, + "mean_feedback_forward_cosine": 0.9991878405694039, + "mean_feedback_forward_relative_error": 0.03967912615306916, + "min_feedback_forward_cosine": 0.9984134435653687 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.15544112352265252 + }, + { + "epoch": 97, + "eval_accuracy": 0.8914, + "eval_loss": 0.37009312133789063, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.054409269243478775, + "mean_feedback_forward_cosine": 0.9992446053412652, + "mean_feedback_forward_relative_error": 0.038267940523162965, + "min_feedback_forward_cosine": 0.9985255599021912 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.15759313933319516 + }, + { + "epoch": 98, + "eval_accuracy": 0.88, + "eval_loss": 0.4166911987304687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05243669077754021, + "mean_feedback_forward_cosine": 0.9992959845450616, + "mean_feedback_forward_relative_error": 0.03694337375101543, + "min_feedback_forward_cosine": 0.9986302256584167 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.1612001913335588 + }, + { + "epoch": 99, + "eval_accuracy": 0.8714, + "eval_loss": 0.4708991271972656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05059374123811722, + "mean_feedback_forward_cosine": 0.9993423665723493, + "mean_feedback_forward_relative_error": 0.035702559405997876, + "min_feedback_forward_cosine": 0.9987248778343201 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.15767538166840872 + }, + { + "epoch": 100, + "eval_accuracy": 0.8808, + "eval_loss": 0.450357470703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04881930351257324, + "mean_feedback_forward_cosine": 0.9993862932728182, + "mean_feedback_forward_relative_error": 0.03448866133487994, + "min_feedback_forward_cosine": 0.9988125562667847 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.15321411666737664 + }, + { + "epoch": 101, + "eval_accuracy": 0.9136, + "eval_loss": 0.29533384399414064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048803918063640594, + "mean_feedback_forward_cosine": 0.9993866758961831, + "mean_feedback_forward_relative_error": 0.03447543967875742, + "min_feedback_forward_cosine": 0.998813271522522 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.09057808789942 + }, + { + "epoch": 102, + "eval_accuracy": 0.9196, + "eval_loss": 0.28288860778808594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048796601593494415, + "mean_feedback_forward_cosine": 0.9993869008556489, + "mean_feedback_forward_relative_error": 0.03446682533549686, + "min_feedback_forward_cosine": 0.9988136291503906 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.06322400026718776 + }, + { + "epoch": 103, + "eval_accuracy": 0.9212, + "eval_loss": 0.2858207183837891, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04879026859998703, + "mean_feedback_forward_cosine": 0.9993870719786613, + "mean_feedback_forward_relative_error": 0.03445892669861355, + "min_feedback_forward_cosine": 0.9988139867782593 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.0535766043484211 + }, + { + "epoch": 104, + "eval_accuracy": 0.9246, + "eval_loss": 0.2870035583496094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04878412187099457, + "mean_feedback_forward_cosine": 0.9993872354107518, + "mean_feedback_forward_relative_error": 0.03445159391530098, + "min_feedback_forward_cosine": 0.9988143444061279 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.04680160608887672 + }, + { + "epoch": 105, + "eval_accuracy": 0.9246, + "eval_loss": 0.29164626159667967, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048778191208839417, + "mean_feedback_forward_cosine": 0.9993874142246861, + "mean_feedback_forward_relative_error": 0.034444511237163696, + "min_feedback_forward_cosine": 0.9988146424293518 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.043014900404877134 + }, + { + "epoch": 106, + "eval_accuracy": 0.9218, + "eval_loss": 0.29402776794433594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04877280443906784, + "mean_feedback_forward_cosine": 0.9993875834249681, + "mean_feedback_forward_relative_error": 0.03443780937983144, + "min_feedback_forward_cosine": 0.9988148212432861 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.03942505518330468 + }, + { + "epoch": 107, + "eval_accuracy": 0.9238, + "eval_loss": 0.30112089538574216, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04876742511987686, + "mean_feedback_forward_cosine": 0.9993877507025196, + "mean_feedback_forward_relative_error": 0.03443125629376981, + "min_feedback_forward_cosine": 0.99881511926651 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.035907337534758776 + }, + { + "epoch": 108, + "eval_accuracy": 0.923, + "eval_loss": 0.3071636474609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04876227304339409, + "mean_feedback_forward_cosine": 0.999387917980071, + "mean_feedback_forward_relative_error": 0.03442491798271095, + "min_feedback_forward_cosine": 0.9988153576850891 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.03364239944087134 + }, + { + "epoch": 109, + "eval_accuracy": 0.922, + "eval_loss": 0.3126777282714844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048756614327430725, + "mean_feedback_forward_cosine": 0.9993880660303177, + "mean_feedback_forward_relative_error": 0.03441854629425272, + "min_feedback_forward_cosine": 0.998815655708313 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.03245734071003066 + }, + { + "epoch": 110, + "eval_accuracy": 0.9204, + "eval_loss": 0.30980113525390623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04875122755765915, + "mean_feedback_forward_cosine": 0.9993882583033654, + "mean_feedback_forward_relative_error": 0.034412265725193486, + "min_feedback_forward_cosine": 0.9988158941268921 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.031768651305966907 + }, + { + "epoch": 111, + "eval_accuracy": 0.924, + "eval_loss": 0.31072572021484374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04874620586633682, + "mean_feedback_forward_cosine": 0.9993884063536121, + "mean_feedback_forward_relative_error": 0.0344061431805453, + "min_feedback_forward_cosine": 0.9988161325454712 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.029296634758843317 + }, + { + "epoch": 112, + "eval_accuracy": 0.9254, + "eval_loss": 0.31651325073242187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0487414114177227, + "mean_feedback_forward_cosine": 0.9993885736311635, + "mean_feedback_forward_relative_error": 0.03440025157385295, + "min_feedback_forward_cosine": 0.9988164901733398 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.026635085297955406 + }, + { + "epoch": 113, + "eval_accuracy": 0.9234, + "eval_loss": 0.3184968017578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04873606190085411, + "mean_feedback_forward_cosine": 0.9993887293723321, + "mean_feedback_forward_relative_error": 0.03439420197279223, + "min_feedback_forward_cosine": 0.9988166093826294 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.02689870360626115 + }, + { + "epoch": 114, + "eval_accuracy": 0.9246, + "eval_loss": 0.3188640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04873174428939819, + "mean_feedback_forward_cosine": 0.9993888601180045, + "mean_feedback_forward_relative_error": 0.03438880129326736, + "min_feedback_forward_cosine": 0.9988168478012085 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.02336271217018366 + }, + { + "epoch": 115, + "eval_accuracy": 0.9254, + "eval_loss": 0.3224325378417969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04872721806168556, + "mean_feedback_forward_cosine": 0.9993890216273646, + "mean_feedback_forward_relative_error": 0.0343832528038371, + "min_feedback_forward_cosine": 0.9988170862197876 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.022886741083198124 + }, + { + "epoch": 116, + "eval_accuracy": 0.9242, + "eval_loss": 0.3264565856933594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048722509294748306, + "mean_feedback_forward_cosine": 0.9993891754458027, + "mean_feedback_forward_relative_error": 0.034377753614417965, + "min_feedback_forward_cosine": 0.9988173246383667 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.02161789578249057 + }, + { + "epoch": 117, + "eval_accuracy": 0.9246, + "eval_loss": 0.3350888854980469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048718128353357315, + "mean_feedback_forward_cosine": 0.9993893273415104, + "mean_feedback_forward_relative_error": 0.034372437234607436, + "min_feedback_forward_cosine": 0.9988175630569458 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.02008457115491231 + }, + { + "epoch": 118, + "eval_accuracy": 0.9222, + "eval_loss": 0.3312977966308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04871320724487305, + "mean_feedback_forward_cosine": 0.9993894561644523, + "mean_feedback_forward_relative_error": 0.034366853925729954, + "min_feedback_forward_cosine": 0.9988176822662354 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.021647869668073125 + }, + { + "epoch": 119, + "eval_accuracy": 0.9222, + "eval_loss": 0.33949063720703127, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0487087108194828, + "mean_feedback_forward_cosine": 0.9993896061374296, + "mean_feedback_forward_relative_error": 0.034361511408802003, + "min_feedback_forward_cosine": 0.998818039894104 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.02027899324049552 + }, + { + "epoch": 120, + "eval_accuracy": 0.921, + "eval_loss": 0.34436965942382813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04870382696390152, + "mean_feedback_forward_cosine": 0.9993897695695201, + "mean_feedback_forward_relative_error": 0.03435609096120442, + "min_feedback_forward_cosine": 0.9988182187080383 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.01882550497353077 + }, + { + "epoch": 121, + "eval_accuracy": 0.9208, + "eval_loss": 0.342112255859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04869933798909187, + "mean_feedback_forward_cosine": 0.9993899022379229, + "mean_feedback_forward_relative_error": 0.034350986412215616, + "min_feedback_forward_cosine": 0.9988184571266174 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.01746830699750119 + }, + { + "epoch": 122, + "eval_accuracy": 0.924, + "eval_loss": 0.34617767333984373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04869527742266655, + "mean_feedback_forward_cosine": 0.999390056056361, + "mean_feedback_forward_relative_error": 0.03434607020068553, + "min_feedback_forward_cosine": 0.9988187551498413 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.016621154091341628 + }, + { + "epoch": 123, + "eval_accuracy": 0.923, + "eval_loss": 0.3514703369140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04869047924876213, + "mean_feedback_forward_cosine": 0.9993902041066077, + "mean_feedback_forward_relative_error": 0.03434080309084346, + "min_feedback_forward_cosine": 0.9988188743591309 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.01763997759670019 + }, + { + "epoch": 124, + "eval_accuracy": 0.9196, + "eval_loss": 0.3572915588378906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04868617281317711, + "mean_feedback_forward_cosine": 0.9993903194704363, + "mean_feedback_forward_relative_error": 0.034335700164158496, + "min_feedback_forward_cosine": 0.9988191723823547 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.01719067797842953 + }, + { + "epoch": 125, + "eval_accuracy": 0.9222, + "eval_loss": 0.35623099365234373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04868151620030403, + "mean_feedback_forward_cosine": 0.9993904636752221, + "mean_feedback_forward_relative_error": 0.034330394599706895, + "min_feedback_forward_cosine": 0.9988192915916443 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.017087505733304552 + }, + { + "epoch": 126, + "eval_accuracy": 0.9236, + "eval_loss": 0.35596282958984377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04867681860923767, + "mean_feedback_forward_cosine": 0.9993906117254688, + "mean_feedback_forward_relative_error": 0.03432521590542409, + "min_feedback_forward_cosine": 0.9988195300102234 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.01596804515255822 + }, + { + "epoch": 127, + "eval_accuracy": 0.9226, + "eval_loss": 0.356564501953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04867197573184967, + "mean_feedback_forward_cosine": 0.9993907713120983, + "mean_feedback_forward_relative_error": 0.034319929628362576, + "min_feedback_forward_cosine": 0.9988197088241577 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.01586283933967352 + }, + { + "epoch": 128, + "eval_accuracy": 0.9218, + "eval_loss": 0.3588898681640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0486680269241333, + "mean_feedback_forward_cosine": 0.9993909020577708, + "mean_feedback_forward_relative_error": 0.034315337787472436, + "min_feedback_forward_cosine": 0.9988199472427368 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.013547314140200615 + }, + { + "epoch": 129, + "eval_accuracy": 0.9224, + "eval_loss": 0.36866251831054686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04866375774145126, + "mean_feedback_forward_cosine": 0.9993910404943651, + "mean_feedback_forward_relative_error": 0.034310508098813794, + "min_feedback_forward_cosine": 0.9988202452659607 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.014398164713051585 + }, + { + "epoch": 130, + "eval_accuracy": 0.9218, + "eval_loss": 0.365063037109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04865895211696625, + "mean_feedback_forward_cosine": 0.9993912231537604, + "mean_feedback_forward_relative_error": 0.03430531098837814, + "min_feedback_forward_cosine": 0.9988205432891846 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.01393503595052494 + }, + { + "epoch": 131, + "eval_accuracy": 0.9254, + "eval_loss": 0.3608143493652344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04865493252873421, + "mean_feedback_forward_cosine": 0.999391338517589, + "mean_feedback_forward_relative_error": 0.03430056653075641, + "min_feedback_forward_cosine": 0.9988206624984741 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.012824879109197192 + }, + { + "epoch": 132, + "eval_accuracy": 0.9224, + "eval_loss": 0.3637261047363281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04865088686347008, + "mean_feedback_forward_cosine": 0.9993914654178004, + "mean_feedback_forward_relative_error": 0.03429574687634745, + "min_feedback_forward_cosine": 0.9988207817077637 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.012399559729629092 + }, + { + "epoch": 133, + "eval_accuracy": 0.9222, + "eval_loss": 0.36333014526367186, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048646971583366394, + "mean_feedback_forward_cosine": 0.9993916269271604, + "mean_feedback_forward_relative_error": 0.03429104835395851, + "min_feedback_forward_cosine": 0.998820960521698 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.012052169627075395 + }, + { + "epoch": 134, + "eval_accuracy": 0.9244, + "eval_loss": 0.3689380126953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048642974346876144, + "mean_feedback_forward_cosine": 0.9993917595955634, + "mean_feedback_forward_relative_error": 0.03428640291695633, + "min_feedback_forward_cosine": 0.9988212585449219 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.01183781708114677 + }, + { + "epoch": 135, + "eval_accuracy": 0.9242, + "eval_loss": 0.36527835083007815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04863850027322769, + "mean_feedback_forward_cosine": 0.9993918884185052, + "mean_feedback_forward_relative_error": 0.034281394594619354, + "min_feedback_forward_cosine": 0.9988214373588562 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.012423885153399573 + }, + { + "epoch": 136, + "eval_accuracy": 0.9228, + "eval_loss": 0.37191195068359373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048634033650159836, + "mean_feedback_forward_cosine": 0.9993920499278653, + "mean_feedback_forward_relative_error": 0.03427618378472905, + "min_feedback_forward_cosine": 0.9988217353820801 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.011888691175646253 + }, + { + "epoch": 137, + "eval_accuracy": 0.9252, + "eval_loss": 0.3752788452148437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04862968251109123, + "mean_feedback_forward_cosine": 0.999392197978112, + "mean_feedback_forward_relative_error": 0.03427136731484244, + "min_feedback_forward_cosine": 0.9988219141960144 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.011469858308798738 + }, + { + "epoch": 138, + "eval_accuracy": 0.9234, + "eval_loss": 0.3746169250488281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04862533137202263, + "mean_feedback_forward_cosine": 0.9993923421828977, + "mean_feedback_forward_relative_error": 0.03426646318046316, + "min_feedback_forward_cosine": 0.9988220930099487 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.011192755055841472 + }, + { + "epoch": 139, + "eval_accuracy": 0.923, + "eval_loss": 0.3725655700683594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04862077161669731, + "mean_feedback_forward_cosine": 0.9993924863876835, + "mean_feedback_forward_relative_error": 0.034261413399250276, + "min_feedback_forward_cosine": 0.9988223910331726 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.011524058209856351 + }, + { + "epoch": 140, + "eval_accuracy": 0.9226, + "eval_loss": 0.37775662841796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04861658811569214, + "mean_feedback_forward_cosine": 0.9993926363606607, + "mean_feedback_forward_relative_error": 0.03425657896385077, + "min_feedback_forward_cosine": 0.9988225698471069 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.010894056089719136 + }, + { + "epoch": 141, + "eval_accuracy": 0.9254, + "eval_loss": 0.3685088928222656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048611849546432495, + "mean_feedback_forward_cosine": 0.9993928036382121, + "mean_feedback_forward_relative_error": 0.03425116128017826, + "min_feedback_forward_cosine": 0.9988226294517517 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.011159929553253784 + }, + { + "epoch": 142, + "eval_accuracy": 0.9258, + "eval_loss": 0.3666699462890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048607733100652695, + "mean_feedback_forward_cosine": 0.9993929363066151, + "mean_feedback_forward_relative_error": 0.03424643322585091, + "min_feedback_forward_cosine": 0.9988230466842651 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.010106286029020944 + }, + { + "epoch": 143, + "eval_accuracy": 0.9222, + "eval_loss": 0.3864923645019531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0486031174659729, + "mean_feedback_forward_cosine": 0.9993930862795922, + "mean_feedback_forward_relative_error": 0.034241410603206006, + "min_feedback_forward_cosine": 0.9988232254981995 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.010455552863805659 + }, + { + "epoch": 144, + "eval_accuracy": 0.923, + "eval_loss": 0.37877412109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04859888553619385, + "mean_feedback_forward_cosine": 0.9993932766299094, + "mean_feedback_forward_relative_error": 0.034236317200045434, + "min_feedback_forward_cosine": 0.998823344707489 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.010130042814215024 + }, + { + "epoch": 145, + "eval_accuracy": 0.9224, + "eval_loss": 0.38209491577148436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048594359308481216, + "mean_feedback_forward_cosine": 0.9993934035301208, + "mean_feedback_forward_relative_error": 0.0342312341916465, + "min_feedback_forward_cosine": 0.9988236427307129 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.009922520908858213 + }, + { + "epoch": 146, + "eval_accuracy": 0.9222, + "eval_loss": 0.3881881774902344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04858988896012306, + "mean_feedback_forward_cosine": 0.9993935381212542, + "mean_feedback_forward_relative_error": 0.034226064540205464, + "min_feedback_forward_cosine": 0.9988237023353577 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.009807878665212127 + }, + { + "epoch": 147, + "eval_accuracy": 0.9248, + "eval_loss": 0.38153494262695314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048584867268800735, + "mean_feedback_forward_cosine": 0.9993937188579191, + "mean_feedback_forward_relative_error": 0.03422067278335171, + "min_feedback_forward_cosine": 0.998823881149292 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.010108151466813353 + }, + { + "epoch": 148, + "eval_accuracy": 0.9246, + "eval_loss": 0.39089484252929685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04858054965734482, + "mean_feedback_forward_cosine": 0.9993938476808609, + "mean_feedback_forward_relative_error": 0.034215778472923467, + "min_feedback_forward_cosine": 0.9988242387771606 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.00944744374089771 + }, + { + "epoch": 149, + "eval_accuracy": 0.9234, + "eval_loss": 0.3983838195800781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857631400227547, + "mean_feedback_forward_cosine": 0.9993940303402562, + "mean_feedback_forward_relative_error": 0.03421078793584339, + "min_feedback_forward_cosine": 0.9988244771957397 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.009071883625868294 + }, + { + "epoch": 150, + "eval_accuracy": 0.922, + "eval_loss": 0.3920445617675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048571620136499405, + "mean_feedback_forward_cosine": 0.9993941918496163, + "mean_feedback_forward_relative_error": 0.03420544635023801, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.009368490175157786 + }, + { + "epoch": 151, + "eval_accuracy": 0.9202, + "eval_loss": 0.3945726989746094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857156425714493, + "mean_feedback_forward_cosine": 0.9993941880041554, + "mean_feedback_forward_relative_error": 0.03420531115825138, + "min_feedback_forward_cosine": 0.9988246560096741 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.008574614812102583 + }, + { + "epoch": 152, + "eval_accuracy": 0.9208, + "eval_loss": 0.3895898193359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857151582837105, + "mean_feedback_forward_cosine": 0.9993941841586944, + "mean_feedback_forward_relative_error": 0.034205187893202225, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.00854987682186895 + }, + { + "epoch": 153, + "eval_accuracy": 0.9218, + "eval_loss": 0.3936317932128906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048571471124887466, + "mean_feedback_forward_cosine": 0.9993941918496163, + "mean_feedback_forward_relative_error": 0.03420507057660049, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.007438057490438223 + }, + { + "epoch": 154, + "eval_accuracy": 0.9226, + "eval_loss": 0.38970969848632814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857144504785538, + "mean_feedback_forward_cosine": 0.9993941976178077, + "mean_feedback_forward_relative_error": 0.034204961972371224, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.007093484061956406 + }, + { + "epoch": 155, + "eval_accuracy": 0.9222, + "eval_loss": 0.3903849304199219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857141524553299, + "mean_feedback_forward_cosine": 0.9993941899268858, + "mean_feedback_forward_relative_error": 0.03420485216643541, + "min_feedback_forward_cosine": 0.9988247752189636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.007574405139767462 + }, + { + "epoch": 156, + "eval_accuracy": 0.923, + "eval_loss": 0.3892175231933594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0485713854432106, + "mean_feedback_forward_cosine": 0.9993941918496163, + "mean_feedback_forward_relative_error": 0.034204739085849256, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.00694707155831986 + }, + { + "epoch": 157, + "eval_accuracy": 0.9224, + "eval_loss": 0.38980198974609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048571355640888214, + "mean_feedback_forward_cosine": 0.9993941880041554, + "mean_feedback_forward_relative_error": 0.034204619486005075, + "min_feedback_forward_cosine": 0.9988247752189636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.007672812810457415 + }, + { + "epoch": 158, + "eval_accuracy": 0.9222, + "eval_loss": 0.3905103515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857131466269493, + "mean_feedback_forward_cosine": 0.9993942110769211, + "mean_feedback_forward_relative_error": 0.03420450631529093, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0080210231008629 + }, + { + "epoch": 159, + "eval_accuracy": 0.924, + "eval_loss": 0.38340865478515623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048571277409791946, + "mean_feedback_forward_cosine": 0.9993941880041554, + "mean_feedback_forward_relative_error": 0.034204392543723504, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.0068947837022650576 + }, + { + "epoch": 160, + "eval_accuracy": 0.9236, + "eval_loss": 0.3892705810546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048571255058050156, + "mean_feedback_forward_cosine": 0.9993941995405382, + "mean_feedback_forward_relative_error": 0.03420428601243804, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.006723156801238656 + }, + { + "epoch": 161, + "eval_accuracy": 0.924, + "eval_loss": 0.387517919921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857123643159866, + "mean_feedback_forward_cosine": 0.9993942110769211, + "mean_feedback_forward_relative_error": 0.03420418536951465, + "min_feedback_forward_cosine": 0.9988246560096741 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.006679223413185941 + }, + { + "epoch": 162, + "eval_accuracy": 0.9234, + "eval_loss": 0.3882124084472656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857119545340538, + "mean_feedback_forward_cosine": 0.9993941956950773, + "mean_feedback_forward_relative_error": 0.0342040729498671, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.007147442474319703 + }, + { + "epoch": 163, + "eval_accuracy": 0.924, + "eval_loss": 0.3867006103515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048571158200502396, + "mean_feedback_forward_cosine": 0.9993942129996515, + "mean_feedback_forward_relative_error": 0.034203966688965595, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.006905054523878627 + }, + { + "epoch": 164, + "eval_accuracy": 0.9226, + "eval_loss": 0.38806248779296876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857112839818001, + "mean_feedback_forward_cosine": 0.9993942091541905, + "mean_feedback_forward_relative_error": 0.03420385676285913, + "min_feedback_forward_cosine": 0.9988246560096741 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.006833887228535281 + }, + { + "epoch": 165, + "eval_accuracy": 0.9228, + "eval_loss": 0.389813525390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857109487056732, + "mean_feedback_forward_cosine": 0.999394218767843, + "mean_feedback_forward_relative_error": 0.03420375596972242, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.006642061247221298 + }, + { + "epoch": 166, + "eval_accuracy": 0.9226, + "eval_loss": 0.390858056640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048571065068244934, + "mean_feedback_forward_cosine": 0.9993942206905734, + "mean_feedback_forward_relative_error": 0.03420365085044215, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.006464248421953784 + }, + { + "epoch": 167, + "eval_accuracy": 0.9242, + "eval_loss": 0.38848478393554686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857103154063225, + "mean_feedback_forward_cosine": 0.9993942283814953, + "mean_feedback_forward_relative_error": 0.03420353828058127, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.006532345111005836 + }, + { + "epoch": 168, + "eval_accuracy": 0.9248, + "eval_loss": 0.38907427978515624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857099801301956, + "mean_feedback_forward_cosine": 0.999394214922382, + "mean_feedback_forward_relative_error": 0.03420343000682131, + "min_feedback_forward_cosine": 0.9988246560096741 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.006440749948720137 + }, + { + "epoch": 169, + "eval_accuracy": 0.9236, + "eval_loss": 0.38989791870117185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570968210697174, + "mean_feedback_forward_cosine": 0.9993942206905734, + "mean_feedback_forward_relative_error": 0.034203322544213266, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.006686322028934956 + }, + { + "epoch": 170, + "eval_accuracy": 0.9236, + "eval_loss": 0.38908255004882814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570938408374786, + "mean_feedback_forward_cosine": 0.9993942091541905, + "mean_feedback_forward_relative_error": 0.03420321913736482, + "min_feedback_forward_cosine": 0.9988246560096741 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.006311853679766257 + }, + { + "epoch": 171, + "eval_accuracy": 0.9238, + "eval_loss": 0.38905315551757813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570916056632996, + "mean_feedback_forward_cosine": 0.9993942341496868, + "mean_feedback_forward_relative_error": 0.034203112335695375, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.006446439210573832 + }, + { + "epoch": 172, + "eval_accuracy": 0.9258, + "eval_loss": 0.3909088073730469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857088252902031, + "mean_feedback_forward_cosine": 0.9993942168451124, + "mean_feedback_forward_relative_error": 0.0342030058945379, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.006494547173836165 + }, + { + "epoch": 173, + "eval_accuracy": 0.9252, + "eval_loss": 0.3877605590820313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570845276117325, + "mean_feedback_forward_cosine": 0.9993942264587649, + "mean_feedback_forward_relative_error": 0.03420290344905469, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.006705886676493618 + }, + { + "epoch": 174, + "eval_accuracy": 0.9242, + "eval_loss": 0.3882752746582031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857081547379494, + "mean_feedback_forward_cosine": 0.9993942303042258, + "mean_feedback_forward_relative_error": 0.034202803166643266, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.00622351783282227 + }, + { + "epoch": 175, + "eval_accuracy": 0.924, + "eval_loss": 0.38828839111328123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570796847343445, + "mean_feedback_forward_cosine": 0.9993942533769915, + "mean_feedback_forward_relative_error": 0.034202703304829135, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.005534759585931897 + }, + { + "epoch": 176, + "eval_accuracy": 0.9254, + "eval_loss": 0.3911351684570312, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570770770311356, + "mean_feedback_forward_cosine": 0.9993942437633392, + "mean_feedback_forward_relative_error": 0.03420259406970393, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.005964737568009231 + }, + { + "epoch": 177, + "eval_accuracy": 0.9248, + "eval_loss": 0.3912963134765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570748418569565, + "mean_feedback_forward_cosine": 0.9993942360724172, + "mean_feedback_forward_relative_error": 0.03420249706194285, + "min_feedback_forward_cosine": 0.9988246560096741 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.005666052417498496 + }, + { + "epoch": 178, + "eval_accuracy": 0.9246, + "eval_loss": 0.39074390869140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857071861624718, + "mean_feedback_forward_cosine": 0.9993942322269562, + "mean_feedback_forward_relative_error": 0.03420238794698831, + "min_feedback_forward_cosine": 0.9988248348236084 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.00602857864888178 + }, + { + "epoch": 179, + "eval_accuracy": 0.9262, + "eval_loss": 0.38614667358398436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570699989795685, + "mean_feedback_forward_cosine": 0.9993942303042258, + "mean_feedback_forward_relative_error": 0.03420228733410759, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.005615825949195358 + }, + { + "epoch": 180, + "eval_accuracy": 0.9244, + "eval_loss": 0.38893945922851564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570673912763596, + "mean_feedback_forward_cosine": 0.9993942572224525, + "mean_feedback_forward_relative_error": 0.03420218726199481, + "min_feedback_forward_cosine": 0.9988247752189636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.006089276160837875 + }, + { + "epoch": 181, + "eval_accuracy": 0.9238, + "eval_loss": 0.3898399291992187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857064411044121, + "mean_feedback_forward_cosine": 0.9993942437633392, + "mean_feedback_forward_relative_error": 0.03420209064478836, + "min_feedback_forward_cosine": 0.9988247752189636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.006377839122795396 + }, + { + "epoch": 182, + "eval_accuracy": 0.923, + "eval_loss": 0.395673095703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857062175869942, + "mean_feedback_forward_cosine": 0.9993942245360343, + "mean_feedback_forward_relative_error": 0.03420198555555074, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.005886330561753777 + }, + { + "epoch": 183, + "eval_accuracy": 0.9234, + "eval_loss": 0.39157060546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857058823108673, + "mean_feedback_forward_cosine": 0.9993942379951477, + "mean_feedback_forward_relative_error": 0.034201881517806364, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.00615320463238491 + }, + { + "epoch": 184, + "eval_accuracy": 0.9234, + "eval_loss": 0.3938642578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857056587934494, + "mean_feedback_forward_cosine": 0.9993942399178782, + "mean_feedback_forward_relative_error": 0.0342017823469735, + "min_feedback_forward_cosine": 0.9988248348236084 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.006089492196589708 + }, + { + "epoch": 185, + "eval_accuracy": 0.9244, + "eval_loss": 0.39137650146484376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570528626441956, + "mean_feedback_forward_cosine": 0.9993942341496868, + "mean_feedback_forward_relative_error": 0.03420168437784718, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.005729687671032217 + }, + { + "epoch": 186, + "eval_accuracy": 0.9238, + "eval_loss": 0.3917858642578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857049137353897, + "mean_feedback_forward_cosine": 0.9993942418406087, + "mean_feedback_forward_relative_error": 0.034201580460273454, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.0060106748471657435 + }, + { + "epoch": 187, + "eval_accuracy": 0.922, + "eval_loss": 0.3935861572265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857047274708748, + "mean_feedback_forward_cosine": 0.9993942360724172, + "mean_feedback_forward_relative_error": 0.034201479006198146, + "min_feedback_forward_cosine": 0.9988248348236084 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.006099106899317768 + }, + { + "epoch": 188, + "eval_accuracy": 0.9238, + "eval_loss": 0.3924464660644531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570454120635986, + "mean_feedback_forward_cosine": 0.9993942629906439, + "mean_feedback_forward_relative_error": 0.03420138022591991, + "min_feedback_forward_cosine": 0.9988248944282532 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.005720732238267859 + }, + { + "epoch": 189, + "eval_accuracy": 0.9258, + "eval_loss": 0.38947210693359374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0485704243183136, + "mean_feedback_forward_cosine": 0.999394255299722, + "mean_feedback_forward_relative_error": 0.03420127672894347, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.0063833029953142 + }, + { + "epoch": 190, + "eval_accuracy": 0.9254, + "eval_loss": 0.3864773376464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857039451599121, + "mean_feedback_forward_cosine": 0.999394255299722, + "mean_feedback_forward_relative_error": 0.03420117773836659, + "min_feedback_forward_cosine": 0.9988246560096741 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.005768505500422584 + }, + { + "epoch": 191, + "eval_accuracy": 0.9246, + "eval_loss": 0.38983641967773436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857035353779793, + "mean_feedback_forward_cosine": 0.9993942476088001, + "mean_feedback_forward_relative_error": 0.03420107397100618, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.006166870318022039 + }, + { + "epoch": 192, + "eval_accuracy": 0.9242, + "eval_loss": 0.3892133483886719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570334911346436, + "mean_feedback_forward_cosine": 0.9993942668361049, + "mean_feedback_forward_relative_error": 0.03420097753405571, + "min_feedback_forward_cosine": 0.9988248348236084 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.005772096807261308 + }, + { + "epoch": 193, + "eval_accuracy": 0.926, + "eval_loss": 0.3912098205566406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570312559604645, + "mean_feedback_forward_cosine": 0.999394259145183, + "mean_feedback_forward_relative_error": 0.03420088317004904, + "min_feedback_forward_cosine": 0.9988247752189636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.005735846981571781 + }, + { + "epoch": 194, + "eval_accuracy": 0.924, + "eval_loss": 0.38991021728515624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570286482572556, + "mean_feedback_forward_cosine": 0.9993942456860696, + "mean_feedback_forward_relative_error": 0.03420078535113604, + "min_feedback_forward_cosine": 0.9988247752189636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.005394473344584306 + }, + { + "epoch": 195, + "eval_accuracy": 0.9234, + "eval_loss": 0.39238130493164064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857025295495987, + "mean_feedback_forward_cosine": 0.9993942456860696, + "mean_feedback_forward_relative_error": 0.03420068224471423, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.005845885767539342 + }, + { + "epoch": 196, + "eval_accuracy": 0.924, + "eval_loss": 0.388215380859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857022315263748, + "mean_feedback_forward_cosine": 0.9993942649133744, + "mean_feedback_forward_relative_error": 0.034200580430126956, + "min_feedback_forward_cosine": 0.9988248348236084 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.005960262824677758 + }, + { + "epoch": 197, + "eval_accuracy": 0.925, + "eval_loss": 0.3863689880371094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857020080089569, + "mean_feedback_forward_cosine": 0.9993942783724877, + "mean_feedback_forward_relative_error": 0.034200483362280557, + "min_feedback_forward_cosine": 0.9988247752189636 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.005644922489176194 + }, + { + "epoch": 198, + "eval_accuracy": 0.9244, + "eval_loss": 0.3886321044921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570167273283005, + "mean_feedback_forward_cosine": 0.9993942649133744, + "mean_feedback_forward_relative_error": 0.034200379955432104, + "min_feedback_forward_cosine": 0.9988248944282532 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.005755332447257307 + }, + { + "epoch": 199, + "eval_accuracy": 0.926, + "eval_loss": 0.3910266357421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.048570141196250916, + "mean_feedback_forward_cosine": 0.999394255299722, + "mean_feedback_forward_relative_error": 0.03420028096485522, + "min_feedback_forward_cosine": 0.9988248348236084 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.006012655192530817 + }, + { + "epoch": 200, + "eval_accuracy": 0.9238, + "eval_loss": 0.39594262084960935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04857011139392853, + "mean_feedback_forward_cosine": 0.9993942726042963, + "mean_feedback_forward_relative_error": 0.034200180532230486, + "min_feedback_forward_cosine": 0.9988247156143188 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.0059044962678311596 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9238, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.39594262084960935, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2325996544, + "peak_memory_reserved_bytes": 3166699520, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 34.060356855392456, + "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": 2877.0025794506073, + "train_wall_s": 2841.6028168201447 + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 615782016000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 615782016000000, + "local_weight_correlation_macs": 619763328000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 619763328000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 2471090688000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json.manifest.json new file mode 100644 index 0000000..16d242d --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "kp", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787027107.8066654, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 2880.570287704468, + "epochs": 200, + "experiment_name": "resnet-r2-clean_kp-d32-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "clean_kp", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d32-lr0p1.json", + "output_exists": true, + "output_sha256": "c8170145b2374f71f7bfdc161acfba10610b8ceb3c561b00b20a6182c369349c", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json new file mode 100644 index 0000000..c51dbc4 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json @@ -0,0 +1,4265 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.9988037016656663, + "feedback_forward_cosine": [ + 0.9995574951171875, + 0.9995169639587402, + 0.9992508888244629, + 0.9991180896759033, + 0.9991408586502075, + 0.9988906979560852, + 0.999122142791748, + 0.9988346099853516, + 0.9992246627807617, + 0.9988415837287903, + 0.9990280866622925, + 0.9985275864601135, + 0.9986708164215088, + 0.9977948069572449, + 0.9980864524841309, + 0.9977025985717773, + 0.9983896017074585, + 0.9977715611457825, + 0.9988182783126831, + 0.9991542100906372, + 0.9989104866981506, + 0.9987741708755493, + 0.9988317489624023, + 0.9986496567726135, + 0.9990257620811462, + 0.9987555146217346, + 0.9989592432975769, + 0.9986085891723633, + 0.9990679621696472, + 0.998760461807251, + 0.9990558624267578, + 0.9986815452575684, + 0.9991776943206787, + 0.9988480806350708, + 0.9991104006767273, + 0.9987267255783081, + 0.9992961883544922, + 0.9995409250259399, + 0.9993419647216797, + 0.9993253946304321, + 0.9993946552276611, + 0.9992978572845459, + 0.9994099736213684, + 0.9992327690124512, + 0.9994537830352783, + 0.9991894960403442, + 0.9993447065353394, + 0.9989761114120483, + 0.9992188215255737, + 0.9986401796340942, + 0.9989041686058044, + 0.997963547706604, + 0.9979575872421265, + 0.9966350197792053, + 0.9999642968177795 + ], + "feedback_forward_norm_ratio": [ + 0.9988628625869751, + 1.00169038772583, + 0.9988455772399902, + 1.0018149614334106, + 0.9983507394790649, + 1.001532793045044, + 0.9981478452682495, + 1.00508713722229, + 0.9976799488067627, + 1.002779483795166, + 0.996472954750061, + 1.0031712055206299, + 0.9947890639305115, + 1.0027819871902466, + 0.9938843250274658, + 1.0071712732315063, + 0.9960893988609314, + 1.0045490264892578, + 0.9968646764755249, + 1.0018043518066406, + 0.9983800649642944, + 1.0021274089813232, + 0.9981029033660889, + 1.002813696861267, + 0.998471736907959, + 1.0024747848510742, + 0.9984707236289978, + 1.00377357006073, + 0.9984996318817139, + 1.003908634185791, + 0.9981768727302551, + 1.00295889377594, + 0.9980287551879883, + 1.003975510597229, + 0.9989507794380188, + 1.003403663635254, + 0.9981476068496704, + 1.0010191202163696, + 0.9987508058547974, + 1.0020904541015625, + 0.9993308186531067, + 1.0017856359481812, + 0.9992122650146484, + 1.0019315481185913, + 0.9992152452468872, + 1.0023112297058105, + 0.9992201924324036, + 1.0025296211242676, + 0.9988380670547485, + 1.002525806427002, + 0.9991728663444519, + 1.003301739692688, + 0.9986709356307983, + 1.003277063369751, + 1.0064365863800049 + ], + "feedback_forward_relative_error": [ + 0.029755709692835808, + 0.03115258552134037, + 0.03870101645588875, + 0.042074669152498245, + 0.04145186021924019, + 0.04716512933373451, + 0.041904985904693604, + 0.04866664856672287, + 0.03940116614103317, + 0.04828055575489998, + 0.04415307566523552, + 0.054445065557956696, + 0.05168891325592995, + 0.06656099855899811, + 0.061974987387657166, + 0.0684041976928711, + 0.05677487328648567, + 0.0670657828450203, + 0.04864046350121498, + 0.041204363107681274, + 0.04667006433010101, + 0.04961398243904114, + 0.04832734540104866, + 0.05211685970425606, + 0.04413510486483574, + 0.05001169815659523, + 0.04561459273099899, + 0.05298462510108948, + 0.043167274445295334, + 0.050040122121572495, + 0.04345225915312767, + 0.05151088535785675, + 0.04056119546294212, + 0.048257168382406235, + 0.04217178001999855, + 0.05066346749663353, + 0.03753187507390976, + 0.030329721048474312, + 0.03627943620085716, + 0.03683214634656906, + 0.034789882600307465, + 0.03754897788167, + 0.03434815630316734, + 0.03925550356507301, + 0.03304633870720863, + 0.040373899042606354, + 0.03619636967778206, + 0.045381058007478714, + 0.039520811289548874, + 0.05227590724825859, + 0.046804919838905334, + 0.06400913000106812, + 0.06388391554355621, + 0.08223697543144226, + 0.01064779981970787 + ], + "innovation_negative_gradient_cosine": [ + 0.9990317821502686, + 0.9989892244338989, + 0.9989813566207886, + 0.9987704753875732, + 0.9989952445030212, + 0.9988008737564087, + 0.9990298748016357, + 0.9987648129463196, + 0.99904465675354, + 0.9989824891090393, + 0.9990545511245728, + 0.9986403584480286, + 0.9990695714950562, + 0.9982811212539673, + 0.9990877509117126, + 0.9978867769241333, + 0.9990891814231873, + 0.9979665279388428, + 0.9990986585617065, + 0.998831033706665, + 0.9991048574447632, + 0.9985469579696655, + 0.9991083145141602, + 0.9985823035240173, + 0.9991117715835571, + 0.9986124634742737, + 0.9991379380226135, + 0.9986610412597656, + 0.9991350173950195, + 0.9989628195762634, + 0.9991565942764282, + 0.9987471699714661, + 0.9991878867149353, + 0.9989113807678223, + 0.9992119073867798, + 0.9989118576049805, + 0.9992275238037109, + 0.9992373585700989, + 0.9992811679840088, + 0.999153733253479, + 0.9993259906768799, + 0.9991576671600342, + 0.9993447661399841, + 0.9992607235908508, + 0.9994027614593506, + 0.9993295073509216, + 0.9994453191757202, + 0.9994139671325684, + 0.9994925260543823, + 0.9995317459106445, + 0.9995478391647339, + 0.9996591210365295, + 0.9996875524520874, + 0.9998064041137695, + 0.7163748741149902 + ], + "mean_feedback_forward_cosine": 0.9988813335245306, + "mean_feedback_forward_relative_error": 0.04600105993449688, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9990317821502686, + 0.9989892244338989, + 0.9989813566207886, + 0.9987704753875732, + 0.9989952445030212, + 0.9988008737564087, + 0.9990298748016357, + 0.9987648129463196, + 0.99904465675354, + 0.9989824891090393, + 0.9990545511245728, + 0.9986403584480286, + 0.9990695714950562, + 0.9982811212539673, + 0.9990877509117126, + 0.9978867769241333, + 0.9990891814231873, + 0.9979665279388428, + 0.9990986585617065, + 0.998831033706665, + 0.9991048574447632, + 0.9985469579696655, + 0.9991083145141602, + 0.9985823035240173, + 0.9991117715835571, + 0.9986124634742737, + 0.9991379380226135, + 0.9986610412597656, + 0.9991350173950195, + 0.9989628195762634, + 0.9991565942764282, + 0.9987471699714661, + 0.9991878867149353, + 0.9989113807678223, + 0.9992119073867798, + 0.9989118576049805, + 0.9992275238037109, + 0.9992373585700989, + 0.9992811679840088, + 0.999153733253479, + 0.9993259906768799, + 0.9991576671600342, + 0.9993447661399841, + 0.9992607235908508, + 0.9994027614593506, + 0.9993295073509216, + 0.9994453191757202, + 0.9994139671325684, + 0.9994925260543823, + 0.9995317459106445, + 0.9995478391647339, + 0.9996591210365295, + 0.9996875524520874, + 0.9998064041137695, + 0.7163748741149902 + ], + "teaching_negative_gradient_cosine": [ + 0.9990317821502686, + 0.9989892244338989, + 0.9989813566207886, + 0.9987704753875732, + 0.9989952445030212, + 0.9988008737564087, + 0.9990298748016357, + 0.9987648129463196, + 0.99904465675354, + 0.9989824891090393, + 0.9990545511245728, + 0.9986403584480286, + 0.9990695714950562, + 0.9982811212539673, + 0.9990877509117126, + 0.9978867769241333, + 0.9990891814231873, + 0.9979665279388428, + 0.9990986585617065, + 0.998831033706665, + 0.9991048574447632, + 0.9985469579696655, + 0.9991083145141602, + 0.9985823035240173, + 0.9991117715835571, + 0.9986124634742737, + 0.9991379380226135, + 0.9986610412597656, + 0.9991350173950195, + 0.9989628195762634, + 0.9991565942764282, + 0.9987471699714661, + 0.9991878867149353, + 0.9989113807678223, + 0.9992119073867798, + 0.9989118576049805, + 0.9992275238037109, + 0.9992373585700989, + 0.9992811679840088, + 0.999153733253479, + 0.9993259906768799, + 0.9991576671600342, + 0.9993447661399841, + 0.9992607235908508, + 0.9994027614593506, + 0.9993295073509216, + 0.9994453191757202, + 0.9994139671325684, + 0.9994925260543823, + 0.9995317459106445, + 0.9995478391647339, + 0.9996591210365295, + 0.9996875524520874, + 0.9998064041137695, + 0.7163748741149902 + ], + "wall_s": 0.11626768112182617 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.2936, + "eval_loss": 2.060192431640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.4729894399642944, + "mean_feedback_forward_cosine": 0.08207923307676207, + "mean_feedback_forward_relative_error": 1.3605215722864323, + "min_feedback_forward_cosine": 0.011852272786200047 + }, + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.123790234099494 + }, + { + "epoch": 2, + "eval_accuracy": 0.4336, + "eval_loss": 1.644146044921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3818473815917969, + "mean_feedback_forward_cosine": 0.1517215799878944, + "mean_feedback_forward_relative_error": 1.306954277645458, + "min_feedback_forward_cosine": 0.04962993040680885 + }, + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.6988532908757528 + }, + { + "epoch": 3, + "eval_accuracy": 0.5284, + "eval_loss": 1.41752919921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3391482830047607, + "mean_feedback_forward_cosine": 0.2346487126567147, + "mean_feedback_forward_relative_error": 1.2409382657571273, + "min_feedback_forward_cosine": 0.11064116656780243 + }, + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.40682961162991 + }, + { + "epoch": 4, + "eval_accuracy": 0.5884, + "eval_loss": 1.1971573974609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3097155094146729, + "mean_feedback_forward_cosine": 0.3149321618405255, + "mean_feedback_forward_relative_error": 1.1737277052619242, + "min_feedback_forward_cosine": 0.14912492036819458 + }, + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 1.1264890681160822 + }, + { + "epoch": 5, + "eval_accuracy": 0.7082, + "eval_loss": 0.8434026123046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2819832563400269, + "mean_feedback_forward_cosine": 0.387133172696287, + "mean_feedback_forward_relative_error": 1.109520400654186, + "min_feedback_forward_cosine": 0.1867155283689499 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.9187799799389309 + }, + { + "epoch": 6, + "eval_accuracy": 0.728, + "eval_loss": 0.8178329467773438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.260962724685669, + "mean_feedback_forward_cosine": 0.43850364685058596, + "mean_feedback_forward_relative_error": 1.0616818113760516, + "min_feedback_forward_cosine": 0.2145915925502777 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7764964687347412 + }, + { + "epoch": 7, + "eval_accuracy": 0.7022, + "eval_loss": 0.9358474975585938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2412327527999878, + "mean_feedback_forward_cosine": 0.4796803523193706, + "mean_feedback_forward_relative_error": 1.0219877210530368, + "min_feedback_forward_cosine": 0.24086108803749084 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6912625069724189 + }, + { + "epoch": 8, + "eval_accuracy": 0.7508, + "eval_loss": 0.762169921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.222935676574707, + "mean_feedback_forward_cosine": 0.5153755480592901, + "mean_feedback_forward_relative_error": 0.986245432767001, + "min_feedback_forward_cosine": 0.26393070816993713 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.6286246035257975 + }, + { + "epoch": 9, + "eval_accuracy": 0.7888, + "eval_loss": 0.6249593627929687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2051806449890137, + "mean_feedback_forward_cosine": 0.5469873997298154, + "mean_feedback_forward_relative_error": 0.9534298192371021, + "min_feedback_forward_cosine": 0.2858530282974243 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5727805726899041 + }, + { + "epoch": 10, + "eval_accuracy": 0.768, + "eval_loss": 0.7355963623046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1889394521713257, + "mean_feedback_forward_cosine": 0.5755576019937342, + "mean_feedback_forward_relative_error": 0.9226999954743819, + "min_feedback_forward_cosine": 0.30642619729042053 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5328205897172292 + }, + { + "epoch": 11, + "eval_accuracy": 0.8112, + "eval_loss": 0.5712257080078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1723517179489136, + "mean_feedback_forward_cosine": 0.6018412416631526, + "mean_feedback_forward_relative_error": 0.8934933689507571, + "min_feedback_forward_cosine": 0.3266719579696655 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.5016598609076606 + }, + { + "epoch": 12, + "eval_accuracy": 0.7896, + "eval_loss": 0.6734730712890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1551438570022583, + "mean_feedback_forward_cosine": 0.625412240895358, + "mean_feedback_forward_relative_error": 0.8661270428787579, + "min_feedback_forward_cosine": 0.34698301553726196 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.47733594166437787 + }, + { + "epoch": 13, + "eval_accuracy": 0.799, + "eval_loss": 0.6092771240234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.138596773147583, + "mean_feedback_forward_cosine": 0.6471591234207154, + "mean_feedback_forward_relative_error": 0.8402352311394431, + "min_feedback_forward_cosine": 0.36682766675949097 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.4463163204775916 + }, + { + "epoch": 14, + "eval_accuracy": 0.8226, + "eval_loss": 0.5384412719726562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.122653603553772, + "mean_feedback_forward_cosine": 0.6674662720073353, + "mean_feedback_forward_relative_error": 0.8154430156404322, + "min_feedback_forward_cosine": 0.3852914571762085 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.4283656439781189 + }, + { + "epoch": 15, + "eval_accuracy": 0.829, + "eval_loss": 0.5133809020996094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1056475639343262, + "mean_feedback_forward_cosine": 0.6869502766565843, + "mean_feedback_forward_relative_error": 0.7906913573091681, + "min_feedback_forward_cosine": 0.4046149253845215 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.4076533457332187 + }, + { + "epoch": 16, + "eval_accuracy": 0.8106, + "eval_loss": 0.6083410766601562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0893713235855103, + "mean_feedback_forward_cosine": 0.7045990922234275, + "mean_feedback_forward_relative_error": 0.7676530827175487, + "min_feedback_forward_cosine": 0.42278337478637695 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.3880259382724762 + }, + { + "epoch": 17, + "eval_accuracy": 0.8292, + "eval_loss": 0.503566552734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0723820924758911, + "mean_feedback_forward_cosine": 0.7214057423851706, + "mean_feedback_forward_relative_error": 0.7450004333799536, + "min_feedback_forward_cosine": 0.44093620777130127 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.3736611439810859 + }, + { + "epoch": 18, + "eval_accuracy": 0.8352, + "eval_loss": 0.5023697204589844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0543376207351685, + "mean_feedback_forward_cosine": 0.737669068033045, + "mean_feedback_forward_relative_error": 0.7226391394029964, + "min_feedback_forward_cosine": 0.4603174924850464 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.364715851465861 + }, + { + "epoch": 19, + "eval_accuracy": 0.8164, + "eval_loss": 0.603892041015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.035783290863037, + "mean_feedback_forward_cosine": 0.7531207366423174, + "mean_feedback_forward_relative_error": 0.70054920993068, + "min_feedback_forward_cosine": 0.4795728921890259 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.3488689161459605 + }, + { + "epoch": 20, + "eval_accuracy": 0.8404, + "eval_loss": 0.5018754943847656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0169116258621216, + "mean_feedback_forward_cosine": 0.7675304158167405, + "mean_feedback_forward_relative_error": 0.6793056907978925, + "min_feedback_forward_cosine": 0.4986239969730377 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.33735975733862983 + }, + { + "epoch": 21, + "eval_accuracy": 0.8446, + "eval_loss": 0.4680168029785156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9975670576095581, + "mean_feedback_forward_cosine": 0.7813035347244957, + "mean_feedback_forward_relative_error": 0.658438462560827, + "min_feedback_forward_cosine": 0.5182281732559204 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.3261894694434272 + }, + { + "epoch": 22, + "eval_accuracy": 0.8524, + "eval_loss": 0.4533489501953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9778302311897278, + "mean_feedback_forward_cosine": 0.7944718913598494, + "mean_feedback_forward_relative_error": 0.6378612987019799, + "min_feedback_forward_cosine": 0.5373923778533936 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3175746521313985 + }, + { + "epoch": 23, + "eval_accuracy": 0.8608, + "eval_loss": 0.4396246276855469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9584059715270996, + "mean_feedback_forward_cosine": 0.8066396474838257, + "mean_feedback_forward_relative_error": 0.6182341255924918, + "min_feedback_forward_cosine": 0.5559208393096924 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.30371766968833075 + }, + { + "epoch": 24, + "eval_accuracy": 0.8596, + "eval_loss": 0.4563602478027344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9386764168739319, + "mean_feedback_forward_cosine": 0.8177748517556624, + "mean_feedback_forward_relative_error": 0.5997207224369049, + "min_feedback_forward_cosine": 0.5744795799255371 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.2900086401462555 + }, + { + "epoch": 25, + "eval_accuracy": 0.8586, + "eval_loss": 0.4243275390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9187594652175903, + "mean_feedback_forward_cosine": 0.8285161896185441, + "mean_feedback_forward_relative_error": 0.5813728538426486, + "min_feedback_forward_cosine": 0.5922850370407104 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.2870960140122308 + }, + { + "epoch": 26, + "eval_accuracy": 0.866, + "eval_loss": 0.41880641479492187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8995776772499084, + "mean_feedback_forward_cosine": 0.8383215871724216, + "mean_feedback_forward_relative_error": 0.5641180030324242, + "min_feedback_forward_cosine": 0.6091452240943909 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.2801684870455 + }, + { + "epoch": 27, + "eval_accuracy": 0.8608, + "eval_loss": 0.45782495727539063, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8790687918663025, + "mean_feedback_forward_cosine": 0.8478951313278892, + "mean_feedback_forward_relative_error": 0.5467334628105164, + "min_feedback_forward_cosine": 0.6270616054534912 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.276506783774164 + }, + { + "epoch": 28, + "eval_accuracy": 0.8738, + "eval_loss": 0.41229105224609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8599552512168884, + "mean_feedback_forward_cosine": 0.8569087667898698, + "mean_feedback_forward_relative_error": 0.5298725553534248, + "min_feedback_forward_cosine": 0.6432315111160278 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.25949633451567755 + }, + { + "epoch": 29, + "eval_accuracy": 0.8548, + "eval_loss": 0.47992633666992185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8409947752952576, + "mean_feedback_forward_cosine": 0.8654873620380055, + "mean_feedback_forward_relative_error": 0.5133874578909441, + "min_feedback_forward_cosine": 0.658990740776062 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.2583834177679486 + }, + { + "epoch": 30, + "eval_accuracy": 0.8498, + "eval_loss": 0.49892302856445314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8211984038352966, + "mean_feedback_forward_cosine": 0.8737081343477423, + "mean_feedback_forward_relative_error": 0.4970983491702513, + "min_feedback_forward_cosine": 0.6749225854873657 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.25416444894737666 + }, + { + "epoch": 31, + "eval_accuracy": 0.8446, + "eval_loss": 0.5110629577636718, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8005052804946899, + "mean_feedback_forward_cosine": 0.8813880790363658, + "mean_feedback_forward_relative_error": 0.4814133836464448, + "min_feedback_forward_cosine": 0.6910892724990845 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.2507732811000612 + }, + { + "epoch": 32, + "eval_accuracy": 0.8732, + "eval_loss": 0.40844691162109376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7807812094688416, + "mean_feedback_forward_cosine": 0.8887567043304443, + "mean_feedback_forward_relative_error": 0.4659483096816323, + "min_feedback_forward_cosine": 0.7062845230102539 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.24385885693762038 + }, + { + "epoch": 33, + "eval_accuracy": 0.8492, + "eval_loss": 0.49456326904296877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7614753842353821, + "mean_feedback_forward_cosine": 0.8955482537096197, + "mean_feedback_forward_relative_error": 0.4511532629078085, + "min_feedback_forward_cosine": 0.720395565032959 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.23739461857477823 + }, + { + "epoch": 34, + "eval_accuracy": 0.8498, + "eval_loss": 0.5223807678222656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.742235541343689, + "mean_feedback_forward_cosine": 0.9018027283928611, + "mean_feedback_forward_relative_error": 0.43716591732068494, + "min_feedback_forward_cosine": 0.7343444228172302 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.23336360233359868 + }, + { + "epoch": 35, + "eval_accuracy": 0.8744, + "eval_loss": 0.41874742431640627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7225374579429626, + "mean_feedback_forward_cosine": 0.9077971696853637, + "mean_feedback_forward_relative_error": 0.4234206673773852, + "min_feedback_forward_cosine": 0.7480750679969788 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.22778456059561836 + }, + { + "epoch": 36, + "eval_accuracy": 0.8776, + "eval_loss": 0.3977263732910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7031345367431641, + "mean_feedback_forward_cosine": 0.9135233012112705, + "mean_feedback_forward_relative_error": 0.4098250017924742, + "min_feedback_forward_cosine": 0.7613703012466431 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.2255130520661672 + }, + { + "epoch": 37, + "eval_accuracy": 0.867, + "eval_loss": 0.4421707946777344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6846021413803101, + "mean_feedback_forward_cosine": 0.9187919194048101, + "mean_feedback_forward_relative_error": 0.39689874730326913, + "min_feedback_forward_cosine": 0.7737734913825989 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.2240035187959671 + }, + { + "epoch": 38, + "eval_accuracy": 0.851, + "eval_loss": 0.4847253173828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6660941243171692, + "mean_feedback_forward_cosine": 0.9238915584304116, + "mean_feedback_forward_relative_error": 0.3839693303812634, + "min_feedback_forward_cosine": 0.7856288552284241 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.21737773218684725 + }, + { + "epoch": 39, + "eval_accuracy": 0.8788, + "eval_loss": 0.41233887939453123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6483564972877502, + "mean_feedback_forward_cosine": 0.9285368594256315, + "mean_feedback_forward_relative_error": 0.3718270332975821, + "min_feedback_forward_cosine": 0.7968446016311646 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.2110462031231986 + }, + { + "epoch": 40, + "eval_accuracy": 0.8744, + "eval_loss": 0.42741173706054686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6304129362106323, + "mean_feedback_forward_cosine": 0.9328895178708163, + "mean_feedback_forward_relative_error": 0.3601139377463948, + "min_feedback_forward_cosine": 0.8078715801239014 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.20277180547714233 + }, + { + "epoch": 41, + "eval_accuracy": 0.881, + "eval_loss": 0.40434840698242186, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6114498376846313, + "mean_feedback_forward_cosine": 0.9371648885987022, + "mean_feedback_forward_relative_error": 0.34830014421181243, + "min_feedback_forward_cosine": 0.8190717101097107 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.20585658853848776 + }, + { + "epoch": 42, + "eval_accuracy": 0.878, + "eval_loss": 0.4233556945800781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5938428640365601, + "mean_feedback_forward_cosine": 0.9411213528026234, + "mean_feedback_forward_relative_error": 0.33697369437326086, + "min_feedback_forward_cosine": 0.8294274210929871 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.2012337659464942 + }, + { + "epoch": 43, + "eval_accuracy": 0.8768, + "eval_loss": 0.41400225830078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5771043300628662, + "mean_feedback_forward_cosine": 0.944767753644423, + "mean_feedback_forward_relative_error": 0.32624045339497654, + "min_feedback_forward_cosine": 0.8388245105743408 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.19511909804344177 + }, + { + "epoch": 44, + "eval_accuracy": 0.8104, + "eval_loss": 0.7237696166992188, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5596101880073547, + "mean_feedback_forward_cosine": 0.948338532447815, + "mean_feedback_forward_relative_error": 0.31540666994723404, + "min_feedback_forward_cosine": 0.8483365774154663 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.1980742193804847 + }, + { + "epoch": 45, + "eval_accuracy": 0.8706, + "eval_loss": 0.46084964599609374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5431535243988037, + "mean_feedback_forward_cosine": 0.9515498302199624, + "mean_feedback_forward_relative_error": 0.3053073207085783, + "min_feedback_forward_cosine": 0.8570070266723633 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.19275552099015977 + }, + { + "epoch": 46, + "eval_accuracy": 0.8814, + "eval_loss": 0.40947244873046873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5276136994361877, + "mean_feedback_forward_cosine": 0.9546002951535312, + "mean_feedback_forward_relative_error": 0.2954358386722478, + "min_feedback_forward_cosine": 0.8650082349777222 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.18924892360104456 + }, + { + "epoch": 47, + "eval_accuracy": 0.8824, + "eval_loss": 0.41092626342773436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5122539401054382, + "mean_feedback_forward_cosine": 0.9575513818047263, + "mean_feedback_forward_relative_error": 0.2855187581344084, + "min_feedback_forward_cosine": 0.8727079629898071 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.18873042728106182 + }, + { + "epoch": 48, + "eval_accuracy": 0.873, + "eval_loss": 0.4387502258300781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.49630874395370483, + "mean_feedback_forward_cosine": 0.9602571855891835, + "mean_feedback_forward_relative_error": 0.2761423014781692, + "min_feedback_forward_cosine": 0.8804479241371155 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.18430980693764157 + }, + { + "epoch": 49, + "eval_accuracy": 0.8618, + "eval_loss": 0.4662376098632813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.48133349418640137, + "mean_feedback_forward_cosine": 0.9627048893408342, + "mean_feedback_forward_relative_error": 0.26738952113823455, + "min_feedback_forward_cosine": 0.8874017000198364 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.1800568481153912 + }, + { + "epoch": 50, + "eval_accuracy": 0.8826, + "eval_loss": 0.4051923889160156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.466120183467865, + "mean_feedback_forward_cosine": 0.9650715307755904, + "mean_feedback_forward_relative_error": 0.2586525908925317, + "min_feedback_forward_cosine": 0.8943054676055908 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.1789180386662483 + }, + { + "epoch": 51, + "eval_accuracy": 0.8702, + "eval_loss": 0.4498435302734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4508410096168518, + "mean_feedback_forward_cosine": 0.9673665068366311, + "mean_feedback_forward_relative_error": 0.24993357875130393, + "min_feedback_forward_cosine": 0.9010016918182373 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.18010127037101323 + }, + { + "epoch": 52, + "eval_accuracy": 0.87, + "eval_loss": 0.477175048828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4368517994880676, + "mean_feedback_forward_cosine": 0.9694177584214644, + "mean_feedback_forward_relative_error": 0.24190838066014378, + "min_feedback_forward_cosine": 0.9070504307746887 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.16981975546148087 + }, + { + "epoch": 53, + "eval_accuracy": 0.8716, + "eval_loss": 0.4523770324707031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4235183298587799, + "mean_feedback_forward_cosine": 0.9713853868571195, + "mean_feedback_forward_relative_error": 0.23388134376569228, + "min_feedback_forward_cosine": 0.9125978946685791 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.17002208354208204 + }, + { + "epoch": 54, + "eval_accuracy": 0.8596, + "eval_loss": 0.5152298583984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40936705470085144, + "mean_feedback_forward_cosine": 0.9733334248716181, + "mean_feedback_forward_relative_error": 0.2257189244031906, + "min_feedback_forward_cosine": 0.9182888865470886 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.17592860356701745 + }, + { + "epoch": 55, + "eval_accuracy": 0.8734, + "eval_loss": 0.4423846618652344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3969172537326813, + "mean_feedback_forward_cosine": 0.9750462586229498, + "mean_feedback_forward_relative_error": 0.21827290789647535, + "min_feedback_forward_cosine": 0.9230918884277344 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.17371374536090428 + }, + { + "epoch": 56, + "eval_accuracy": 0.8784, + "eval_loss": 0.4110608825683594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38447803258895874, + "mean_feedback_forward_cosine": 0.976641688563607, + "mean_feedback_forward_relative_error": 0.2111172491176562, + "min_feedback_forward_cosine": 0.9277975559234619 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.16767991562949286 + }, + { + "epoch": 57, + "eval_accuracy": 0.8784, + "eval_loss": 0.4192649353027344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3719860017299652, + "mean_feedback_forward_cosine": 0.97813284072009, + "mean_feedback_forward_relative_error": 0.2041995076293295, + "min_feedback_forward_cosine": 0.9323537349700928 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.16788513456980386 + }, + { + "epoch": 58, + "eval_accuracy": 0.886, + "eval_loss": 0.37670152587890626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3604535162448883, + "mean_feedback_forward_cosine": 0.9794968269088051, + "mean_feedback_forward_relative_error": 0.1976644179360433, + "min_feedback_forward_cosine": 0.9364805221557617 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.16070309614075554 + }, + { + "epoch": 59, + "eval_accuracy": 0.8636, + "eval_loss": 0.4858129577636719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3487560451030731, + "mean_feedback_forward_cosine": 0.9808387929742987, + "mean_feedback_forward_relative_error": 0.1910235660997304, + "min_feedback_forward_cosine": 0.9404898881912231 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.1631975841826863 + }, + { + "epoch": 60, + "eval_accuracy": 0.8766, + "eval_loss": 0.4174827453613281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33727237582206726, + "mean_feedback_forward_cosine": 0.982091445272619, + "mean_feedback_forward_relative_error": 0.18464546623555098, + "min_feedback_forward_cosine": 0.9443246722221375 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.1633899968147278 + }, + { + "epoch": 61, + "eval_accuracy": 0.873, + "eval_loss": 0.452418115234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32584843039512634, + "mean_feedback_forward_cosine": 0.9832355997779153, + "mean_feedback_forward_relative_error": 0.17862626761198044, + "min_feedback_forward_cosine": 0.9479833245277405 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.159811972117424 + }, + { + "epoch": 62, + "eval_accuracy": 0.8828, + "eval_loss": 0.4084390075683594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.31567618250846863, + "mean_feedback_forward_cosine": 0.9843033476309343, + "mean_feedback_forward_relative_error": 0.17278472164815123, + "min_feedback_forward_cosine": 0.9511356353759766 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.15469107734097376 + }, + { + "epoch": 63, + "eval_accuracy": 0.878, + "eval_loss": 0.44356014404296873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30494773387908936, + "mean_feedback_forward_cosine": 0.9853211641311646, + "mean_feedback_forward_relative_error": 0.16705804108218714, + "min_feedback_forward_cosine": 0.9543746709823608 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.1573526954491933 + }, + { + "epoch": 64, + "eval_accuracy": 0.8818, + "eval_loss": 0.4164936279296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29509037733078003, + "mean_feedback_forward_cosine": 0.9862557031891562, + "mean_feedback_forward_relative_error": 0.16163663694804364, + "min_feedback_forward_cosine": 0.9572689533233643 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.15185034220880933 + }, + { + "epoch": 65, + "eval_accuracy": 0.8884, + "eval_loss": 0.429320068359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28480038046836853, + "mean_feedback_forward_cosine": 0.9871627710082315, + "mean_feedback_forward_relative_error": 0.15618729252706876, + "min_feedback_forward_cosine": 0.9601784944534302 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.15663046832614475 + }, + { + "epoch": 66, + "eval_accuracy": 0.8818, + "eval_loss": 0.3846602600097656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27500370144844055, + "mean_feedback_forward_cosine": 0.9880119280381636, + "mean_feedback_forward_relative_error": 0.15090371072292327, + "min_feedback_forward_cosine": 0.9628419280052185 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.152567342710495 + }, + { + "epoch": 67, + "eval_accuracy": 0.8742, + "eval_loss": 0.4532151306152344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2653149366378784, + "mean_feedback_forward_cosine": 0.9888372052799571, + "mean_feedback_forward_relative_error": 0.14560815332965418, + "min_feedback_forward_cosine": 0.9653981924057007 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.15156706712577078 + }, + { + "epoch": 68, + "eval_accuracy": 0.8854, + "eval_loss": 0.413903759765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2560727298259735, + "mean_feedback_forward_cosine": 0.9895687645131891, + "mean_feedback_forward_relative_error": 0.1407341560179537, + "min_feedback_forward_cosine": 0.9677327871322632 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.15645641708903843 + }, + { + "epoch": 69, + "eval_accuracy": 0.8872, + "eval_loss": 0.40043238525390623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24709388613700867, + "mean_feedback_forward_cosine": 0.9902782396836715, + "mean_feedback_forward_relative_error": 0.13584166609428144, + "min_feedback_forward_cosine": 0.9699423909187317 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.15490477414660983 + }, + { + "epoch": 70, + "eval_accuracy": 0.885, + "eval_loss": 0.397725732421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23907920718193054, + "mean_feedback_forward_cosine": 0.9909099470485341, + "mean_feedback_forward_relative_error": 0.13134042986414649, + "min_feedback_forward_cosine": 0.9718361496925354 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.1508726365831163 + }, + { + "epoch": 71, + "eval_accuracy": 0.8558, + "eval_loss": 0.5760824829101563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2312377542257309, + "mean_feedback_forward_cosine": 0.9915095892819491, + "mean_feedback_forward_relative_error": 0.12692420821298253, + "min_feedback_forward_cosine": 0.9736427068710327 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.14491609041425918 + }, + { + "epoch": 72, + "eval_accuracy": 0.8696, + "eval_loss": 0.458574853515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22312340140342712, + "mean_feedback_forward_cosine": 0.9920836979692632, + "mean_feedback_forward_relative_error": 0.12254367403008722, + "min_feedback_forward_cosine": 0.9754534363746643 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.1479260725710127 + }, + { + "epoch": 73, + "eval_accuracy": 0.8892, + "eval_loss": 0.40083897705078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21591386198997498, + "mean_feedback_forward_cosine": 0.9926146171309731, + "mean_feedback_forward_relative_error": 0.11832766709002582, + "min_feedback_forward_cosine": 0.976998507976532 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.14577782358063593 + }, + { + "epoch": 74, + "eval_accuracy": 0.8836, + "eval_loss": 0.4026210205078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20882253348827362, + "mean_feedback_forward_cosine": 0.9930870424617421, + "mean_feedback_forward_relative_error": 0.11447676802900704, + "min_feedback_forward_cosine": 0.9784666299819946 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.14426382974121305 + }, + { + "epoch": 75, + "eval_accuracy": 0.883, + "eval_loss": 0.42882922973632814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20159201323986053, + "mean_feedback_forward_cosine": 0.9935504837469621, + "mean_feedback_forward_relative_error": 0.11056213744662025, + "min_feedback_forward_cosine": 0.9799209237098694 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.14628605412244797 + }, + { + "epoch": 76, + "eval_accuracy": 0.878, + "eval_loss": 0.41888806762695313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1942160427570343, + "mean_feedback_forward_cosine": 0.9940022219311108, + "mean_feedback_forward_relative_error": 0.10662056024778974, + "min_feedback_forward_cosine": 0.9813616275787354 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.15107803481419882 + }, + { + "epoch": 77, + "eval_accuracy": 0.892, + "eval_loss": 0.3812136474609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18739287555217743, + "mean_feedback_forward_cosine": 0.9944140856916254, + "mean_feedback_forward_relative_error": 0.10289184376597404, + "min_feedback_forward_cosine": 0.982636034488678 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.145390626361635 + }, + { + "epoch": 78, + "eval_accuracy": 0.8826, + "eval_loss": 0.4390092895507812, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1810525357723236, + "mean_feedback_forward_cosine": 0.9947815277359703, + "mean_feedback_forward_relative_error": 0.0994456759230657, + "min_feedback_forward_cosine": 0.9837870597839355 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.14267062601778244 + }, + { + "epoch": 79, + "eval_accuracy": 0.8756, + "eval_loss": 0.41855242919921876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1744450032711029, + "mean_feedback_forward_cosine": 0.9951367963444103, + "mean_feedback_forward_relative_error": 0.09600104195150462, + "min_feedback_forward_cosine": 0.9849445819854736 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.14812842531204223 + }, + { + "epoch": 80, + "eval_accuracy": 0.8682, + "eval_loss": 0.4982807067871094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16858580708503723, + "mean_feedback_forward_cosine": 0.9954568223519759, + "mean_feedback_forward_relative_error": 0.0927825734696605, + "min_feedback_forward_cosine": 0.985930323600769 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.14163471700880262 + }, + { + "epoch": 81, + "eval_accuracy": 0.8788, + "eval_loss": 0.4318050842285156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1628764420747757, + "mean_feedback_forward_cosine": 0.9957529859109359, + "mean_feedback_forward_relative_error": 0.08970520404929465, + "min_feedback_forward_cosine": 0.9868632555007935 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.14164452673594158 + }, + { + "epoch": 82, + "eval_accuracy": 0.8824, + "eval_loss": 0.4356259216308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15716975927352905, + "mean_feedback_forward_cosine": 0.9960444970564408, + "mean_feedback_forward_relative_error": 0.08655340471728282, + "min_feedback_forward_cosine": 0.9877631068229675 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.1429812172730764 + }, + { + "epoch": 83, + "eval_accuracy": 0.8834, + "eval_loss": 0.4211099670410156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1516208052635193, + "mean_feedback_forward_cosine": 0.996317475492304, + "mean_feedback_forward_relative_error": 0.08351170965893702, + "min_feedback_forward_cosine": 0.9886099696159363 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.14329515885247124 + }, + { + "epoch": 84, + "eval_accuracy": 0.9018, + "eval_loss": 0.35779424438476565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14630314707756042, + "mean_feedback_forward_cosine": 0.9965671257539229, + "mean_feedback_forward_relative_error": 0.08063399812037295, + "min_feedback_forward_cosine": 0.9893922805786133 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.14221931150092018 + }, + { + "epoch": 85, + "eval_accuracy": 0.8702, + "eval_loss": 0.4682876647949219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14148655533790588, + "mean_feedback_forward_cosine": 0.9967920650135387, + "mean_feedback_forward_relative_error": 0.0779483347792517, + "min_feedback_forward_cosine": 0.9900739192962646 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.137186794895596 + }, + { + "epoch": 86, + "eval_accuracy": 0.8932, + "eval_loss": 0.3757369873046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1365615576505661, + "mean_feedback_forward_cosine": 0.9970066233114763, + "mean_feedback_forward_relative_error": 0.07529233558611437, + "min_feedback_forward_cosine": 0.9907506704330444 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.1364652813633283 + }, + { + "epoch": 87, + "eval_accuracy": 0.8844, + "eval_loss": 0.4202875732421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13171401619911194, + "mean_feedback_forward_cosine": 0.9972147313031283, + "mean_feedback_forward_relative_error": 0.07261722778732127, + "min_feedback_forward_cosine": 0.9913944602012634 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.14023481232060325 + }, + { + "epoch": 88, + "eval_accuracy": 0.8998, + "eval_loss": 0.3637839233398438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1271563023328781, + "mean_feedback_forward_cosine": 0.997396829995242, + "mean_feedback_forward_relative_error": 0.07020163962786848, + "min_feedback_forward_cosine": 0.9919781684875488 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.13764463415410783 + }, + { + "epoch": 89, + "eval_accuracy": 0.8942, + "eval_loss": 0.37466764526367186, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12278088927268982, + "mean_feedback_forward_cosine": 0.9975727265531367, + "mean_feedback_forward_relative_error": 0.06778270504014058, + "min_feedback_forward_cosine": 0.9925185441970825 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.13379682774146398 + }, + { + "epoch": 90, + "eval_accuracy": 0.8878, + "eval_loss": 0.41289153442382814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11860892176628113, + "mean_feedback_forward_cosine": 0.9977324615825306, + "mean_feedback_forward_relative_error": 0.06551586030559106, + "min_feedback_forward_cosine": 0.9930169582366943 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.13367654723591274 + }, + { + "epoch": 91, + "eval_accuracy": 0.8796, + "eval_loss": 0.4534392883300781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11397954821586609, + "mean_feedback_forward_cosine": 0.9978942480954257, + "mean_feedback_forward_relative_error": 0.06314405470409176, + "min_feedback_forward_cosine": 0.9935492277145386 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.1437890195131302 + }, + { + "epoch": 92, + "eval_accuracy": 0.877, + "eval_loss": 0.4637198852539062, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11020797491073608, + "mean_feedback_forward_cosine": 0.9980307849970731, + "mean_feedback_forward_relative_error": 0.061057206615805625, + "min_feedback_forward_cosine": 0.9939678311347961 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.1297376984834671 + }, + { + "epoch": 93, + "eval_accuracy": 0.8974, + "eval_loss": 0.38976353759765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10636366158723831, + "mean_feedback_forward_cosine": 0.9981618870388378, + "mean_feedback_forward_relative_error": 0.05898658186197281, + "min_feedback_forward_cosine": 0.9943795204162598 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.13253259558942582 + }, + { + "epoch": 94, + "eval_accuracy": 0.881, + "eval_loss": 0.45050140991210935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1024697870016098, + "mean_feedback_forward_cosine": 0.9982858592813665, + "mean_feedback_forward_relative_error": 0.05696909615600651, + "min_feedback_forward_cosine": 0.9947818517684937 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.13139092605246439 + }, + { + "epoch": 95, + "eval_accuracy": 0.884, + "eval_loss": 0.42125306396484374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09880731254816055, + "mean_feedback_forward_cosine": 0.9984017632224343, + "mean_feedback_forward_relative_error": 0.05500872118229216, + "min_feedback_forward_cosine": 0.9951469302177429 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.13053015309174856 + }, + { + "epoch": 96, + "eval_accuracy": 0.8994, + "eval_loss": 0.37246392211914064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09534408897161484, + "mean_feedback_forward_cosine": 0.9985089713876898, + "mean_feedback_forward_relative_error": 0.053137221051888034, + "min_feedback_forward_cosine": 0.9954792261123657 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.13170791505177815 + }, + { + "epoch": 97, + "eval_accuracy": 0.8916, + "eval_loss": 0.37829818115234376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09184065461158752, + "mean_feedback_forward_cosine": 0.9986103144558993, + "mean_feedback_forward_relative_error": 0.051301561533050107, + "min_feedback_forward_cosine": 0.9958046674728394 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.1322530888001124 + }, + { + "epoch": 98, + "eval_accuracy": 0.8822, + "eval_loss": 0.44099720458984376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08854187279939651, + "mean_feedback_forward_cosine": 0.9987025206739252, + "mean_feedback_forward_relative_error": 0.04957084210420197, + "min_feedback_forward_cosine": 0.9960996508598328 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.1289476733631558 + }, + { + "epoch": 99, + "eval_accuracy": 0.898, + "eval_loss": 0.3525207153320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08554445952177048, + "mean_feedback_forward_cosine": 0.9987886905670166, + "mean_feedback_forward_relative_error": 0.047894603725184094, + "min_feedback_forward_cosine": 0.9963589906692505 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.1272515431880951 + }, + { + "epoch": 100, + "eval_accuracy": 0.8954, + "eval_loss": 0.400292919921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08254595100879669, + "mean_feedback_forward_cosine": 0.9988718823953109, + "mean_feedback_forward_relative_error": 0.04622269711372527, + "min_feedback_forward_cosine": 0.9966089129447937 + }, + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.1293600611527761 + }, + { + "epoch": 101, + "eval_accuracy": 0.9198, + "eval_loss": 0.28196166381835935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08252088725566864, + "mean_feedback_forward_cosine": 0.9988725673068654, + "mean_feedback_forward_relative_error": 0.04620732959698547, + "min_feedback_forward_cosine": 0.9966108798980713 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.06943065629336569 + }, + { + "epoch": 102, + "eval_accuracy": 0.9222, + "eval_loss": 0.28367977294921876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08251013606786728, + "mean_feedback_forward_cosine": 0.9988728794184598, + "mean_feedback_forward_relative_error": 0.04619971408762715, + "min_feedback_forward_cosine": 0.9966118931770325 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.0442628238108423 + }, + { + "epoch": 103, + "eval_accuracy": 0.9236, + "eval_loss": 0.27802401123046877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08250072598457336, + "mean_feedback_forward_cosine": 0.9988731460137801, + "mean_feedback_forward_relative_error": 0.046192759241570125, + "min_feedback_forward_cosine": 0.9966127872467041 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.03693747533584634 + }, + { + "epoch": 104, + "eval_accuracy": 0.925, + "eval_loss": 0.2826873077392578, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0824924111366272, + "mean_feedback_forward_cosine": 0.9988733931021256, + "mean_feedback_forward_relative_error": 0.04618666246533394, + "min_feedback_forward_cosine": 0.9966134428977966 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.030649015731364488 + }, + { + "epoch": 105, + "eval_accuracy": 0.9262, + "eval_loss": 0.2913733154296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.082485131919384, + "mean_feedback_forward_cosine": 0.9988736239346591, + "mean_feedback_forward_relative_error": 0.04618093506856398, + "min_feedback_forward_cosine": 0.9966140389442444 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.02735501567953163 + }, + { + "epoch": 106, + "eval_accuracy": 0.9246, + "eval_loss": 0.2939286346435547, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08247771114110947, + "mean_feedback_forward_cosine": 0.9988738623532382, + "mean_feedback_forward_relative_error": 0.04617522387680682, + "min_feedback_forward_cosine": 0.9966146349906921 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.024195579751663737 + }, + { + "epoch": 107, + "eval_accuracy": 0.9272, + "eval_loss": 0.29316375732421873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08247034251689911, + "mean_feedback_forward_cosine": 0.9988740758462386, + "mean_feedback_forward_relative_error": 0.046169685081324796, + "min_feedback_forward_cosine": 0.9966151714324951 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.022561528134014873 + }, + { + "epoch": 108, + "eval_accuracy": 0.9244, + "eval_loss": 0.30675077514648436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08246370404958725, + "mean_feedback_forward_cosine": 0.9988743045113303, + "mean_feedback_forward_relative_error": 0.04616451940753243, + "min_feedback_forward_cosine": 0.9966158270835876 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.02048250459664398 + }, + { + "epoch": 109, + "eval_accuracy": 0.9252, + "eval_loss": 0.3081547302246094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08245643228292465, + "mean_feedback_forward_cosine": 0.9988745136694475, + "mean_feedback_forward_relative_error": 0.04615932416848161, + "min_feedback_forward_cosine": 0.9966166019439697 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.019332495719856686 + }, + { + "epoch": 110, + "eval_accuracy": 0.9258, + "eval_loss": 0.3063327392578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08244861662387848, + "mean_feedback_forward_cosine": 0.9988747369159352, + "mean_feedback_forward_relative_error": 0.04615418922833421, + "min_feedback_forward_cosine": 0.9966172575950623 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.018840931750006146 + }, + { + "epoch": 111, + "eval_accuracy": 0.926, + "eval_loss": 0.30996431884765624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08244211226701736, + "mean_feedback_forward_cosine": 0.9988749330694026, + "mean_feedback_forward_relative_error": 0.04614953860979189, + "min_feedback_forward_cosine": 0.9966177344322205 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.016484283983376292 + }, + { + "epoch": 112, + "eval_accuracy": 0.9262, + "eval_loss": 0.31271492309570315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0824364498257637, + "mean_feedback_forward_cosine": 0.9988750945438039, + "mean_feedback_forward_relative_error": 0.04614522804252126, + "min_feedback_forward_cosine": 0.9966182708740234 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.014451664542489582 + }, + { + "epoch": 113, + "eval_accuracy": 0.926, + "eval_loss": 0.322545263671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08243008702993393, + "mean_feedback_forward_cosine": 0.9988752787763422, + "mean_feedback_forward_relative_error": 0.04614073788239197, + "min_feedback_forward_cosine": 0.9966187477111816 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.01427282026078966 + }, + { + "epoch": 114, + "eval_accuracy": 0.9246, + "eval_loss": 0.32204202270507815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08242438733577728, + "mean_feedback_forward_cosine": 0.9988754662600431, + "mean_feedback_forward_relative_error": 0.04613645566119389, + "min_feedback_forward_cosine": 0.9966192245483398 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.01288464905437496 + }, + { + "epoch": 115, + "eval_accuracy": 0.9238, + "eval_loss": 0.32415628662109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08241870999336243, + "mean_feedback_forward_cosine": 0.9988756277344444, + "mean_feedback_forward_relative_error": 0.046132276952266696, + "min_feedback_forward_cosine": 0.996619701385498 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.012184910022715727 + }, + { + "epoch": 116, + "eval_accuracy": 0.927, + "eval_loss": 0.32800870971679685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0824129655957222, + "mean_feedback_forward_cosine": 0.9988758065483787, + "mean_feedback_forward_relative_error": 0.0461282127621499, + "min_feedback_forward_cosine": 0.9966202974319458 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.011643429762952857 + }, + { + "epoch": 117, + "eval_accuracy": 0.9266, + "eval_loss": 0.3297094970703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08240707963705063, + "mean_feedback_forward_cosine": 0.9988759918646379, + "mean_feedback_forward_relative_error": 0.04612408006055788, + "min_feedback_forward_cosine": 0.996620774269104 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.011566649926371044 + }, + { + "epoch": 118, + "eval_accuracy": 0.9274, + "eval_loss": 0.33520316162109376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0824013203382492, + "mean_feedback_forward_cosine": 0.9988761804320595, + "mean_feedback_forward_relative_error": 0.04611968616531654, + "min_feedback_forward_cosine": 0.996621310710907 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.01146308294420855 + }, + { + "epoch": 119, + "eval_accuracy": 0.9252, + "eval_loss": 0.3379811279296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08239556103944778, + "mean_feedback_forward_cosine": 0.9988763310692527, + "mean_feedback_forward_relative_error": 0.04611563034016978, + "min_feedback_forward_cosine": 0.9966216087341309 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.010218534137308598 + }, + { + "epoch": 120, + "eval_accuracy": 0.9264, + "eval_loss": 0.33973890380859373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08238988369703293, + "mean_feedback_forward_cosine": 0.9988765250552785, + "mean_feedback_forward_relative_error": 0.04611164289103313, + "min_feedback_forward_cosine": 0.9966222047805786 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.00960727816954669 + }, + { + "epoch": 121, + "eval_accuracy": 0.9266, + "eval_loss": 0.34343662719726564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08238489925861359, + "mean_feedback_forward_cosine": 0.998876690864563, + "mean_feedback_forward_relative_error": 0.046107920445501806, + "min_feedback_forward_cosine": 0.9966225624084473 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.009085430883988738 + }, + { + "epoch": 122, + "eval_accuracy": 0.9246, + "eval_loss": 0.34254222412109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08237940073013306, + "mean_feedback_forward_cosine": 0.9988768653436141, + "mean_feedback_forward_relative_error": 0.04610400257462805, + "min_feedback_forward_cosine": 0.996623158454895 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.008975493807718157 + }, + { + "epoch": 123, + "eval_accuracy": 0.9276, + "eval_loss": 0.34295497436523437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0823746845126152, + "mean_feedback_forward_cosine": 0.9988770127296448, + "mean_feedback_forward_relative_error": 0.04610043037682772, + "min_feedback_forward_cosine": 0.9966235756874084 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.008150759787774749 + }, + { + "epoch": 124, + "eval_accuracy": 0.9246, + "eval_loss": 0.3488459045410156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08236976712942123, + "mean_feedback_forward_cosine": 0.998877163366838, + "mean_feedback_forward_relative_error": 0.04609683183106509, + "min_feedback_forward_cosine": 0.9966238737106323 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.008191198907544215 + }, + { + "epoch": 125, + "eval_accuracy": 0.9276, + "eval_loss": 0.34850970458984376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08236398547887802, + "mean_feedback_forward_cosine": 0.9988773215900768, + "mean_feedback_forward_relative_error": 0.0460928878154267, + "min_feedback_forward_cosine": 0.9966244101524353 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.008717926258511014 + }, + { + "epoch": 126, + "eval_accuracy": 0.9266, + "eval_loss": 0.3463415405273437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08235813677310944, + "mean_feedback_forward_cosine": 0.9988774852319198, + "mean_feedback_forward_relative_error": 0.04608909797600724, + "min_feedback_forward_cosine": 0.9966248273849487 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.008424199743734465 + }, + { + "epoch": 127, + "eval_accuracy": 0.9242, + "eval_loss": 0.3489424072265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08235276490449905, + "mean_feedback_forward_cosine": 0.9988776391202753, + "mean_feedback_forward_relative_error": 0.046085428514263845, + "min_feedback_forward_cosine": 0.9966253042221069 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.00735124247815046 + }, + { + "epoch": 128, + "eval_accuracy": 0.9268, + "eval_loss": 0.3469889953613281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08234743773937225, + "mean_feedback_forward_cosine": 0.9988778103481639, + "mean_feedback_forward_relative_error": 0.04608147801323371, + "min_feedback_forward_cosine": 0.9966257810592651 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.0076482207431561415 + }, + { + "epoch": 129, + "eval_accuracy": 0.9268, + "eval_loss": 0.3487642822265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08234293758869171, + "mean_feedback_forward_cosine": 0.9988779599016363, + "mean_feedback_forward_relative_error": 0.04607815569774671, + "min_feedback_forward_cosine": 0.9966261386871338 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.006547223501776655 + }, + { + "epoch": 130, + "eval_accuracy": 0.9266, + "eval_loss": 0.35033135986328123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08233887702226639, + "mean_feedback_forward_cosine": 0.9988781018690629, + "mean_feedback_forward_relative_error": 0.04607508634640412, + "min_feedback_forward_cosine": 0.9966265559196472 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.006076416883534855 + }, + { + "epoch": 131, + "eval_accuracy": 0.9282, + "eval_loss": 0.35761566162109376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08233317732810974, + "mean_feedback_forward_cosine": 0.998878276348114, + "mean_feedback_forward_relative_error": 0.04607112404297699, + "min_feedback_forward_cosine": 0.9966269731521606 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.007004095608658261 + }, + { + "epoch": 132, + "eval_accuracy": 0.9256, + "eval_loss": 0.35786424560546876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08232785016298294, + "mean_feedback_forward_cosine": 0.9988784291527488, + "mean_feedback_forward_relative_error": 0.046067386340688576, + "min_feedback_forward_cosine": 0.9966274499893188 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.00657963326341576 + }, + { + "epoch": 133, + "eval_accuracy": 0.927, + "eval_loss": 0.3533356994628906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08232413232326508, + "mean_feedback_forward_cosine": 0.9988785591992465, + "mean_feedback_forward_relative_error": 0.04606463924388994, + "min_feedback_forward_cosine": 0.996627688407898 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.005323932779717466 + }, + { + "epoch": 134, + "eval_accuracy": 0.9276, + "eval_loss": 0.35352208251953127, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08231943100690842, + "mean_feedback_forward_cosine": 0.9988786946643483, + "mean_feedback_forward_relative_error": 0.04606130861423232, + "min_feedback_forward_cosine": 0.9966281056404114 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.005774464671355155 + }, + { + "epoch": 135, + "eval_accuracy": 0.927, + "eval_loss": 0.3579586608886719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08231444656848907, + "mean_feedback_forward_cosine": 0.9988788355480541, + "mean_feedback_forward_relative_error": 0.046057646366005595, + "min_feedback_forward_cosine": 0.9966285228729248 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.006122449112104045 + }, + { + "epoch": 136, + "eval_accuracy": 0.9266, + "eval_loss": 0.3627355712890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08231008797883987, + "mean_feedback_forward_cosine": 0.9988789666782726, + "mean_feedback_forward_relative_error": 0.046054657260802656, + "min_feedback_forward_cosine": 0.9966288805007935 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.005008096188720729 + }, + { + "epoch": 137, + "eval_accuracy": 0.9254, + "eval_loss": 0.36479898681640627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08230576664209366, + "mean_feedback_forward_cosine": 0.9988791205666282, + "mean_feedback_forward_relative_error": 0.04605141476812688, + "min_feedback_forward_cosine": 0.9966292381286621 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.00519936425000843 + }, + { + "epoch": 138, + "eval_accuracy": 0.9272, + "eval_loss": 0.36759619140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08230125904083252, + "mean_feedback_forward_cosine": 0.9988792657852172, + "mean_feedback_forward_relative_error": 0.04604820312762802, + "min_feedback_forward_cosine": 0.9966295957565308 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.005097926148772239 + }, + { + "epoch": 139, + "eval_accuracy": 0.928, + "eval_loss": 0.367848046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08229579031467438, + "mean_feedback_forward_cosine": 0.9988794348456643, + "mean_feedback_forward_relative_error": 0.04604436362331564, + "min_feedback_forward_cosine": 0.9966301321983337 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.00578090578098264 + }, + { + "epoch": 140, + "eval_accuracy": 0.9258, + "eval_loss": 0.3670384765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08228998631238937, + "mean_feedback_forward_cosine": 0.9988796158270402, + "mean_feedback_forward_relative_error": 0.04604042116552591, + "min_feedback_forward_cosine": 0.9966304898262024 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.005641720619507962 + }, + { + "epoch": 141, + "eval_accuracy": 0.9266, + "eval_loss": 0.36997833251953127, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08228462934494019, + "mean_feedback_forward_cosine": 0.998879792473533, + "mean_feedback_forward_relative_error": 0.04603636854074218, + "min_feedback_forward_cosine": 0.9966309070587158 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.0057345138311800025 + }, + { + "epoch": 142, + "eval_accuracy": 0.9278, + "eval_loss": 0.36403873291015626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08227932453155518, + "mean_feedback_forward_cosine": 0.998879955031655, + "mean_feedback_forward_relative_error": 0.046032756346870554, + "min_feedback_forward_cosine": 0.9966314435005188 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.004907455729775958 + }, + { + "epoch": 143, + "eval_accuracy": 0.928, + "eval_loss": 0.36518594360351564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08227457106113434, + "mean_feedback_forward_cosine": 0.9988800861618735, + "mean_feedback_forward_relative_error": 0.046029586328024216, + "min_feedback_forward_cosine": 0.9966318607330322 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.004633074706503087 + }, + { + "epoch": 144, + "eval_accuracy": 0.9288, + "eval_loss": 0.36308513793945313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822695642709732, + "mean_feedback_forward_cosine": 0.9988802649758078, + "mean_feedback_forward_relative_error": 0.04602613601494919, + "min_feedback_forward_cosine": 0.9966320991516113 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.004616090032748051 + }, + { + "epoch": 145, + "eval_accuracy": 0.928, + "eval_loss": 0.3633923767089844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08226427435874939, + "mean_feedback_forward_cosine": 0.998880432952534, + "mean_feedback_forward_relative_error": 0.04602236039936543, + "min_feedback_forward_cosine": 0.9966326951980591 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.005115178668167856 + }, + { + "epoch": 146, + "eval_accuracy": 0.9282, + "eval_loss": 0.3682091247558594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822596624493599, + "mean_feedback_forward_cosine": 0.9988805814222856, + "mean_feedback_forward_relative_error": 0.0460187761282379, + "min_feedback_forward_cosine": 0.9966330528259277 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.004550867236633268 + }, + { + "epoch": 147, + "eval_accuracy": 0.927, + "eval_loss": 0.3651136474609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08225411176681519, + "mean_feedback_forward_cosine": 0.99888076023622, + "mean_feedback_forward_relative_error": 0.04601478727365082, + "min_feedback_forward_cosine": 0.9966334700584412 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.004593415929294295 + }, + { + "epoch": 148, + "eval_accuracy": 0.9262, + "eval_loss": 0.3646269836425781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08224842697381973, + "mean_feedback_forward_cosine": 0.9988809282129462, + "mean_feedback_forward_relative_error": 0.046011200902814214, + "min_feedback_forward_cosine": 0.9966338872909546 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.004782757617574599 + }, + { + "epoch": 149, + "eval_accuracy": 0.9272, + "eval_loss": 0.365397119140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822431892156601, + "mean_feedback_forward_cosine": 0.9988811276175759, + "mean_feedback_forward_relative_error": 0.04600701702928001, + "min_feedback_forward_cosine": 0.9966343641281128 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.004701021136467655 + }, + { + "epoch": 150, + "eval_accuracy": 0.9304, + "eval_loss": 0.3664459228515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223860710859299, + "mean_feedback_forward_cosine": 0.9988812652501193, + "mean_feedback_forward_relative_error": 0.0460035984150388, + "min_feedback_forward_cosine": 0.9966347217559814 + }, + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.004233766572032538 + }, + { + "epoch": 151, + "eval_accuracy": 0.9292, + "eval_loss": 0.36865162353515624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223855495452881, + "mean_feedback_forward_cosine": 0.9988812836733731, + "mean_feedback_forward_relative_error": 0.046003527346659794, + "min_feedback_forward_cosine": 0.9966347217559814 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.0036866345710224574 + }, + { + "epoch": 152, + "eval_accuracy": 0.93, + "eval_loss": 0.36565570068359377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223850280046463, + "mean_feedback_forward_cosine": 0.9988812869245356, + "mean_feedback_forward_relative_error": 0.046003469282930545, + "min_feedback_forward_cosine": 0.9966347217559814 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.003957245495936109 + }, + { + "epoch": 153, + "eval_accuracy": 0.929, + "eval_loss": 0.3695786437988281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223845064640045, + "mean_feedback_forward_cosine": 0.9988812869245356, + "mean_feedback_forward_relative_error": 0.04600340781564062, + "min_feedback_forward_cosine": 0.9966347813606262 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.003925301420285056 + }, + { + "epoch": 154, + "eval_accuracy": 0.9302, + "eval_loss": 0.36479189453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223842084407806, + "mean_feedback_forward_cosine": 0.998881278254769, + "mean_feedback_forward_relative_error": 0.046003351631489664, + "min_feedback_forward_cosine": 0.9966347813606262 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.0035732608032723267 + }, + { + "epoch": 155, + "eval_accuracy": 0.9302, + "eval_loss": 0.3620485961914062, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223836869001389, + "mean_feedback_forward_cosine": 0.9988812825896524, + "mean_feedback_forward_relative_error": 0.04600329211151058, + "min_feedback_forward_cosine": 0.9966347217559814 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.0037292424756826628 + }, + { + "epoch": 156, + "eval_accuracy": 0.9308, + "eval_loss": 0.3636377868652344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822383388876915, + "mean_feedback_forward_cosine": 0.9988812825896524, + "mean_feedback_forward_relative_error": 0.046003235639496286, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.0035863490228851638 + }, + { + "epoch": 157, + "eval_accuracy": 0.93, + "eval_loss": 0.36617817993164065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223830908536911, + "mean_feedback_forward_cosine": 0.9988812869245356, + "mean_feedback_forward_relative_error": 0.04600318036973476, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.0034266716464733085 + }, + { + "epoch": 158, + "eval_accuracy": 0.9284, + "eval_loss": 0.36658130493164065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223827183246613, + "mean_feedback_forward_cosine": 0.9988812825896524, + "mean_feedback_forward_relative_error": 0.04600312345745888, + "min_feedback_forward_cosine": 0.9966347217559814 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0035902588734713693 + }, + { + "epoch": 159, + "eval_accuracy": 0.93, + "eval_loss": 0.3643918701171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223822712898254, + "mean_feedback_forward_cosine": 0.998881290175698, + "mean_feedback_forward_relative_error": 0.04600307231938297, + "min_feedback_forward_cosine": 0.9966347813606262 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.003073369382456359 + }, + { + "epoch": 160, + "eval_accuracy": 0.9296, + "eval_loss": 0.3648558898925781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223821222782135, + "mean_feedback_forward_cosine": 0.998881290175698, + "mean_feedback_forward_relative_error": 0.04600302441553636, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.0029175158710322445 + }, + { + "epoch": 161, + "eval_accuracy": 0.9304, + "eval_loss": 0.367481298828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223816752433777, + "mean_feedback_forward_cosine": 0.9988812869245356, + "mean_feedback_forward_relative_error": 0.04600296841764992, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.0034142314993465942 + }, + { + "epoch": 162, + "eval_accuracy": 0.9284, + "eval_loss": 0.3675195251464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223813772201538, + "mean_feedback_forward_cosine": 0.9988812934268605, + "mean_feedback_forward_relative_error": 0.04600291731344028, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.0030324965446359583 + }, + { + "epoch": 163, + "eval_accuracy": 0.9292, + "eval_loss": 0.36677179565429685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822380930185318, + "mean_feedback_forward_cosine": 0.998881302096627, + "mean_feedback_forward_relative_error": 0.04600286566737023, + "min_feedback_forward_cosine": 0.9966347217559814 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.0034874971714284686 + }, + { + "epoch": 164, + "eval_accuracy": 0.9286, + "eval_loss": 0.36896765747070315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223806321620941, + "mean_feedback_forward_cosine": 0.998881302096627, + "mean_feedback_forward_relative_error": 0.046002815003422175, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.0029989460522101986 + }, + { + "epoch": 165, + "eval_accuracy": 0.9306, + "eval_loss": 0.366157421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223803341388702, + "mean_feedback_forward_cosine": 0.9988813031803477, + "mean_feedback_forward_relative_error": 0.04600276467813687, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.0031791578263044356 + }, + { + "epoch": 166, + "eval_accuracy": 0.9296, + "eval_loss": 0.3687070068359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223800361156464, + "mean_feedback_forward_cosine": 0.9988812880082564, + "mean_feedback_forward_relative_error": 0.0460027188570662, + "min_feedback_forward_cosine": 0.9966347813606262 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.002584952381066978 + }, + { + "epoch": 167, + "eval_accuracy": 0.93, + "eval_loss": 0.3658307983398438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223797380924225, + "mean_feedback_forward_cosine": 0.9988812923431396, + "mean_feedback_forward_relative_error": 0.04600266805765304, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.0029853353516198695 + }, + { + "epoch": 168, + "eval_accuracy": 0.9294, + "eval_loss": 0.36633271484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223794400691986, + "mean_feedback_forward_cosine": 0.9988813064315102, + "mean_feedback_forward_relative_error": 0.04600261669944633, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.0029499868230480285 + }, + { + "epoch": 169, + "eval_accuracy": 0.9298, + "eval_loss": 0.36940147705078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223791420459747, + "mean_feedback_forward_cosine": 0.9988812934268605, + "mean_feedback_forward_relative_error": 0.046002566916021434, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.002881501671630475 + }, + { + "epoch": 170, + "eval_accuracy": 0.9294, + "eval_loss": 0.369974560546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822378620505333, + "mean_feedback_forward_cosine": 0.9988812999291854, + "mean_feedback_forward_relative_error": 0.04600251288237897, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.0030441269177322585 + }, + { + "epoch": 171, + "eval_accuracy": 0.9282, + "eval_loss": 0.36931307983398437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223782479763031, + "mean_feedback_forward_cosine": 0.9988813096826726, + "mean_feedback_forward_relative_error": 0.04600245668129487, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.0033884187957892815 + }, + { + "epoch": 172, + "eval_accuracy": 0.9286, + "eval_loss": 0.36912783813476563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223780989646912, + "mean_feedback_forward_cosine": 0.9988813031803477, + "mean_feedback_forward_relative_error": 0.04600240898064591, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.0030959877320668764 + }, + { + "epoch": 173, + "eval_accuracy": 0.929, + "eval_loss": 0.36873092041015626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223778009414673, + "mean_feedback_forward_cosine": 0.9988813042640686, + "mean_feedback_forward_relative_error": 0.04600236126306382, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.0028156025466819605 + }, + { + "epoch": 174, + "eval_accuracy": 0.9298, + "eval_loss": 0.36716961669921877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223774284124374, + "mean_feedback_forward_cosine": 0.9988812934268605, + "mean_feedback_forward_relative_error": 0.04600231291895563, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.0027567072859654824 + }, + { + "epoch": 175, + "eval_accuracy": 0.9292, + "eval_loss": 0.3688115600585937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223771303892136, + "mean_feedback_forward_cosine": 0.9988813085989519, + "mean_feedback_forward_relative_error": 0.04600225873291493, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.003222914586795701 + }, + { + "epoch": 176, + "eval_accuracy": 0.9296, + "eval_loss": 0.36736928100585936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223768323659897, + "mean_feedback_forward_cosine": 0.9988813031803477, + "mean_feedback_forward_relative_error": 0.04600221047347242, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.0028732660255498354 + }, + { + "epoch": 177, + "eval_accuracy": 0.9294, + "eval_loss": 0.372470849609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223765343427658, + "mean_feedback_forward_cosine": 0.9988813107663935, + "mean_feedback_forward_relative_error": 0.04600216378881173, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.0025538987044348483 + }, + { + "epoch": 178, + "eval_accuracy": 0.9284, + "eval_loss": 0.36910101928710937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822376236319542, + "mean_feedback_forward_cosine": 0.9988813064315102, + "mean_feedback_forward_relative_error": 0.04600211475044489, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.002970595129744874 + }, + { + "epoch": 179, + "eval_accuracy": 0.9298, + "eval_loss": 0.37037894287109374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822376161813736, + "mean_feedback_forward_cosine": 0.998881314017556, + "mean_feedback_forward_relative_error": 0.046002069691365416, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.0025246295870385237 + }, + { + "epoch": 180, + "eval_accuracy": 0.9294, + "eval_loss": 0.36845943603515624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223757892847061, + "mean_feedback_forward_cosine": 0.9988813075152311, + "mean_feedback_forward_relative_error": 0.0460020206191323, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.002713063888189693 + }, + { + "epoch": 181, + "eval_accuracy": 0.9286, + "eval_loss": 0.3703472412109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223755657672882, + "mean_feedback_forward_cosine": 0.9988813118501143, + "mean_feedback_forward_relative_error": 0.04600197120823644, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.0025600299073590175 + }, + { + "epoch": 182, + "eval_accuracy": 0.9292, + "eval_loss": 0.3755940856933594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223753422498703, + "mean_feedback_forward_cosine": 0.9988812988454645, + "mean_feedback_forward_relative_error": 0.0460019274699417, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.0027186543648027714 + }, + { + "epoch": 183, + "eval_accuracy": 0.9292, + "eval_loss": 0.3694640441894531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223750442266464, + "mean_feedback_forward_cosine": 0.9988813085989519, + "mean_feedback_forward_relative_error": 0.04600187953222882, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.0028076297329221334 + }, + { + "epoch": 184, + "eval_accuracy": 0.9292, + "eval_loss": 0.3719008056640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223746716976166, + "mean_feedback_forward_cosine": 0.9988813118501143, + "mean_feedback_forward_relative_error": 0.046001831357452005, + "min_feedback_forward_cosine": 0.9966347813606262 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.00280507246243457 + }, + { + "epoch": 185, + "eval_accuracy": 0.9292, + "eval_loss": 0.3697983947753906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223744481801987, + "mean_feedback_forward_cosine": 0.9988813129338351, + "mean_feedback_forward_relative_error": 0.046001782928678125, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.002665952594392002 + }, + { + "epoch": 186, + "eval_accuracy": 0.9294, + "eval_loss": 0.3683447998046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223741501569748, + "mean_feedback_forward_cosine": 0.9988813216036017, + "mean_feedback_forward_relative_error": 0.04600173328071833, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.0028820886498110163 + }, + { + "epoch": 187, + "eval_accuracy": 0.9282, + "eval_loss": 0.37149351806640624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223740011453629, + "mean_feedback_forward_cosine": 0.9988813172687184, + "mean_feedback_forward_relative_error": 0.04600168620659546, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.00259548195157614 + }, + { + "epoch": 188, + "eval_accuracy": 0.9298, + "eval_loss": 0.37394226684570314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822373703122139, + "mean_feedback_forward_cosine": 0.9988813085989519, + "mean_feedback_forward_relative_error": 0.0460016364401037, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.0025710957238243687 + }, + { + "epoch": 189, + "eval_accuracy": 0.9294, + "eval_loss": 0.3728668823242188, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223734050989151, + "mean_feedback_forward_cosine": 0.9988813161849975, + "mean_feedback_forward_relative_error": 0.04600158809599551, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.002441268061970671 + }, + { + "epoch": 190, + "eval_accuracy": 0.9278, + "eval_loss": 0.37325433959960935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223730325698853, + "mean_feedback_forward_cosine": 0.9988813237710432, + "mean_feedback_forward_relative_error": 0.04600153495980935, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.0030400358228633803 + }, + { + "epoch": 191, + "eval_accuracy": 0.929, + "eval_loss": 0.3715138000488281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223726600408554, + "mean_feedback_forward_cosine": 0.9988813443617387, + "mean_feedback_forward_relative_error": 0.04600149122151462, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.0025101503582723025 + }, + { + "epoch": 192, + "eval_accuracy": 0.9292, + "eval_loss": 0.3713754638671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223722875118256, + "mean_feedback_forward_cosine": 0.9988813129338351, + "mean_feedback_forward_relative_error": 0.04600143927064809, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.0031009860323948993 + }, + { + "epoch": 193, + "eval_accuracy": 0.9292, + "eval_loss": 0.3713826843261719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223719894886017, + "mean_feedback_forward_cosine": 0.9988813270222057, + "mean_feedback_forward_relative_error": 0.046001392823051324, + "min_feedback_forward_cosine": 0.996634840965271 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.0025039921342084807 + }, + { + "epoch": 194, + "eval_accuracy": 0.9298, + "eval_loss": 0.3682757446289063, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223717659711838, + "mean_feedback_forward_cosine": 0.9988813324408098, + "mean_feedback_forward_relative_error": 0.046001344089480964, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.002788148091774848 + }, + { + "epoch": 195, + "eval_accuracy": 0.9296, + "eval_loss": 0.37077115478515627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0822371393442154, + "mean_feedback_forward_cosine": 0.9988813226873224, + "mean_feedback_forward_relative_error": 0.046001296236433765, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.002563142611583074 + }, + { + "epoch": 196, + "eval_accuracy": 0.9296, + "eval_loss": 0.37052415771484376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223710209131241, + "mean_feedback_forward_cosine": 0.9988813378594138, + "mean_feedback_forward_relative_error": 0.04600124660540711, + "min_feedback_forward_cosine": 0.9966350197792053 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.0027777536426153446 + }, + { + "epoch": 197, + "eval_accuracy": 0.929, + "eval_loss": 0.36994708862304687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223707228899002, + "mean_feedback_forward_cosine": 0.9988813497803428, + "mean_feedback_forward_relative_error": 0.04600120046260682, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.0028230392361473706 + }, + { + "epoch": 198, + "eval_accuracy": 0.9292, + "eval_loss": 0.3723174560546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223704993724823, + "mean_feedback_forward_cosine": 0.9988813291896473, + "mean_feedback_forward_relative_error": 0.04600115440447222, + "min_feedback_forward_cosine": 0.9966349005699158 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.0022861307460400795 + }, + { + "epoch": 199, + "eval_accuracy": 0.9296, + "eval_loss": 0.36884978637695315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223701268434525, + "mean_feedback_forward_cosine": 0.9988813216036017, + "mean_feedback_forward_relative_error": 0.046001107211817396, + "min_feedback_forward_cosine": 0.9966349601745605 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.0030935465381791193 + }, + { + "epoch": 200, + "eval_accuracy": 0.9274, + "eval_loss": 0.3740764099121094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08223697543144226, + "mean_feedback_forward_cosine": 0.9988813335245306, + "mean_feedback_forward_relative_error": 0.04600105993449688, + "min_feedback_forward_cosine": 0.9966350197792053 + }, + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.002820015812975665 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9274, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.3740764099121094, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 3256025088, + "peak_memory_reserved_bytes": 4517265408, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 60.171127796173096, + "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": 5054.333840370178, + "train_wall_s": 4991.6432955265045 + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1125389952000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1125389952000000, + "local_weight_correlation_macs": 1129371264000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1129371264000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 4509522432000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json.manifest.json new file mode 100644 index 0000000..8e5d9a6 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "kp", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787020819.994194, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 5064.757298946381, + "epochs": 200, + "experiment_name": "resnet-r2-clean_kp-d56-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "clean_kp", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-clean_kp-d56-lr0p1.json", + "output_exists": true, + "output_sha256": "69d1c59511cfe98a72af9c1be079d4fa581096630b2e93f4ad028a4a935bc9fa", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json b/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json new file mode 100644 index 0000000..43fff4c --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json @@ -0,0 +1,2604 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "lr": 0.01, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": -4.417266851911942e-05, + "innovation_negative_gradient_cosine": [ + -0.00042737211333587766, + 0.0018140280153602362, + 0.0009411051869392395, + -0.002272804267704487, + -0.001042853808030486, + 0.0007228609756566584, + -0.0014636360574513674, + 0.004689148627221584, + 0.00022076242021284997, + 0.0029502036049962044, + 0.006819497793912888, + 0.005044919438660145, + 0.007197106722742319, + 0.019712356850504875, + 0.021992990747094154, + 0.02051355317234993, + 0.026908088475465775, + 0.029072165489196777, + 0.039473552256822586 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.00042737211333587766, + 0.0018140280153602362, + 0.0009411051869392395, + -0.002272804267704487, + -0.001042853808030486, + 0.0007228609756566584, + -0.0014636360574513674, + 0.004689148627221584, + 0.00022076242021284997, + 0.0029502036049962044, + 0.006819497793912888, + 0.005044919438660145, + 0.007197106722742319, + 0.019712356850504875, + 0.021992990747094154, + 0.02051355317234993, + 0.026908088475465775, + 0.029072165489196777, + 0.039473552256822586 + ], + "teaching_negative_gradient_cosine": [ + -0.00042737211333587766, + 0.0018140280153602362, + 0.0009411051869392395, + -0.002272804267704487, + -0.001042853808030486, + 0.0007228609756566584, + -0.0014636360574513674, + 0.004689148627221584, + 0.00022076242021284997, + 0.0029502036049962044, + 0.006819497793912888, + 0.005044919438660145, + 0.007197106722742319, + 0.019712356850504875, + 0.021992990747094154, + 0.02051355317234993, + 0.026908088475465775, + 0.029072165489196777, + 0.039473552256822586 + ], + "wall_s": 0.13636183738708496 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.2384, + "eval_loss": 2.13388974609375, + "eval_split": "validation", + "lr": 0.002, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.196710070376926 + }, + { + "epoch": 2, + "eval_accuracy": 0.219, + "eval_loss": 2.118686376953125, + "eval_split": "validation", + "lr": 0.004, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 2.1524469988505044 + }, + { + "epoch": 3, + "eval_accuracy": 0.2358, + "eval_loss": 2.120386865234375, + "eval_split": "validation", + "lr": 0.006, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 2.16505956264072 + }, + { + "epoch": 4, + "eval_accuracy": 0.2538, + "eval_loss": 2.0546119140625, + "eval_split": "validation", + "lr": 0.008, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 2.250287850613064 + }, + { + "epoch": 5, + "eval_accuracy": 0.2272, + "eval_loss": 2.4211458984375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 2.4122211878458657 + }, + { + "epoch": 6, + "eval_accuracy": 0.2418, + "eval_loss": 2.48471376953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 2.5066193411933053 + }, + { + "epoch": 7, + "eval_accuracy": 0.2642, + "eval_loss": 2.71391015625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 3.1259462588416205 + }, + { + "epoch": 8, + "eval_accuracy": 0.2838, + "eval_loss": 3.04355810546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 3.7678212508307563 + }, + { + "epoch": 9, + "eval_accuracy": 0.2724, + "eval_loss": 3.169311083984375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 3.511760868496365 + }, + { + "epoch": 10, + "eval_accuracy": 0.2212, + "eval_loss": 7.113378125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 4.115778273179796 + }, + { + "epoch": 11, + "eval_accuracy": 0.2276, + "eval_loss": 3.861981689453125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 5.214946193271213 + }, + { + "epoch": 12, + "eval_accuracy": 0.3098, + "eval_loss": 3.261718896484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 5.978883670467801 + }, + { + "epoch": 13, + "eval_accuracy": 0.2664, + "eval_loss": 4.601353125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 6.004014733632405 + }, + { + "epoch": 14, + "eval_accuracy": 0.2076, + "eval_loss": 8.8213029296875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 6.477945835537381 + }, + { + "epoch": 15, + "eval_accuracy": 0.1582, + "eval_loss": 10.4473234375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 8.101313818952772 + }, + { + "epoch": 16, + "eval_accuracy": 0.3258, + "eval_loss": 4.61436240234375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 8.165552450985379 + }, + { + "epoch": 17, + "eval_accuracy": 0.3088, + "eval_loss": 10.7543853515625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 9.495957034132216 + }, + { + "epoch": 18, + "eval_accuracy": 0.2616, + "eval_loss": 12.048984375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 10.429058261617024 + }, + { + "epoch": 19, + "eval_accuracy": 0.2384, + "eval_loss": 11.6567517578125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 10.979766198221842 + }, + { + "epoch": 20, + "eval_accuracy": 0.2934, + "eval_loss": 14.16575390625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 12.528717605929904 + }, + { + "epoch": 21, + "eval_accuracy": 0.232, + "eval_loss": 26.35266015625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 14.306350407918295 + }, + { + "epoch": 22, + "eval_accuracy": 0.2194, + "eval_loss": 22.857603125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 14.619046181911893 + }, + { + "epoch": 23, + "eval_accuracy": 0.2256, + "eval_loss": 11.67859609375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 16.98607453087701 + }, + { + "epoch": 24, + "eval_accuracy": 0.2354, + "eval_loss": 17.170334375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 19.04719298078749 + }, + { + "epoch": 25, + "eval_accuracy": 0.2598, + "eval_loss": 15.6980568359375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 20.507712957763673 + }, + { + "epoch": 26, + "eval_accuracy": 0.2124, + "eval_loss": 24.550714453125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 20.977451344129776 + }, + { + "epoch": 27, + "eval_accuracy": 0.2686, + "eval_loss": 34.49674140625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 22.611929234483508 + }, + { + "epoch": 28, + "eval_accuracy": 0.2596, + "eval_loss": 18.92392265625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 22.959516687011718 + }, + { + "epoch": 29, + "eval_accuracy": 0.2792, + "eval_loss": 27.662224609375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 24.995158879937065 + }, + { + "epoch": 30, + "eval_accuracy": 0.2662, + "eval_loss": 33.141328515625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 29.480212257554797 + }, + { + "epoch": 31, + "eval_accuracy": 0.226, + "eval_loss": 27.803342578125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 26.761602039252388 + }, + { + "epoch": 32, + "eval_accuracy": 0.277, + "eval_loss": 20.175499609375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 30.244316629367404 + }, + { + "epoch": 33, + "eval_accuracy": 0.3044, + "eval_loss": 29.908916796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 31.322398626708985 + }, + { + "epoch": 34, + "eval_accuracy": 0.245, + "eval_loss": 45.0069875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 33.19071531711155 + }, + { + "epoch": 35, + "eval_accuracy": 0.3816, + "eval_loss": 12.8741255859375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 35.12927552693685 + }, + { + "epoch": 36, + "eval_accuracy": 0.2418, + "eval_loss": 62.24336953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 38.10827041354709 + }, + { + "epoch": 37, + "eval_accuracy": 0.295, + "eval_loss": 36.87111953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 42.39294600694444 + }, + { + "epoch": 38, + "eval_accuracy": 0.1248, + "eval_loss": 83.8959890625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 42.81453029039171 + }, + { + "epoch": 39, + "eval_accuracy": 0.289, + "eval_loss": 85.4705140625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 46.396105011664496 + }, + { + "epoch": 40, + "eval_accuracy": 0.224, + "eval_loss": 59.92510703125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 46.26691556532118 + }, + { + "epoch": 41, + "eval_accuracy": 0.2546, + "eval_loss": 37.7355953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 50.41342158237033 + }, + { + "epoch": 42, + "eval_accuracy": 0.2456, + "eval_loss": 77.8837359375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 50.42762740614149 + }, + { + "epoch": 43, + "eval_accuracy": 0.2972, + "eval_loss": 37.9858015625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 55.54701350640191 + }, + { + "epoch": 44, + "eval_accuracy": 0.3668, + "eval_loss": 43.6133109375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 62.59512099134657 + }, + { + "epoch": 45, + "eval_accuracy": 0.2564, + "eval_loss": 53.872640625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 56.61767745632596 + }, + { + "epoch": 46, + "eval_accuracy": 0.1912, + "eval_loss": 132.7386125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 70.22152931179471 + }, + { + "epoch": 47, + "eval_accuracy": 0.368, + "eval_loss": 35.1759109375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 70.48875649888781 + }, + { + "epoch": 48, + "eval_accuracy": 0.365, + "eval_loss": 30.53370625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 69.93644878607856 + }, + { + "epoch": 49, + "eval_accuracy": 0.283, + "eval_loss": 57.0330375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 77.78397584296333 + }, + { + "epoch": 50, + "eval_accuracy": 0.2704, + "eval_loss": 118.6991953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 75.2017873196072 + }, + { + "epoch": 51, + "eval_accuracy": 0.316, + "eval_loss": 63.6082796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 83.56330694444445 + }, + { + "epoch": 52, + "eval_accuracy": 0.2914, + "eval_loss": 108.32306875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 86.26330736897786 + }, + { + "epoch": 53, + "eval_accuracy": 0.334, + "eval_loss": 46.92616015625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 87.97140079684787 + }, + { + "epoch": 54, + "eval_accuracy": 0.2518, + "eval_loss": 104.82296875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 93.9687024102105 + }, + { + "epoch": 55, + "eval_accuracy": 0.3186, + "eval_loss": 61.5724078125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 96.0325408203125 + }, + { + "epoch": 56, + "eval_accuracy": 0.3622, + "eval_loss": 54.3312796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 103.135226820204 + }, + { + "epoch": 57, + "eval_accuracy": 0.2204, + "eval_loss": 214.8653625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 112.44556560058594 + }, + { + "epoch": 58, + "eval_accuracy": 0.336, + "eval_loss": 76.492653125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 97.48807627766926 + }, + { + "epoch": 59, + "eval_accuracy": 0.295, + "eval_loss": 134.529453125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 115.6962455078125 + }, + { + "epoch": 60, + "eval_accuracy": 0.3338, + "eval_loss": 80.4154484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 114.1658334906684 + }, + { + "epoch": 61, + "eval_accuracy": 0.2612, + "eval_loss": 82.935240625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 120.49449004584419 + }, + { + "epoch": 62, + "eval_accuracy": 0.2506, + "eval_loss": 177.4444, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 129.52251632215712 + }, + { + "epoch": 63, + "eval_accuracy": 0.2992, + "eval_loss": 158.935075, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 126.56348369276259 + }, + { + "epoch": 64, + "eval_accuracy": 0.2318, + "eval_loss": 224.766190625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 129.15274677734374 + }, + { + "epoch": 65, + "eval_accuracy": 0.258, + "eval_loss": 149.426728125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 149.42977838812934 + }, + { + "epoch": 66, + "eval_accuracy": 0.1468, + "eval_loss": 236.724253125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 137.12026451009115 + }, + { + "epoch": 67, + "eval_accuracy": 0.288, + "eval_loss": 133.139496875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 164.30003099907768 + }, + { + "epoch": 68, + "eval_accuracy": 0.2536, + "eval_loss": 175.942903125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 146.02838853624132 + }, + { + "epoch": 69, + "eval_accuracy": 0.2536, + "eval_loss": 117.812740625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 170.32516332465278 + }, + { + "epoch": 70, + "eval_accuracy": 0.3408, + "eval_loss": 139.700415625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 167.26187187771268 + }, + { + "epoch": 71, + "eval_accuracy": 0.275, + "eval_loss": 271.8581, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 150.94737664930557 + }, + { + "epoch": 72, + "eval_accuracy": 0.2768, + "eval_loss": 109.6359796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 178.7975426405165 + }, + { + "epoch": 73, + "eval_accuracy": 0.2492, + "eval_loss": 153.839534375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 181.78071877712674 + }, + { + "epoch": 74, + "eval_accuracy": 0.2398, + "eval_loss": 362.14533125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 179.54143683810764 + }, + { + "epoch": 75, + "eval_accuracy": 0.333, + "eval_loss": 109.10584375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 228.60323557128908 + }, + { + "epoch": 76, + "eval_accuracy": 0.1564, + "eval_loss": 465.6028625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 178.3261564453125 + }, + { + "epoch": 77, + "eval_accuracy": 0.3438, + "eval_loss": 100.3464203125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 208.42907971191406 + }, + { + "epoch": 78, + "eval_accuracy": 0.2746, + "eval_loss": 294.67795, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 210.42169096137152 + }, + { + "epoch": 79, + "eval_accuracy": 0.276, + "eval_loss": 190.441990625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 224.93065251193576 + }, + { + "epoch": 80, + "eval_accuracy": 0.3246, + "eval_loss": 248.689990625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 222.3671440999349 + }, + { + "epoch": 81, + "eval_accuracy": 0.271, + "eval_loss": 167.538346875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 235.71425759006075 + }, + { + "epoch": 82, + "eval_accuracy": 0.1452, + "eval_loss": 355.31036875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 222.38550850423178 + }, + { + "epoch": 83, + "eval_accuracy": 0.2744, + "eval_loss": 190.79715625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 234.47545372721353 + }, + { + "epoch": 84, + "eval_accuracy": 0.2446, + "eval_loss": 268.720678125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 239.26784101833766 + }, + { + "epoch": 85, + "eval_accuracy": 0.2664, + "eval_loss": 219.119975, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 277.5253189887153 + }, + { + "epoch": 86, + "eval_accuracy": 0.2812, + "eval_loss": 186.918509375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 267.0558938340929 + }, + { + "epoch": 87, + "eval_accuracy": 0.275, + "eval_loss": 291.21615625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 253.1571425998264 + }, + { + "epoch": 88, + "eval_accuracy": 0.4022, + "eval_loss": 197.7845375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 271.55762882486977 + }, + { + "epoch": 89, + "eval_accuracy": 0.2136, + "eval_loss": 384.08820625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 272.41770344509547 + }, + { + "epoch": 90, + "eval_accuracy": 0.2894, + "eval_loss": 266.3597, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 279.4282325954861 + }, + { + "epoch": 91, + "eval_accuracy": 0.3094, + "eval_loss": 177.42945, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 307.67940743001304 + }, + { + "epoch": 92, + "eval_accuracy": 0.1138, + "eval_loss": 565.199775, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 308.0857044731988 + }, + { + "epoch": 93, + "eval_accuracy": 0.2484, + "eval_loss": 306.35499375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 296.103281765408 + }, + { + "epoch": 94, + "eval_accuracy": 0.3628, + "eval_loss": 315.681325, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 322.4415712673611 + }, + { + "epoch": 95, + "eval_accuracy": 0.3176, + "eval_loss": 212.66410625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 327.5514536295573 + }, + { + "epoch": 96, + "eval_accuracy": 0.345, + "eval_loss": 154.168359375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 375.67241025661895 + }, + { + "epoch": 97, + "eval_accuracy": 0.364, + "eval_loss": 292.67804375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 345.29349494900174 + }, + { + "epoch": 98, + "eval_accuracy": 0.3088, + "eval_loss": 412.27108125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 366.83919205729165 + }, + { + "epoch": 99, + "eval_accuracy": 0.2158, + "eval_loss": 538.530225, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 346.40431893988716 + }, + { + "epoch": 100, + "eval_accuracy": 0.2714, + "eval_loss": 350.2196375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 364.6515966362847 + }, + { + "epoch": 101, + "eval_accuracy": 0.4338, + "eval_loss": 72.9980640625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 97.5824759250217 + }, + { + "epoch": 102, + "eval_accuracy": 0.3946, + "eval_loss": 63.793271875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 68.66056476982965 + }, + { + "epoch": 103, + "eval_accuracy": 0.3972, + "eval_loss": 54.701396875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 58.456144613308375 + }, + { + "epoch": 104, + "eval_accuracy": 0.395, + "eval_loss": 45.80713984375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 52.66843123033312 + }, + { + "epoch": 105, + "eval_accuracy": 0.3904, + "eval_loss": 42.960628125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 47.554370183648004 + }, + { + "epoch": 106, + "eval_accuracy": 0.3928, + "eval_loss": 37.1106109375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 43.874717364501954 + }, + { + "epoch": 107, + "eval_accuracy": 0.4084, + "eval_loss": 33.58920625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 42.896230186631946 + }, + { + "epoch": 108, + "eval_accuracy": 0.3816, + "eval_loss": 35.4541578125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 40.49658525119358 + }, + { + "epoch": 109, + "eval_accuracy": 0.3904, + "eval_loss": 31.881768359375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 38.327581633843316 + }, + { + "epoch": 110, + "eval_accuracy": 0.3458, + "eval_loss": 38.830978125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 35.389524947781034 + }, + { + "epoch": 111, + "eval_accuracy": 0.3544, + "eval_loss": 36.43687421875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 34.1201142171224 + }, + { + "epoch": 112, + "eval_accuracy": 0.3744, + "eval_loss": 27.77460859375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 32.8625382297092 + }, + { + "epoch": 113, + "eval_accuracy": 0.3666, + "eval_loss": 34.01153984375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 32.72871079440647 + }, + { + "epoch": 114, + "eval_accuracy": 0.3624, + "eval_loss": 29.06900859375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 31.676572363620334 + }, + { + "epoch": 115, + "eval_accuracy": 0.2994, + "eval_loss": 38.16306328125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 30.014509815131294 + }, + { + "epoch": 116, + "eval_accuracy": 0.348, + "eval_loss": 27.01443515625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 30.27423641289605 + }, + { + "epoch": 117, + "eval_accuracy": 0.3002, + "eval_loss": 45.1408640625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 30.936224088541667 + }, + { + "epoch": 118, + "eval_accuracy": 0.3548, + "eval_loss": 34.59181484375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 30.146907942708335 + }, + { + "epoch": 119, + "eval_accuracy": 0.3376, + "eval_loss": 32.60268515625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 28.432915776231553 + }, + { + "epoch": 120, + "eval_accuracy": 0.3364, + "eval_loss": 34.4127140625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 29.565025434705944 + }, + { + "epoch": 121, + "eval_accuracy": 0.3672, + "eval_loss": 21.1856453125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 28.357878097195094 + }, + { + "epoch": 122, + "eval_accuracy": 0.3018, + "eval_loss": 31.5717625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 29.79608405388726 + }, + { + "epoch": 123, + "eval_accuracy": 0.3136, + "eval_loss": 28.0333640625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 32.39813755900065 + }, + { + "epoch": 124, + "eval_accuracy": 0.33, + "eval_loss": 25.2027203125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 28.785149905225964 + }, + { + "epoch": 125, + "eval_accuracy": 0.365, + "eval_loss": 50.01820859375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 30.066769323052302 + }, + { + "epoch": 126, + "eval_accuracy": 0.2992, + "eval_loss": 38.751096875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 28.825669424777562 + }, + { + "epoch": 127, + "eval_accuracy": 0.3188, + "eval_loss": 31.16794609375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 31.951569648573134 + }, + { + "epoch": 128, + "eval_accuracy": 0.3536, + "eval_loss": 26.708534375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 28.16607522176107 + }, + { + "epoch": 129, + "eval_accuracy": 0.3712, + "eval_loss": 22.56696171875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 30.68563767361111 + }, + { + "epoch": 130, + "eval_accuracy": 0.3722, + "eval_loss": 27.211113671875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 26.697583384874132 + }, + { + "epoch": 131, + "eval_accuracy": 0.3952, + "eval_loss": 15.98902734375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 28.169558378770617 + }, + { + "epoch": 132, + "eval_accuracy": 0.345, + "eval_loss": 27.111884375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 29.73483961995443 + }, + { + "epoch": 133, + "eval_accuracy": 0.3092, + "eval_loss": 33.95772265625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 35.36185829772949 + }, + { + "epoch": 134, + "eval_accuracy": 0.3584, + "eval_loss": 20.575744921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 30.647867134602865 + }, + { + "epoch": 135, + "eval_accuracy": 0.3062, + "eval_loss": 25.4248328125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 26.337036939154732 + }, + { + "epoch": 136, + "eval_accuracy": 0.4092, + "eval_loss": 15.70295703125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 33.879409608968096 + }, + { + "epoch": 137, + "eval_accuracy": 0.2672, + "eval_loss": 40.63083671875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 30.210656193712023 + }, + { + "epoch": 138, + "eval_accuracy": 0.2778, + "eval_loss": 34.45653125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 34.42727603759766 + }, + { + "epoch": 139, + "eval_accuracy": 0.3152, + "eval_loss": 41.16488515625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 30.716559368896483 + }, + { + "epoch": 140, + "eval_accuracy": 0.293, + "eval_loss": 29.28139296875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 30.197615693155925 + }, + { + "epoch": 141, + "eval_accuracy": 0.2916, + "eval_loss": 32.38239140625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 28.30102506205241 + }, + { + "epoch": 142, + "eval_accuracy": 0.2882, + "eval_loss": 31.02913203125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 33.08134716661241 + }, + { + "epoch": 143, + "eval_accuracy": 0.3546, + "eval_loss": 26.13366640625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 35.202170224338104 + }, + { + "epoch": 144, + "eval_accuracy": 0.4018, + "eval_loss": 18.689922265625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 28.16269857449002 + }, + { + "epoch": 145, + "eval_accuracy": 0.2002, + "eval_loss": 45.4462328125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 29.617123545328777 + }, + { + "epoch": 146, + "eval_accuracy": 0.3582, + "eval_loss": 24.901905078125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 34.85614291212294 + }, + { + "epoch": 147, + "eval_accuracy": 0.3712, + "eval_loss": 39.63477265625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 32.07841902872721 + }, + { + "epoch": 148, + "eval_accuracy": 0.3844, + "eval_loss": 23.235046875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 34.80584049445258 + }, + { + "epoch": 149, + "eval_accuracy": 0.3602, + "eval_loss": 31.175896484375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 32.27111835700141 + }, + { + "epoch": 150, + "eval_accuracy": 0.2914, + "eval_loss": 28.874794921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 31.977565850151908 + }, + { + "epoch": 151, + "eval_accuracy": 0.4476, + "eval_loss": 10.8708392578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 11.651757261149088 + }, + { + "epoch": 152, + "eval_accuracy": 0.446, + "eval_loss": 9.7358521484375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 9.99678972998725 + }, + { + "epoch": 153, + "eval_accuracy": 0.4404, + "eval_loss": 8.9593001953125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 9.319264670477972 + }, + { + "epoch": 154, + "eval_accuracy": 0.4428, + "eval_loss": 8.37555732421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 8.697566080050999 + }, + { + "epoch": 155, + "eval_accuracy": 0.4502, + "eval_loss": 7.80713427734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 8.319801394992405 + }, + { + "epoch": 156, + "eval_accuracy": 0.4344, + "eval_loss": 7.308902734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 7.8531677338494195 + }, + { + "epoch": 157, + "eval_accuracy": 0.4406, + "eval_loss": 6.8099802734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 7.516104418097602 + }, + { + "epoch": 158, + "eval_accuracy": 0.4278, + "eval_loss": 6.6249916015625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 7.2278142467922635 + }, + { + "epoch": 159, + "eval_accuracy": 0.4346, + "eval_loss": 6.33050771484375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 6.960628581576877 + }, + { + "epoch": 160, + "eval_accuracy": 0.4234, + "eval_loss": 6.38564306640625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 6.7058130749172635 + }, + { + "epoch": 161, + "eval_accuracy": 0.4256, + "eval_loss": 5.79991982421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 6.504179000430637 + }, + { + "epoch": 162, + "eval_accuracy": 0.408, + "eval_loss": 5.9061658203125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 6.2047558926052515 + }, + { + "epoch": 163, + "eval_accuracy": 0.4004, + "eval_loss": 6.0522150390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 6.164733709801568 + }, + { + "epoch": 164, + "eval_accuracy": 0.4006, + "eval_loss": 5.61034873046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 5.856545046403673 + }, + { + "epoch": 165, + "eval_accuracy": 0.3968, + "eval_loss": 5.4717162109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 5.771546636369493 + }, + { + "epoch": 166, + "eval_accuracy": 0.3862, + "eval_loss": 5.516412109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 5.6266596131218805 + }, + { + "epoch": 167, + "eval_accuracy": 0.4108, + "eval_loss": 4.80401162109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 5.476546478610569 + }, + { + "epoch": 168, + "eval_accuracy": 0.3934, + "eval_loss": 5.0674400390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 5.386792821163601 + }, + { + "epoch": 169, + "eval_accuracy": 0.3716, + "eval_loss": 5.19190400390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 5.217026385074192 + }, + { + "epoch": 170, + "eval_accuracy": 0.4034, + "eval_loss": 4.618946875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 5.097870582156711 + }, + { + "epoch": 171, + "eval_accuracy": 0.3978, + "eval_loss": 4.6441703125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 4.992071345859104 + }, + { + "epoch": 172, + "eval_accuracy": 0.415, + "eval_loss": 4.309574365234375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 5.006023575761583 + }, + { + "epoch": 173, + "eval_accuracy": 0.3728, + "eval_loss": 4.7692791015625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 4.765749631924099 + }, + { + "epoch": 174, + "eval_accuracy": 0.3972, + "eval_loss": 4.273984814453125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 4.669230929056804 + }, + { + "epoch": 175, + "eval_accuracy": 0.3868, + "eval_loss": 4.346198095703125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 4.767588217163086 + }, + { + "epoch": 176, + "eval_accuracy": 0.4138, + "eval_loss": 4.04446513671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 4.574014286422729 + }, + { + "epoch": 177, + "eval_accuracy": 0.3986, + "eval_loss": 4.01624287109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 4.483139426252577 + }, + { + "epoch": 178, + "eval_accuracy": 0.379, + "eval_loss": 4.116311572265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 4.440480245547825 + }, + { + "epoch": 179, + "eval_accuracy": 0.411, + "eval_loss": 3.7937888671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 4.464287152438693 + }, + { + "epoch": 180, + "eval_accuracy": 0.3336, + "eval_loss": 4.47425, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 4.374478776211209 + }, + { + "epoch": 181, + "eval_accuracy": 0.3926, + "eval_loss": 3.835095654296875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 4.33659748111301 + }, + { + "epoch": 182, + "eval_accuracy": 0.3928, + "eval_loss": 3.6995392578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 4.252237670898437 + }, + { + "epoch": 183, + "eval_accuracy": 0.3914, + "eval_loss": 3.63160205078125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 4.279515455034044 + }, + { + "epoch": 184, + "eval_accuracy": 0.3968, + "eval_loss": 3.59172255859375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 4.1540915985107425 + }, + { + "epoch": 185, + "eval_accuracy": 0.3818, + "eval_loss": 3.789047998046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 4.0526812603420685 + }, + { + "epoch": 186, + "eval_accuracy": 0.361, + "eval_loss": 4.066609326171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 4.036473704528809 + }, + { + "epoch": 187, + "eval_accuracy": 0.3994, + "eval_loss": 3.52081162109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 4.109895914713542 + }, + { + "epoch": 188, + "eval_accuracy": 0.3908, + "eval_loss": 3.533601025390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 4.0574144656711155 + }, + { + "epoch": 189, + "eval_accuracy": 0.3962, + "eval_loss": 3.42502001953125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 4.032029030354818 + }, + { + "epoch": 190, + "eval_accuracy": 0.4156, + "eval_loss": 3.239801220703125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 3.874290327750312 + }, + { + "epoch": 191, + "eval_accuracy": 0.3698, + "eval_loss": 3.478581982421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 3.7694188574896916 + }, + { + "epoch": 192, + "eval_accuracy": 0.3578, + "eval_loss": 3.897442578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 3.7616168645222983 + }, + { + "epoch": 193, + "eval_accuracy": 0.3774, + "eval_loss": 3.65839599609375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 3.8300268634372285 + }, + { + "epoch": 194, + "eval_accuracy": 0.3948, + "eval_loss": 3.29561572265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 3.8239806078592937 + }, + { + "epoch": 195, + "eval_accuracy": 0.3592, + "eval_loss": 3.819576904296875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 3.7138434155782063 + }, + { + "epoch": 196, + "eval_accuracy": 0.4016, + "eval_loss": 3.35995888671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 3.6911021950191922 + }, + { + "epoch": 197, + "eval_accuracy": 0.3778, + "eval_loss": 3.279795556640625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 3.6857616521199543 + }, + { + "epoch": 198, + "eval_accuracy": 0.3356, + "eval_loss": 4.24252001953125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 3.709500290086534 + }, + { + "epoch": 199, + "eval_accuracy": 0.3088, + "eval_loss": 3.90576396484375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 3.7073160080379908 + }, + { + "epoch": 200, + "eval_accuracy": 0.3492, + "eval_loss": 3.6838392578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 3.5896300240410697 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.3492, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 3.6838392578125, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2018397696, + "peak_memory_reserved_bytes": 2480930816, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 21.157899856567383, + "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": 1183.4196124076843, + "train_wall_s": 1162.2410652637482 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 16957440000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 364959360000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 364959360000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 746876160000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json.manifest.json new file mode 100644 index 0000000..1882f01 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json.manifest.json @@ -0,0 +1,132 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "dfa", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.01", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--vectorizer_mode", + "spatial_template" + ], + "completed_unix_time": 1787035384.2526608, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 1187.1560876369476, + "epochs": 200, + "experiment_name": "resnet-r2-dfa-d20-lr0p01", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "dfa", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d20-lr0p01.json", + "output_exists": true, + "output_sha256": "638ace440ba8224c18e4e0f09f2dbac291ec32e4f54bc9efadbbff58ddf2ac45", + "rate": 0.01, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json b/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json new file mode 100644 index 0000000..3b96f18 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json @@ -0,0 +1,2700 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "lr": 0.01, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.00031910948309814556, + "innovation_negative_gradient_cosine": [ + 0.0026197603438049555, + 0.0006819418631494045, + 0.00017868289432954043, + 0.0009858061093837023, + -0.001170104369521141, + 0.0005506974994204938, + 0.00034609687281772494, + 5.504034925252199e-05, + -0.0002867505536414683, + -0.0007700761780142784, + 0.0017122221179306507, + 0.004392784088850021, + 0.0056973351165652275, + -0.00015289167640730739, + 0.0028711319901049137, + 0.003501438070088625, + 0.005120588466525078, + 0.003510239999741316, + 0.00014667602954432368, + 0.0016334315296262503, + 0.0033825868740677834, + 0.011951932683587074, + 0.016150185838341713, + 0.017887040972709656, + 0.010713891126215458, + 0.017346883192658424, + 0.025565773248672485, + 0.023414287716150284, + 0.025312259793281555, + 0.020356435328722, + 0.04579668864607811 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0026197603438049555, + 0.0006819418631494045, + 0.00017868289432954043, + 0.0009858061093837023, + -0.001170104369521141, + 0.0005506974994204938, + 0.00034609687281772494, + 5.504034925252199e-05, + -0.0002867505536414683, + -0.0007700761780142784, + 0.0017122221179306507, + 0.004392784088850021, + 0.0056973351165652275, + -0.00015289167640730739, + 0.0028711319901049137, + 0.003501438070088625, + 0.005120588466525078, + 0.003510239999741316, + 0.00014667602954432368, + 0.0016334315296262503, + 0.0033825868740677834, + 0.011951932683587074, + 0.016150185838341713, + 0.017887040972709656, + 0.010713891126215458, + 0.017346883192658424, + 0.025565773248672485, + 0.023414287716150284, + 0.025312259793281555, + 0.020356435328722, + 0.04579668864607811 + ], + "teaching_negative_gradient_cosine": [ + 0.0026197603438049555, + 0.0006819418631494045, + 0.00017868289432954043, + 0.0009858061093837023, + -0.001170104369521141, + 0.0005506974994204938, + 0.00034609687281772494, + 5.504034925252199e-05, + -0.0002867505536414683, + -0.0007700761780142784, + 0.0017122221179306507, + 0.004392784088850021, + 0.0056973351165652275, + -0.00015289167640730739, + 0.0028711319901049137, + 0.003501438070088625, + 0.005120588466525078, + 0.003510239999741316, + 0.00014667602954432368, + 0.0016334315296262503, + 0.0033825868740677834, + 0.011951932683587074, + 0.016150185838341713, + 0.017887040972709656, + 0.010713891126215458, + 0.017346883192658424, + 0.025565773248672485, + 0.023414287716150284, + 0.025312259793281555, + 0.020356435328722, + 0.04579668864607811 + ], + "wall_s": 0.1506354808807373 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.1968, + "eval_loss": 2.13564140625, + "eval_split": "validation", + "lr": 0.002, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.2802412104288736 + }, + { + "epoch": 2, + "eval_accuracy": 0.1762, + "eval_loss": 2.403104150390625, + "eval_split": "validation", + "lr": 0.004, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 2.3138541755676267 + }, + { + "epoch": 3, + "eval_accuracy": 0.166, + "eval_loss": 2.668672314453125, + "eval_split": "validation", + "lr": 0.006, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 3.1551799483829073 + }, + { + "epoch": 4, + "eval_accuracy": 0.1524, + "eval_loss": 3.85100947265625, + "eval_split": "validation", + "lr": 0.008, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 6.567851277033488 + }, + { + "epoch": 5, + "eval_accuracy": 0.1702, + "eval_loss": 9.0174177734375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 8.539597244093153 + }, + { + "epoch": 6, + "eval_accuracy": 0.2132, + "eval_loss": 6.2054705078125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 8.005350156317817 + }, + { + "epoch": 7, + "eval_accuracy": 0.2122, + "eval_loss": 6.59003876953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 8.283805978224013 + }, + { + "epoch": 8, + "eval_accuracy": 0.1796, + "eval_loss": 10.077484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 8.76988885184394 + }, + { + "epoch": 9, + "eval_accuracy": 0.194, + "eval_loss": 12.0967720703125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 9.657652395799424 + }, + { + "epoch": 10, + "eval_accuracy": 0.244, + "eval_loss": 5.8700173828125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 10.630738082038032 + }, + { + "epoch": 11, + "eval_accuracy": 0.229, + "eval_loss": 9.34055859375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 11.782854916551377 + }, + { + "epoch": 12, + "eval_accuracy": 0.257, + "eval_loss": 13.1749107421875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 12.472624915907119 + }, + { + "epoch": 13, + "eval_accuracy": 0.2512, + "eval_loss": 15.2414419921875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 16.400539314439563 + }, + { + "epoch": 14, + "eval_accuracy": 0.2382, + "eval_loss": 13.90708046875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 15.362716340637206 + }, + { + "epoch": 15, + "eval_accuracy": 0.1612, + "eval_loss": 20.7498, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 15.200875530666774 + }, + { + "epoch": 16, + "eval_accuracy": 0.1846, + "eval_loss": 27.910105078125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 18.039595056830514 + }, + { + "epoch": 17, + "eval_accuracy": 0.268, + "eval_loss": 27.628282421875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 19.268495307752822 + }, + { + "epoch": 18, + "eval_accuracy": 0.2842, + "eval_loss": 22.1950453125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 19.945682086859808 + }, + { + "epoch": 19, + "eval_accuracy": 0.2712, + "eval_loss": 20.5347578125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 22.335486112467446 + }, + { + "epoch": 20, + "eval_accuracy": 0.2604, + "eval_loss": 22.29314609375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 25.390615134006076 + }, + { + "epoch": 21, + "eval_accuracy": 0.1656, + "eval_loss": 40.85045546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 24.838533505588106 + }, + { + "epoch": 22, + "eval_accuracy": 0.2032, + "eval_loss": 59.94360859375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 30.458272143554687 + }, + { + "epoch": 23, + "eval_accuracy": 0.3094, + "eval_loss": 24.10346015625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 31.5929201107449 + }, + { + "epoch": 24, + "eval_accuracy": 0.2626, + "eval_loss": 27.9797484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 35.06654516364203 + }, + { + "epoch": 25, + "eval_accuracy": 0.2232, + "eval_loss": 40.38759375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 34.452620131429036 + }, + { + "epoch": 26, + "eval_accuracy": 0.2136, + "eval_loss": 38.92077890625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 39.40891842041015 + }, + { + "epoch": 27, + "eval_accuracy": 0.2128, + "eval_loss": 36.42456640625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 42.401445976426864 + }, + { + "epoch": 28, + "eval_accuracy": 0.3382, + "eval_loss": 41.78907578125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 48.708959084743924 + }, + { + "epoch": 29, + "eval_accuracy": 0.3258, + "eval_loss": 47.034034375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 47.791174917263454 + }, + { + "epoch": 30, + "eval_accuracy": 0.2702, + "eval_loss": 50.88515546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 59.97448439670139 + }, + { + "epoch": 31, + "eval_accuracy": 0.2674, + "eval_loss": 62.69679375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 55.62923475884332 + }, + { + "epoch": 32, + "eval_accuracy": 0.2688, + "eval_loss": 91.7570109375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 56.54406637641059 + }, + { + "epoch": 33, + "eval_accuracy": 0.25, + "eval_loss": 53.91616796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 56.0292227613661 + }, + { + "epoch": 34, + "eval_accuracy": 0.2872, + "eval_loss": 108.42245, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 65.2930051079644 + }, + { + "epoch": 35, + "eval_accuracy": 0.3082, + "eval_loss": 52.419546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 62.557755900065104 + }, + { + "epoch": 36, + "eval_accuracy": 0.21, + "eval_loss": 248.62119375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 71.82943644748264 + }, + { + "epoch": 37, + "eval_accuracy": 0.3456, + "eval_loss": 104.5305140625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 80.15177752956814 + }, + { + "epoch": 38, + "eval_accuracy": 0.2798, + "eval_loss": 63.1959390625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 74.20427698228625 + }, + { + "epoch": 39, + "eval_accuracy": 0.2526, + "eval_loss": 67.9075953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 86.0083361219618 + }, + { + "epoch": 40, + "eval_accuracy": 0.1372, + "eval_loss": 213.52529375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 90.54625876193576 + }, + { + "epoch": 41, + "eval_accuracy": 0.291, + "eval_loss": 85.5634140625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 97.03032708468967 + }, + { + "epoch": 42, + "eval_accuracy": 0.287, + "eval_loss": 132.8560890625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 102.40799818793403 + }, + { + "epoch": 43, + "eval_accuracy": 0.226, + "eval_loss": 219.480903125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 110.62748044162326 + }, + { + "epoch": 44, + "eval_accuracy": 0.3476, + "eval_loss": 139.70363125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 121.17438940836588 + }, + { + "epoch": 45, + "eval_accuracy": 0.3148, + "eval_loss": 115.74535625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 117.96440033772787 + }, + { + "epoch": 46, + "eval_accuracy": 0.2498, + "eval_loss": 161.649003125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 130.1123986111111 + }, + { + "epoch": 47, + "eval_accuracy": 0.2454, + "eval_loss": 309.7527625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 132.362695046658 + }, + { + "epoch": 48, + "eval_accuracy": 0.2128, + "eval_loss": 173.7661875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 141.5597231499566 + }, + { + "epoch": 49, + "eval_accuracy": 0.323, + "eval_loss": 87.1395671875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 139.6630462632921 + }, + { + "epoch": 50, + "eval_accuracy": 0.2816, + "eval_loss": 232.42523125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 163.6217903415256 + }, + { + "epoch": 51, + "eval_accuracy": 0.1946, + "eval_loss": 188.797478125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 164.3333604844835 + }, + { + "epoch": 52, + "eval_accuracy": 0.2566, + "eval_loss": 239.084840625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 178.35102162000868 + }, + { + "epoch": 53, + "eval_accuracy": 0.3502, + "eval_loss": 123.9788515625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 175.44256982150608 + }, + { + "epoch": 54, + "eval_accuracy": 0.3146, + "eval_loss": 206.992815625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 184.28907900390624 + }, + { + "epoch": 55, + "eval_accuracy": 0.3146, + "eval_loss": 163.399478125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 207.67702931043837 + }, + { + "epoch": 56, + "eval_accuracy": 0.2962, + "eval_loss": 223.243553125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 205.11963244628907 + }, + { + "epoch": 57, + "eval_accuracy": 0.3784, + "eval_loss": 137.992953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 214.7449376844618 + }, + { + "epoch": 58, + "eval_accuracy": 0.2166, + "eval_loss": 476.17575, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 239.8989421766493 + }, + { + "epoch": 59, + "eval_accuracy": 0.2722, + "eval_loss": 290.41935625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 239.53743671875 + }, + { + "epoch": 60, + "eval_accuracy": 0.316, + "eval_loss": 164.1238, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 245.6346437608507 + }, + { + "epoch": 61, + "eval_accuracy": 0.1988, + "eval_loss": 365.0298, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 251.1119849609375 + }, + { + "epoch": 62, + "eval_accuracy": 0.2056, + "eval_loss": 389.20041875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 266.53731458333334 + }, + { + "epoch": 63, + "eval_accuracy": 0.3122, + "eval_loss": 268.056540625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 275.1557644124349 + }, + { + "epoch": 64, + "eval_accuracy": 0.22, + "eval_loss": 282.2733, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 276.21525709092884 + }, + { + "epoch": 65, + "eval_accuracy": 0.3348, + "eval_loss": 217.802796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 293.0577929416233 + }, + { + "epoch": 66, + "eval_accuracy": 0.2684, + "eval_loss": 286.4739, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 333.1723864176432 + }, + { + "epoch": 67, + "eval_accuracy": 0.3034, + "eval_loss": 199.15364375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 319.33838850911457 + }, + { + "epoch": 68, + "eval_accuracy": 0.2108, + "eval_loss": 308.4932, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 284.5920985188802 + }, + { + "epoch": 69, + "eval_accuracy": 0.2392, + "eval_loss": 458.07015, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 337.78335944281685 + }, + { + "epoch": 70, + "eval_accuracy": 0.2808, + "eval_loss": 321.44930625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 315.11677765570744 + }, + { + "epoch": 71, + "eval_accuracy": 0.225, + "eval_loss": 421.11063125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 358.4162332628038 + }, + { + "epoch": 72, + "eval_accuracy": 0.325, + "eval_loss": 251.6319625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 358.8881578938802 + }, + { + "epoch": 73, + "eval_accuracy": 0.3256, + "eval_loss": 257.94384375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 403.11145792100695 + }, + { + "epoch": 74, + "eval_accuracy": 0.312, + "eval_loss": 222.5570625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 384.9969331814236 + }, + { + "epoch": 75, + "eval_accuracy": 0.2346, + "eval_loss": 340.03154375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 401.10929453396267 + }, + { + "epoch": 76, + "eval_accuracy": 0.2106, + "eval_loss": 1018.6606125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 399.87897292751734 + }, + { + "epoch": 77, + "eval_accuracy": 0.3102, + "eval_loss": 404.8272375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 429.4044539496528 + }, + { + "epoch": 78, + "eval_accuracy": 0.2868, + "eval_loss": 436.610575, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 457.54408018663196 + }, + { + "epoch": 79, + "eval_accuracy": 0.2612, + "eval_loss": 579.3421, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 483.91433331163194 + }, + { + "epoch": 80, + "eval_accuracy": 0.2456, + "eval_loss": 575.697925, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 465.56115231119793 + }, + { + "epoch": 81, + "eval_accuracy": 0.22, + "eval_loss": 1274.014525, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 491.67860532769095 + }, + { + "epoch": 82, + "eval_accuracy": 0.181, + "eval_loss": 725.602575, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 509.37582771809895 + }, + { + "epoch": 83, + "eval_accuracy": 0.2862, + "eval_loss": 433.20714375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 523.651078266059 + }, + { + "epoch": 84, + "eval_accuracy": 0.306, + "eval_loss": 331.74839375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 505.1366246799045 + }, + { + "epoch": 85, + "eval_accuracy": 0.2364, + "eval_loss": 566.8942875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 579.8360815646702 + }, + { + "epoch": 86, + "eval_accuracy": 0.2528, + "eval_loss": 943.22735, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 574.1495292317709 + }, + { + "epoch": 87, + "eval_accuracy": 0.276, + "eval_loss": 709.0178, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 538.2599395833333 + }, + { + "epoch": 88, + "eval_accuracy": 0.2866, + "eval_loss": 380.433375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 624.6833575466579 + }, + { + "epoch": 89, + "eval_accuracy": 0.2842, + "eval_loss": 482.03155625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 608.3776178927951 + }, + { + "epoch": 90, + "eval_accuracy": 0.3342, + "eval_loss": 464.89378125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 647.8050713107639 + }, + { + "epoch": 91, + "eval_accuracy": 0.3126, + "eval_loss": 836.0266375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 613.0687445529514 + }, + { + "epoch": 92, + "eval_accuracy": 0.3, + "eval_loss": 486.60365, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 650.7298907063802 + }, + { + "epoch": 93, + "eval_accuracy": 0.298, + "eval_loss": 469.757075, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 620.019738937717 + }, + { + "epoch": 94, + "eval_accuracy": 0.2938, + "eval_loss": 415.58044375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 746.7210650390625 + }, + { + "epoch": 95, + "eval_accuracy": 0.3562, + "eval_loss": 411.128125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 742.6021716959635 + }, + { + "epoch": 96, + "eval_accuracy": 0.2962, + "eval_loss": 554.3056, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 746.4939790690104 + }, + { + "epoch": 97, + "eval_accuracy": 0.2416, + "eval_loss": 825.879775, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 697.7495639756944 + }, + { + "epoch": 98, + "eval_accuracy": 0.3696, + "eval_loss": 489.983625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 766.2687794487847 + }, + { + "epoch": 99, + "eval_accuracy": 0.249, + "eval_loss": 866.526825, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 707.4956386827257 + }, + { + "epoch": 100, + "eval_accuracy": 0.2272, + "eval_loss": 736.6153625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 860.6614454047309 + }, + { + "epoch": 101, + "eval_accuracy": 0.3856, + "eval_loss": 170.472615625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 196.49338107910157 + }, + { + "epoch": 102, + "eval_accuracy": 0.3778, + "eval_loss": 125.7411875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 141.02132633734809 + }, + { + "epoch": 103, + "eval_accuracy": 0.356, + "eval_loss": 110.767259375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 115.80608062608506 + }, + { + "epoch": 104, + "eval_accuracy": 0.343, + "eval_loss": 106.7759078125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 103.91751080322265 + }, + { + "epoch": 105, + "eval_accuracy": 0.3468, + "eval_loss": 96.4980328125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 96.93153265652127 + }, + { + "epoch": 106, + "eval_accuracy": 0.3528, + "eval_loss": 81.44060625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 92.24946930202908 + }, + { + "epoch": 107, + "eval_accuracy": 0.353, + "eval_loss": 77.130496875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 87.91589212239583 + }, + { + "epoch": 108, + "eval_accuracy": 0.3476, + "eval_loss": 75.27756875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 82.24052918836806 + }, + { + "epoch": 109, + "eval_accuracy": 0.3438, + "eval_loss": 74.3921109375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 80.62625283610026 + }, + { + "epoch": 110, + "eval_accuracy": 0.3346, + "eval_loss": 82.432540625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 74.16640807834202 + }, + { + "epoch": 111, + "eval_accuracy": 0.3414, + "eval_loss": 68.23063359375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 72.85556735432942 + }, + { + "epoch": 112, + "eval_accuracy": 0.3604, + "eval_loss": 60.68131328125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 71.44236545952691 + }, + { + "epoch": 113, + "eval_accuracy": 0.3438, + "eval_loss": 63.771634375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 66.29341225043403 + }, + { + "epoch": 114, + "eval_accuracy": 0.359, + "eval_loss": 59.54249375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 68.79256528998481 + }, + { + "epoch": 115, + "eval_accuracy": 0.3114, + "eval_loss": 64.330353125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 64.35555018920898 + }, + { + "epoch": 116, + "eval_accuracy": 0.2398, + "eval_loss": 98.6124921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 63.517505746799046 + }, + { + "epoch": 117, + "eval_accuracy": 0.3264, + "eval_loss": 54.6644921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 64.07564553019206 + }, + { + "epoch": 118, + "eval_accuracy": 0.3136, + "eval_loss": 73.9800765625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 63.32237302517361 + }, + { + "epoch": 119, + "eval_accuracy": 0.3248, + "eval_loss": 56.1960296875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 62.23008333536784 + }, + { + "epoch": 120, + "eval_accuracy": 0.3352, + "eval_loss": 76.2859921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 64.19782397460938 + }, + { + "epoch": 121, + "eval_accuracy": 0.325, + "eval_loss": 52.52334140625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 62.10008289048937 + }, + { + "epoch": 122, + "eval_accuracy": 0.3006, + "eval_loss": 64.4516078125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 69.5385602355957 + }, + { + "epoch": 123, + "eval_accuracy": 0.3222, + "eval_loss": 59.18459375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 68.50677893473308 + }, + { + "epoch": 124, + "eval_accuracy": 0.3454, + "eval_loss": 66.70076875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 64.05500425075955 + }, + { + "epoch": 125, + "eval_accuracy": 0.299, + "eval_loss": 96.0758625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 67.54195568440755 + }, + { + "epoch": 126, + "eval_accuracy": 0.3506, + "eval_loss": 50.27091875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 63.50760168999566 + }, + { + "epoch": 127, + "eval_accuracy": 0.3584, + "eval_loss": 42.83064453125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 61.460221661376956 + }, + { + "epoch": 128, + "eval_accuracy": 0.3492, + "eval_loss": 46.741225, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 68.67775972290039 + }, + { + "epoch": 129, + "eval_accuracy": 0.3366, + "eval_loss": 55.25401015625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 65.46874283650716 + }, + { + "epoch": 130, + "eval_accuracy": 0.3758, + "eval_loss": 35.30758203125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 60.136354113091365 + }, + { + "epoch": 131, + "eval_accuracy": 0.381, + "eval_loss": 46.256021875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 71.47179066297743 + }, + { + "epoch": 132, + "eval_accuracy": 0.3672, + "eval_loss": 39.7463421875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 62.817046925184464 + }, + { + "epoch": 133, + "eval_accuracy": 0.3436, + "eval_loss": 65.4880921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 70.35553123440212 + }, + { + "epoch": 134, + "eval_accuracy": 0.3516, + "eval_loss": 61.6984265625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 64.2483411376953 + }, + { + "epoch": 135, + "eval_accuracy": 0.3028, + "eval_loss": 63.19440390625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 66.96016059977214 + }, + { + "epoch": 136, + "eval_accuracy": 0.3524, + "eval_loss": 51.108896875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 62.91782479654948 + }, + { + "epoch": 137, + "eval_accuracy": 0.216, + "eval_loss": 78.7760421875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 70.46476604546442 + }, + { + "epoch": 138, + "eval_accuracy": 0.368, + "eval_loss": 79.4366125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 72.66207418416342 + }, + { + "epoch": 139, + "eval_accuracy": 0.3388, + "eval_loss": 49.279284375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 61.06694605577257 + }, + { + "epoch": 140, + "eval_accuracy": 0.4082, + "eval_loss": 31.25261875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 59.71975162150065 + }, + { + "epoch": 141, + "eval_accuracy": 0.3726, + "eval_loss": 47.8163640625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 69.24733488498264 + }, + { + "epoch": 142, + "eval_accuracy": 0.2152, + "eval_loss": 96.3982875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 61.999896885850696 + }, + { + "epoch": 143, + "eval_accuracy": 0.418, + "eval_loss": 33.0312359375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 75.65242304958767 + }, + { + "epoch": 144, + "eval_accuracy": 0.356, + "eval_loss": 51.88513671875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 66.79744441663954 + }, + { + "epoch": 145, + "eval_accuracy": 0.3446, + "eval_loss": 57.6310875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 68.28875555216472 + }, + { + "epoch": 146, + "eval_accuracy": 0.3326, + "eval_loss": 53.0010203125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 68.38004661187065 + }, + { + "epoch": 147, + "eval_accuracy": 0.4004, + "eval_loss": 39.793740625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 72.77143829074436 + }, + { + "epoch": 148, + "eval_accuracy": 0.3338, + "eval_loss": 62.11200625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 63.826129070366754 + }, + { + "epoch": 149, + "eval_accuracy": 0.2826, + "eval_loss": 98.2671859375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 71.38788469509548 + }, + { + "epoch": 150, + "eval_accuracy": 0.3802, + "eval_loss": 62.30995546875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 69.05508831041124 + }, + { + "epoch": 151, + "eval_accuracy": 0.4546, + "eval_loss": 23.298765625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 25.48181363050673 + }, + { + "epoch": 152, + "eval_accuracy": 0.4396, + "eval_loss": 21.12952734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 22.125638820054796 + }, + { + "epoch": 153, + "eval_accuracy": 0.4406, + "eval_loss": 19.194971875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 20.11697272983127 + }, + { + "epoch": 154, + "eval_accuracy": 0.4418, + "eval_loss": 17.68252578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 18.992026596917047 + }, + { + "epoch": 155, + "eval_accuracy": 0.438, + "eval_loss": 16.4560037109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 17.603953161621092 + }, + { + "epoch": 156, + "eval_accuracy": 0.427, + "eval_loss": 15.984819140625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 16.884023179796007 + }, + { + "epoch": 157, + "eval_accuracy": 0.427, + "eval_loss": 14.9954673828125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 16.29205541856554 + }, + { + "epoch": 158, + "eval_accuracy": 0.4164, + "eval_loss": 14.2277224609375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 15.487207347954644 + }, + { + "epoch": 159, + "eval_accuracy": 0.4124, + "eval_loss": 14.00275390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 15.221728933207194 + }, + { + "epoch": 160, + "eval_accuracy": 0.3914, + "eval_loss": 14.5430076171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 14.397079717508952 + }, + { + "epoch": 161, + "eval_accuracy": 0.4068, + "eval_loss": 13.4264064453125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 14.152302168443468 + }, + { + "epoch": 162, + "eval_accuracy": 0.403, + "eval_loss": 12.5586671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 13.672727896457248 + }, + { + "epoch": 163, + "eval_accuracy": 0.391, + "eval_loss": 12.6078431640625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 13.379090177917481 + }, + { + "epoch": 164, + "eval_accuracy": 0.396, + "eval_loss": 11.770783203125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 12.801284927707249 + }, + { + "epoch": 165, + "eval_accuracy": 0.3832, + "eval_loss": 11.81651875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 12.41921732228597 + }, + { + "epoch": 166, + "eval_accuracy": 0.3774, + "eval_loss": 11.469639453125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 12.336118894958496 + }, + { + "epoch": 167, + "eval_accuracy": 0.403, + "eval_loss": 10.486908984375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 11.962089084540473 + }, + { + "epoch": 168, + "eval_accuracy": 0.3634, + "eval_loss": 11.5825072265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 11.826469014824761 + }, + { + "epoch": 169, + "eval_accuracy": 0.3782, + "eval_loss": 11.0532982421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 11.664914200168186 + }, + { + "epoch": 170, + "eval_accuracy": 0.3866, + "eval_loss": 10.220952734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 11.301445455593534 + }, + { + "epoch": 171, + "eval_accuracy": 0.363, + "eval_loss": 11.0618185546875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 11.134456986490886 + }, + { + "epoch": 172, + "eval_accuracy": 0.3632, + "eval_loss": 11.0839998046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 10.998900364685058 + }, + { + "epoch": 173, + "eval_accuracy": 0.3562, + "eval_loss": 10.4976865234375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 10.807568714396158 + }, + { + "epoch": 174, + "eval_accuracy": 0.3702, + "eval_loss": 11.3861921875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 10.480875332811143 + }, + { + "epoch": 175, + "eval_accuracy": 0.3412, + "eval_loss": 10.180069140625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 10.594809376186795 + }, + { + "epoch": 176, + "eval_accuracy": 0.3464, + "eval_loss": 9.7542810546875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 10.234108125474718 + }, + { + "epoch": 177, + "eval_accuracy": 0.3656, + "eval_loss": 9.5753017578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 10.208396461486817 + }, + { + "epoch": 178, + "eval_accuracy": 0.3656, + "eval_loss": 8.915300390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 10.092519539896648 + }, + { + "epoch": 179, + "eval_accuracy": 0.3652, + "eval_loss": 8.97712861328125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 10.079463104248047 + }, + { + "epoch": 180, + "eval_accuracy": 0.2898, + "eval_loss": 10.8206451171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 9.76496174299452 + }, + { + "epoch": 181, + "eval_accuracy": 0.3582, + "eval_loss": 9.1681966796875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 10.213150364345974 + }, + { + "epoch": 182, + "eval_accuracy": 0.365, + "eval_loss": 8.5061740234375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 9.497418525526259 + }, + { + "epoch": 183, + "eval_accuracy": 0.3328, + "eval_loss": 10.3511533203125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 9.881835655890571 + }, + { + "epoch": 184, + "eval_accuracy": 0.3122, + "eval_loss": 10.184031640625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 9.944425131225586 + }, + { + "epoch": 185, + "eval_accuracy": 0.384, + "eval_loss": 9.4912837890625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 9.53767821384006 + }, + { + "epoch": 186, + "eval_accuracy": 0.3214, + "eval_loss": 9.5392822265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 9.37409217291938 + }, + { + "epoch": 187, + "eval_accuracy": 0.3552, + "eval_loss": 8.4641732421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 9.708195000712077 + }, + { + "epoch": 188, + "eval_accuracy": 0.3766, + "eval_loss": 7.79647529296875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 9.330968306647407 + }, + { + "epoch": 189, + "eval_accuracy": 0.3472, + "eval_loss": 8.42413623046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 9.682495934889051 + }, + { + "epoch": 190, + "eval_accuracy": 0.3468, + "eval_loss": 8.903707421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 9.455760409037271 + }, + { + "epoch": 191, + "eval_accuracy": 0.3394, + "eval_loss": 9.9961833984375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 9.311902354431153 + }, + { + "epoch": 192, + "eval_accuracy": 0.3286, + "eval_loss": 8.53327626953125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 9.558651632351346 + }, + { + "epoch": 193, + "eval_accuracy": 0.3126, + "eval_loss": 9.6648263671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 9.197669339497883 + }, + { + "epoch": 194, + "eval_accuracy": 0.373, + "eval_loss": 7.50697080078125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 9.651352426656088 + }, + { + "epoch": 195, + "eval_accuracy": 0.3362, + "eval_loss": 9.22571796875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 8.665485123782688 + }, + { + "epoch": 196, + "eval_accuracy": 0.3336, + "eval_loss": 8.859921875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 8.966691216362847 + }, + { + "epoch": 197, + "eval_accuracy": 0.3446, + "eval_loss": 7.8034123046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 8.929122918023003 + }, + { + "epoch": 198, + "eval_accuracy": 0.2802, + "eval_loss": 12.0830345703125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 9.355501737297907 + }, + { + "epoch": 199, + "eval_accuracy": 0.3612, + "eval_loss": 7.60638193359375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 9.219078210618761 + }, + { + "epoch": 200, + "eval_accuracy": 0.3874, + "eval_loss": 6.95157197265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 8.747222507137723 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.3874, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 6.95157197265625, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2338062336, + "peak_memory_reserved_bytes": 3078619136, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 34.20064926147461, + "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": 1875.3221385478973, + "train_wall_s": 1841.099274635315 + }, + "work": { + "apical_macs_per_example": 3031040, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 27279360000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 619763328000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 619763328000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 1266806016000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json.manifest.json new file mode 100644 index 0000000..90116fd --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json.manifest.json @@ -0,0 +1,132 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "dfa", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.01", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--vectorizer_mode", + "spatial_template" + ], + "completed_unix_time": 1787030547.6867511, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 1879.2923183441162, + "epochs": 200, + "experiment_name": "resnet-r2-dfa-d32-lr0p01", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "dfa", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json", + "output_exists": true, + "output_sha256": "fdc45fa5f2e27326288de0b06db43af4ff8b0d46164473daf9f1b9a114dcb474", + "rate": 0.01, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json b/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json new file mode 100644 index 0000000..fa08611 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json @@ -0,0 +1,2892 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "lr": 0.01, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.00035024905325068784, + "innovation_negative_gradient_cosine": [ + -0.0008385714027099311, + 0.002124390099197626, + -0.0008659408194944263, + -0.00014567584730684757, + -0.0005021363031119108, + 0.001265335944481194, + 0.004038718529045582, + 0.0005503536085598171, + 0.0014425356639549136, + -0.0012388156028464437, + 0.0004629977629519999, + -0.0005257049342617393, + 0.0005172170931473374, + 0.0012648290721699595, + -0.0009433561353944242, + 0.00048206010251306, + -0.00037864106707274914, + -0.00040511280531063676, + -0.0019009984098374844, + 0.002641211496666074, + 0.000786549411714077, + -0.0003178365295752883, + 0.0014189683133736253, + 0.00026458967477083206, + 0.0013584874104708433, + 8.317664469359443e-05, + 0.0022656084038317204, + 0.0007815274875611067, + -0.0014294714201241732, + 0.003384759183973074, + 0.0010553388856351376, + 0.0004074107273481786, + -0.0002539380802772939, + 7.720990106463432e-05, + -0.0019574719481170177, + 0.0013086958788335323, + -0.001038939692080021, + 0.009168915450572968, + 0.007564560975879431, + 0.005770865827798843, + 0.008899712935090065, + 0.007965256460011005, + 0.01344345323741436, + 0.009990139864385128, + 0.012594806961715221, + 0.009420137852430344, + 0.007213872857391834, + 0.003913925029337406, + 0.006504267454147339, + 0.006452905014157295, + 0.017092103138566017, + 0.006244623102247715, + 0.013663386926054955, + 0.01786511018872261, + 0.02073792740702629 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0008385714027099311, + 0.002124390099197626, + -0.0008659408194944263, + -0.00014567584730684757, + -0.0005021363031119108, + 0.001265335944481194, + 0.004038718529045582, + 0.0005503536085598171, + 0.0014425356639549136, + -0.0012388156028464437, + 0.0004629977629519999, + -0.0005257049342617393, + 0.0005172170931473374, + 0.0012648290721699595, + -0.0009433561353944242, + 0.00048206010251306, + -0.00037864106707274914, + -0.00040511280531063676, + -0.0019009984098374844, + 0.002641211496666074, + 0.000786549411714077, + -0.0003178365295752883, + 0.0014189683133736253, + 0.00026458967477083206, + 0.0013584874104708433, + 8.317664469359443e-05, + 0.0022656084038317204, + 0.0007815274875611067, + -0.0014294714201241732, + 0.003384759183973074, + 0.0010553388856351376, + 0.0004074107273481786, + -0.0002539380802772939, + 7.720990106463432e-05, + -0.0019574719481170177, + 0.0013086958788335323, + -0.001038939692080021, + 0.009168915450572968, + 0.007564560975879431, + 0.005770865827798843, + 0.008899712935090065, + 0.007965256460011005, + 0.01344345323741436, + 0.009990139864385128, + 0.012594806961715221, + 0.009420137852430344, + 0.007213872857391834, + 0.003913925029337406, + 0.006504267454147339, + 0.006452905014157295, + 0.017092103138566017, + 0.006244623102247715, + 0.013663386926054955, + 0.01786511018872261, + 0.02073792740702629 + ], + "teaching_negative_gradient_cosine": [ + -0.0008385714027099311, + 0.002124390099197626, + -0.0008659408194944263, + -0.00014567584730684757, + -0.0005021363031119108, + 0.001265335944481194, + 0.004038718529045582, + 0.0005503536085598171, + 0.0014425356639549136, + -0.0012388156028464437, + 0.0004629977629519999, + -0.0005257049342617393, + 0.0005172170931473374, + 0.0012648290721699595, + -0.0009433561353944242, + 0.00048206010251306, + -0.00037864106707274914, + -0.00040511280531063676, + -0.0019009984098374844, + 0.002641211496666074, + 0.000786549411714077, + -0.0003178365295752883, + 0.0014189683133736253, + 0.00026458967477083206, + 0.0013584874104708433, + 8.317664469359443e-05, + 0.0022656084038317204, + 0.0007815274875611067, + -0.0014294714201241732, + 0.003384759183973074, + 0.0010553388856351376, + 0.0004074107273481786, + -0.0002539380802772939, + 7.720990106463432e-05, + -0.0019574719481170177, + 0.0013086958788335323, + -0.001038939692080021, + 0.009168915450572968, + 0.007564560975879431, + 0.005770865827798843, + 0.008899712935090065, + 0.007965256460011005, + 0.01344345323741436, + 0.009990139864385128, + 0.012594806961715221, + 0.009420137852430344, + 0.007213872857391834, + 0.003913925029337406, + 0.006504267454147339, + 0.006452905014157295, + 0.017092103138566017, + 0.006244623102247715, + 0.013663386926054955, + 0.01786511018872261, + 0.02073792740702629 + ], + "wall_s": 0.17446684837341309 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.1736, + "eval_loss": 2.3541751953125, + "eval_split": "validation", + "lr": 0.002, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.479923766793145 + }, + { + "epoch": 2, + "eval_accuracy": 0.171, + "eval_loss": 9.3748625, + "eval_split": "validation", + "lr": 0.004, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 9.503815193684895 + }, + { + "epoch": 3, + "eval_accuracy": 0.1162, + "eval_loss": 15.8507943359375, + "eval_split": "validation", + "lr": 0.006, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 17.57388380262587 + }, + { + "epoch": 4, + "eval_accuracy": 0.1476, + "eval_loss": 18.62470546875, + "eval_split": "validation", + "lr": 0.008, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 20.294757021755643 + }, + { + "epoch": 5, + "eval_accuracy": 0.2192, + "eval_loss": 17.0168251953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 22.494775368584527 + }, + { + "epoch": 6, + "eval_accuracy": 0.2456, + "eval_loss": 12.4064818359375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 18.441068319702147 + }, + { + "epoch": 7, + "eval_accuracy": 0.2018, + "eval_loss": 13.71370859375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 18.688175535922582 + }, + { + "epoch": 8, + "eval_accuracy": 0.1932, + "eval_loss": 24.330255078125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 20.849487725830077 + }, + { + "epoch": 9, + "eval_accuracy": 0.2196, + "eval_loss": 15.015166796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 20.860968447875976 + }, + { + "epoch": 10, + "eval_accuracy": 0.2212, + "eval_loss": 22.310587890625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 22.45043119913737 + }, + { + "epoch": 11, + "eval_accuracy": 0.1606, + "eval_loss": 24.616283203125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 21.248531058417427 + }, + { + "epoch": 12, + "eval_accuracy": 0.1408, + "eval_loss": 38.6989578125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 23.867157828097874 + }, + { + "epoch": 13, + "eval_accuracy": 0.1746, + "eval_loss": 33.787591796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 25.00604586452908 + }, + { + "epoch": 14, + "eval_accuracy": 0.1706, + "eval_loss": 35.00100546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 29.429117143419052 + }, + { + "epoch": 15, + "eval_accuracy": 0.1468, + "eval_loss": 49.8456734375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 35.39160467122396 + }, + { + "epoch": 16, + "eval_accuracy": 0.1636, + "eval_loss": 35.58575703125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 31.424257394070096 + }, + { + "epoch": 17, + "eval_accuracy": 0.2082, + "eval_loss": 38.31974375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 41.756677141655814 + }, + { + "epoch": 18, + "eval_accuracy": 0.1674, + "eval_loss": 55.60708984375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 41.471936461046006 + }, + { + "epoch": 19, + "eval_accuracy": 0.2752, + "eval_loss": 26.146371484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 47.22188578592936 + }, + { + "epoch": 20, + "eval_accuracy": 0.246, + "eval_loss": 49.61728359375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 48.96045647583008 + }, + { + "epoch": 21, + "eval_accuracy": 0.3108, + "eval_loss": 28.35482109375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 54.232426001315645 + }, + { + "epoch": 22, + "eval_accuracy": 0.193, + "eval_loss": 78.530896875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 56.057253458658856 + }, + { + "epoch": 23, + "eval_accuracy": 0.239, + "eval_loss": 56.023084375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 61.60340079481337 + }, + { + "epoch": 24, + "eval_accuracy": 0.2034, + "eval_loss": 140.8517625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 68.93976577148437 + }, + { + "epoch": 25, + "eval_accuracy": 0.2146, + "eval_loss": 98.26845, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 65.16236778971354 + }, + { + "epoch": 26, + "eval_accuracy": 0.1006, + "eval_loss": 183.918659375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 83.24809537217882 + }, + { + "epoch": 27, + "eval_accuracy": 0.2808, + "eval_loss": 129.1784984375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 84.530659375 + }, + { + "epoch": 28, + "eval_accuracy": 0.2766, + "eval_loss": 76.771484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 95.1158966044108 + }, + { + "epoch": 29, + "eval_accuracy": 0.2638, + "eval_loss": 93.3139046875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 95.88311271023221 + }, + { + "epoch": 30, + "eval_accuracy": 0.2536, + "eval_loss": 129.6387484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 109.07512716064453 + }, + { + "epoch": 31, + "eval_accuracy": 0.1254, + "eval_loss": 210.85293125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 115.62372081434462 + }, + { + "epoch": 32, + "eval_accuracy": 0.2206, + "eval_loss": 185.76780625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 133.46289936523436 + }, + { + "epoch": 33, + "eval_accuracy": 0.225, + "eval_loss": 185.7995, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 137.48025703667534 + }, + { + "epoch": 34, + "eval_accuracy": 0.1878, + "eval_loss": 260.351153125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 157.19087931043836 + }, + { + "epoch": 35, + "eval_accuracy": 0.1858, + "eval_loss": 387.71186875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 155.11019038628473 + }, + { + "epoch": 36, + "eval_accuracy": 0.1748, + "eval_loss": 393.25378125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 170.66294742838542 + }, + { + "epoch": 37, + "eval_accuracy": 0.3108, + "eval_loss": 177.023921875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 191.58247672254774 + }, + { + "epoch": 38, + "eval_accuracy": 0.2834, + "eval_loss": 254.260515625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 209.28699675835503 + }, + { + "epoch": 39, + "eval_accuracy": 0.2852, + "eval_loss": 224.945253125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 235.84405463324651 + }, + { + "epoch": 40, + "eval_accuracy": 0.1732, + "eval_loss": 513.14443125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 239.26524459092883 + }, + { + "epoch": 41, + "eval_accuracy": 0.231, + "eval_loss": 222.885778125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 254.86575007052951 + }, + { + "epoch": 42, + "eval_accuracy": 0.2524, + "eval_loss": 229.431778125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 261.95233366970484 + }, + { + "epoch": 43, + "eval_accuracy": 0.3022, + "eval_loss": 391.91845625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 324.3291342556424 + }, + { + "epoch": 44, + "eval_accuracy": 0.2278, + "eval_loss": 556.278825, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 312.7284970106337 + }, + { + "epoch": 45, + "eval_accuracy": 0.1716, + "eval_loss": 764.924875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 358.8140581542969 + }, + { + "epoch": 46, + "eval_accuracy": 0.1858, + "eval_loss": 484.53405, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 334.96564144422746 + }, + { + "epoch": 47, + "eval_accuracy": 0.1762, + "eval_loss": 656.2453375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 390.17689079318575 + }, + { + "epoch": 48, + "eval_accuracy": 0.2056, + "eval_loss": 441.7227625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 406.4874663411458 + }, + { + "epoch": 49, + "eval_accuracy": 0.2152, + "eval_loss": 727.2657625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 465.0384421766493 + }, + { + "epoch": 50, + "eval_accuracy": 0.2126, + "eval_loss": 541.7233875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 481.0208101074219 + }, + { + "epoch": 51, + "eval_accuracy": 0.1636, + "eval_loss": 1147.0993, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 481.1614602701823 + }, + { + "epoch": 52, + "eval_accuracy": 0.132, + "eval_loss": 840.8927625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 548.6089418836806 + }, + { + "epoch": 53, + "eval_accuracy": 0.2612, + "eval_loss": 661.988025, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 678.038278624132 + }, + { + "epoch": 54, + "eval_accuracy": 0.1514, + "eval_loss": 2164.9526, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 609.5724734592014 + }, + { + "epoch": 55, + "eval_accuracy": 0.2108, + "eval_loss": 1088.262275, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 725.6776028862847 + }, + { + "epoch": 56, + "eval_accuracy": 0.331, + "eval_loss": 319.69473125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 736.4521655978733 + }, + { + "epoch": 57, + "eval_accuracy": 0.2552, + "eval_loss": 764.159225, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 651.7272706271701 + }, + { + "epoch": 58, + "eval_accuracy": 0.2334, + "eval_loss": 1569.883025, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 754.1738535807292 + }, + { + "epoch": 59, + "eval_accuracy": 0.2296, + "eval_loss": 1001.5583875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 898.3274509548611 + }, + { + "epoch": 60, + "eval_accuracy": 0.2058, + "eval_loss": 942.179375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 842.5285305989584 + }, + { + "epoch": 61, + "eval_accuracy": 0.2784, + "eval_loss": 641.63215, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 922.3867413628473 + }, + { + "epoch": 62, + "eval_accuracy": 0.2018, + "eval_loss": 1669.65185, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 936.2655243923612 + }, + { + "epoch": 63, + "eval_accuracy": 0.1514, + "eval_loss": 2189.349025, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 1097.9668812065972 + }, + { + "epoch": 64, + "eval_accuracy": 0.1242, + "eval_loss": 1848.996975, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 904.8054986979166 + }, + { + "epoch": 65, + "eval_accuracy": 0.2676, + "eval_loss": 945.107175, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 1117.1221231336806 + }, + { + "epoch": 66, + "eval_accuracy": 0.2208, + "eval_loss": 877.7712375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 1260.3183142415364 + }, + { + "epoch": 67, + "eval_accuracy": 0.1668, + "eval_loss": 1989.77125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 1177.6693628472221 + }, + { + "epoch": 68, + "eval_accuracy": 0.2552, + "eval_loss": 2615.7875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 1239.7195300347223 + }, + { + "epoch": 69, + "eval_accuracy": 0.2904, + "eval_loss": 1145.899225, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 1413.3063585828993 + }, + { + "epoch": 70, + "eval_accuracy": 0.1982, + "eval_loss": 1340.564725, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 1432.3528199001737 + }, + { + "epoch": 71, + "eval_accuracy": 0.2026, + "eval_loss": 2114.94005, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 1649.5298797743055 + }, + { + "epoch": 72, + "eval_accuracy": 0.2952, + "eval_loss": 817.0742125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 1511.4559931857639 + }, + { + "epoch": 73, + "eval_accuracy": 0.1986, + "eval_loss": 5055.5722, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 1597.8940986979167 + }, + { + "epoch": 74, + "eval_accuracy": 0.2644, + "eval_loss": 1979.6754, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 1545.4152473090278 + }, + { + "epoch": 75, + "eval_accuracy": 0.2184, + "eval_loss": 1628.492025, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 1909.3885863715277 + }, + { + "epoch": 76, + "eval_accuracy": 0.115, + "eval_loss": 3524.5311, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 1744.735998372396 + }, + { + "epoch": 77, + "eval_accuracy": 0.2232, + "eval_loss": 1634.16325, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 1917.358931282552 + }, + { + "epoch": 78, + "eval_accuracy": 0.2678, + "eval_loss": 1975.21855, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 1775.3661026584202 + }, + { + "epoch": 79, + "eval_accuracy": 0.12, + "eval_loss": 5105.1117, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 2197.024656922743 + }, + { + "epoch": 80, + "eval_accuracy": 0.2726, + "eval_loss": 1750.003225, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 2032.038926204427 + }, + { + "epoch": 81, + "eval_accuracy": 0.1904, + "eval_loss": 2697.03705, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 2351.129897764757 + }, + { + "epoch": 82, + "eval_accuracy": 0.1592, + "eval_loss": 3450.82265, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 2180.3887864583335 + }, + { + "epoch": 83, + "eval_accuracy": 0.3354, + "eval_loss": 1321.29855, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 2424.0763131944445 + }, + { + "epoch": 84, + "eval_accuracy": 0.2848, + "eval_loss": 3661.75795, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 2325.0078926649308 + }, + { + "epoch": 85, + "eval_accuracy": 0.2994, + "eval_loss": 1387.98565, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 2394.1546099826387 + }, + { + "epoch": 86, + "eval_accuracy": 0.2162, + "eval_loss": 5530.6169, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 2654.00433984375 + }, + { + "epoch": 87, + "eval_accuracy": 0.1348, + "eval_loss": 4538.7721, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 2801.9152317708335 + }, + { + "epoch": 88, + "eval_accuracy": 0.2436, + "eval_loss": 3880.88055, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 2592.33646875 + }, + { + "epoch": 89, + "eval_accuracy": 0.2474, + "eval_loss": 3858.4386, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 2924.044119314236 + }, + { + "epoch": 90, + "eval_accuracy": 0.1872, + "eval_loss": 4376.7555, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 2877.3811128689235 + }, + { + "epoch": 91, + "eval_accuracy": 0.227, + "eval_loss": 3983.18415, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 3019.5719048611113 + }, + { + "epoch": 92, + "eval_accuracy": 0.1876, + "eval_loss": 4822.9704, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 3324.442807378472 + }, + { + "epoch": 93, + "eval_accuracy": 0.1002, + "eval_loss": 8843.1426, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 3076.9131136284723 + }, + { + "epoch": 94, + "eval_accuracy": 0.2588, + "eval_loss": 2395.90985, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 3655.9118378038193 + }, + { + "epoch": 95, + "eval_accuracy": 0.3154, + "eval_loss": 2502.6303, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 3351.3480279513888 + }, + { + "epoch": 96, + "eval_accuracy": 0.2102, + "eval_loss": 2872.629, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 3646.8260487413195 + }, + { + "epoch": 97, + "eval_accuracy": 0.2226, + "eval_loss": 3102.1834, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 3670.6937383246527 + }, + { + "epoch": 98, + "eval_accuracy": 0.204, + "eval_loss": 5511.3655, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 3852.314850607639 + }, + { + "epoch": 99, + "eval_accuracy": 0.2012, + "eval_loss": 3771.54135, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 3782.515890190972 + }, + { + "epoch": 100, + "eval_accuracy": 0.224, + "eval_loss": 5541.5195, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 4224.866707638889 + }, + { + "epoch": 101, + "eval_accuracy": 0.3192, + "eval_loss": 954.1664625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 1023.3584813802083 + }, + { + "epoch": 102, + "eval_accuracy": 0.3176, + "eval_loss": 666.8220625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 674.6085620225695 + }, + { + "epoch": 103, + "eval_accuracy": 0.3044, + "eval_loss": 627.003725, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 585.7811496582032 + }, + { + "epoch": 104, + "eval_accuracy": 0.3124, + "eval_loss": 584.7140875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 545.9971641438802 + }, + { + "epoch": 105, + "eval_accuracy": 0.3118, + "eval_loss": 552.403825, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 524.543052484809 + }, + { + "epoch": 106, + "eval_accuracy": 0.32, + "eval_loss": 489.4699125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 491.8761935221354 + }, + { + "epoch": 107, + "eval_accuracy": 0.3254, + "eval_loss": 419.153225, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 467.915876936849 + }, + { + "epoch": 108, + "eval_accuracy": 0.3148, + "eval_loss": 416.48076875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 445.60208774414065 + }, + { + "epoch": 109, + "eval_accuracy": 0.3162, + "eval_loss": 417.7529375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 418.3163262369792 + }, + { + "epoch": 110, + "eval_accuracy": 0.3118, + "eval_loss": 381.2394, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 394.4938233561198 + }, + { + "epoch": 111, + "eval_accuracy": 0.2974, + "eval_loss": 393.32450625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 393.7556446126302 + }, + { + "epoch": 112, + "eval_accuracy": 0.3308, + "eval_loss": 316.0081375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 375.48029161783853 + }, + { + "epoch": 113, + "eval_accuracy": 0.33, + "eval_loss": 293.41238125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 355.54808799913195 + }, + { + "epoch": 114, + "eval_accuracy": 0.3242, + "eval_loss": 295.97705625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 340.8098280056424 + }, + { + "epoch": 115, + "eval_accuracy": 0.272, + "eval_loss": 356.230325, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 324.02832730577256 + }, + { + "epoch": 116, + "eval_accuracy": 0.2806, + "eval_loss": 388.42944375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 317.93263887261287 + }, + { + "epoch": 117, + "eval_accuracy": 0.2912, + "eval_loss": 320.12799375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 328.08235987413195 + }, + { + "epoch": 118, + "eval_accuracy": 0.2682, + "eval_loss": 365.60305, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 316.4193104817708 + }, + { + "epoch": 119, + "eval_accuracy": 0.3076, + "eval_loss": 293.89094375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 315.2604624891493 + }, + { + "epoch": 120, + "eval_accuracy": 0.2916, + "eval_loss": 338.8138, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 311.2781292751736 + }, + { + "epoch": 121, + "eval_accuracy": 0.2784, + "eval_loss": 359.22824375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 308.00667970920136 + }, + { + "epoch": 122, + "eval_accuracy": 0.3256, + "eval_loss": 303.96686875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 306.2332032118056 + }, + { + "epoch": 123, + "eval_accuracy": 0.246, + "eval_loss": 472.69735625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 326.6572144422743 + }, + { + "epoch": 124, + "eval_accuracy": 0.3092, + "eval_loss": 303.3160625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 327.3754111029731 + }, + { + "epoch": 125, + "eval_accuracy": 0.3476, + "eval_loss": 250.8448625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 333.8108040581597 + }, + { + "epoch": 126, + "eval_accuracy": 0.3196, + "eval_loss": 226.150090625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 308.0572482096354 + }, + { + "epoch": 127, + "eval_accuracy": 0.3436, + "eval_loss": 242.99584375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 302.3140259521484 + }, + { + "epoch": 128, + "eval_accuracy": 0.3344, + "eval_loss": 201.0353, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 297.69994912923175 + }, + { + "epoch": 129, + "eval_accuracy": 0.2382, + "eval_loss": 500.881875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 310.16952446831596 + }, + { + "epoch": 130, + "eval_accuracy": 0.2756, + "eval_loss": 516.33688125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 327.5882095594618 + }, + { + "epoch": 131, + "eval_accuracy": 0.3174, + "eval_loss": 267.765934375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 350.485053257921 + }, + { + "epoch": 132, + "eval_accuracy": 0.2736, + "eval_loss": 290.18941875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 317.8272129801432 + }, + { + "epoch": 133, + "eval_accuracy": 0.3064, + "eval_loss": 302.160225, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 305.53068083767363 + }, + { + "epoch": 134, + "eval_accuracy": 0.3088, + "eval_loss": 261.52979375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 303.8541805284288 + }, + { + "epoch": 135, + "eval_accuracy": 0.3396, + "eval_loss": 214.982534375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 330.44600800238715 + }, + { + "epoch": 136, + "eval_accuracy": 0.3392, + "eval_loss": 236.768690625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 304.66661754014757 + }, + { + "epoch": 137, + "eval_accuracy": 0.3212, + "eval_loss": 232.519890625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 349.3437628499349 + }, + { + "epoch": 138, + "eval_accuracy": 0.2252, + "eval_loss": 339.20315, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 296.8857865234375 + }, + { + "epoch": 139, + "eval_accuracy": 0.3216, + "eval_loss": 316.37996875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 366.07723196614586 + }, + { + "epoch": 140, + "eval_accuracy": 0.2446, + "eval_loss": 391.34220625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 324.58141136610243 + }, + { + "epoch": 141, + "eval_accuracy": 0.2642, + "eval_loss": 596.2132, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 298.179056000434 + }, + { + "epoch": 142, + "eval_accuracy": 0.2394, + "eval_loss": 403.21318125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 344.9279659396701 + }, + { + "epoch": 143, + "eval_accuracy": 0.2662, + "eval_loss": 385.6030375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 337.4383267252604 + }, + { + "epoch": 144, + "eval_accuracy": 0.3, + "eval_loss": 316.413725, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 322.35182532552085 + }, + { + "epoch": 145, + "eval_accuracy": 0.2924, + "eval_loss": 302.7712375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 328.3091288248698 + }, + { + "epoch": 146, + "eval_accuracy": 0.3518, + "eval_loss": 263.23478125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 357.0935332411024 + }, + { + "epoch": 147, + "eval_accuracy": 0.2666, + "eval_loss": 348.51591875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 302.8438815375434 + }, + { + "epoch": 148, + "eval_accuracy": 0.3322, + "eval_loss": 237.38480625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 351.8545443223741 + }, + { + "epoch": 149, + "eval_accuracy": 0.2848, + "eval_loss": 410.83615, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 337.79270292426213 + }, + { + "epoch": 150, + "eval_accuracy": 0.3464, + "eval_loss": 277.364209375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 314.32726003689237 + }, + { + "epoch": 151, + "eval_accuracy": 0.4056, + "eval_loss": 109.8672171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 116.31430750325521 + }, + { + "epoch": 152, + "eval_accuracy": 0.388, + "eval_loss": 104.958153125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 105.8084699001736 + }, + { + "epoch": 153, + "eval_accuracy": 0.3916, + "eval_loss": 96.0502734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 100.08656565348308 + }, + { + "epoch": 154, + "eval_accuracy": 0.3864, + "eval_loss": 92.389784375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 93.1440177951389 + }, + { + "epoch": 155, + "eval_accuracy": 0.3882, + "eval_loss": 86.4438921875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 88.13311031901041 + }, + { + "epoch": 156, + "eval_accuracy": 0.3774, + "eval_loss": 84.1961578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 84.23837499864366 + }, + { + "epoch": 157, + "eval_accuracy": 0.378, + "eval_loss": 76.8194421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 81.50850522867839 + }, + { + "epoch": 158, + "eval_accuracy": 0.3542, + "eval_loss": 76.5250671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 77.57814040527344 + }, + { + "epoch": 159, + "eval_accuracy": 0.3394, + "eval_loss": 79.1525265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 76.05412509901258 + }, + { + "epoch": 160, + "eval_accuracy": 0.351, + "eval_loss": 73.306775, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 72.83907199164497 + }, + { + "epoch": 161, + "eval_accuracy": 0.3548, + "eval_loss": 67.07043828125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 71.70867970987956 + }, + { + "epoch": 162, + "eval_accuracy": 0.3582, + "eval_loss": 63.2703171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 68.49551196695964 + }, + { + "epoch": 163, + "eval_accuracy": 0.3382, + "eval_loss": 64.855196875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 67.76478578287761 + }, + { + "epoch": 164, + "eval_accuracy": 0.3164, + "eval_loss": 68.83066171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 66.98758099500868 + }, + { + "epoch": 165, + "eval_accuracy": 0.3122, + "eval_loss": 76.994125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 65.67458069729275 + }, + { + "epoch": 166, + "eval_accuracy": 0.3226, + "eval_loss": 64.54301171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 66.97706856689453 + }, + { + "epoch": 167, + "eval_accuracy": 0.345, + "eval_loss": 66.2377984375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 65.27456371934679 + }, + { + "epoch": 168, + "eval_accuracy": 0.3366, + "eval_loss": 58.94499375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 64.11031446533202 + }, + { + "epoch": 169, + "eval_accuracy": 0.3396, + "eval_loss": 63.12518359375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 64.05919189724392 + }, + { + "epoch": 170, + "eval_accuracy": 0.3254, + "eval_loss": 60.609440625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 61.94588626302083 + }, + { + "epoch": 171, + "eval_accuracy": 0.3226, + "eval_loss": 63.21683515625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 61.45856361355252 + }, + { + "epoch": 172, + "eval_accuracy": 0.3216, + "eval_loss": 61.59688359375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 62.51791443006727 + }, + { + "epoch": 173, + "eval_accuracy": 0.3212, + "eval_loss": 60.5227796875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 61.21382712944879 + }, + { + "epoch": 174, + "eval_accuracy": 0.3418, + "eval_loss": 56.33953046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 60.35559227159288 + }, + { + "epoch": 175, + "eval_accuracy": 0.3166, + "eval_loss": 66.64052890625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 59.094696812608504 + }, + { + "epoch": 176, + "eval_accuracy": 0.33, + "eval_loss": 58.2246609375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 59.71331578233507 + }, + { + "epoch": 177, + "eval_accuracy": 0.3022, + "eval_loss": 60.79954296875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 59.35683131239149 + }, + { + "epoch": 178, + "eval_accuracy": 0.3458, + "eval_loss": 53.1319875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 57.96395812988281 + }, + { + "epoch": 179, + "eval_accuracy": 0.3396, + "eval_loss": 51.00816796875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 58.248469355604385 + }, + { + "epoch": 180, + "eval_accuracy": 0.3052, + "eval_loss": 56.26803671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 58.54176217244466 + }, + { + "epoch": 181, + "eval_accuracy": 0.3184, + "eval_loss": 54.643240625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 57.55351157226563 + }, + { + "epoch": 182, + "eval_accuracy": 0.2988, + "eval_loss": 57.60181640625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 57.566943718804254 + }, + { + "epoch": 183, + "eval_accuracy": 0.301, + "eval_loss": 56.3104078125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 57.762457067192926 + }, + { + "epoch": 184, + "eval_accuracy": 0.3196, + "eval_loss": 56.75072109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 59.09771092393663 + }, + { + "epoch": 185, + "eval_accuracy": 0.3478, + "eval_loss": 49.8344, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 56.15469523654514 + }, + { + "epoch": 186, + "eval_accuracy": 0.2954, + "eval_loss": 53.380528125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 57.07655614352756 + }, + { + "epoch": 187, + "eval_accuracy": 0.3048, + "eval_loss": 59.88317734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 56.7686597418891 + }, + { + "epoch": 188, + "eval_accuracy": 0.3098, + "eval_loss": 53.8027515625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 56.743705487060545 + }, + { + "epoch": 189, + "eval_accuracy": 0.3278, + "eval_loss": 49.3894125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 56.37033995768229 + }, + { + "epoch": 190, + "eval_accuracy": 0.2882, + "eval_loss": 57.20807890625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 56.76239995727539 + }, + { + "epoch": 191, + "eval_accuracy": 0.3118, + "eval_loss": 52.07348515625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 56.76262446221246 + }, + { + "epoch": 192, + "eval_accuracy": 0.3122, + "eval_loss": 51.81205, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 55.770336328125 + }, + { + "epoch": 193, + "eval_accuracy": 0.2786, + "eval_loss": 55.8960453125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 55.33157925279406 + }, + { + "epoch": 194, + "eval_accuracy": 0.2784, + "eval_loss": 69.81824453125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 55.13422055121528 + }, + { + "epoch": 195, + "eval_accuracy": 0.3014, + "eval_loss": 52.612559375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 56.69760437011719 + }, + { + "epoch": 196, + "eval_accuracy": 0.2598, + "eval_loss": 55.4450078125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 54.42104867485894 + }, + { + "epoch": 197, + "eval_accuracy": 0.2926, + "eval_loss": 61.0219375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 55.378040165201824 + }, + { + "epoch": 198, + "eval_accuracy": 0.277, + "eval_loss": 58.84813046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 56.909702324761284 + }, + { + "epoch": 199, + "eval_accuracy": 0.2776, + "eval_loss": 59.28854296875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 57.713160746934676 + }, + { + "epoch": 200, + "eval_accuracy": 0.284, + "eval_loss": 56.89802421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 56.61742428792318 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.284, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 56.89802421875, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 3276921856, + "peak_memory_reserved_bytes": 4416602112, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 60.47297954559326, + "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": 3241.7450346946716, + "train_wall_s": 3181.2487540245056 + }, + "work": { + "apical_macs_per_example": 5324800, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 47923200000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 1129371264000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1129371264000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 2306665728000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json.manifest.json new file mode 100644 index 0000000..0427c07 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json.manifest.json @@ -0,0 +1,132 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "dfa", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.01", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--vectorizer_mode", + "spatial_template" + ], + "completed_unix_time": 1787022844.8435245, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 3246.3810822963715, + "epochs": 200, + "experiment_name": "resnet-r2-dfa-d56-lr0p01", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "dfa", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json", + "output_exists": true, + "output_sha256": "f646a557aca62c9dbfb2672fedd765f0cd51d11256c50995933d43d90a18c8f9", + "rate": 0.01, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json b/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json new file mode 100644 index 0000000..1d1bf92 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json @@ -0,0 +1,2965 @@ +{ + "architecture": { + "base_width": 16, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 269722, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 200, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.05, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "method": "dualprop", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json", + "output_lr": 0.05, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.6441188268873426, + "train_seconds": 245.87645721435547, + "validation": { + "accuracy": 0.4738, + "examples": 5000, + "loss": 1.4967913818359375 + }, + "validation_seconds": 0.12893104553222656 + }, + { + "epoch": 2, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.1635693959765965, + "train_seconds": 245.52287817001343, + "validation": { + "accuracy": 0.6026, + "examples": 5000, + "loss": 1.153321435546875 + }, + "validation_seconds": 0.10486650466918945 + }, + { + "epoch": 3, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.9352410600874159, + "train_seconds": 245.50975513458252, + "validation": { + "accuracy": 0.6018, + "examples": 5000, + "loss": 1.329039013671875 + }, + "validation_seconds": 0.10483121871948242 + }, + { + "epoch": 4, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7935926506890191, + "train_seconds": 245.52499318122864, + "validation": { + "accuracy": 0.7, + "examples": 5000, + "loss": 0.8777848876953125 + }, + "validation_seconds": 0.10486555099487305 + }, + { + "epoch": 5, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7077589258935716, + "train_seconds": 245.52429103851318, + "validation": { + "accuracy": 0.7442, + "examples": 5000, + "loss": 0.7581129028320313 + }, + "validation_seconds": 0.1048576831817627 + }, + { + "epoch": 6, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6486876674969991, + "train_seconds": 245.51856327056885, + "validation": { + "accuracy": 0.7254, + "examples": 5000, + "loss": 0.8741846923828125 + }, + "validation_seconds": 0.10486769676208496 + }, + { + "epoch": 7, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6074715187284682, + "train_seconds": 245.71568083763123, + "validation": { + "accuracy": 0.7682, + "examples": 5000, + "loss": 0.6674996459960938 + }, + "validation_seconds": 0.10516047477722168 + }, + { + "epoch": 8, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5710463250690037, + "train_seconds": 245.7029790878296, + "validation": { + "accuracy": 0.7782, + "examples": 5000, + "loss": 0.6623569702148437 + }, + "validation_seconds": 0.10492897033691406 + }, + { + "epoch": 9, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5336540970166525, + "train_seconds": 245.72987985610962, + "validation": { + "accuracy": 0.786, + "examples": 5000, + "loss": 0.640069384765625 + }, + "validation_seconds": 0.10491442680358887 + }, + { + "epoch": 10, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5106024785730574, + "train_seconds": 245.71102261543274, + "validation": { + "accuracy": 0.7826, + "examples": 5000, + "loss": 0.6581781616210938 + }, + "validation_seconds": 0.104949951171875 + }, + { + "epoch": 11, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.488464806482527, + "train_seconds": 245.70360898971558, + "validation": { + "accuracy": 0.7776, + "examples": 5000, + "loss": 0.6763726928710937 + }, + "validation_seconds": 0.10500192642211914 + }, + { + "epoch": 12, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4659737181398604, + "train_seconds": 245.7096302509308, + "validation": { + "accuracy": 0.8018, + "examples": 5000, + "loss": 0.603849560546875 + }, + "validation_seconds": 0.10489654541015625 + }, + { + "epoch": 13, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4473519527753194, + "train_seconds": 245.70638012886047, + "validation": { + "accuracy": 0.7406, + "examples": 5000, + "loss": 0.8407753784179688 + }, + "validation_seconds": 0.10495471954345703 + }, + { + "epoch": 14, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4388091464042664, + "train_seconds": 245.70655918121338, + "validation": { + "accuracy": 0.8338, + "examples": 5000, + "loss": 0.48709592895507814 + }, + "validation_seconds": 0.10503792762756348 + }, + { + "epoch": 15, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.42087091799841986, + "train_seconds": 245.71817874908447, + "validation": { + "accuracy": 0.7902, + "examples": 5000, + "loss": 0.6748830810546875 + }, + "validation_seconds": 0.1049962043762207 + }, + { + "epoch": 16, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4085377493328518, + "train_seconds": 245.71288919448853, + "validation": { + "accuracy": 0.8178, + "examples": 5000, + "loss": 0.5664857482910156 + }, + "validation_seconds": 0.10489010810852051 + }, + { + "epoch": 17, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.39353587019178604, + "train_seconds": 245.71254181861877, + "validation": { + "accuracy": 0.8128, + "examples": 5000, + "loss": 0.5877752197265625 + }, + "validation_seconds": 0.1049199104309082 + }, + { + "epoch": 18, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.38593393679724797, + "train_seconds": 245.7317178249359, + "validation": { + "accuracy": 0.8052, + "examples": 5000, + "loss": 0.5994394897460937 + }, + "validation_seconds": 0.1049954891204834 + }, + { + "epoch": 19, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.37112634809547, + "train_seconds": 245.7781913280487, + "validation": { + "accuracy": 0.8426, + "examples": 5000, + "loss": 0.4838470458984375 + }, + "validation_seconds": 0.10508608818054199 + }, + { + "epoch": 20, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3595076397365994, + "train_seconds": 246.36518168449402, + "validation": { + "accuracy": 0.8094, + "examples": 5000, + "loss": 0.6470749267578125 + }, + "validation_seconds": 0.10510373115539551 + }, + { + "epoch": 21, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3523521614180671, + "train_seconds": 246.52931666374207, + "validation": { + "accuracy": 0.8242, + "examples": 5000, + "loss": 0.5845088745117187 + }, + "validation_seconds": 0.10520172119140625 + }, + { + "epoch": 22, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.34563857612080046, + "train_seconds": 246.01910829544067, + "validation": { + "accuracy": 0.8096, + "examples": 5000, + "loss": 0.62537509765625 + }, + "validation_seconds": 0.10520815849304199 + }, + { + "epoch": 23, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.33976192728678384, + "train_seconds": 245.8921320438385, + "validation": { + "accuracy": 0.8444, + "examples": 5000, + "loss": 0.484679150390625 + }, + "validation_seconds": 0.1049337387084961 + }, + { + "epoch": 24, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3225088294665019, + "train_seconds": 245.83658075332642, + "validation": { + "accuracy": 0.8396, + "examples": 5000, + "loss": 0.50716240234375 + }, + "validation_seconds": 0.1049196720123291 + }, + { + "epoch": 25, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.32271037464141844, + "train_seconds": 245.84278297424316, + "validation": { + "accuracy": 0.8406, + "examples": 5000, + "loss": 0.493395458984375 + }, + "validation_seconds": 0.10499238967895508 + }, + { + "epoch": 26, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3157399925496843, + "train_seconds": 245.84543251991272, + "validation": { + "accuracy": 0.858, + "examples": 5000, + "loss": 0.44974503173828123 + }, + "validation_seconds": 0.10491776466369629 + }, + { + "epoch": 27, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.30543111317952476, + "train_seconds": 245.86448335647583, + "validation": { + "accuracy": 0.8416, + "examples": 5000, + "loss": 0.5209959777832032 + }, + "validation_seconds": 0.10499191284179688 + }, + { + "epoch": 28, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.301433762105306, + "train_seconds": 245.840149641037, + "validation": { + "accuracy": 0.8244, + "examples": 5000, + "loss": 0.5745289489746094 + }, + "validation_seconds": 0.10509037971496582 + }, + { + "epoch": 29, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.29656535551018187, + "train_seconds": 245.8558783531189, + "validation": { + "accuracy": 0.8144, + "examples": 5000, + "loss": 0.6412824096679688 + }, + "validation_seconds": 0.10495448112487793 + }, + { + "epoch": 30, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.29441430922614203, + "train_seconds": 245.80975580215454, + "validation": { + "accuracy": 0.8356, + "examples": 5000, + "loss": 0.5149374877929688 + }, + "validation_seconds": 0.10511445999145508 + }, + { + "epoch": 31, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.28566161390410527, + "train_seconds": 245.73937225341797, + "validation": { + "accuracy": 0.842, + "examples": 5000, + "loss": 0.5199394226074219 + }, + "validation_seconds": 0.10500478744506836 + }, + { + "epoch": 32, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2805537829717, + "train_seconds": 245.7104251384735, + "validation": { + "accuracy": 0.8688, + "examples": 5000, + "loss": 0.4206529296875 + }, + "validation_seconds": 0.10506677627563477 + }, + { + "epoch": 33, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2817669762743844, + "train_seconds": 245.7130422592163, + "validation": { + "accuracy": 0.8566, + "examples": 5000, + "loss": 0.43376856689453125 + }, + "validation_seconds": 0.1049799919128418 + }, + { + "epoch": 34, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.27807320489618514, + "train_seconds": 245.71902012825012, + "validation": { + "accuracy": 0.8558, + "examples": 5000, + "loss": 0.46831655883789064 + }, + "validation_seconds": 0.10493636131286621 + }, + { + "epoch": 35, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.263952144712872, + "train_seconds": 245.93091106414795, + "validation": { + "accuracy": 0.8554, + "examples": 5000, + "loss": 0.4892894958496094 + }, + "validation_seconds": 0.10500526428222656 + }, + { + "epoch": 36, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2642927113586002, + "train_seconds": 246.30246591567993, + "validation": { + "accuracy": 0.8482, + "examples": 5000, + "loss": 0.47789179077148436 + }, + "validation_seconds": 0.10546326637268066 + }, + { + "epoch": 37, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2638369460807906, + "train_seconds": 246.64353728294373, + "validation": { + "accuracy": 0.8326, + "examples": 5000, + "loss": 0.5493978393554687 + }, + "validation_seconds": 0.10544991493225098 + }, + { + "epoch": 38, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.25538891750971476, + "train_seconds": 246.44886422157288, + "validation": { + "accuracy": 0.8574, + "examples": 5000, + "loss": 0.451583837890625 + }, + "validation_seconds": 0.10479283332824707 + }, + { + "epoch": 39, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.25059934278064305, + "train_seconds": 246.8432900905609, + "validation": { + "accuracy": 0.8636, + "examples": 5000, + "loss": 0.44829965209960937 + }, + "validation_seconds": 0.10497689247131348 + }, + { + "epoch": 40, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.24968606785933176, + "train_seconds": 246.86512732505798, + "validation": { + "accuracy": 0.857, + "examples": 5000, + "loss": 0.47868128662109377 + }, + "validation_seconds": 0.10499143600463867 + }, + { + "epoch": 41, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.24340666038990022, + "train_seconds": 246.9741039276123, + "validation": { + "accuracy": 0.863, + "examples": 5000, + "loss": 0.43970326538085935 + }, + "validation_seconds": 0.10489535331726074 + }, + { + "epoch": 42, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2426151712603039, + "train_seconds": 246.62085914611816, + "validation": { + "accuracy": 0.8678, + "examples": 5000, + "loss": 0.43893058471679686 + }, + "validation_seconds": 0.10530686378479004 + }, + { + "epoch": 43, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.23919208281305102, + "train_seconds": 246.47220969200134, + "validation": { + "accuracy": 0.8564, + "examples": 5000, + "loss": 0.4502086669921875 + }, + "validation_seconds": 0.1048734188079834 + }, + { + "epoch": 44, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2358168052620358, + "train_seconds": 246.44223856925964, + "validation": { + "accuracy": 0.8584, + "examples": 5000, + "loss": 0.4598034118652344 + }, + "validation_seconds": 0.10493111610412598 + }, + { + "epoch": 45, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.23206949197981092, + "train_seconds": 245.91084361076355, + "validation": { + "accuracy": 0.8434, + "examples": 5000, + "loss": 0.5504330444335938 + }, + "validation_seconds": 0.10497212409973145 + }, + { + "epoch": 46, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.23059860911899144, + "train_seconds": 246.39553570747375, + "validation": { + "accuracy": 0.8514, + "examples": 5000, + "loss": 0.4961731567382813 + }, + "validation_seconds": 0.10496854782104492 + }, + { + "epoch": 47, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2259720978895823, + "train_seconds": 246.28038716316223, + "validation": { + "accuracy": 0.8612, + "examples": 5000, + "loss": 0.4870090087890625 + }, + "validation_seconds": 0.10482335090637207 + }, + { + "epoch": 48, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2301974679734972, + "train_seconds": 245.90750241279602, + "validation": { + "accuracy": 0.8584, + "examples": 5000, + "loss": 0.4882284912109375 + }, + "validation_seconds": 0.10474634170532227 + }, + { + "epoch": 49, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.22358488881323071, + "train_seconds": 245.81437158584595, + "validation": { + "accuracy": 0.8672, + "examples": 5000, + "loss": 0.46109835205078126 + }, + "validation_seconds": 0.10482645034790039 + }, + { + "epoch": 50, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2217408281246821, + "train_seconds": 245.80533456802368, + "validation": { + "accuracy": 0.8702, + "examples": 5000, + "loss": 0.42463759765625 + }, + "validation_seconds": 0.10474348068237305 + }, + { + "epoch": 51, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.21557051228152382, + "train_seconds": 245.94883275032043, + "validation": { + "accuracy": 0.8518, + "examples": 5000, + "loss": 0.48058946533203123 + }, + "validation_seconds": 0.10477089881896973 + }, + { + "epoch": 52, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2120681472301483, + "train_seconds": 245.94964838027954, + "validation": { + "accuracy": 0.865, + "examples": 5000, + "loss": 0.44984417114257813 + }, + "validation_seconds": 0.10477733612060547 + }, + { + "epoch": 53, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20977391286161212, + "train_seconds": 245.92657470703125, + "validation": { + "accuracy": 0.8518, + "examples": 5000, + "loss": 0.53985625 + }, + "validation_seconds": 0.10475325584411621 + }, + { + "epoch": 54, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2089330201546351, + "train_seconds": 245.93967461585999, + "validation": { + "accuracy": 0.856, + "examples": 5000, + "loss": 0.4909478576660156 + }, + "validation_seconds": 0.10521721839904785 + }, + { + "epoch": 55, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.21272749128341675, + "train_seconds": 246.59948468208313, + "validation": { + "accuracy": 0.8546, + "examples": 5000, + "loss": 0.4906638977050781 + }, + "validation_seconds": 0.10508394241333008 + }, + { + "epoch": 56, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20717158926592932, + "train_seconds": 246.48116064071655, + "validation": { + "accuracy": 0.8504, + "examples": 5000, + "loss": 0.5430794372558594 + }, + "validation_seconds": 0.10493278503417969 + }, + { + "epoch": 57, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20313986417717403, + "train_seconds": 246.57427525520325, + "validation": { + "accuracy": 0.842, + "examples": 5000, + "loss": 0.5557522338867188 + }, + "validation_seconds": 0.1049644947052002 + }, + { + "epoch": 58, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2027928885936737, + "train_seconds": 246.51094436645508, + "validation": { + "accuracy": 0.8666, + "examples": 5000, + "loss": 0.4270877563476563 + }, + "validation_seconds": 0.1048426628112793 + }, + { + "epoch": 59, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.20086540544033052, + "train_seconds": 245.954576253891, + "validation": { + "accuracy": 0.8526, + "examples": 5000, + "loss": 0.4917937927246094 + }, + "validation_seconds": 0.10494685173034668 + }, + { + "epoch": 60, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19744458295504252, + "train_seconds": 246.04650807380676, + "validation": { + "accuracy": 0.8634, + "examples": 5000, + "loss": 0.4751939880371094 + }, + "validation_seconds": 0.10476160049438477 + }, + { + "epoch": 61, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19625209110842812, + "train_seconds": 246.03907656669617, + "validation": { + "accuracy": 0.855, + "examples": 5000, + "loss": 0.5089188537597656 + }, + "validation_seconds": 0.10500311851501465 + }, + { + "epoch": 62, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19375401494238112, + "train_seconds": 245.91973066329956, + "validation": { + "accuracy": 0.8516, + "examples": 5000, + "loss": 0.5393552185058593 + }, + "validation_seconds": 0.10500383377075195 + }, + { + "epoch": 63, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1935203471104304, + "train_seconds": 245.89782786369324, + "validation": { + "accuracy": 0.861, + "examples": 5000, + "loss": 0.4727430725097656 + }, + "validation_seconds": 0.1048274040222168 + }, + { + "epoch": 64, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18534126149018607, + "train_seconds": 245.9205825328827, + "validation": { + "accuracy": 0.8472, + "examples": 5000, + "loss": 0.5769324584960938 + }, + "validation_seconds": 0.10484862327575684 + }, + { + "epoch": 65, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.19316486716535355, + "train_seconds": 245.91525053977966, + "validation": { + "accuracy": 0.8758, + "examples": 5000, + "loss": 0.4404202941894531 + }, + "validation_seconds": 0.10477232933044434 + }, + { + "epoch": 66, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18819561854203543, + "train_seconds": 245.94995212554932, + "validation": { + "accuracy": 0.8632, + "examples": 5000, + "loss": 0.4956598876953125 + }, + "validation_seconds": 0.10492324829101562 + }, + { + "epoch": 67, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1861683264070087, + "train_seconds": 245.92055249214172, + "validation": { + "accuracy": 0.8396, + "examples": 5000, + "loss": 0.564496875 + }, + "validation_seconds": 0.10488319396972656 + }, + { + "epoch": 68, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18245379467010497, + "train_seconds": 245.90807962417603, + "validation": { + "accuracy": 0.855, + "examples": 5000, + "loss": 0.516592431640625 + }, + "validation_seconds": 0.10478520393371582 + }, + { + "epoch": 69, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18739715305169424, + "train_seconds": 245.9303948879242, + "validation": { + "accuracy": 0.8826, + "examples": 5000, + "loss": 0.4084698120117187 + }, + "validation_seconds": 0.10473227500915527 + }, + { + "epoch": 70, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18380963235431247, + "train_seconds": 245.92044949531555, + "validation": { + "accuracy": 0.8758, + "examples": 5000, + "loss": 0.43710973510742185 + }, + "validation_seconds": 0.1049046516418457 + }, + { + "epoch": 71, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.18311760312716166, + "train_seconds": 245.96293902397156, + "validation": { + "accuracy": 0.8716, + "examples": 5000, + "loss": 0.4604240112304688 + }, + "validation_seconds": 0.1050562858581543 + }, + { + "epoch": 72, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17762095175319248, + "train_seconds": 245.93334341049194, + "validation": { + "accuracy": 0.8606, + "examples": 5000, + "loss": 0.5003765075683594 + }, + "validation_seconds": 0.10496139526367188 + }, + { + "epoch": 73, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1763817933294508, + "train_seconds": 245.9250099658966, + "validation": { + "accuracy": 0.882, + "examples": 5000, + "loss": 0.41763900756835937 + }, + "validation_seconds": 0.10477662086486816 + }, + { + "epoch": 74, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17314121738274893, + "train_seconds": 245.8988173007965, + "validation": { + "accuracy": 0.8886, + "examples": 5000, + "loss": 0.37928946533203123 + }, + "validation_seconds": 0.10474419593811035 + }, + { + "epoch": 75, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17797526009877523, + "train_seconds": 246.13520216941833, + "validation": { + "accuracy": 0.8752, + "examples": 5000, + "loss": 0.4501698974609375 + }, + "validation_seconds": 0.10515260696411133 + }, + { + "epoch": 76, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17506222959889306, + "train_seconds": 246.67396068572998, + "validation": { + "accuracy": 0.8754, + "examples": 5000, + "loss": 0.4668177490234375 + }, + "validation_seconds": 0.10496044158935547 + }, + { + "epoch": 77, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1717323899322086, + "train_seconds": 246.7168323993683, + "validation": { + "accuracy": 0.8626, + "examples": 5000, + "loss": 0.5042861206054687 + }, + "validation_seconds": 0.10485363006591797 + }, + { + "epoch": 78, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17638354170189963, + "train_seconds": 246.71074175834656, + "validation": { + "accuracy": 0.8708, + "examples": 5000, + "loss": 0.478639453125 + }, + "validation_seconds": 0.10538220405578613 + }, + { + "epoch": 79, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1677387358956867, + "train_seconds": 246.7130355834961, + "validation": { + "accuracy": 0.8472, + "examples": 5000, + "loss": 0.5276833251953125 + }, + "validation_seconds": 0.10472655296325684 + }, + { + "epoch": 80, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16351781239774493, + "train_seconds": 246.68999099731445, + "validation": { + "accuracy": 0.8682, + "examples": 5000, + "loss": 0.48907069091796873 + }, + "validation_seconds": 0.10485410690307617 + }, + { + "epoch": 81, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.17571784110599095, + "train_seconds": 246.70585751533508, + "validation": { + "accuracy": 0.868, + "examples": 5000, + "loss": 0.4847575805664063 + }, + "validation_seconds": 0.10532069206237793 + }, + { + "epoch": 82, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1694504910853174, + "train_seconds": 246.70015454292297, + "validation": { + "accuracy": 0.8556, + "examples": 5000, + "loss": 0.488367724609375 + }, + "validation_seconds": 0.10495138168334961 + }, + { + "epoch": 83, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16532643840048047, + "train_seconds": 246.68881845474243, + "validation": { + "accuracy": 0.8416, + "examples": 5000, + "loss": 0.5911368530273438 + }, + "validation_seconds": 0.10479950904846191 + }, + { + "epoch": 84, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1684850957579083, + "train_seconds": 246.71207571029663, + "validation": { + "accuracy": 0.8732, + "examples": 5000, + "loss": 0.46909482421875 + }, + "validation_seconds": 0.1052088737487793 + }, + { + "epoch": 85, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1677253027809991, + "train_seconds": 246.69193530082703, + "validation": { + "accuracy": 0.8696, + "examples": 5000, + "loss": 0.4961885070800781 + }, + "validation_seconds": 0.10480213165283203 + }, + { + "epoch": 86, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16023339446915522, + "train_seconds": 246.63249588012695, + "validation": { + "accuracy": 0.8818, + "examples": 5000, + "loss": 0.4365656616210937 + }, + "validation_seconds": 0.1050879955291748 + }, + { + "epoch": 87, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1696844631936815, + "train_seconds": 246.65419721603394, + "validation": { + "accuracy": 0.8864, + "examples": 5000, + "loss": 0.394526025390625 + }, + "validation_seconds": 0.10512447357177734 + }, + { + "epoch": 88, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16108827577696908, + "train_seconds": 246.70431089401245, + "validation": { + "accuracy": 0.8798, + "examples": 5000, + "loss": 0.45328688354492186 + }, + "validation_seconds": 0.10512113571166992 + }, + { + "epoch": 89, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16032158904737898, + "train_seconds": 246.6812767982483, + "validation": { + "accuracy": 0.8816, + "examples": 5000, + "loss": 0.44019458618164065 + }, + "validation_seconds": 0.10508894920349121 + }, + { + "epoch": 90, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16225236432817247, + "train_seconds": 246.68167066574097, + "validation": { + "accuracy": 0.8624, + "examples": 5000, + "loss": 0.4962203369140625 + }, + "validation_seconds": 0.10485458374023438 + }, + { + "epoch": 91, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16403424759838317, + "train_seconds": 246.61265230178833, + "validation": { + "accuracy": 0.8774, + "examples": 5000, + "loss": 0.43822531127929687 + }, + "validation_seconds": 0.10491776466369629 + }, + { + "epoch": 92, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1596482249458631, + "train_seconds": 246.5586667060852, + "validation": { + "accuracy": 0.8592, + "examples": 5000, + "loss": 0.5626332885742188 + }, + "validation_seconds": 0.10481786727905273 + }, + { + "epoch": 93, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15875176428423987, + "train_seconds": 246.6671712398529, + "validation": { + "accuracy": 0.8762, + "examples": 5000, + "loss": 0.45965128173828124 + }, + "validation_seconds": 0.10487985610961914 + }, + { + "epoch": 94, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.16234073853757647, + "train_seconds": 246.63065123558044, + "validation": { + "accuracy": 0.868, + "examples": 5000, + "loss": 0.46680614624023437 + }, + "validation_seconds": 0.10476207733154297 + }, + { + "epoch": 95, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1534736484448115, + "train_seconds": 246.65962553024292, + "validation": { + "accuracy": 0.8672, + "examples": 5000, + "loss": 0.46624967041015625 + }, + "validation_seconds": 0.10492587089538574 + }, + { + "epoch": 96, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15440344122515784, + "train_seconds": 246.66373991966248, + "validation": { + "accuracy": 0.8628, + "examples": 5000, + "loss": 0.4870942626953125 + }, + "validation_seconds": 0.10490155220031738 + }, + { + "epoch": 97, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15565202848911286, + "train_seconds": 246.66854810714722, + "validation": { + "accuracy": 0.8712, + "examples": 5000, + "loss": 0.4484557861328125 + }, + "validation_seconds": 0.10535168647766113 + }, + { + "epoch": 98, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.15759743853674996, + "train_seconds": 246.67762279510498, + "validation": { + "accuracy": 0.8688, + "examples": 5000, + "loss": 0.49867626953125 + }, + "validation_seconds": 0.10507059097290039 + }, + { + "epoch": 99, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.1574016169892417, + "train_seconds": 246.6524622440338, + "validation": { + "accuracy": 0.8764, + "examples": 5000, + "loss": 0.4572004699707031 + }, + "validation_seconds": 0.1049041748046875 + }, + { + "epoch": 100, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.14737063789367677, + "train_seconds": 246.63342666625977, + "validation": { + "accuracy": 0.8754, + "examples": 5000, + "loss": 0.4618093322753906 + }, + "validation_seconds": 0.10494613647460938 + }, + { + "epoch": 101, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.09404547926055061, + "train_seconds": 246.65588760375977, + "validation": { + "accuracy": 0.906, + "examples": 5000, + "loss": 0.3335665466308594 + }, + "validation_seconds": 0.10500240325927734 + }, + { + "epoch": 102, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07055688915782504, + "train_seconds": 246.65637946128845, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.3303562255859375 + }, + "validation_seconds": 0.10512924194335938 + }, + { + "epoch": 103, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.06393063375221358, + "train_seconds": 246.56659412384033, + "validation": { + "accuracy": 0.9106, + "examples": 5000, + "loss": 0.33267850341796873 + }, + "validation_seconds": 0.10535550117492676 + }, + { + "epoch": 104, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.056251141385237374, + "train_seconds": 246.58122849464417, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.34011267700195313 + }, + "validation_seconds": 0.10490632057189941 + }, + { + "epoch": 105, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.056818765364752874, + "train_seconds": 246.65290474891663, + "validation": { + "accuracy": 0.906, + "examples": 5000, + "loss": 0.3483814758300781 + }, + "validation_seconds": 0.10506415367126465 + }, + { + "epoch": 106, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.052452130284574294, + "train_seconds": 246.61904573440552, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.345201025390625 + }, + "validation_seconds": 0.10527157783508301 + }, + { + "epoch": 107, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.05003148424790965, + "train_seconds": 246.64277410507202, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.3456508056640625 + }, + "validation_seconds": 0.10487556457519531 + }, + { + "epoch": 108, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04837106784184774, + "train_seconds": 246.62830758094788, + "validation": { + "accuracy": 0.9082, + "examples": 5000, + "loss": 0.3555091491699219 + }, + "validation_seconds": 0.10479950904846191 + }, + { + "epoch": 109, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.046110330545239976, + "train_seconds": 246.64310789108276, + "validation": { + "accuracy": 0.9074, + "examples": 5000, + "loss": 0.36177589721679687 + }, + "validation_seconds": 0.10510897636413574 + }, + { + "epoch": 110, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04422715940343009, + "train_seconds": 246.65255451202393, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.35412333984375 + }, + "validation_seconds": 0.10495305061340332 + }, + { + "epoch": 111, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04302270753913456, + "train_seconds": 246.68583822250366, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.36001808471679686 + }, + "validation_seconds": 0.10499882698059082 + }, + { + "epoch": 112, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04221789358986749, + "train_seconds": 246.63890480995178, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.3641726318359375 + }, + "validation_seconds": 0.10499691963195801 + }, + { + "epoch": 113, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.04019610329998864, + "train_seconds": 246.6600148677826, + "validation": { + "accuracy": 0.9112, + "examples": 5000, + "loss": 0.3682267639160156 + }, + "validation_seconds": 0.10518097877502441 + }, + { + "epoch": 114, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.038186527136299345, + "train_seconds": 246.6488721370697, + "validation": { + "accuracy": 0.907, + "examples": 5000, + "loss": 0.37181755981445314 + }, + "validation_seconds": 0.10516762733459473 + }, + { + "epoch": 115, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03772044310834673, + "train_seconds": 246.58393955230713, + "validation": { + "accuracy": 0.908, + "examples": 5000, + "loss": 0.37084036865234377 + }, + "validation_seconds": 0.10496854782104492 + }, + { + "epoch": 116, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03771770534714063, + "train_seconds": 246.46069836616516, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.3673696716308594 + }, + "validation_seconds": 0.10475349426269531 + }, + { + "epoch": 117, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03610480902194977, + "train_seconds": 246.45283031463623, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.3759578430175781 + }, + "validation_seconds": 0.10521340370178223 + }, + { + "epoch": 118, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03562721369995011, + "train_seconds": 246.5085368156433, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.377801953125 + }, + "validation_seconds": 0.10504674911499023 + }, + { + "epoch": 119, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0351684531306227, + "train_seconds": 246.4897859096527, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.3793407165527344 + }, + "validation_seconds": 0.1051180362701416 + }, + { + "epoch": 120, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03166843252413803, + "train_seconds": 246.504944562912, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.3807004577636719 + }, + "validation_seconds": 0.10520434379577637 + }, + { + "epoch": 121, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.032848191968599956, + "train_seconds": 246.5223376750946, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.3911223388671875 + }, + "validation_seconds": 0.10503935813903809 + }, + { + "epoch": 122, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.032460003905495007, + "train_seconds": 246.50015473365784, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.3859025024414062 + }, + "validation_seconds": 0.10502910614013672 + }, + { + "epoch": 123, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03331895860698488, + "train_seconds": 246.51648592948914, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.3818461975097656 + }, + "validation_seconds": 0.10492110252380371 + }, + { + "epoch": 124, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.030804785005582702, + "train_seconds": 246.51632118225098, + "validation": { + "accuracy": 0.9082, + "examples": 5000, + "loss": 0.384043212890625 + }, + "validation_seconds": 0.10508346557617188 + }, + { + "epoch": 125, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.03100608646472295, + "train_seconds": 246.48151516914368, + "validation": { + "accuracy": 0.9068, + "examples": 5000, + "loss": 0.3950748840332031 + }, + "validation_seconds": 0.10526251792907715 + }, + { + "epoch": 126, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.030607278982798258, + "train_seconds": 246.49509143829346, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.38821942749023436 + }, + "validation_seconds": 0.1049349308013916 + }, + { + "epoch": 127, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02984628584285577, + "train_seconds": 246.48997020721436, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.39522205810546873 + }, + "validation_seconds": 0.1047677993774414 + }, + { + "epoch": 128, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.028140256262818974, + "train_seconds": 246.5008955001831, + "validation": { + "accuracy": 0.9058, + "examples": 5000, + "loss": 0.3940588623046875 + }, + "validation_seconds": 0.10520315170288086 + }, + { + "epoch": 129, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.029524326505263647, + "train_seconds": 246.49133825302124, + "validation": { + "accuracy": 0.906, + "examples": 5000, + "loss": 0.40357665405273435 + }, + "validation_seconds": 0.10487580299377441 + }, + { + "epoch": 130, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.027381085895167457, + "train_seconds": 246.53742814064026, + "validation": { + "accuracy": 0.9058, + "examples": 5000, + "loss": 0.40291564331054686 + }, + "validation_seconds": 0.10510921478271484 + }, + { + "epoch": 131, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02712565425733725, + "train_seconds": 246.5171241760254, + "validation": { + "accuracy": 0.9066, + "examples": 5000, + "loss": 0.39782152099609375 + }, + "validation_seconds": 0.10495710372924805 + }, + { + "epoch": 132, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.027452782110704316, + "train_seconds": 246.53909969329834, + "validation": { + "accuracy": 0.9072, + "examples": 5000, + "loss": 0.40547529907226565 + }, + "validation_seconds": 0.1050260066986084 + }, + { + "epoch": 133, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02682856381237507, + "train_seconds": 246.52162146568298, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.400748779296875 + }, + "validation_seconds": 0.10499382019042969 + }, + { + "epoch": 134, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02624693241649204, + "train_seconds": 246.5282382965088, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4016249328613281 + }, + "validation_seconds": 0.10514974594116211 + }, + { + "epoch": 135, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02609049816512399, + "train_seconds": 246.50827550888062, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.40999682006835936 + }, + "validation_seconds": 0.10509872436523438 + }, + { + "epoch": 136, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02532903560532464, + "train_seconds": 246.4754238128662, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.40476075439453124 + }, + "validation_seconds": 0.10525250434875488 + }, + { + "epoch": 137, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.025431149927112793, + "train_seconds": 246.53441071510315, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.40455454711914063 + }, + "validation_seconds": 0.10527610778808594 + }, + { + "epoch": 138, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02547713861929046, + "train_seconds": 246.49636459350586, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.4073498596191406 + }, + "validation_seconds": 0.10490679740905762 + }, + { + "epoch": 139, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02374524391161071, + "train_seconds": 246.48351001739502, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4119003784179687 + }, + "validation_seconds": 0.1051790714263916 + }, + { + "epoch": 140, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.023575902736849254, + "train_seconds": 246.4875295162201, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4121861877441406 + }, + "validation_seconds": 0.10484886169433594 + }, + { + "epoch": 141, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.025533000514242385, + "train_seconds": 246.53514981269836, + "validation": { + "accuracy": 0.911, + "examples": 5000, + "loss": 0.40740313720703125 + }, + "validation_seconds": 0.1048581600189209 + }, + { + "epoch": 142, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.023307849234673713, + "train_seconds": 246.49939465522766, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4205747985839844 + }, + "validation_seconds": 0.10505390167236328 + }, + { + "epoch": 143, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.024081955377260843, + "train_seconds": 246.50040888786316, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.42347489013671874 + }, + "validation_seconds": 0.10478520393371582 + }, + { + "epoch": 144, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02217308875951502, + "train_seconds": 246.50287771224976, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.4230736328125 + }, + "validation_seconds": 0.1047658920288086 + }, + { + "epoch": 145, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.021583292372690306, + "train_seconds": 246.49319005012512, + "validation": { + "accuracy": 0.9068, + "examples": 5000, + "loss": 0.4246301025390625 + }, + "validation_seconds": 0.10511493682861328 + }, + { + "epoch": 146, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.024299849601421092, + "train_seconds": 246.4877734184265, + "validation": { + "accuracy": 0.9064, + "examples": 5000, + "loss": 0.4284990051269531 + }, + "validation_seconds": 0.10503721237182617 + }, + { + "epoch": 147, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.02203905286722713, + "train_seconds": 246.4707317352295, + "validation": { + "accuracy": 0.9056, + "examples": 5000, + "loss": 0.43494686279296874 + }, + "validation_seconds": 0.10479950904846191 + }, + { + "epoch": 148, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0230322131584088, + "train_seconds": 246.5043556690216, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4222664306640625 + }, + "validation_seconds": 0.1047523021697998 + }, + { + "epoch": 149, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.021366101835833654, + "train_seconds": 246.5034623146057, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.4292074523925781 + }, + "validation_seconds": 0.10499930381774902 + }, + { + "epoch": 150, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.021379361391067504, + "train_seconds": 246.49155592918396, + "validation": { + "accuracy": 0.9062, + "examples": 5000, + "loss": 0.4323758117675781 + }, + "validation_seconds": 0.10504913330078125 + }, + { + "epoch": 151, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.019433662343025206, + "train_seconds": 246.48384475708008, + "validation": { + "accuracy": 0.9076, + "examples": 5000, + "loss": 0.4278642761230469 + }, + "validation_seconds": 0.10531210899353027 + }, + { + "epoch": 152, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.019419856379098363, + "train_seconds": 246.4657862186432, + "validation": { + "accuracy": 0.908, + "examples": 5000, + "loss": 0.4250206359863281 + }, + "validation_seconds": 0.10491490364074707 + }, + { + "epoch": 153, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01779731034901407, + "train_seconds": 246.51931881904602, + "validation": { + "accuracy": 0.907, + "examples": 5000, + "loss": 0.4278398681640625 + }, + "validation_seconds": 0.10487937927246094 + }, + { + "epoch": 154, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01879941195083989, + "train_seconds": 246.67745804786682, + "validation": { + "accuracy": 0.9066, + "examples": 5000, + "loss": 0.42561909790039065 + }, + "validation_seconds": 0.10490798950195312 + }, + { + "epoch": 155, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017670769177211654, + "train_seconds": 246.60387182235718, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.42608068237304686 + }, + "validation_seconds": 0.10499191284179688 + }, + { + "epoch": 156, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01898041370842192, + "train_seconds": 246.49284768104553, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.42472913818359376 + }, + "validation_seconds": 0.10478997230529785 + }, + { + "epoch": 157, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01866564043296708, + "train_seconds": 246.5001609325409, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42693660888671875 + }, + "validation_seconds": 0.10493803024291992 + }, + { + "epoch": 158, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.018088148384003175, + "train_seconds": 246.44772601127625, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.424379541015625 + }, + "validation_seconds": 0.1050882339477539 + }, + { + "epoch": 159, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01760909207902021, + "train_seconds": 246.48644709587097, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.4213317810058594 + }, + "validation_seconds": 0.10483264923095703 + }, + { + "epoch": 160, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01838235599050919, + "train_seconds": 246.49702405929565, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4235311767578125 + }, + "validation_seconds": 0.10515522956848145 + }, + { + "epoch": 161, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017624795015901328, + "train_seconds": 246.47225213050842, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42268819580078126 + }, + "validation_seconds": 0.10508155822753906 + }, + { + "epoch": 162, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016919317315518857, + "train_seconds": 246.48047471046448, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.42747427368164065 + }, + "validation_seconds": 0.10546684265136719 + }, + { + "epoch": 163, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.018289856963025198, + "train_seconds": 246.46548342704773, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.424314111328125 + }, + "validation_seconds": 0.10488176345825195 + }, + { + "epoch": 164, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01622216137184037, + "train_seconds": 246.47679567337036, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.42815198974609375 + }, + "validation_seconds": 0.10511016845703125 + }, + { + "epoch": 165, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01683462748808993, + "train_seconds": 246.48844981193542, + "validation": { + "accuracy": 0.9082, + "examples": 5000, + "loss": 0.42897827758789064 + }, + "validation_seconds": 0.10507750511169434 + }, + { + "epoch": 166, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01676006626519892, + "train_seconds": 246.48650288581848, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.4275100830078125 + }, + "validation_seconds": 0.1049661636352539 + }, + { + "epoch": 167, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016608415773842068, + "train_seconds": 246.47876834869385, + "validation": { + "accuracy": 0.908, + "examples": 5000, + "loss": 0.4286995666503906 + }, + "validation_seconds": 0.10474586486816406 + }, + { + "epoch": 168, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017247727344764603, + "train_seconds": 246.4671652317047, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.4272683166503906 + }, + "validation_seconds": 0.10498309135437012 + }, + { + "epoch": 169, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016528946716917885, + "train_seconds": 246.58322262763977, + "validation": { + "accuracy": 0.9106, + "examples": 5000, + "loss": 0.4269004150390625 + }, + "validation_seconds": 0.10540938377380371 + }, + { + "epoch": 170, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016474853175216252, + "train_seconds": 246.69525337219238, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.42872039794921873 + }, + "validation_seconds": 0.10507774353027344 + }, + { + "epoch": 171, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016024321196476617, + "train_seconds": 246.60345554351807, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4304836059570312 + }, + "validation_seconds": 0.10527491569519043 + }, + { + "epoch": 172, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017347405864298344, + "train_seconds": 246.6121952533722, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.42888103637695313 + }, + "validation_seconds": 0.10518264770507812 + }, + { + "epoch": 173, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016893391016125678, + "train_seconds": 246.6187880039215, + "validation": { + "accuracy": 0.9098, + "examples": 5000, + "loss": 0.4287686279296875 + }, + "validation_seconds": 0.1053009033203125 + }, + { + "epoch": 174, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016590854315045808, + "train_seconds": 246.58989000320435, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.42554988403320315 + }, + "validation_seconds": 0.10519289970397949 + }, + { + "epoch": 175, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01584643976688385, + "train_seconds": 246.5360791683197, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.425059716796875 + }, + "validation_seconds": 0.10481643676757812 + }, + { + "epoch": 176, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016978252725137605, + "train_seconds": 246.5840287208557, + "validation": { + "accuracy": 0.9094, + "examples": 5000, + "loss": 0.4239813842773438 + }, + "validation_seconds": 0.1047203540802002 + }, + { + "epoch": 177, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015669997653530703, + "train_seconds": 246.60086369514465, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.4284854431152344 + }, + "validation_seconds": 0.10500645637512207 + }, + { + "epoch": 178, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017344997499841783, + "train_seconds": 246.62754106521606, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4284616516113281 + }, + "validation_seconds": 0.10516667366027832 + }, + { + "epoch": 179, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016015476820866268, + "train_seconds": 246.58459210395813, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.4248768615722656 + }, + "validation_seconds": 0.1052248477935791 + }, + { + "epoch": 180, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016324492257336776, + "train_seconds": 246.54819989204407, + "validation": { + "accuracy": 0.9108, + "examples": 5000, + "loss": 0.42777464599609377 + }, + "validation_seconds": 0.10559630393981934 + }, + { + "epoch": 181, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01666753611365954, + "train_seconds": 246.57083463668823, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.42751864013671875 + }, + "validation_seconds": 0.10503649711608887 + }, + { + "epoch": 182, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016868565215335952, + "train_seconds": 246.51821660995483, + "validation": { + "accuracy": 0.91, + "examples": 5000, + "loss": 0.43049243774414064 + }, + "validation_seconds": 0.1048879623413086 + }, + { + "epoch": 183, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017401854633788266, + "train_seconds": 246.54656767845154, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.43035255126953126 + }, + "validation_seconds": 0.10518288612365723 + }, + { + "epoch": 184, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.01609395571152369, + "train_seconds": 246.47771430015564, + "validation": { + "accuracy": 0.9088, + "examples": 5000, + "loss": 0.43076982421875 + }, + "validation_seconds": 0.10487246513366699 + }, + { + "epoch": 185, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015626564153035483, + "train_seconds": 246.0274088382721, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.43607783813476564 + }, + "validation_seconds": 0.10477232933044434 + }, + { + "epoch": 186, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016511785845789644, + "train_seconds": 246.0552146434784, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.43122318115234376 + }, + "validation_seconds": 0.10496950149536133 + }, + { + "epoch": 187, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016176714078916445, + "train_seconds": 246.01906418800354, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.43309248046875 + }, + "validation_seconds": 0.10488486289978027 + }, + { + "epoch": 188, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017060648574762874, + "train_seconds": 246.00013732910156, + "validation": { + "accuracy": 0.9078, + "examples": 5000, + "loss": 0.43197459106445313 + }, + "validation_seconds": 0.1047358512878418 + }, + { + "epoch": 189, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015920067586998144, + "train_seconds": 246.03866243362427, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.42890040283203124 + }, + "validation_seconds": 0.10486721992492676 + }, + { + "epoch": 190, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015953422194719315, + "train_seconds": 246.05727362632751, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42850841064453127 + }, + "validation_seconds": 0.1049807071685791 + }, + { + "epoch": 191, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015881843958629502, + "train_seconds": 246.0005452632904, + "validation": { + "accuracy": 0.9084, + "examples": 5000, + "loss": 0.4348256774902344 + }, + "validation_seconds": 0.10516357421875 + }, + { + "epoch": 192, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.014872584097584088, + "train_seconds": 246.02225160598755, + "validation": { + "accuracy": 0.9092, + "examples": 5000, + "loss": 0.43110437622070313 + }, + "validation_seconds": 0.10500121116638184 + }, + { + "epoch": 193, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016930411442783143, + "train_seconds": 246.41416931152344, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.4323291320800781 + }, + "validation_seconds": 0.1049497127532959 + }, + { + "epoch": 194, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015967371432648765, + "train_seconds": 246.44947600364685, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.4302525146484375 + }, + "validation_seconds": 0.10496187210083008 + }, + { + "epoch": 195, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015971045525537596, + "train_seconds": 246.44577145576477, + "validation": { + "accuracy": 0.9086, + "examples": 5000, + "loss": 0.42978770141601563 + }, + "validation_seconds": 0.10512423515319824 + }, + { + "epoch": 196, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015177000864677959, + "train_seconds": 246.432710647583, + "validation": { + "accuracy": 0.909, + "examples": 5000, + "loss": 0.432722900390625 + }, + "validation_seconds": 0.10512948036193848 + }, + { + "epoch": 197, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.017064794731885196, + "train_seconds": 246.43957209587097, + "validation": { + "accuracy": 0.9098, + "examples": 5000, + "loss": 0.43294736938476563 + }, + "validation_seconds": 0.10485196113586426 + }, + { + "epoch": 198, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015426826933191883, + "train_seconds": 246.46257519721985, + "validation": { + "accuracy": 0.9102, + "examples": 5000, + "loss": 0.4334692504882813 + }, + "validation_seconds": 0.10505104064941406 + }, + { + "epoch": 199, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.016626012229753864, + "train_seconds": 246.4922273159027, + "validation": { + "accuracy": 0.9096, + "examples": 5000, + "loss": 0.430858544921875 + }, + "validation_seconds": 0.10512781143188477 + }, + { + "epoch": 200, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.015557368935479057, + "train_seconds": 246.5367796421051, + "validation": { + "accuracy": 0.9104, + "examples": 5000, + "loss": 0.43468799438476563 + }, + "validation_seconds": 0.10501646995544434 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.9104, + "evaluation_seconds": 0.10538721084594727, + "examples": 5000, + "finite": true, + "loss": 0.43468799438476563 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2008608768, + "peak_memory_reserved_bytes": 2808086528, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 49286.44217157364, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 200, + "feedforward_example_passes": 10005000, + "forward_macs_per_example": 40551040, + "forward_parameter_count": 269722, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 144000000, + "logical_task_loss_queries": 0, + "num_trainable_layers": 20, + "ordinary_training_examples": 9000000, + "ordinary_validation_examples": 1005000, + "relaxation_example_passes": 144000000, + "training_example_presentations": 9000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json.manifest.json new file mode 100644 index 0000000..48aff8e --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json.manifest.json @@ -0,0 +1,132 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "dualprop", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.05", + "--output_lr", + "0.05", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--dp_alpha", + "0.0", + "--dp_beta", + "0.1", + "--dp_inference_passes", + "16" + ], + "completed_unix_time": 1786999539.905762, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 49289.51700615883, + "epochs": 200, + "experiment_name": "resnet-r2-dualprop-d20-lr0p05", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "dualprop", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d20-lr0p05.json", + "output_exists": true, + "output_sha256": "62576fdd2cce8c465f80416e8eba0f449a025fb0c04e3d13b6aacf6ef5699cd3", + "rate": 0.05, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json b/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json new file mode 100644 index 0000000..31332f5 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json @@ -0,0 +1,2965 @@ +{ + "architecture": { + "base_width": 16, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 464154, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 200, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.05, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "method": "dualprop", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json", + "output_lr": 0.05, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.8004995579401653, + "train_seconds": 563.7639303207397, + "validation": { + "accuracy": 0.4328, + "examples": 5000, + "loss": 1.557549609375 + }, + "validation_seconds": 0.19327449798583984 + }, + { + "epoch": 2, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.4176780060662164, + "train_seconds": 558.1841831207275, + "validation": { + "accuracy": 0.4978, + "examples": 5000, + "loss": 1.464878515625 + }, + "validation_seconds": 0.17017483711242676 + }, + { + "epoch": 3, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.2581868873596191, + "train_seconds": 558.1408138275146, + "validation": { + "accuracy": 0.5734, + "examples": 5000, + "loss": 1.216384326171875 + }, + "validation_seconds": 0.17085552215576172 + }, + { + "epoch": 4, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.1536001446617974, + "train_seconds": 558.4430482387543, + "validation": { + "accuracy": 0.605, + "examples": 5000, + "loss": 1.1569667236328125 + }, + "validation_seconds": 0.17072129249572754 + }, + { + "epoch": 5, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 1.0694282343122694, + "train_seconds": 558.2857687473297, + "validation": { + "accuracy": 0.5586, + "examples": 5000, + "loss": 1.3289294189453125 + }, + "validation_seconds": 0.1704881191253662 + }, + { + "epoch": 6, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.9969480787065294, + "train_seconds": 558.2259268760681, + "validation": { + "accuracy": 0.6168, + "examples": 5000, + "loss": 1.1593406494140626 + }, + "validation_seconds": 0.17039990425109863 + }, + { + "epoch": 7, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.9467010908550686, + "train_seconds": 558.310210943222, + "validation": { + "accuracy": 0.672, + "examples": 5000, + "loss": 0.9539438110351562 + }, + "validation_seconds": 0.17075109481811523 + }, + { + "epoch": 8, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.9019010075887044, + "train_seconds": 558.3344357013702, + "validation": { + "accuracy": 0.6492, + "examples": 5000, + "loss": 0.9964987182617188 + }, + "validation_seconds": 0.1706087589263916 + }, + { + "epoch": 9, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.8510134382459852, + "train_seconds": 558.2307453155518, + "validation": { + "accuracy": 0.6824, + "examples": 5000, + "loss": 0.8969764892578125 + }, + "validation_seconds": 0.17046070098876953 + }, + { + "epoch": 10, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.8201242864820693, + "train_seconds": 558.233719587326, + "validation": { + "accuracy": 0.6836, + "examples": 5000, + "loss": 0.9454015991210938 + }, + "validation_seconds": 0.170670747756958 + }, + { + "epoch": 11, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7837703473091125, + "train_seconds": 558.2044539451599, + "validation": { + "accuracy": 0.7118, + "examples": 5000, + "loss": 0.86032119140625 + }, + "validation_seconds": 0.17042040824890137 + }, + { + "epoch": 12, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7580736145867242, + "train_seconds": 557.9444570541382, + "validation": { + "accuracy": 0.7284, + "examples": 5000, + "loss": 0.7895968017578125 + }, + "validation_seconds": 0.1704864501953125 + }, + { + "epoch": 13, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7373642958958944, + "train_seconds": 557.9286274909973, + "validation": { + "accuracy": 0.679, + "examples": 5000, + "loss": 0.9507829223632812 + }, + "validation_seconds": 0.17061734199523926 + }, + { + "epoch": 14, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.7064299224853515, + "train_seconds": 557.9418158531189, + "validation": { + "accuracy": 0.7248, + "examples": 5000, + "loss": 0.7915084350585937 + }, + "validation_seconds": 0.17058253288269043 + }, + { + "epoch": 15, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6900307644420199, + "train_seconds": 557.9299385547638, + "validation": { + "accuracy": 0.7308, + "examples": 5000, + "loss": 0.8004146240234375 + }, + "validation_seconds": 0.1706540584564209 + }, + { + "epoch": 16, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.673870362080468, + "train_seconds": 557.9290549755096, + "validation": { + "accuracy": 0.715, + "examples": 5000, + "loss": 0.8556584228515625 + }, + "validation_seconds": 0.17044925689697266 + }, + { + "epoch": 17, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6590127436743842, + "train_seconds": 557.9305055141449, + "validation": { + "accuracy": 0.7504, + "examples": 5000, + "loss": 0.7366830810546875 + }, + "validation_seconds": 0.17050766944885254 + }, + { + "epoch": 18, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6435344869931539, + "train_seconds": 557.9347696304321, + "validation": { + "accuracy": 0.736, + "examples": 5000, + "loss": 0.7978809814453125 + }, + "validation_seconds": 0.17049336433410645 + }, + { + "epoch": 19, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6259501564449734, + "train_seconds": 557.9711971282959, + "validation": { + "accuracy": 0.761, + "examples": 5000, + "loss": 0.7134017700195312 + }, + "validation_seconds": 0.1704099178314209 + }, + { + "epoch": 20, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.6067184276686775, + "train_seconds": 557.9173040390015, + "validation": { + "accuracy": 0.7484, + "examples": 5000, + "loss": 0.754651025390625 + }, + "validation_seconds": 0.17058968544006348 + }, + { + "epoch": 21, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5990790719456143, + "train_seconds": 557.9328565597534, + "validation": { + "accuracy": 0.7502, + "examples": 5000, + "loss": 0.7732291381835937 + }, + "validation_seconds": 0.17064690589904785 + }, + { + "epoch": 22, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5874932703706953, + "train_seconds": 558.1986618041992, + "validation": { + "accuracy": 0.7624, + "examples": 5000, + "loss": 0.7070720458984375 + }, + "validation_seconds": 0.17105531692504883 + }, + { + "epoch": 23, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5748521372265286, + "train_seconds": 558.8032100200653, + "validation": { + "accuracy": 0.7744, + "examples": 5000, + "loss": 0.6826259155273438 + }, + "validation_seconds": 0.1711256504058838 + }, + { + "epoch": 24, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5692189562638601, + "train_seconds": 559.0503284931183, + "validation": { + "accuracy": 0.7674, + "examples": 5000, + "loss": 0.7043788940429687 + }, + "validation_seconds": 0.17097234725952148 + }, + { + "epoch": 25, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5517877299308777, + "train_seconds": 558.8799324035645, + "validation": { + "accuracy": 0.7764, + "examples": 5000, + "loss": 0.6821715209960938 + }, + "validation_seconds": 0.17085647583007812 + }, + { + "epoch": 26, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5391471517350939, + "train_seconds": 558.8605720996857, + "validation": { + "accuracy": 0.7406, + "examples": 5000, + "loss": 0.8436376220703125 + }, + "validation_seconds": 0.17091798782348633 + }, + { + "epoch": 27, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5402941966957516, + "train_seconds": 558.9005839824677, + "validation": { + "accuracy": 0.7802, + "examples": 5000, + "loss": 0.6626584350585938 + }, + "validation_seconds": 0.17087268829345703 + }, + { + "epoch": 28, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5282208977381389, + "train_seconds": 558.775890827179, + "validation": { + "accuracy": 0.7828, + "examples": 5000, + "loss": 0.6616478881835938 + }, + "validation_seconds": 0.1705186367034912 + }, + { + "epoch": 29, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5246573151694404, + "train_seconds": 558.2551503181458, + "validation": { + "accuracy": 0.7682, + "examples": 5000, + "loss": 0.7028174072265625 + }, + "validation_seconds": 0.17051959037780762 + }, + { + "epoch": 30, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5210510110855102, + "train_seconds": 558.6386096477509, + "validation": { + "accuracy": 0.7812, + "examples": 5000, + "loss": 0.6577355712890625 + }, + "validation_seconds": 0.1710948944091797 + }, + { + "epoch": 31, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5098838945123885, + "train_seconds": 558.713506937027, + "validation": { + "accuracy": 0.7962, + "examples": 5000, + "loss": 0.6501040649414063 + }, + "validation_seconds": 0.17058014869689941 + }, + { + "epoch": 32, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.5057243042204115, + "train_seconds": 558.045622587204, + "validation": { + "accuracy": 0.7628, + "examples": 5000, + "loss": 0.725605029296875 + }, + "validation_seconds": 0.17093896865844727 + }, + { + "epoch": 33, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.49384519081115724, + "train_seconds": 558.0422115325928, + "validation": { + "accuracy": 0.795, + "examples": 5000, + "loss": 0.6151902099609375 + }, + "validation_seconds": 0.17077398300170898 + }, + { + "epoch": 34, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4975082278781467, + "train_seconds": 558.0692346096039, + "validation": { + "accuracy": 0.7904, + "examples": 5000, + "loss": 0.6437848266601562 + }, + "validation_seconds": 0.17070817947387695 + }, + { + "epoch": 35, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4914001432524787, + "train_seconds": 558.0532667636871, + "validation": { + "accuracy": 0.7844, + "examples": 5000, + "loss": 0.6282599609375 + }, + "validation_seconds": 0.1710524559020996 + }, + { + "epoch": 36, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.48799099100960625, + "train_seconds": 558.0403628349304, + "validation": { + "accuracy": 0.7858, + "examples": 5000, + "loss": 0.664295068359375 + }, + "validation_seconds": 0.17072486877441406 + }, + { + "epoch": 37, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.48119476664861044, + "train_seconds": 558.0373411178589, + "validation": { + "accuracy": 0.7984, + "examples": 5000, + "loss": 0.6241986450195313 + }, + "validation_seconds": 0.17064595222473145 + }, + { + "epoch": 38, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4726532228257921, + "train_seconds": 558.2699501514435, + "validation": { + "accuracy": 0.755, + "examples": 5000, + "loss": 0.750040283203125 + }, + "validation_seconds": 0.17100858688354492 + }, + { + "epoch": 39, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.45894244771533543, + "train_seconds": 558.2855563163757, + "validation": { + "accuracy": 0.8098, + "examples": 5000, + "loss": 0.5754578002929688 + }, + "validation_seconds": 0.1708691120147705 + }, + { + "epoch": 40, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4561460572242737, + "train_seconds": 558.2716417312622, + "validation": { + "accuracy": 0.7994, + "examples": 5000, + "loss": 0.6093698974609375 + }, + "validation_seconds": 0.17070937156677246 + }, + { + "epoch": 41, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.45154505163828534, + "train_seconds": 558.3074398040771, + "validation": { + "accuracy": 0.8016, + "examples": 5000, + "loss": 0.58370234375 + }, + "validation_seconds": 0.17050838470458984 + }, + { + "epoch": 42, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.44932130374908447, + "train_seconds": 558.3013200759888, + "validation": { + "accuracy": 0.7796, + "examples": 5000, + "loss": 0.6969513793945312 + }, + "validation_seconds": 0.17067265510559082 + }, + { + "epoch": 43, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4478407272444831, + "train_seconds": 558.3348424434662, + "validation": { + "accuracy": 0.794, + "examples": 5000, + "loss": 0.6354197021484375 + }, + "validation_seconds": 0.1706705093383789 + }, + { + "epoch": 44, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4487147048632304, + "train_seconds": 558.3120241165161, + "validation": { + "accuracy": 0.7848, + "examples": 5000, + "loss": 0.6651848510742188 + }, + "validation_seconds": 0.17068982124328613 + }, + { + "epoch": 45, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.44171588740878637, + "train_seconds": 558.7507667541504, + "validation": { + "accuracy": 0.7996, + "examples": 5000, + "loss": 0.6317643798828125 + }, + "validation_seconds": 0.17115306854248047 + }, + { + "epoch": 46, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.44115461015701296, + "train_seconds": 558.9619481563568, + "validation": { + "accuracy": 0.8178, + "examples": 5000, + "loss": 0.5661478759765625 + }, + "validation_seconds": 0.17086172103881836 + }, + { + "epoch": 47, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.42366895433531865, + "train_seconds": 558.5156924724579, + "validation": { + "accuracy": 0.8084, + "examples": 5000, + "loss": 0.6045284912109375 + }, + "validation_seconds": 0.17072129249572754 + }, + { + "epoch": 48, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4294327813837263, + "train_seconds": 558.0816631317139, + "validation": { + "accuracy": 0.8078, + "examples": 5000, + "loss": 0.6047329711914062 + }, + "validation_seconds": 0.17075610160827637 + }, + { + "epoch": 49, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.42285685943497553, + "train_seconds": 558.0882160663605, + "validation": { + "accuracy": 0.7926, + "examples": 5000, + "loss": 0.6357721313476562 + }, + "validation_seconds": 0.17081046104431152 + }, + { + "epoch": 50, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.42114123702579076, + "train_seconds": 558.0787107944489, + "validation": { + "accuracy": 0.7966, + "examples": 5000, + "loss": 0.62530205078125 + }, + "validation_seconds": 0.17049264907836914 + }, + { + "epoch": 51, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4282205765194363, + "train_seconds": 558.0904269218445, + "validation": { + "accuracy": 0.7872, + "examples": 5000, + "loss": 0.6708918823242187 + }, + "validation_seconds": 0.17061495780944824 + }, + { + "epoch": 52, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4141006850772434, + "train_seconds": 558.082554101944, + "validation": { + "accuracy": 0.8066, + "examples": 5000, + "loss": 0.5993174438476563 + }, + "validation_seconds": 0.17061448097229004 + }, + { + "epoch": 53, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4119393820762634, + "train_seconds": 558.0787973403931, + "validation": { + "accuracy": 0.8058, + "examples": 5000, + "loss": 0.6037708984375 + }, + "validation_seconds": 0.1706247329711914 + }, + { + "epoch": 54, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4145192951838175, + "train_seconds": 558.0709445476532, + "validation": { + "accuracy": 0.7988, + "examples": 5000, + "loss": 0.6170823974609375 + }, + "validation_seconds": 0.17055344581604004 + }, + { + "epoch": 55, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4110671861436632, + "train_seconds": 558.0375108718872, + "validation": { + "accuracy": 0.7954, + "examples": 5000, + "loss": 0.64954423828125 + }, + "validation_seconds": 0.17034626007080078 + }, + { + "epoch": 56, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4194034888267517, + "train_seconds": 558.2900795936584, + "validation": { + "accuracy": 0.7774, + "examples": 5000, + "loss": 0.6947837890625 + }, + "validation_seconds": 0.17063689231872559 + }, + { + "epoch": 57, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.40363140162361993, + "train_seconds": 558.2407958507538, + "validation": { + "accuracy": 0.7928, + "examples": 5000, + "loss": 0.6374149169921876 + }, + "validation_seconds": 0.17067933082580566 + }, + { + "epoch": 58, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.4071178272035387, + "train_seconds": 558.2327244281769, + "validation": { + "accuracy": 0.8138, + "examples": 5000, + "loss": 0.5728923095703125 + }, + "validation_seconds": 0.17045927047729492 + }, + { + "epoch": 59, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.40012233917978074, + "train_seconds": 558.2001922130585, + "validation": { + "accuracy": 0.7878, + "examples": 5000, + "loss": 0.6841786987304688 + }, + "validation_seconds": 0.17034292221069336 + }, + { + "epoch": 60, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3900605029688941, + "train_seconds": 558.22704911232, + "validation": { + "accuracy": 0.8002, + "examples": 5000, + "loss": 0.6395064331054687 + }, + "validation_seconds": 0.17066001892089844 + }, + { + "epoch": 61, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.38680489319165545, + "train_seconds": 558.1992979049683, + "validation": { + "accuracy": 0.7962, + "examples": 5000, + "loss": 0.6516251831054688 + }, + "validation_seconds": 0.1707756519317627 + }, + { + "epoch": 62, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.37478828299310474, + "train_seconds": 558.2286744117737, + "validation": { + "accuracy": 0.813, + "examples": 5000, + "loss": 0.5861128295898438 + }, + "validation_seconds": 0.17053627967834473 + }, + { + "epoch": 63, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3731063287470076, + "train_seconds": 558.2254464626312, + "validation": { + "accuracy": 0.8112, + "examples": 5000, + "loss": 0.6001865600585937 + }, + "validation_seconds": 0.17064380645751953 + }, + { + "epoch": 64, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3759114433394538, + "train_seconds": 558.2426578998566, + "validation": { + "accuracy": 0.8086, + "examples": 5000, + "loss": 0.6028027587890625 + }, + "validation_seconds": 0.1704556941986084 + }, + { + "epoch": 65, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.37589247471491494, + "train_seconds": 558.2081379890442, + "validation": { + "accuracy": 0.8192, + "examples": 5000, + "loss": 0.573157470703125 + }, + "validation_seconds": 0.17031216621398926 + }, + { + "epoch": 66, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.37520403162903254, + "train_seconds": 557.9686496257782, + "validation": { + "accuracy": 0.807, + "examples": 5000, + "loss": 0.6031291137695313 + }, + "validation_seconds": 0.17033934593200684 + }, + { + "epoch": 67, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.37030814722379046, + "train_seconds": 557.9458205699921, + "validation": { + "accuracy": 0.8084, + "examples": 5000, + "loss": 0.61266845703125 + }, + "validation_seconds": 0.17030119895935059 + }, + { + "epoch": 68, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3623477321412828, + "train_seconds": 558.0508515834808, + "validation": { + "accuracy": 0.8238, + "examples": 5000, + "loss": 0.541462060546875 + }, + "validation_seconds": 0.17065834999084473 + }, + { + "epoch": 69, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3633347564008501, + "train_seconds": 558.0455734729767, + "validation": { + "accuracy": 0.8018, + "examples": 5000, + "loss": 0.6289015258789062 + }, + "validation_seconds": 0.17070508003234863 + }, + { + "epoch": 70, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.35279572947820026, + "train_seconds": 558.0403199195862, + "validation": { + "accuracy": 0.8192, + "examples": 5000, + "loss": 0.5510507446289062 + }, + "validation_seconds": 0.1707603931427002 + }, + { + "epoch": 71, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.35029824435975815, + "train_seconds": 558.0362615585327, + "validation": { + "accuracy": 0.8078, + "examples": 5000, + "loss": 0.6152504028320313 + }, + "validation_seconds": 0.17045378684997559 + }, + { + "epoch": 72, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.36111382049984403, + "train_seconds": 558.0379478931427, + "validation": { + "accuracy": 0.7906, + "examples": 5000, + "loss": 0.6859040771484375 + }, + "validation_seconds": 0.17042279243469238 + }, + { + "epoch": 73, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3539845959769355, + "train_seconds": 558.021562576294, + "validation": { + "accuracy": 0.8212, + "examples": 5000, + "loss": 0.5775150268554687 + }, + "validation_seconds": 0.1705188751220703 + }, + { + "epoch": 74, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3474092450671726, + "train_seconds": 558.0308134555817, + "validation": { + "accuracy": 0.8092, + "examples": 5000, + "loss": 0.6013641967773438 + }, + "validation_seconds": 0.17047381401062012 + }, + { + "epoch": 75, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.35469483111169603, + "train_seconds": 558.0403017997742, + "validation": { + "accuracy": 0.8254, + "examples": 5000, + "loss": 0.54343388671875 + }, + "validation_seconds": 0.17022395133972168 + }, + { + "epoch": 76, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3471669368902842, + "train_seconds": 558.0478513240814, + "validation": { + "accuracy": 0.807, + "examples": 5000, + "loss": 0.631876904296875 + }, + "validation_seconds": 0.1706843376159668 + }, + { + "epoch": 77, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3457748140070174, + "train_seconds": 557.9572274684906, + "validation": { + "accuracy": 0.8212, + "examples": 5000, + "loss": 0.5879268798828124 + }, + "validation_seconds": 0.17055583000183105 + }, + { + "epoch": 78, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.33029527435302736, + "train_seconds": 557.9223172664642, + "validation": { + "accuracy": 0.828, + "examples": 5000, + "loss": 0.5529513061523438 + }, + "validation_seconds": 0.1705477237701416 + }, + { + "epoch": 79, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.33166916541523406, + "train_seconds": 557.9095616340637, + "validation": { + "accuracy": 0.7824, + "examples": 5000, + "loss": 0.6858224365234376 + }, + "validation_seconds": 0.17069387435913086 + }, + { + "epoch": 80, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.33432486742072637, + "train_seconds": 557.9043171405792, + "validation": { + "accuracy": 0.8154, + "examples": 5000, + "loss": 0.5964239135742188 + }, + "validation_seconds": 0.1705014705657959 + }, + { + "epoch": 81, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.32564821405410765, + "train_seconds": 557.9098527431488, + "validation": { + "accuracy": 0.8184, + "examples": 5000, + "loss": 0.564143408203125 + }, + "validation_seconds": 0.1705763339996338 + }, + { + "epoch": 82, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.31590132386419506, + "train_seconds": 557.9197072982788, + "validation": { + "accuracy": 0.8242, + "examples": 5000, + "loss": 0.5693760864257813 + }, + "validation_seconds": 0.17065834999084473 + }, + { + "epoch": 83, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3159837660683526, + "train_seconds": 557.9388198852539, + "validation": { + "accuracy": 0.8126, + "examples": 5000, + "loss": 0.6296174926757813 + }, + "validation_seconds": 0.17052793502807617 + }, + { + "epoch": 84, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.32203944091796877, + "train_seconds": 557.9468953609467, + "validation": { + "accuracy": 0.806, + "examples": 5000, + "loss": 0.6376169067382812 + }, + "validation_seconds": 0.17077326774597168 + }, + { + "epoch": 85, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.32088813879224987, + "train_seconds": 557.9684026241302, + "validation": { + "accuracy": 0.8202, + "examples": 5000, + "loss": 0.5882717407226562 + }, + "validation_seconds": 0.17091774940490723 + }, + { + "epoch": 86, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.315234685060713, + "train_seconds": 557.9197709560394, + "validation": { + "accuracy": 0.829, + "examples": 5000, + "loss": 0.5767408813476562 + }, + "validation_seconds": 0.17058467864990234 + }, + { + "epoch": 87, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3200781061384413, + "train_seconds": 557.932461977005, + "validation": { + "accuracy": 0.8266, + "examples": 5000, + "loss": 0.5719342407226562 + }, + "validation_seconds": 0.17056751251220703 + }, + { + "epoch": 88, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.32161954258812797, + "train_seconds": 557.9289646148682, + "validation": { + "accuracy": 0.8242, + "examples": 5000, + "loss": 0.5660660522460937 + }, + "validation_seconds": 0.17087936401367188 + }, + { + "epoch": 89, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.31135784103075664, + "train_seconds": 557.9694781303406, + "validation": { + "accuracy": 0.7986, + "examples": 5000, + "loss": 0.6810950317382812 + }, + "validation_seconds": 0.17078852653503418 + }, + { + "epoch": 90, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.31059243212805854, + "train_seconds": 557.9276149272919, + "validation": { + "accuracy": 0.8162, + "examples": 5000, + "loss": 0.6226161254882813 + }, + "validation_seconds": 0.17064809799194336 + }, + { + "epoch": 91, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.303583119614919, + "train_seconds": 557.93887758255, + "validation": { + "accuracy": 0.8278, + "examples": 5000, + "loss": 0.575564697265625 + }, + "validation_seconds": 0.17057085037231445 + }, + { + "epoch": 92, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.2989853093253242, + "train_seconds": 557.9392566680908, + "validation": { + "accuracy": 0.8348, + "examples": 5000, + "loss": 0.5497850830078125 + }, + "validation_seconds": 0.17045283317565918 + }, + { + "epoch": 93, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.30290545211368136, + "train_seconds": 558.1350507736206, + "validation": { + "accuracy": 0.829, + "examples": 5000, + "loss": 0.564298681640625 + }, + "validation_seconds": 0.17111921310424805 + }, + { + "epoch": 94, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.31313663847181533, + "train_seconds": 559.4218375682831, + "validation": { + "accuracy": 0.8368, + "examples": 5000, + "loss": 0.5624496704101563 + }, + "validation_seconds": 0.17076325416564941 + }, + { + "epoch": 95, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.29203504636817507, + "train_seconds": 558.3938927650452, + "validation": { + "accuracy": 0.8092, + "examples": 5000, + "loss": 0.6693233520507812 + }, + "validation_seconds": 0.17063307762145996 + }, + { + "epoch": 96, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3094335092279646, + "train_seconds": 558.233589887619, + "validation": { + "accuracy": 0.8352, + "examples": 5000, + "loss": 0.5715654907226563 + }, + "validation_seconds": 0.1705608367919922 + }, + { + "epoch": 97, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.3069044820573595, + "train_seconds": 558.2597603797913, + "validation": { + "accuracy": 0.8294, + "examples": 5000, + "loss": 0.5501206298828125 + }, + "validation_seconds": 0.1707000732421875 + }, + { + "epoch": 98, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.30671100243462457, + "train_seconds": 558.1544392108917, + "validation": { + "accuracy": 0.8368, + "examples": 5000, + "loss": 0.5354385192871094 + }, + "validation_seconds": 0.17098045349121094 + }, + { + "epoch": 99, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.30218536727163525, + "train_seconds": 557.9338235855103, + "validation": { + "accuracy": 0.8324, + "examples": 5000, + "loss": 0.5434078918457032 + }, + "validation_seconds": 0.17069149017333984 + }, + { + "epoch": 100, + "lr": 0.05, + "output_lr": 0.05, + "train_examples": 45000, + "train_loss": 0.28684145905971525, + "train_seconds": 558.2469670772552, + "validation": { + "accuracy": 0.8372, + "examples": 5000, + "loss": 0.5330762023925781 + }, + "validation_seconds": 0.17123103141784668 + }, + { + "epoch": 101, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.20102937622732586, + "train_seconds": 560.1686954498291, + "validation": { + "accuracy": 0.8634, + "examples": 5000, + "loss": 0.4263051818847656 + }, + "validation_seconds": 0.1707172393798828 + }, + { + "epoch": 102, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.1707019625398848, + "train_seconds": 560.6133296489716, + "validation": { + "accuracy": 0.8636, + "examples": 5000, + "loss": 0.42772237548828124 + }, + "validation_seconds": 0.17063069343566895 + }, + { + "epoch": 103, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.16009664902024798, + "train_seconds": 560.8625993728638, + "validation": { + "accuracy": 0.8652, + "examples": 5000, + "loss": 0.4311399169921875 + }, + "validation_seconds": 0.17083454132080078 + }, + { + "epoch": 104, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.1511841663281123, + "train_seconds": 559.8500211238861, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.42962298583984376 + }, + "validation_seconds": 0.1706254482269287 + }, + { + "epoch": 105, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.1450319301393297, + "train_seconds": 559.5080921649933, + "validation": { + "accuracy": 0.8668, + "examples": 5000, + "loss": 0.43597459716796877 + }, + "validation_seconds": 0.17025041580200195 + }, + { + "epoch": 106, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.1385588534726037, + "train_seconds": 558.8428866863251, + "validation": { + "accuracy": 0.8688, + "examples": 5000, + "loss": 0.44206187744140624 + }, + "validation_seconds": 0.17053508758544922 + }, + { + "epoch": 107, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.1349012790547477, + "train_seconds": 558.7124433517456, + "validation": { + "accuracy": 0.8644, + "examples": 5000, + "loss": 0.44303953857421874 + }, + "validation_seconds": 0.1705188751220703 + }, + { + "epoch": 108, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.13633077380127376, + "train_seconds": 558.8155248165131, + "validation": { + "accuracy": 0.866, + "examples": 5000, + "loss": 0.4498537658691406 + }, + "validation_seconds": 0.17037415504455566 + }, + { + "epoch": 109, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.12727142566045124, + "train_seconds": 559.8420617580414, + "validation": { + "accuracy": 0.8666, + "examples": 5000, + "loss": 0.4562872497558594 + }, + "validation_seconds": 0.17095494270324707 + }, + { + "epoch": 110, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.12273772934277852, + "train_seconds": 560.2172212600708, + "validation": { + "accuracy": 0.869, + "examples": 5000, + "loss": 0.46257111206054685 + }, + "validation_seconds": 0.1706855297088623 + }, + { + "epoch": 111, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.1186966257651647, + "train_seconds": 559.0523889064789, + "validation": { + "accuracy": 0.8694, + "examples": 5000, + "loss": 0.46149122924804686 + }, + "validation_seconds": 0.1704082489013672 + }, + { + "epoch": 112, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.1151852872159746, + "train_seconds": 558.8536169528961, + "validation": { + "accuracy": 0.8686, + "examples": 5000, + "loss": 0.468364404296875 + }, + "validation_seconds": 0.17055869102478027 + }, + { + "epoch": 113, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.11532456085681915, + "train_seconds": 558.7769916057587, + "validation": { + "accuracy": 0.871, + "examples": 5000, + "loss": 0.46807992553710936 + }, + "validation_seconds": 0.17038846015930176 + }, + { + "epoch": 114, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.11298640092081494, + "train_seconds": 558.7878985404968, + "validation": { + "accuracy": 0.8696, + "examples": 5000, + "loss": 0.4718349487304688 + }, + "validation_seconds": 0.1701805591583252 + }, + { + "epoch": 115, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.11400147623750899, + "train_seconds": 558.6093730926514, + "validation": { + "accuracy": 0.8706, + "examples": 5000, + "loss": 0.4778865234375 + }, + "validation_seconds": 0.1703321933746338 + }, + { + "epoch": 116, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.10852504232459598, + "train_seconds": 558.69162774086, + "validation": { + "accuracy": 0.8702, + "examples": 5000, + "loss": 0.48102982788085935 + }, + "validation_seconds": 0.16998052597045898 + }, + { + "epoch": 117, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.10526954917377895, + "train_seconds": 558.6054136753082, + "validation": { + "accuracy": 0.8664, + "examples": 5000, + "loss": 0.48121639404296873 + }, + "validation_seconds": 0.17051982879638672 + }, + { + "epoch": 118, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.10441256948245896, + "train_seconds": 559.6752879619598, + "validation": { + "accuracy": 0.8698, + "examples": 5000, + "loss": 0.4906882080078125 + }, + "validation_seconds": 0.17052960395812988 + }, + { + "epoch": 119, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.10324348965088527, + "train_seconds": 560.4540963172913, + "validation": { + "accuracy": 0.8706, + "examples": 5000, + "loss": 0.49455729370117185 + }, + "validation_seconds": 0.1706228256225586 + }, + { + "epoch": 120, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.10344991790453593, + "train_seconds": 560.4171199798584, + "validation": { + "accuracy": 0.8678, + "examples": 5000, + "loss": 0.4905825256347656 + }, + "validation_seconds": 0.17073750495910645 + }, + { + "epoch": 121, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.09832412128316032, + "train_seconds": 560.4106814861298, + "validation": { + "accuracy": 0.871, + "examples": 5000, + "loss": 0.498076904296875 + }, + "validation_seconds": 0.1704578399658203 + }, + { + "epoch": 122, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.09772718087302314, + "train_seconds": 560.4327030181885, + "validation": { + "accuracy": 0.8688, + "examples": 5000, + "loss": 0.5086375427246094 + }, + "validation_seconds": 0.17103815078735352 + }, + { + "epoch": 123, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.10162255337900586, + "train_seconds": 560.3050816059113, + "validation": { + "accuracy": 0.8672, + "examples": 5000, + "loss": 0.5079470642089844 + }, + "validation_seconds": 0.17122697830200195 + }, + { + "epoch": 124, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.09721829018063015, + "train_seconds": 560.3977375030518, + "validation": { + "accuracy": 0.8696, + "examples": 5000, + "loss": 0.5183314514160157 + }, + "validation_seconds": 0.1710338592529297 + }, + { + "epoch": 125, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.093445472296079, + "train_seconds": 560.1785962581635, + "validation": { + "accuracy": 0.8664, + "examples": 5000, + "loss": 0.5179838073730468 + }, + "validation_seconds": 0.17027044296264648 + }, + { + "epoch": 126, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.09554820556905534, + "train_seconds": 560.3105957508087, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.5202593566894531 + }, + "validation_seconds": 0.1709599494934082 + }, + { + "epoch": 127, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0952698231961992, + "train_seconds": 560.3248987197876, + "validation": { + "accuracy": 0.8676, + "examples": 5000, + "loss": 0.5269124206542969 + }, + "validation_seconds": 0.17055344581604004 + }, + { + "epoch": 128, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08956516365210215, + "train_seconds": 560.349618434906, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.5317488220214843 + }, + "validation_seconds": 0.17071056365966797 + }, + { + "epoch": 129, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.09056334316995408, + "train_seconds": 560.2911763191223, + "validation": { + "accuracy": 0.866, + "examples": 5000, + "loss": 0.5350216003417969 + }, + "validation_seconds": 0.1710660457611084 + }, + { + "epoch": 130, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08700342583258946, + "train_seconds": 560.1963536739349, + "validation": { + "accuracy": 0.8686, + "examples": 5000, + "loss": 0.5305935913085937 + }, + "validation_seconds": 0.17052793502807617 + }, + { + "epoch": 131, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08819638755056593, + "train_seconds": 560.2379078865051, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.5385918884277344 + }, + "validation_seconds": 0.1703329086303711 + }, + { + "epoch": 132, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0854943878531456, + "train_seconds": 560.2816424369812, + "validation": { + "accuracy": 0.8658, + "examples": 5000, + "loss": 0.5405256225585937 + }, + "validation_seconds": 0.17044830322265625 + }, + { + "epoch": 133, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08349375572767523, + "train_seconds": 560.3219213485718, + "validation": { + "accuracy": 0.8636, + "examples": 5000, + "loss": 0.5356781799316406 + }, + "validation_seconds": 0.17093181610107422 + }, + { + "epoch": 134, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08320598353147507, + "train_seconds": 560.3251914978027, + "validation": { + "accuracy": 0.8656, + "examples": 5000, + "loss": 0.5479776062011719 + }, + "validation_seconds": 0.17060375213623047 + }, + { + "epoch": 135, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08140208751625484, + "train_seconds": 560.2988884449005, + "validation": { + "accuracy": 0.8656, + "examples": 5000, + "loss": 0.5495034912109374 + }, + "validation_seconds": 0.1703953742980957 + }, + { + "epoch": 136, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08287404755486383, + "train_seconds": 559.8910458087921, + "validation": { + "accuracy": 0.8678, + "examples": 5000, + "loss": 0.5502274353027343 + }, + "validation_seconds": 0.17048001289367676 + }, + { + "epoch": 137, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08143085203435686, + "train_seconds": 559.9673924446106, + "validation": { + "accuracy": 0.8664, + "examples": 5000, + "loss": 0.5595277648925782 + }, + "validation_seconds": 0.17099213600158691 + }, + { + "epoch": 138, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.08175256986684269, + "train_seconds": 559.9922814369202, + "validation": { + "accuracy": 0.865, + "examples": 5000, + "loss": 0.5596325988769532 + }, + "validation_seconds": 0.17058587074279785 + }, + { + "epoch": 139, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0801470733165741, + "train_seconds": 560.0086665153503, + "validation": { + "accuracy": 0.865, + "examples": 5000, + "loss": 0.5506817626953125 + }, + "validation_seconds": 0.17049789428710938 + }, + { + "epoch": 140, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07910232688056099, + "train_seconds": 559.9355964660645, + "validation": { + "accuracy": 0.8666, + "examples": 5000, + "loss": 0.5706710083007812 + }, + "validation_seconds": 0.1705949306488037 + }, + { + "epoch": 141, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07886751850545407, + "train_seconds": 559.9587299823761, + "validation": { + "accuracy": 0.8652, + "examples": 5000, + "loss": 0.5789751953125 + }, + "validation_seconds": 0.17045021057128906 + }, + { + "epoch": 142, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07730617905788952, + "train_seconds": 560.0075674057007, + "validation": { + "accuracy": 0.8618, + "examples": 5000, + "loss": 0.5797901489257813 + }, + "validation_seconds": 0.170365571975708 + }, + { + "epoch": 143, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07442515372832616, + "train_seconds": 560.0367701053619, + "validation": { + "accuracy": 0.8684, + "examples": 5000, + "loss": 0.5737571044921875 + }, + "validation_seconds": 0.17035984992980957 + }, + { + "epoch": 144, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07612440371645822, + "train_seconds": 559.9988217353821, + "validation": { + "accuracy": 0.8664, + "examples": 5000, + "loss": 0.5886032470703125 + }, + "validation_seconds": 0.17034530639648438 + }, + { + "epoch": 145, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0729590860320462, + "train_seconds": 559.9792611598969, + "validation": { + "accuracy": 0.8696, + "examples": 5000, + "loss": 0.5738935424804688 + }, + "validation_seconds": 0.17038536071777344 + }, + { + "epoch": 146, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07158674071364933, + "train_seconds": 559.9309771060944, + "validation": { + "accuracy": 0.87, + "examples": 5000, + "loss": 0.5678085327148438 + }, + "validation_seconds": 0.17050409317016602 + }, + { + "epoch": 147, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07183046047290166, + "train_seconds": 559.9765605926514, + "validation": { + "accuracy": 0.8648, + "examples": 5000, + "loss": 0.57717275390625 + }, + "validation_seconds": 0.17051148414611816 + }, + { + "epoch": 148, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07182821125984191, + "train_seconds": 559.96338057518, + "validation": { + "accuracy": 0.8672, + "examples": 5000, + "loss": 0.5847402465820313 + }, + "validation_seconds": 0.1704258918762207 + }, + { + "epoch": 149, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.0710466639969084, + "train_seconds": 559.9064841270447, + "validation": { + "accuracy": 0.864, + "examples": 5000, + "loss": 0.58584228515625 + }, + "validation_seconds": 0.17047476768493652 + }, + { + "epoch": 150, + "lr": 0.005000000000000001, + "output_lr": 0.005000000000000001, + "train_examples": 45000, + "train_loss": 0.07163201610512204, + "train_seconds": 559.985518693924, + "validation": { + "accuracy": 0.862, + "examples": 5000, + "loss": 0.60404404296875 + }, + "validation_seconds": 0.1704697608947754 + }, + { + "epoch": 151, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.06554258665839831, + "train_seconds": 559.9303710460663, + "validation": { + "accuracy": 0.8634, + "examples": 5000, + "loss": 0.59264501953125 + }, + "validation_seconds": 0.17035746574401855 + }, + { + "epoch": 152, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.06308317802614635, + "train_seconds": 560.0243787765503, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.5929781616210937 + }, + "validation_seconds": 0.17056035995483398 + }, + { + "epoch": 153, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.0585381213956409, + "train_seconds": 560.1924431324005, + "validation": { + "accuracy": 0.8636, + "examples": 5000, + "loss": 0.58992890625 + }, + "validation_seconds": 0.17081665992736816 + }, + { + "epoch": 154, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.057798803909619646, + "train_seconds": 559.9080834388733, + "validation": { + "accuracy": 0.8642, + "examples": 5000, + "loss": 0.58432783203125 + }, + "validation_seconds": 0.17063117027282715 + }, + { + "epoch": 155, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05727993083530002, + "train_seconds": 559.9425733089447, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.5870113159179687 + }, + "validation_seconds": 0.17041635513305664 + }, + { + "epoch": 156, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.058408254322740764, + "train_seconds": 559.9141864776611, + "validation": { + "accuracy": 0.866, + "examples": 5000, + "loss": 0.5888887939453125 + }, + "validation_seconds": 0.17054033279418945 + }, + { + "epoch": 157, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05870731733375126, + "train_seconds": 559.904869556427, + "validation": { + "accuracy": 0.8642, + "examples": 5000, + "loss": 0.5919999145507813 + }, + "validation_seconds": 0.1707630157470703 + }, + { + "epoch": 158, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.056921730032894344, + "train_seconds": 559.9407722949982, + "validation": { + "accuracy": 0.8642, + "examples": 5000, + "loss": 0.592196240234375 + }, + "validation_seconds": 0.17051410675048828 + }, + { + "epoch": 159, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.057444885372453264, + "train_seconds": 560.0611102581024, + "validation": { + "accuracy": 0.866, + "examples": 5000, + "loss": 0.58419599609375 + }, + "validation_seconds": 0.1703627109527588 + }, + { + "epoch": 160, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05547764422363705, + "train_seconds": 560.2958824634552, + "validation": { + "accuracy": 0.8638, + "examples": 5000, + "loss": 0.5911913940429687 + }, + "validation_seconds": 0.17037415504455566 + }, + { + "epoch": 161, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.055644365504052905, + "train_seconds": 560.2289936542511, + "validation": { + "accuracy": 0.8638, + "examples": 5000, + "loss": 0.578126025390625 + }, + "validation_seconds": 0.1710362434387207 + }, + { + "epoch": 162, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05510484569966793, + "train_seconds": 560.129026889801, + "validation": { + "accuracy": 0.8656, + "examples": 5000, + "loss": 0.5924788818359376 + }, + "validation_seconds": 0.17116570472717285 + }, + { + "epoch": 163, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05579974476231469, + "train_seconds": 560.2356305122375, + "validation": { + "accuracy": 0.865, + "examples": 5000, + "loss": 0.5858259155273438 + }, + "validation_seconds": 0.1709749698638916 + }, + { + "epoch": 164, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05521068836980396, + "train_seconds": 560.1183788776398, + "validation": { + "accuracy": 0.8666, + "examples": 5000, + "loss": 0.5908378173828125 + }, + "validation_seconds": 0.17043590545654297 + }, + { + "epoch": 165, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05445151421162817, + "train_seconds": 560.0445311069489, + "validation": { + "accuracy": 0.8652, + "examples": 5000, + "loss": 0.586676708984375 + }, + "validation_seconds": 0.17090559005737305 + }, + { + "epoch": 166, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.055989591648843554, + "train_seconds": 559.4602563381195, + "validation": { + "accuracy": 0.8666, + "examples": 5000, + "loss": 0.5927463989257813 + }, + "validation_seconds": 0.17037081718444824 + }, + { + "epoch": 167, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05388901816672749, + "train_seconds": 558.9057474136353, + "validation": { + "accuracy": 0.8634, + "examples": 5000, + "loss": 0.5915533813476562 + }, + "validation_seconds": 0.17028379440307617 + }, + { + "epoch": 168, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.054578808952702416, + "train_seconds": 558.9692997932434, + "validation": { + "accuracy": 0.8652, + "examples": 5000, + "loss": 0.5952635986328125 + }, + "validation_seconds": 0.17091751098632812 + }, + { + "epoch": 169, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.054068607613775466, + "train_seconds": 558.840797662735, + "validation": { + "accuracy": 0.8646, + "examples": 5000, + "loss": 0.589640966796875 + }, + "validation_seconds": 0.1702558994293213 + }, + { + "epoch": 170, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.053210495739512975, + "train_seconds": 559.8298916816711, + "validation": { + "accuracy": 0.8664, + "examples": 5000, + "loss": 0.5905986572265625 + }, + "validation_seconds": 0.17023777961730957 + }, + { + "epoch": 171, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.052864870564805135, + "train_seconds": 559.8400263786316, + "validation": { + "accuracy": 0.8632, + "examples": 5000, + "loss": 0.5897403686523438 + }, + "validation_seconds": 0.17044568061828613 + }, + { + "epoch": 172, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.052678663874997034, + "train_seconds": 559.9006729125977, + "validation": { + "accuracy": 0.8634, + "examples": 5000, + "loss": 0.5942950317382812 + }, + "validation_seconds": 0.17066121101379395 + }, + { + "epoch": 173, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05529955416785346, + "train_seconds": 560.0638835430145, + "validation": { + "accuracy": 0.8628, + "examples": 5000, + "loss": 0.5956375244140625 + }, + "validation_seconds": 0.17058396339416504 + }, + { + "epoch": 174, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.053033065217733384, + "train_seconds": 560.1647348403931, + "validation": { + "accuracy": 0.864, + "examples": 5000, + "loss": 0.5948213745117188 + }, + "validation_seconds": 0.17058897018432617 + }, + { + "epoch": 175, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05241316869656245, + "train_seconds": 560.0912938117981, + "validation": { + "accuracy": 0.8656, + "examples": 5000, + "loss": 0.5923770812988282 + }, + "validation_seconds": 0.17046594619750977 + }, + { + "epoch": 176, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05160712323255009, + "train_seconds": 560.2796442508698, + "validation": { + "accuracy": 0.8624, + "examples": 5000, + "loss": 0.5936520629882812 + }, + "validation_seconds": 0.17072343826293945 + }, + { + "epoch": 177, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.051562978375620315, + "train_seconds": 560.2672243118286, + "validation": { + "accuracy": 0.8648, + "examples": 5000, + "loss": 0.5952009033203125 + }, + "validation_seconds": 0.17060518264770508 + }, + { + "epoch": 178, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.053683729785018494, + "train_seconds": 560.2481393814087, + "validation": { + "accuracy": 0.8644, + "examples": 5000, + "loss": 0.5962612548828125 + }, + "validation_seconds": 0.1711122989654541 + }, + { + "epoch": 179, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05219864928192563, + "train_seconds": 560.473662853241, + "validation": { + "accuracy": 0.8634, + "examples": 5000, + "loss": 0.5932921630859375 + }, + "validation_seconds": 0.1708521842956543 + }, + { + "epoch": 180, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.053021824158562554, + "train_seconds": 560.7065186500549, + "validation": { + "accuracy": 0.8632, + "examples": 5000, + "loss": 0.5973969970703125 + }, + "validation_seconds": 0.17085528373718262 + }, + { + "epoch": 181, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05194481589926614, + "train_seconds": 560.7485203742981, + "validation": { + "accuracy": 0.8646, + "examples": 5000, + "loss": 0.597285546875 + }, + "validation_seconds": 0.17076420783996582 + }, + { + "epoch": 182, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.0498733964893553, + "train_seconds": 560.4968025684357, + "validation": { + "accuracy": 0.8638, + "examples": 5000, + "loss": 0.5942079223632812 + }, + "validation_seconds": 0.1707749366760254 + }, + { + "epoch": 183, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05121098554680745, + "train_seconds": 560.602189540863, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.5904847534179688 + }, + "validation_seconds": 0.17096948623657227 + }, + { + "epoch": 184, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.0532610271414121, + "train_seconds": 560.6874899864197, + "validation": { + "accuracy": 0.8646, + "examples": 5000, + "loss": 0.5949594970703125 + }, + "validation_seconds": 0.17095732688903809 + }, + { + "epoch": 185, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05066566181447771, + "train_seconds": 560.6240856647491, + "validation": { + "accuracy": 0.866, + "examples": 5000, + "loss": 0.598093505859375 + }, + "validation_seconds": 0.1707146167755127 + }, + { + "epoch": 186, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.050586108859380086, + "train_seconds": 560.5915801525116, + "validation": { + "accuracy": 0.8656, + "examples": 5000, + "loss": 0.5995073974609375 + }, + "validation_seconds": 0.17087697982788086 + }, + { + "epoch": 187, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.04926515153249105, + "train_seconds": 560.5769941806793, + "validation": { + "accuracy": 0.8646, + "examples": 5000, + "loss": 0.6008431518554688 + }, + "validation_seconds": 0.1711268424987793 + }, + { + "epoch": 188, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05084701956609885, + "train_seconds": 560.4484589099884, + "validation": { + "accuracy": 0.8644, + "examples": 5000, + "loss": 0.5961071411132812 + }, + "validation_seconds": 0.1709892749786377 + }, + { + "epoch": 189, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05086444434059991, + "train_seconds": 560.4446725845337, + "validation": { + "accuracy": 0.8664, + "examples": 5000, + "loss": 0.595338525390625 + }, + "validation_seconds": 0.17079830169677734 + }, + { + "epoch": 190, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05069162193536758, + "train_seconds": 560.5038962364197, + "validation": { + "accuracy": 0.8638, + "examples": 5000, + "loss": 0.5992982055664062 + }, + "validation_seconds": 0.17066144943237305 + }, + { + "epoch": 191, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05099646294448111, + "train_seconds": 560.6068978309631, + "validation": { + "accuracy": 0.8638, + "examples": 5000, + "loss": 0.6020646362304688 + }, + "validation_seconds": 0.17072200775146484 + }, + { + "epoch": 192, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05033617229064306, + "train_seconds": 560.5441524982452, + "validation": { + "accuracy": 0.865, + "examples": 5000, + "loss": 0.6090272827148437 + }, + "validation_seconds": 0.1708674430847168 + }, + { + "epoch": 193, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.05045138215753767, + "train_seconds": 560.5373227596283, + "validation": { + "accuracy": 0.862, + "examples": 5000, + "loss": 0.6072682006835938 + }, + "validation_seconds": 0.17099285125732422 + }, + { + "epoch": 194, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.049500737303495405, + "train_seconds": 560.6162390708923, + "validation": { + "accuracy": 0.8644, + "examples": 5000, + "loss": 0.597724462890625 + }, + "validation_seconds": 0.17102837562561035 + }, + { + "epoch": 195, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.04977997776005003, + "train_seconds": 560.5924124717712, + "validation": { + "accuracy": 0.8648, + "examples": 5000, + "loss": 0.6030875610351563 + }, + "validation_seconds": 0.17075753211975098 + }, + { + "epoch": 196, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.04957166252202458, + "train_seconds": 560.5982604026794, + "validation": { + "accuracy": 0.8652, + "examples": 5000, + "loss": 0.604028369140625 + }, + "validation_seconds": 0.17163372039794922 + }, + { + "epoch": 197, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.050121073783106276, + "train_seconds": 560.0612788200378, + "validation": { + "accuracy": 0.864, + "examples": 5000, + "loss": 0.6068662475585938 + }, + "validation_seconds": 0.17065048217773438 + }, + { + "epoch": 198, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.04878396252923541, + "train_seconds": 559.8991024494171, + "validation": { + "accuracy": 0.8662, + "examples": 5000, + "loss": 0.6060625732421875 + }, + "validation_seconds": 0.1706082820892334 + }, + { + "epoch": 199, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.0506157725599077, + "train_seconds": 559.9076845645905, + "validation": { + "accuracy": 0.8642, + "examples": 5000, + "loss": 0.6069480834960937 + }, + "validation_seconds": 0.17104268074035645 + }, + { + "epoch": 200, + "lr": 0.0005000000000000001, + "output_lr": 0.0005000000000000001, + "train_examples": 45000, + "train_loss": 0.050066802283128105, + "train_seconds": 559.9151027202606, + "validation": { + "accuracy": 0.8654, + "examples": 5000, + "loss": 0.6110852416992187 + }, + "validation_seconds": 0.1712048053741455 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.8654, + "evaluation_seconds": 0.1711885929107666, + "examples": 5000, + "finite": true, + "loss": 0.6110852416992187 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2322300928, + "peak_memory_reserved_bytes": 3586129920, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 111860.19822311401, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 200, + "feedforward_example_passes": 10005000, + "forward_macs_per_example": 68862592, + "forward_parameter_count": 464154, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 144000000, + "logical_task_loss_queries": 0, + "num_trainable_layers": 32, + "ordinary_training_examples": 9000000, + "ordinary_validation_examples": 1005000, + "relaxation_example_passes": 144000000, + "training_example_presentations": 9000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json.manifest.json new file mode 100644 index 0000000..d0094ca --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json.manifest.json @@ -0,0 +1,132 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "dualprop", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.05", + "--output_lr", + "0.05", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--dp_alpha", + "0.0", + "--dp_beta", + "0.1", + "--dp_inference_passes", + "16" + ], + "completed_unix_time": 1787014926.9230115, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 111870.60429954529, + "epochs": 200, + "experiment_name": "resnet-r2-dualprop-d32-lr0p05", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-9cbd0c6c-f3fb-633d-be44-101a79992530", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "dualprop", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d32-lr0p05.json", + "output_exists": true, + "output_sha256": "d45682de145f1650b61be527d44ba9e38495ef5203112166a19b72ec7d8f0917", + "rate": 0.05, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-dualprop-d56-lr0p05.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-dualprop-d56-lr0p05.json.manifest.json new file mode 100644 index 0000000..d562f01 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dualprop-d56-lr0p05.json.manifest.json @@ -0,0 +1,132 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "dualprop", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d56-lr0p05.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.05", + "--output_lr", + "0.05", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--dp_alpha", + "0.0", + "--dp_beta", + "0.1", + "--dp_inference_passes", + "16" + ], + "completed_unix_time": 1786380961.7231288, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 172800.46599197388, + "epochs": 200, + "experiment_name": "resnet-r2-dualprop-d56-lr0p05", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-b520421c-b120-70f0-eae8-1ff79ad97bf5", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "dualprop", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dualprop-d56-lr0p05.json", + "output_exists": false, + "output_sha256": null, + "rate": 0.05, + "return_code": null, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "timeout", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json b/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json new file mode 100644 index 0000000..e1e4072 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json @@ -0,0 +1,1565 @@ +{ + "architecture": { + "base_width": 16, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 269722, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 100, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.0003, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "constant", + "max_steps": 0, + "method": "ep", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json", + "output_lr": 0.0003, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.10000092970662647, + "train_seconds": 139.11669993400574, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09817654571533203 + }, + "validation_seconds": 11.159880876541138 + }, + { + "epoch": 2, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.096760009001361, + "train_seconds": 138.82202982902527, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09676226196289063 + }, + "validation_seconds": 11.14545726776123 + }, + { + "epoch": 3, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09596625119845073, + "train_seconds": 138.83978295326233, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0960817398071289 + }, + "validation_seconds": 11.141392230987549 + }, + { + "epoch": 4, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09608556209140354, + "train_seconds": 138.88129901885986, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09456661682128906 + }, + "validation_seconds": 11.150422811508179 + }, + { + "epoch": 5, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.096144191037284, + "train_seconds": 138.87270593643188, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09416954193115235 + }, + "validation_seconds": 11.146359920501709 + }, + { + "epoch": 6, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0957507728377978, + "train_seconds": 138.86185312271118, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09589024200439453 + }, + "validation_seconds": 11.14265751838684 + }, + { + "epoch": 7, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609245043992996, + "train_seconds": 138.87552165985107, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0969051010131836 + }, + "validation_seconds": 11.147951364517212 + }, + { + "epoch": 8, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09637515972455342, + "train_seconds": 139.2255883216858, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0969967056274414 + }, + "validation_seconds": 11.152616500854492 + }, + { + "epoch": 9, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09704533698028989, + "train_seconds": 139.07272458076477, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09479008636474609 + }, + "validation_seconds": 11.14713978767395 + }, + { + "epoch": 10, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0958041417002678, + "train_seconds": 138.8842315673828, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09533480987548829 + }, + "validation_seconds": 11.152060270309448 + }, + { + "epoch": 11, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09585828256342147, + "train_seconds": 138.86268758773804, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.096821923828125 + }, + "validation_seconds": 11.150679111480713 + }, + { + "epoch": 12, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.095636143067148, + "train_seconds": 138.91375017166138, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09711122283935547 + }, + "validation_seconds": 11.14675760269165 + }, + { + "epoch": 13, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09573306566079458, + "train_seconds": 138.89357042312622, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09591180725097656 + }, + "validation_seconds": 11.144380331039429 + }, + { + "epoch": 14, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09582586610449685, + "train_seconds": 138.9151952266693, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09458838500976563 + }, + "validation_seconds": 11.150224208831787 + }, + { + "epoch": 15, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.095475492699941, + "train_seconds": 138.97902250289917, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0952644287109375 + }, + "validation_seconds": 11.146376609802246 + }, + { + "epoch": 16, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09557586745686002, + "train_seconds": 138.93276834487915, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09532229919433594 + }, + "validation_seconds": 11.151374340057373 + }, + { + "epoch": 17, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09595397739410401, + "train_seconds": 138.96198630332947, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09608304290771484 + }, + "validation_seconds": 11.14755392074585 + }, + { + "epoch": 18, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09530874093108707, + "train_seconds": 138.92328310012817, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09483186340332031 + }, + "validation_seconds": 11.149696826934814 + }, + { + "epoch": 19, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09545710382726458, + "train_seconds": 138.9205515384674, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09437860565185546 + }, + "validation_seconds": 11.14361047744751 + }, + { + "epoch": 20, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09598765107393264, + "train_seconds": 138.96228528022766, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09477271881103516 + }, + "validation_seconds": 11.148051023483276 + }, + { + "epoch": 21, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09561405028104782, + "train_seconds": 138.91624879837036, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09785403137207031 + }, + "validation_seconds": 11.150966167449951 + }, + { + "epoch": 22, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09561793030500412, + "train_seconds": 138.9595160484314, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09515677032470703 + }, + "validation_seconds": 11.145541429519653 + }, + { + "epoch": 23, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09509115369982189, + "train_seconds": 138.9069902896881, + "validation": { + "accuracy": 0.0922, + "examples": 5000, + "loss": 0.09626204528808593 + }, + "validation_seconds": 11.147856712341309 + }, + { + "epoch": 24, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09560073957708147, + "train_seconds": 138.90601563453674, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09640469055175781 + }, + "validation_seconds": 11.146676540374756 + }, + { + "epoch": 25, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09574450379742516, + "train_seconds": 138.99443531036377, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09708086853027344 + }, + "validation_seconds": 11.152676582336426 + }, + { + "epoch": 26, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09567417042917675, + "train_seconds": 139.10978770256042, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09664617004394531 + }, + "validation_seconds": 11.15486764907837 + }, + { + "epoch": 27, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09516358087592655, + "train_seconds": 139.1539888381958, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09485475616455079 + }, + "validation_seconds": 11.144922256469727 + }, + { + "epoch": 28, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09542587406900194, + "train_seconds": 138.9068100452423, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.094023095703125 + }, + "validation_seconds": 11.144463777542114 + }, + { + "epoch": 29, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09558091937568453, + "train_seconds": 138.85909581184387, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0952291488647461 + }, + "validation_seconds": 11.140281677246094 + }, + { + "epoch": 30, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09604495452377532, + "train_seconds": 138.83931016921997, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09689130554199218 + }, + "validation_seconds": 11.138436555862427 + }, + { + "epoch": 31, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09616093474494086, + "train_seconds": 138.60884046554565, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09545914764404297 + }, + "validation_seconds": 11.133517980575562 + }, + { + "epoch": 32, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0956987985306316, + "train_seconds": 138.62059688568115, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09510513305664063 + }, + "validation_seconds": 11.137269735336304 + }, + { + "epoch": 33, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09554892805814744, + "train_seconds": 138.61837697029114, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0953763916015625 + }, + "validation_seconds": 11.135218620300293 + }, + { + "epoch": 34, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09605118489927715, + "train_seconds": 138.62338376045227, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09536750946044922 + }, + "validation_seconds": 11.138956546783447 + }, + { + "epoch": 35, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09561948593987359, + "train_seconds": 138.6473937034607, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09514895172119141 + }, + "validation_seconds": 11.136772155761719 + }, + { + "epoch": 36, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09602282423973084, + "train_seconds": 138.74485182762146, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09477901153564453 + }, + "validation_seconds": 11.140544414520264 + }, + { + "epoch": 37, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09655213383701113, + "train_seconds": 138.73414587974548, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09620673370361328 + }, + "validation_seconds": 11.137867212295532 + }, + { + "epoch": 38, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09576946388880411, + "train_seconds": 138.76311469078064, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09446058502197266 + }, + "validation_seconds": 11.140277862548828 + }, + { + "epoch": 39, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.095974667693509, + "train_seconds": 138.73441815376282, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09710462036132812 + }, + "validation_seconds": 11.13795518875122 + }, + { + "epoch": 40, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609547713862525, + "train_seconds": 138.68924736976624, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09623699340820313 + }, + "validation_seconds": 11.141454458236694 + }, + { + "epoch": 41, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09539313185214997, + "train_seconds": 138.66758728027344, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09325079956054687 + }, + "validation_seconds": 11.137190341949463 + }, + { + "epoch": 42, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0956763015402688, + "train_seconds": 138.69396257400513, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0959182632446289 + }, + "validation_seconds": 11.139819622039795 + }, + { + "epoch": 43, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09580025514364243, + "train_seconds": 138.75355863571167, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09510519409179688 + }, + "validation_seconds": 11.140304565429688 + }, + { + "epoch": 44, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09580142854981952, + "train_seconds": 138.7804079055786, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09460178375244141 + }, + "validation_seconds": 11.142134666442871 + }, + { + "epoch": 45, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0952616327073839, + "train_seconds": 138.74207544326782, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09451842498779296 + }, + "validation_seconds": 11.138129711151123 + }, + { + "epoch": 46, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09581543866263495, + "train_seconds": 138.78086709976196, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09661761474609375 + }, + "validation_seconds": 11.14474630355835 + }, + { + "epoch": 47, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09611166249513627, + "train_seconds": 138.81940126419067, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09566448211669921 + }, + "validation_seconds": 11.141072750091553 + }, + { + "epoch": 48, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09542014650636249, + "train_seconds": 138.77664518356323, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09565098419189454 + }, + "validation_seconds": 11.144866228103638 + }, + { + "epoch": 49, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609724575148688, + "train_seconds": 138.78950238227844, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09689692687988281 + }, + "validation_seconds": 11.142502546310425 + }, + { + "epoch": 50, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09599001576503118, + "train_seconds": 138.87918281555176, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09705646209716796 + }, + "validation_seconds": 11.141917705535889 + }, + { + "epoch": 51, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09611067996554905, + "train_seconds": 138.81238508224487, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09507562713623047 + }, + "validation_seconds": 11.139949560165405 + }, + { + "epoch": 52, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09620131080547968, + "train_seconds": 138.90105414390564, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09800862884521484 + }, + "validation_seconds": 11.146653175354004 + }, + { + "epoch": 53, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09654972725576824, + "train_seconds": 138.98731064796448, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09603324127197266 + }, + "validation_seconds": 11.144920110702515 + }, + { + "epoch": 54, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09627177377939224, + "train_seconds": 138.9740378856659, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0954896240234375 + }, + "validation_seconds": 11.148634433746338 + }, + { + "epoch": 55, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09626585129234526, + "train_seconds": 138.9596757888794, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09900887756347657 + }, + "validation_seconds": 11.143959760665894 + }, + { + "epoch": 56, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09648386177884208, + "train_seconds": 138.9568064212799, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09393155670166016 + }, + "validation_seconds": 11.14871096611023 + }, + { + "epoch": 57, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09645076555146112, + "train_seconds": 138.95467948913574, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0980503646850586 + }, + "validation_seconds": 11.14472484588623 + }, + { + "epoch": 58, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09627769343058268, + "train_seconds": 139.03842067718506, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09474974517822266 + }, + "validation_seconds": 11.150963068008423 + }, + { + "epoch": 59, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09605537111361821, + "train_seconds": 139.1419117450714, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09802635192871094 + }, + "validation_seconds": 11.146832466125488 + }, + { + "epoch": 60, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09588881915277905, + "train_seconds": 139.09739804267883, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0964722396850586 + }, + "validation_seconds": 11.148901224136353 + }, + { + "epoch": 61, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09624453692171309, + "train_seconds": 139.0157241821289, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09603090057373047 + }, + "validation_seconds": 11.145693302154541 + }, + { + "epoch": 62, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09638927124738693, + "train_seconds": 139.01228952407837, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09630657653808594 + }, + "validation_seconds": 11.15030837059021 + }, + { + "epoch": 63, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09640581868754493, + "train_seconds": 139.06430864334106, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0956555206298828 + }, + "validation_seconds": 11.145128965377808 + }, + { + "epoch": 64, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09614854072729746, + "train_seconds": 139.0981409549713, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09355887908935547 + }, + "validation_seconds": 11.149465322494507 + }, + { + "epoch": 65, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09623182185226016, + "train_seconds": 139.18455290794373, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09463775634765625 + }, + "validation_seconds": 11.150774478912354 + }, + { + "epoch": 66, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09593160262770123, + "train_seconds": 139.1528673171997, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0963241683959961 + }, + "validation_seconds": 11.157290697097778 + }, + { + "epoch": 67, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09562942595614328, + "train_seconds": 139.32904410362244, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09660213775634766 + }, + "validation_seconds": 11.155967235565186 + }, + { + "epoch": 68, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09585837714936998, + "train_seconds": 139.3654706478119, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09532363433837891 + }, + "validation_seconds": 11.157169580459595 + }, + { + "epoch": 69, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09573291099998686, + "train_seconds": 139.34520745277405, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09697261047363281 + }, + "validation_seconds": 11.158665418624878 + }, + { + "epoch": 70, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09632284501658546, + "train_seconds": 139.34444737434387, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09541180267333985 + }, + "validation_seconds": 11.157097578048706 + }, + { + "epoch": 71, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09558174745374255, + "train_seconds": 139.3179748058319, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09908675689697266 + }, + "validation_seconds": 11.1538667678833 + }, + { + "epoch": 72, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09581099202367994, + "train_seconds": 139.2078824043274, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0985763900756836 + }, + "validation_seconds": 11.154834270477295 + }, + { + "epoch": 73, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09601372849543889, + "train_seconds": 139.2479133605957, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09627315826416015 + }, + "validation_seconds": 11.15363883972168 + }, + { + "epoch": 74, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09621802689234416, + "train_seconds": 139.25913310050964, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0955810287475586 + }, + "validation_seconds": 11.156188011169434 + }, + { + "epoch": 75, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09564000614616606, + "train_seconds": 139.2822630405426, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09619723358154297 + }, + "validation_seconds": 11.155544519424438 + }, + { + "epoch": 76, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0960162280175421, + "train_seconds": 139.34670281410217, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09646122741699219 + }, + "validation_seconds": 11.155498504638672 + }, + { + "epoch": 77, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09561879307958815, + "train_seconds": 139.26200103759766, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09832343292236329 + }, + "validation_seconds": 11.154755353927612 + }, + { + "epoch": 78, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09588940723737081, + "train_seconds": 139.21251392364502, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09473636779785156 + }, + "validation_seconds": 11.15525221824646 + }, + { + "epoch": 79, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09603877027299669, + "train_seconds": 139.1633071899414, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09692667694091797 + }, + "validation_seconds": 11.1514732837677 + }, + { + "epoch": 80, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09629039370351368, + "train_seconds": 139.21222305297852, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0954868438720703 + }, + "validation_seconds": 11.154970169067383 + }, + { + "epoch": 81, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09558552947839101, + "train_seconds": 139.18993878364563, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09273733062744141 + }, + "validation_seconds": 11.155423641204834 + }, + { + "epoch": 82, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09602079207499822, + "train_seconds": 139.26600408554077, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09672743530273438 + }, + "validation_seconds": 11.157419919967651 + }, + { + "epoch": 83, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09581270765331057, + "train_seconds": 139.23123788833618, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09539513549804687 + }, + "validation_seconds": 11.153031349182129 + }, + { + "epoch": 84, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09577875078784095, + "train_seconds": 139.1194291114807, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09731991729736328 + }, + "validation_seconds": 11.154894828796387 + }, + { + "epoch": 85, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09578391640848584, + "train_seconds": 139.0757772922516, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09821551055908204 + }, + "validation_seconds": 11.155132532119751 + }, + { + "epoch": 86, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09557791891230477, + "train_seconds": 139.0469081401825, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09509595336914063 + }, + "validation_seconds": 11.150206327438354 + }, + { + "epoch": 87, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09617045336696836, + "train_seconds": 138.9863407611847, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0955680892944336 + }, + "validation_seconds": 11.152897119522095 + }, + { + "epoch": 88, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09557901961538527, + "train_seconds": 139.05934143066406, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09560486907958984 + }, + "validation_seconds": 11.151857137680054 + }, + { + "epoch": 89, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09592116408877903, + "train_seconds": 139.1188690662384, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0936294677734375 + }, + "validation_seconds": 11.1571364402771 + }, + { + "epoch": 90, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09588112912840313, + "train_seconds": 139.1143081188202, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09687020568847657 + }, + "validation_seconds": 11.151924133300781 + }, + { + "epoch": 91, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09571159390211105, + "train_seconds": 139.09063410758972, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09342134704589844 + }, + "validation_seconds": 11.154841661453247 + }, + { + "epoch": 92, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09590040313800176, + "train_seconds": 139.09315967559814, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0954732666015625 + }, + "validation_seconds": 11.153255462646484 + }, + { + "epoch": 93, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09588631826241811, + "train_seconds": 139.12650752067566, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09767638397216796 + }, + "validation_seconds": 11.1552095413208 + }, + { + "epoch": 94, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0961217908249961, + "train_seconds": 139.14415001869202, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09442243499755859 + }, + "validation_seconds": 11.151734828948975 + }, + { + "epoch": 95, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09611847210062875, + "train_seconds": 139.14410185813904, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09692444000244141 + }, + "validation_seconds": 11.152769804000854 + }, + { + "epoch": 96, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0958005420896742, + "train_seconds": 139.11920261383057, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09459645385742188 + }, + "validation_seconds": 11.155317544937134 + }, + { + "epoch": 97, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09605646768146091, + "train_seconds": 139.06602239608765, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09579430389404296 + }, + "validation_seconds": 11.153225183486938 + }, + { + "epoch": 98, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09623256181875865, + "train_seconds": 139.11849069595337, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09721367797851563 + }, + "validation_seconds": 11.151837587356567 + }, + { + "epoch": 99, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09644517441458172, + "train_seconds": 139.11283159255981, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09481856689453125 + }, + "validation_seconds": 11.155541896820068 + }, + { + "epoch": 100, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0958042881541782, + "train_seconds": 139.26254963874817, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09437472686767578 + }, + "validation_seconds": 11.155186176300049 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.1, + "evaluation_seconds": 11.155261516571045, + "examples": 5000, + "finite": true, + "loss": 0.09437472686767578 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 4845611008, + "peak_memory_reserved_bytes": 5939134464, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 15025.534816026688, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 100, + "feedforward_example_passes": 5005000, + "forward_macs_per_example": 40551040, + "forward_parameter_count": 269722, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 118100000, + "logical_task_loss_queries": 0, + "num_trainable_layers": 20, + "ordinary_training_examples": 4500000, + "ordinary_validation_examples": 505000, + "relaxation_example_passes": 118100000, + "training_example_presentations": 4500000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json.manifest.json new file mode 100644 index 0000000..3a951bf --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json.manifest.json @@ -0,0 +1,134 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "ep", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "100", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.0003", + "--output_lr", + "0.0003", + "--lr_schedule", + "constant", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--ep_beta", + "0.5", + "--ep_dt", + "0.5", + "--ep_free_steps", + "20", + "--ep_nudge_steps", + "4" + ], + "completed_unix_time": 1786950247.4193482, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 15028.455446720123, + "epochs": 100, + "experiment_name": "resnet-r2-ep-d20-lr0p0003", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "constant", + "method": "ep", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d20-lr0p0003.json", + "output_exists": true, + "output_sha256": "2be303594e67846889489474ed9bde767bf0abcdea660716ddf421bfc455060f", + "rate": 0.0003, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json b/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json new file mode 100644 index 0000000..b703ba2 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json @@ -0,0 +1,1565 @@ +{ + "architecture": { + "base_width": 16, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 464154, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 100, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.0003, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "constant", + "max_steps": 0, + "method": "ep", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json", + "output_lr": 0.0003, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.10584368345869911, + "train_seconds": 233.75673270225525, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09748240203857422 + }, + "validation_seconds": 18.687715530395508 + }, + { + "epoch": 2, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09607491334146924, + "train_seconds": 227.22966647148132, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09624678039550781 + }, + "validation_seconds": 18.22874617576599 + }, + { + "epoch": 3, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09619754118786918, + "train_seconds": 227.6295883655548, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09637385711669921 + }, + "validation_seconds": 18.230604887008667 + }, + { + "epoch": 4, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.096190148557557, + "train_seconds": 227.70378065109253, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09466078643798828 + }, + "validation_seconds": 18.22860026359558 + }, + { + "epoch": 5, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09619633632633422, + "train_seconds": 227.67504620552063, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0939787109375 + }, + "validation_seconds": 18.228357076644897 + }, + { + "epoch": 6, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09575496319664849, + "train_seconds": 227.9477686882019, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0957251968383789 + }, + "validation_seconds": 18.234267473220825 + }, + { + "epoch": 7, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0960714242219925, + "train_seconds": 227.73594498634338, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09659960479736328 + }, + "validation_seconds": 18.239380836486816 + }, + { + "epoch": 8, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09630487336715063, + "train_seconds": 228.03009343147278, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0964680160522461 + }, + "validation_seconds": 18.236380100250244 + }, + { + "epoch": 9, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0967883091436492, + "train_seconds": 228.04988503456116, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0942939926147461 + }, + "validation_seconds": 18.238703966140747 + }, + { + "epoch": 10, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09595642296738095, + "train_seconds": 227.94507455825806, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09557775573730469 + }, + "validation_seconds": 18.23998737335205 + }, + { + "epoch": 11, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09607761137882868, + "train_seconds": 227.97127604484558, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09727976989746094 + }, + "validation_seconds": 18.244027853012085 + }, + { + "epoch": 12, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09592779109875361, + "train_seconds": 228.01011180877686, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09815475616455079 + }, + "validation_seconds": 18.242377281188965 + }, + { + "epoch": 13, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09610559370915095, + "train_seconds": 228.01637840270996, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09631928253173828 + }, + "validation_seconds": 18.243257522583008 + }, + { + "epoch": 14, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0962134348101086, + "train_seconds": 227.9632499217987, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09476177368164063 + }, + "validation_seconds": 18.235589265823364 + }, + { + "epoch": 15, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0959621706340048, + "train_seconds": 228.08780241012573, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09589082336425782 + }, + "validation_seconds": 18.23856282234192 + }, + { + "epoch": 16, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09578963030576707, + "train_seconds": 228.17340397834778, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09558954162597656 + }, + "validation_seconds": 18.241756200790405 + }, + { + "epoch": 17, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09626044083568785, + "train_seconds": 228.17240142822266, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09684005889892579 + }, + "validation_seconds": 18.241687536239624 + }, + { + "epoch": 18, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09598865276575089, + "train_seconds": 228.0541534423828, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0955846206665039 + }, + "validation_seconds": 18.231510877609253 + }, + { + "epoch": 19, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609488620228238, + "train_seconds": 227.87761616706848, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09553548736572266 + }, + "validation_seconds": 18.23040270805359 + }, + { + "epoch": 20, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09648366617361705, + "train_seconds": 227.81606435775757, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09536498565673829 + }, + "validation_seconds": 18.23584270477295 + }, + { + "epoch": 21, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09604000683757993, + "train_seconds": 227.8779010772705, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09821461181640626 + }, + "validation_seconds": 18.240376234054565 + }, + { + "epoch": 22, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09619036858214272, + "train_seconds": 228.08918619155884, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09566349029541016 + }, + "validation_seconds": 18.23198366165161 + }, + { + "epoch": 23, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09555974514219497, + "train_seconds": 227.9332685470581, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09647967376708984 + }, + "validation_seconds": 18.237937688827515 + }, + { + "epoch": 24, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09629934613572226, + "train_seconds": 227.74945855140686, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09690995330810546 + }, + "validation_seconds": 18.240371465682983 + }, + { + "epoch": 25, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09639721555842294, + "train_seconds": 227.9227864742279, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09780321807861328 + }, + "validation_seconds": 18.22849941253662 + }, + { + "epoch": 26, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09636331896252102, + "train_seconds": 227.874342918396, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09749327545166016 + }, + "validation_seconds": 18.230159044265747 + }, + { + "epoch": 27, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0959015838517083, + "train_seconds": 227.95591282844543, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09564348297119141 + }, + "validation_seconds": 18.233003854751587 + }, + { + "epoch": 28, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609772173298729, + "train_seconds": 227.70959520339966, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0948007080078125 + }, + "validation_seconds": 18.234058380126953 + }, + { + "epoch": 29, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09593701247109307, + "train_seconds": 227.7418727874756, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09526193695068359 + }, + "validation_seconds": 18.231982469558716 + }, + { + "epoch": 30, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09594838575919469, + "train_seconds": 227.90904927253723, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09659895324707031 + }, + "validation_seconds": 18.23198628425598 + }, + { + "epoch": 31, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09625638297796249, + "train_seconds": 228.0143768787384, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09601235046386719 + }, + "validation_seconds": 18.23534321784973 + }, + { + "epoch": 32, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09599113212161595, + "train_seconds": 227.99561190605164, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0954147232055664 + }, + "validation_seconds": 18.23655915260315 + }, + { + "epoch": 33, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09575922750896877, + "train_seconds": 227.84987020492554, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09525252075195312 + }, + "validation_seconds": 18.23592734336853 + }, + { + "epoch": 34, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09635789152251349, + "train_seconds": 227.81707239151, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09551718444824218 + }, + "validation_seconds": 18.234346628189087 + }, + { + "epoch": 35, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09587441050873863, + "train_seconds": 227.86760759353638, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09489838562011718 + }, + "validation_seconds": 18.23371911048889 + }, + { + "epoch": 36, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09623189973698722, + "train_seconds": 227.76344990730286, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09447245178222656 + }, + "validation_seconds": 18.233267545700073 + }, + { + "epoch": 37, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09652466434902615, + "train_seconds": 228.0108609199524, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09592162322998046 + }, + "validation_seconds": 18.23678684234619 + }, + { + "epoch": 38, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09560628210968442, + "train_seconds": 228.01479268074036, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09405255126953126 + }, + "validation_seconds": 18.23764657974243 + }, + { + "epoch": 39, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0959057019048267, + "train_seconds": 228.06421947479248, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09734141845703125 + }, + "validation_seconds": 18.23655915260315 + }, + { + "epoch": 40, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09606925669776069, + "train_seconds": 228.01972579956055, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09611746368408203 + }, + "validation_seconds": 18.233552932739258 + }, + { + "epoch": 41, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09539408498605093, + "train_seconds": 227.97406220436096, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09368319549560547 + }, + "validation_seconds": 18.23234748840332 + }, + { + "epoch": 42, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09593300796614752, + "train_seconds": 227.96903944015503, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09569558410644531 + }, + "validation_seconds": 18.239282608032227 + }, + { + "epoch": 43, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09585642129712635, + "train_seconds": 227.84337186813354, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09548786163330078 + }, + "validation_seconds": 18.231594562530518 + }, + { + "epoch": 44, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09586924339665306, + "train_seconds": 227.81595945358276, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0945896484375 + }, + "validation_seconds": 18.23236846923828 + }, + { + "epoch": 45, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09525853757328458, + "train_seconds": 227.8766598701477, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09463276062011719 + }, + "validation_seconds": 18.23231053352356 + }, + { + "epoch": 46, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09574040567874909, + "train_seconds": 228.08118081092834, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09664346923828125 + }, + "validation_seconds": 18.23355007171631 + }, + { + "epoch": 47, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09600410343408584, + "train_seconds": 228.07431888580322, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09555250701904297 + }, + "validation_seconds": 18.234946966171265 + }, + { + "epoch": 48, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09547720072666804, + "train_seconds": 228.05487155914307, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0959607162475586 + }, + "validation_seconds": 18.235946893692017 + }, + { + "epoch": 49, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09616076918972863, + "train_seconds": 228.20362186431885, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09696863861083985 + }, + "validation_seconds": 18.243194818496704 + }, + { + "epoch": 50, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09633234109481176, + "train_seconds": 228.4779417514801, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09757501220703126 + }, + "validation_seconds": 18.253101587295532 + }, + { + "epoch": 51, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09652728734413783, + "train_seconds": 228.59612488746643, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09503995056152344 + }, + "validation_seconds": 18.25339436531067 + }, + { + "epoch": 52, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09643947992589738, + "train_seconds": 228.75814485549927, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09830532989501953 + }, + "validation_seconds": 18.267754793167114 + }, + { + "epoch": 53, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09715808055930668, + "train_seconds": 228.82663297653198, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09692757568359375 + }, + "validation_seconds": 18.26055932044983 + }, + { + "epoch": 54, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09683998356130388, + "train_seconds": 228.8337893486023, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0960301010131836 + }, + "validation_seconds": 18.252885818481445 + }, + { + "epoch": 55, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09688982499837875, + "train_seconds": 228.71393632888794, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09920754699707031 + }, + "validation_seconds": 18.25892972946167 + }, + { + "epoch": 56, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09654736138449775, + "train_seconds": 228.71237993240356, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0933240005493164 + }, + "validation_seconds": 18.253551483154297 + }, + { + "epoch": 57, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09635696367555194, + "train_seconds": 228.71023201942444, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09789161682128907 + }, + "validation_seconds": 18.252516269683838 + }, + { + "epoch": 58, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09628176087935765, + "train_seconds": 228.71620917320251, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09455877990722657 + }, + "validation_seconds": 18.252784967422485 + }, + { + "epoch": 59, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09615534450610479, + "train_seconds": 228.703430891037, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09805198974609375 + }, + "validation_seconds": 18.256057262420654 + }, + { + "epoch": 60, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09615144361654918, + "train_seconds": 228.6930000782013, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09711163940429687 + }, + "validation_seconds": 18.25197148323059 + }, + { + "epoch": 61, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09662816899882423, + "train_seconds": 228.6902298927307, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09641593322753907 + }, + "validation_seconds": 18.250696659088135 + }, + { + "epoch": 62, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09674948098129696, + "train_seconds": 228.68463706970215, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09669037780761719 + }, + "validation_seconds": 18.25190758705139 + }, + { + "epoch": 63, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09671043863296509, + "train_seconds": 228.63868236541748, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0957095199584961 + }, + "validation_seconds": 18.234554290771484 + }, + { + "epoch": 64, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09636272977458106, + "train_seconds": 228.15654468536377, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09310142059326172 + }, + "validation_seconds": 18.236981868743896 + }, + { + "epoch": 65, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09650074934429592, + "train_seconds": 228.1613838672638, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09531402130126954 + }, + "validation_seconds": 18.24026322364807 + }, + { + "epoch": 66, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09634359340137906, + "train_seconds": 228.17927384376526, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09655115051269532 + }, + "validation_seconds": 18.235257625579834 + }, + { + "epoch": 67, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09597892744541169, + "train_seconds": 228.24445581436157, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09803577880859375 + }, + "validation_seconds": 18.258594512939453 + }, + { + "epoch": 68, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09632101210488213, + "train_seconds": 228.73210763931274, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09556730346679687 + }, + "validation_seconds": 18.272592306137085 + }, + { + "epoch": 69, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09618460634814369, + "train_seconds": 228.6269886493683, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09790851898193359 + }, + "validation_seconds": 18.23514723777771 + }, + { + "epoch": 70, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09652581750684314, + "train_seconds": 228.1185097694397, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09527283630371093 + }, + "validation_seconds": 18.23189115524292 + }, + { + "epoch": 71, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0956978734254837, + "train_seconds": 228.124103307724, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09942215576171876 + }, + "validation_seconds": 18.23433780670166 + }, + { + "epoch": 72, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09645447168217765, + "train_seconds": 228.13148307800293, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09943707580566406 + }, + "validation_seconds": 18.238800525665283 + }, + { + "epoch": 73, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09633775558339225, + "train_seconds": 228.1336727142334, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09661944732666015 + }, + "validation_seconds": 18.240188598632812 + }, + { + "epoch": 74, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09634426296022203, + "train_seconds": 228.14839053153992, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09583654479980469 + }, + "validation_seconds": 18.23649311065674 + }, + { + "epoch": 75, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09580984490447574, + "train_seconds": 228.13686203956604, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09728650207519532 + }, + "validation_seconds": 18.236411094665527 + }, + { + "epoch": 76, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09626505046553081, + "train_seconds": 228.14012694358826, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09705371398925781 + }, + "validation_seconds": 18.238210678100586 + }, + { + "epoch": 77, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09598139480352402, + "train_seconds": 228.11646938323975, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09914067993164062 + }, + "validation_seconds": 18.237971544265747 + }, + { + "epoch": 78, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0964213231338395, + "train_seconds": 228.11607599258423, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09502608642578125 + }, + "validation_seconds": 18.238613843917847 + }, + { + "epoch": 79, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09655453516244888, + "train_seconds": 228.1036672592163, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09777397918701172 + }, + "validation_seconds": 18.237144470214844 + }, + { + "epoch": 80, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09686183725330565, + "train_seconds": 228.09968042373657, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09582723693847656 + }, + "validation_seconds": 18.23502254486084 + }, + { + "epoch": 81, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09596207317643696, + "train_seconds": 228.0861895084381, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09246314697265624 + }, + "validation_seconds": 18.233213663101196 + }, + { + "epoch": 82, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09665712188747194, + "train_seconds": 228.10383939743042, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09760987091064453 + }, + "validation_seconds": 18.23534655570984 + }, + { + "epoch": 83, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09619577255249023, + "train_seconds": 228.10573935508728, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09589176177978516 + }, + "validation_seconds": 18.236335515975952 + }, + { + "epoch": 84, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09611746475564109, + "train_seconds": 228.09352922439575, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09739808807373047 + }, + "validation_seconds": 18.236633777618408 + }, + { + "epoch": 85, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09611809586021636, + "train_seconds": 228.1817696094513, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09837869110107422 + }, + "validation_seconds": 18.24014949798584 + }, + { + "epoch": 86, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09562978496021694, + "train_seconds": 228.21705293655396, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09497322692871094 + }, + "validation_seconds": 18.23759365081787 + }, + { + "epoch": 87, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09583760881291495, + "train_seconds": 228.14998769760132, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0954147964477539 + }, + "validation_seconds": 18.23929476737976 + }, + { + "epoch": 88, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09554384915033977, + "train_seconds": 228.06396865844727, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09424874877929687 + }, + "validation_seconds": 18.236170053482056 + }, + { + "epoch": 89, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09572083793216281, + "train_seconds": 228.18749070167542, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09340807495117187 + }, + "validation_seconds": 18.244277715682983 + }, + { + "epoch": 90, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09604131587876213, + "train_seconds": 228.1858491897583, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09667099304199218 + }, + "validation_seconds": 18.237085342407227 + }, + { + "epoch": 91, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09581272191471524, + "train_seconds": 228.2162148952484, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09316986083984374 + }, + "validation_seconds": 18.237507104873657 + }, + { + "epoch": 92, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09599215721024407, + "train_seconds": 228.16782188415527, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09519173889160157 + }, + "validation_seconds": 18.24071955680847 + }, + { + "epoch": 93, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09601165684858957, + "train_seconds": 228.21594524383545, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09708834075927734 + }, + "validation_seconds": 18.237923622131348 + }, + { + "epoch": 94, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0963773810784022, + "train_seconds": 228.19517064094543, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09413018035888672 + }, + "validation_seconds": 18.24626135826111 + }, + { + "epoch": 95, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09622333311769697, + "train_seconds": 228.20594668388367, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09718131561279297 + }, + "validation_seconds": 18.239275217056274 + }, + { + "epoch": 96, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09596092004113727, + "train_seconds": 228.1629090309143, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09380335388183594 + }, + "validation_seconds": 18.23921275138855 + }, + { + "epoch": 97, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0961335700088077, + "train_seconds": 228.2011206150055, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09569081115722657 + }, + "validation_seconds": 18.25113034248352 + }, + { + "epoch": 98, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0961867148703999, + "train_seconds": 228.19716882705688, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09790823364257813 + }, + "validation_seconds": 18.23853349685669 + }, + { + "epoch": 99, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09642944700850381, + "train_seconds": 228.1831932067871, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09380260467529297 + }, + "validation_seconds": 18.241806268692017 + }, + { + "epoch": 100, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09570178092850579, + "train_seconds": 228.12436485290527, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09370457153320312 + }, + "validation_seconds": 18.23480200767517 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.1, + "evaluation_seconds": 18.23977279663086, + "examples": 5000, + "finite": true, + "loss": 0.09370457153320312 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 6988224512, + "peak_memory_reserved_bytes": 8430551040, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 24661.099477529526, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 100, + "feedforward_example_passes": 5005000, + "forward_macs_per_example": 68862592, + "forward_parameter_count": 464154, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 118100000, + "logical_task_loss_queries": 0, + "num_trainable_layers": 32, + "ordinary_training_examples": 4500000, + "ordinary_validation_examples": 505000, + "relaxation_example_passes": 118100000, + "training_example_presentations": 4500000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json.manifest.json new file mode 100644 index 0000000..b2c97eb --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json.manifest.json @@ -0,0 +1,134 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "ep", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "100", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.0003", + "--output_lr", + "0.0003", + "--lr_schedule", + "constant", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--ep_beta", + "0.5", + "--ep_dt", + "0.5", + "--ep_free_steps", + "20", + "--ep_nudge_steps", + "4" + ], + "completed_unix_time": 1786927727.498714, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 24671.18036866188, + "epochs": 100, + "experiment_name": "resnet-r2-ep-d32-lr0p0003", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-2ddb8c2b-c2d4-96ac-fdd7-f2ded951f824", + "total_memory_mib": 49140 + }, + "lr_schedule": "constant", + "method": "ep", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d32-lr0p0003.json", + "output_exists": true, + "output_sha256": "421c661bd6c13d7df5840943805ae1f72ccfd629b8d41b994fe9ca5e2ed65771", + "rate": 0.0003, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json b/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json new file mode 100644 index 0000000..1b617af --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json @@ -0,0 +1,1565 @@ +{ + "architecture": { + "base_width": 16, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 853018, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 100, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.0003, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "constant", + "max_steps": 0, + "method": "ep", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json", + "output_lr": 0.0003, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.1049093120151096, + "train_seconds": 470.40172600746155, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09854823455810546 + }, + "validation_seconds": 42.355618476867676 + }, + { + "epoch": 2, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09679424016343223, + "train_seconds": 466.0533540248871, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09679457397460937 + }, + "validation_seconds": 42.082069396972656 + }, + { + "epoch": 3, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09715077778763241, + "train_seconds": 465.50510478019714, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09756388244628907 + }, + "validation_seconds": 41.84662628173828 + }, + { + "epoch": 4, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0970374315155877, + "train_seconds": 464.9111385345459, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0956327865600586 + }, + "validation_seconds": 42.19015383720398 + }, + { + "epoch": 5, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09693956523074045, + "train_seconds": 465.1891815662384, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09490829162597657 + }, + "validation_seconds": 41.88029718399048 + }, + { + "epoch": 6, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09654380642440584, + "train_seconds": 465.1731870174408, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0966841812133789 + }, + "validation_seconds": 42.02381706237793 + }, + { + "epoch": 7, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09670941620535321, + "train_seconds": 465.4678430557251, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09708750610351563 + }, + "validation_seconds": 41.63112282752991 + }, + { + "epoch": 8, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09702321206993526, + "train_seconds": 464.5645751953125, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09724942169189453 + }, + "validation_seconds": 41.95182228088379 + }, + { + "epoch": 9, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0974015151196056, + "train_seconds": 464.3173403739929, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0945250732421875 + }, + "validation_seconds": 41.758512020111084 + }, + { + "epoch": 10, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09567246304617988, + "train_seconds": 464.4738943576813, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09481480102539062 + }, + "validation_seconds": 41.882903814315796 + }, + { + "epoch": 11, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09578449822399351, + "train_seconds": 464.7627408504486, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09709003295898437 + }, + "validation_seconds": 41.83140850067139 + }, + { + "epoch": 12, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09586407614946366, + "train_seconds": 464.4240849018097, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0982633773803711 + }, + "validation_seconds": 41.97194838523865 + }, + { + "epoch": 13, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09627521215942171, + "train_seconds": 464.32663130760193, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09660293426513672 + }, + "validation_seconds": 41.93223762512207 + }, + { + "epoch": 14, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09617627680831485, + "train_seconds": 464.46084427833557, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09428795776367188 + }, + "validation_seconds": 42.017765522003174 + }, + { + "epoch": 15, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09564328297244178, + "train_seconds": 465.19329810142517, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09600251922607422 + }, + "validation_seconds": 42.234293937683105 + }, + { + "epoch": 16, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09563761240641276, + "train_seconds": 464.90938997268677, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09587730102539062 + }, + "validation_seconds": 42.10805058479309 + }, + { + "epoch": 17, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0960964433060752, + "train_seconds": 465.1038143634796, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09685572204589844 + }, + "validation_seconds": 42.27157926559448 + }, + { + "epoch": 18, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609683181047439, + "train_seconds": 466.1774206161499, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09549898681640626 + }, + "validation_seconds": 42.174476623535156 + }, + { + "epoch": 19, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09612310879892773, + "train_seconds": 466.1518986225128, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09561274871826173 + }, + "validation_seconds": 42.16247892379761 + }, + { + "epoch": 20, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09647270717223486, + "train_seconds": 465.53555250167847, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09512040252685547 + }, + "validation_seconds": 42.129403829574585 + }, + { + "epoch": 21, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09616433023346795, + "train_seconds": 465.08661556243896, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0983990463256836 + }, + "validation_seconds": 42.04495024681091 + }, + { + "epoch": 22, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09624761987659666, + "train_seconds": 464.83717226982117, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0956104736328125 + }, + "validation_seconds": 42.02618384361267 + }, + { + "epoch": 23, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09572103581825891, + "train_seconds": 465.7274680137634, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0967745132446289 + }, + "validation_seconds": 42.163888931274414 + }, + { + "epoch": 24, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09640631458759308, + "train_seconds": 465.5332636833191, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09754253997802734 + }, + "validation_seconds": 42.26589584350586 + }, + { + "epoch": 25, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09652167629798253, + "train_seconds": 465.39615631103516, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09833413848876953 + }, + "validation_seconds": 42.07328724861145 + }, + { + "epoch": 26, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09637802203363842, + "train_seconds": 465.0871801376343, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09762132720947266 + }, + "validation_seconds": 41.8577835559845 + }, + { + "epoch": 27, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09588909540838665, + "train_seconds": 465.2824354171753, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09559941711425782 + }, + "validation_seconds": 42.20577144622803 + }, + { + "epoch": 28, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09600038005246056, + "train_seconds": 465.70421624183655, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09474451599121093 + }, + "validation_seconds": 41.94756460189819 + }, + { + "epoch": 29, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09606572427219814, + "train_seconds": 465.30998396873474, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09541961975097656 + }, + "validation_seconds": 42.07582426071167 + }, + { + "epoch": 30, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09589181250201331, + "train_seconds": 465.0965838432312, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09653666229248047 + }, + "validation_seconds": 42.29110312461853 + }, + { + "epoch": 31, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09631023094124264, + "train_seconds": 465.18078541755676, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0960676986694336 + }, + "validation_seconds": 42.05308723449707 + }, + { + "epoch": 32, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09603364160723156, + "train_seconds": 464.56597423553467, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09548921508789063 + }, + "validation_seconds": 41.80490756034851 + }, + { + "epoch": 33, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09560389386018117, + "train_seconds": 464.77473068237305, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09516982727050781 + }, + "validation_seconds": 42.142924308776855 + }, + { + "epoch": 34, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09607277266846763, + "train_seconds": 464.5014383792877, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09546457977294921 + }, + "validation_seconds": 42.0104558467865 + }, + { + "epoch": 35, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09576119128863017, + "train_seconds": 464.63925290107727, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09528033142089844 + }, + "validation_seconds": 42.067020416259766 + }, + { + "epoch": 36, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09611127299070359, + "train_seconds": 465.3532955646515, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09493963012695313 + }, + "validation_seconds": 41.93064761161804 + }, + { + "epoch": 37, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09682440170579486, + "train_seconds": 464.56911444664, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09636844940185547 + }, + "validation_seconds": 41.72228002548218 + }, + { + "epoch": 38, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09608428453736835, + "train_seconds": 463.5885651111603, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09487273864746094 + }, + "validation_seconds": 41.65004539489746 + }, + { + "epoch": 39, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0963151536266009, + "train_seconds": 464.0708773136139, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09748384704589844 + }, + "validation_seconds": 41.8119695186615 + }, + { + "epoch": 40, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09639640080663893, + "train_seconds": 464.3524947166443, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0965095932006836 + }, + "validation_seconds": 41.94960570335388 + }, + { + "epoch": 41, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0956977930717998, + "train_seconds": 464.94561409950256, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0935846435546875 + }, + "validation_seconds": 42.05508899688721 + }, + { + "epoch": 42, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09583505643606186, + "train_seconds": 465.431129693985, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09561628875732422 + }, + "validation_seconds": 41.988160371780396 + }, + { + "epoch": 43, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09605752380688985, + "train_seconds": 465.50451612472534, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09550503692626953 + }, + "validation_seconds": 42.18235206604004 + }, + { + "epoch": 44, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0956310129072931, + "train_seconds": 465.6041624546051, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09446280822753907 + }, + "validation_seconds": 42.08299422264099 + }, + { + "epoch": 45, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09536773497660955, + "train_seconds": 465.80098724365234, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09453959808349609 + }, + "validation_seconds": 42.09350252151489 + }, + { + "epoch": 46, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0958818547460768, + "train_seconds": 466.4351236820221, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09704439849853516 + }, + "validation_seconds": 42.08637857437134 + }, + { + "epoch": 47, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09645931057135264, + "train_seconds": 466.09434366226196, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0961579849243164 + }, + "validation_seconds": 42.06746745109558 + }, + { + "epoch": 48, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0954085854821735, + "train_seconds": 465.25619769096375, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0969421142578125 + }, + "validation_seconds": 41.731086015701294 + }, + { + "epoch": 49, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09619432701004876, + "train_seconds": 465.3983533382416, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0965331298828125 + }, + "validation_seconds": 41.816789865493774 + }, + { + "epoch": 50, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09600847476058536, + "train_seconds": 465.3787190914154, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09710611572265625 + }, + "validation_seconds": 42.01187825202942 + }, + { + "epoch": 51, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09619079148901834, + "train_seconds": 465.54860496520996, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0951966293334961 + }, + "validation_seconds": 42.02596402168274 + }, + { + "epoch": 52, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09627778605222702, + "train_seconds": 465.53260564804077, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09787473602294922 + }, + "validation_seconds": 41.86947846412659 + }, + { + "epoch": 53, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09646207979785071, + "train_seconds": 465.1533234119415, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09574561614990235 + }, + "validation_seconds": 41.8131160736084 + }, + { + "epoch": 54, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09629596228996913, + "train_seconds": 465.05003571510315, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0952306869506836 + }, + "validation_seconds": 41.844584703445435 + }, + { + "epoch": 55, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09640434854692884, + "train_seconds": 465.1228246688843, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09881563262939454 + }, + "validation_seconds": 41.718056440353394 + }, + { + "epoch": 56, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0962243594262335, + "train_seconds": 464.6118927001953, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09332275543212891 + }, + "validation_seconds": 41.883305311203 + }, + { + "epoch": 57, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09593700580994288, + "train_seconds": 464.6503734588623, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0978786376953125 + }, + "validation_seconds": 42.113295555114746 + }, + { + "epoch": 58, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609533179865944, + "train_seconds": 464.24342918395996, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09458008117675781 + }, + "validation_seconds": 41.631030797958374 + }, + { + "epoch": 59, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09571239232884513, + "train_seconds": 464.2138133049011, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09771599426269531 + }, + "validation_seconds": 41.696253299713135 + }, + { + "epoch": 60, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09557220001882977, + "train_seconds": 464.86264300346375, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09608076782226563 + }, + "validation_seconds": 41.83806109428406 + }, + { + "epoch": 61, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09603700024021997, + "train_seconds": 464.84575176239014, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09600220947265625 + }, + "validation_seconds": 41.66805815696716 + }, + { + "epoch": 62, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09620499748918745, + "train_seconds": 464.15450048446655, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09603590545654297 + }, + "validation_seconds": 41.74782752990723 + }, + { + "epoch": 63, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0962453338013755, + "train_seconds": 464.59076714515686, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09555384216308593 + }, + "validation_seconds": 41.62419104576111 + }, + { + "epoch": 64, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09611023871501287, + "train_seconds": 464.39524030685425, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09319230194091797 + }, + "validation_seconds": 41.61292552947998 + }, + { + "epoch": 65, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09617606501976649, + "train_seconds": 464.5210211277008, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09480837554931641 + }, + "validation_seconds": 41.89604687690735 + }, + { + "epoch": 66, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0961561086707645, + "train_seconds": 464.26877069473267, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09668070220947265 + }, + "validation_seconds": 41.73206186294556 + }, + { + "epoch": 67, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09578220612340503, + "train_seconds": 464.26437520980835, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09633714294433594 + }, + "validation_seconds": 41.749727964401245 + }, + { + "epoch": 68, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09601921813752916, + "train_seconds": 464.677282333374, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0954210693359375 + }, + "validation_seconds": 41.653220653533936 + }, + { + "epoch": 69, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0958134253554874, + "train_seconds": 464.6186845302582, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09663692932128906 + }, + "validation_seconds": 41.856159687042236 + }, + { + "epoch": 70, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09630409976376428, + "train_seconds": 463.54873037338257, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09525140838623047 + }, + "validation_seconds": 41.317835092544556 + }, + { + "epoch": 71, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09535665365589989, + "train_seconds": 462.5900402069092, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09930306091308594 + }, + "validation_seconds": 41.25684332847595 + }, + { + "epoch": 72, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09590638970666461, + "train_seconds": 462.9886476993561, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09919052581787109 + }, + "validation_seconds": 41.03996682167053 + }, + { + "epoch": 73, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09605056209696664, + "train_seconds": 462.34330105781555, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09603641967773438 + }, + "validation_seconds": 41.09140729904175 + }, + { + "epoch": 74, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09621421782573064, + "train_seconds": 462.92142152786255, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09542268829345703 + }, + "validation_seconds": 41.37644982337952 + }, + { + "epoch": 75, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09559676223993302, + "train_seconds": 462.5050027370453, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.095793212890625 + }, + "validation_seconds": 41.256924629211426 + }, + { + "epoch": 76, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09604653212229411, + "train_seconds": 462.49127101898193, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09619957275390625 + }, + "validation_seconds": 41.06153750419617 + }, + { + "epoch": 77, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09564897516965866, + "train_seconds": 462.3650414943695, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09792278289794921 + }, + "validation_seconds": 41.19801068305969 + }, + { + "epoch": 78, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09598682165808148, + "train_seconds": 462.66747975349426, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09466236419677734 + }, + "validation_seconds": 41.33188843727112 + }, + { + "epoch": 79, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09604800218343734, + "train_seconds": 463.091757774353, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09714752655029296 + }, + "validation_seconds": 41.29967379570007 + }, + { + "epoch": 80, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0963094130622016, + "train_seconds": 462.95998096466064, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09555538482666015 + }, + "validation_seconds": 41.277440547943115 + }, + { + "epoch": 81, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09557393687168757, + "train_seconds": 462.97270131111145, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09269241485595703 + }, + "validation_seconds": 41.38614511489868 + }, + { + "epoch": 82, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09606576175159878, + "train_seconds": 462.94847536087036, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09701722106933594 + }, + "validation_seconds": 41.294069051742554 + }, + { + "epoch": 83, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09575901180903117, + "train_seconds": 463.94762802124023, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09515934143066407 + }, + "validation_seconds": 41.48845863342285 + }, + { + "epoch": 84, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09559025944206449, + "train_seconds": 463.4072983264923, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09719501342773437 + }, + "validation_seconds": 41.55417251586914 + }, + { + "epoch": 85, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09569342220359378, + "train_seconds": 462.94681215286255, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09763941040039062 + }, + "validation_seconds": 41.40012240409851 + }, + { + "epoch": 86, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09554593718979094, + "train_seconds": 463.18996381759644, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09514545745849609 + }, + "validation_seconds": 41.31304478645325 + }, + { + "epoch": 87, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09627738233407339, + "train_seconds": 463.1316273212433, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.0954114501953125 + }, + "validation_seconds": 41.61716270446777 + }, + { + "epoch": 88, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09551246789958742, + "train_seconds": 463.3723156452179, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09582576599121094 + }, + "validation_seconds": 41.4737389087677 + }, + { + "epoch": 89, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09581275145345264, + "train_seconds": 463.5044505596161, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09365298461914062 + }, + "validation_seconds": 41.32333970069885 + }, + { + "epoch": 90, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09594039207034641, + "train_seconds": 463.22664976119995, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09755116119384766 + }, + "validation_seconds": 41.452555894851685 + }, + { + "epoch": 91, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09589890376726787, + "train_seconds": 463.3689317703247, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09362464294433594 + }, + "validation_seconds": 41.49372625350952 + }, + { + "epoch": 92, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09604071430762609, + "train_seconds": 463.23748564720154, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09576106414794922 + }, + "validation_seconds": 41.29102540016174 + }, + { + "epoch": 93, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0958403738392724, + "train_seconds": 462.9280080795288, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09767513275146485 + }, + "validation_seconds": 41.327802896499634 + }, + { + "epoch": 94, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09607556787464354, + "train_seconds": 463.41283416748047, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09459663238525391 + }, + "validation_seconds": 41.50932788848877 + }, + { + "epoch": 95, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09609565815925598, + "train_seconds": 463.442754983902, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09667721099853516 + }, + "validation_seconds": 41.346176624298096 + }, + { + "epoch": 96, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09569309736887614, + "train_seconds": 462.8477804660797, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09482781524658203 + }, + "validation_seconds": 41.56019306182861 + }, + { + "epoch": 97, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09579816060596043, + "train_seconds": 463.14385962486267, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09566875915527344 + }, + "validation_seconds": 41.318729877471924 + }, + { + "epoch": 98, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.0960154987997479, + "train_seconds": 463.26160764694214, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09674229278564453 + }, + "validation_seconds": 41.44693160057068 + }, + { + "epoch": 99, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09639488592677646, + "train_seconds": 462.9610106945038, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09510955657958985 + }, + "validation_seconds": 41.40132117271423 + }, + { + "epoch": 100, + "lr": 0.0003, + "output_lr": 0.0003, + "train_examples": 45000, + "train_loss": 0.09564637204408645, + "train_seconds": 463.3234283924103, + "validation": { + "accuracy": 0.1, + "examples": 5000, + "loss": 0.09450665283203125 + }, + "validation_seconds": 41.3101544380188 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.1, + "evaluation_seconds": 41.240410566329956, + "examples": 5000, + "finite": true, + "loss": 0.09450665283203125 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 11683346432, + "peak_memory_reserved_bytes": 13419675648, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 50663.67151045799, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 100, + "feedforward_example_passes": 5005000, + "forward_macs_per_example": 125485696, + "forward_parameter_count": 853018, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 118100000, + "logical_task_loss_queries": 0, + "num_trainable_layers": 56, + "ordinary_training_examples": 4500000, + "ordinary_validation_examples": 505000, + "relaxation_example_passes": 118100000, + "training_example_presentations": 4500000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json.manifest.json new file mode 100644 index 0000000..c191ccc --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json.manifest.json @@ -0,0 +1,134 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "ep", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "100", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.0003", + "--output_lr", + "0.0003", + "--lr_schedule", + "constant", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--ep_beta", + "0.5", + "--ep_dt", + "0.5", + "--ep_free_steps", + "20", + "--ep_nudge_steps", + "4" + ], + "completed_unix_time": 1786208159.8132813, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 50669.00653362274, + "epochs": 100, + "experiment_name": "resnet-r2-ep-d56-lr0p0003", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-b520421c-b120-70f0-eae8-1ff79ad97bf5", + "total_memory_mib": 49140 + }, + "lr_schedule": "constant", + "method": "ep", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ep-d56-lr0p0003.json", + "output_exists": true, + "output_sha256": "f582283d07e8d090b09b62270031ef5fe81d0e79c3da4b1ff4e23084265ec957", + "rate": 0.0003, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json new file mode 100644 index 0000000..1096d53 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json @@ -0,0 +1,2669 @@ +{ + "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": 267904, + "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": 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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "hfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.011225155787542462, + "feedback_forward_cosine": [ + 0.06530784070491791, + 0.23644882440567017, + 0.14832958579063416, + 0.27313339710235596, + 0.11167817562818527, + 0.22827664017677307, + 0.11369006335735321, + 0.18833361566066742, + 0.08932708203792572, + 0.2524581551551819, + 0.10242670029401779, + 0.24816101789474487, + 0.08819255232810974, + 0.19018340110778809, + 0.08682424575090408, + 0.2735845446586609, + 0.09671223163604736, + 0.26411014795303345, + 0.7856658697128296 + ], + "feedback_forward_norm_ratio": [ + 1.5932362079620361, + 1.5769914388656616, + 1.9141095876693726, + 1.9809224605560303, + 2.1602015495300293, + 2.258635997772217, + 2.286642551422119, + 1.7141746282577515, + 2.1372995376586914, + 1.9748387336730957, + 2.4078481197357178, + 2.3249430656433105, + 2.3330516815185547, + 1.5594871044158936, + 1.7516019344329834, + 1.3729456663131714, + 1.6120082139968872, + 1.1900783777236938, + 0.22846350073814392 + ], + "feedback_forward_relative_error": [ + 1.8249109983444214, + 1.6556408405303955, + 2.0238521099090576, + 1.9600871801376343, + 2.276834726333618, + 2.2517213821411133, + 2.3893086910247803, + 1.8145856857299805, + 2.2773256301879883, + 1.9755656719207764, + 2.5108718872070312, + 2.2916018962860107, + 2.455934762954712, + 1.6848807334899902, + 1.940089225769043, + 1.4607349634170532, + 1.812944769859314, + 1.3370351791381836, + 0.8325884938240051 + ], + "innovation_negative_gradient_cosine": [ + 0.0030611385591328144, + 0.006984669715166092, + 0.015605964697897434, + 0.010937267914414406, + 0.02078421413898468, + 0.009977679699659348, + 0.03010556846857071, + 0.02181582897901535, + 0.06369832903146744, + 0.040703386068344116, + 0.09983091056346893, + 0.06209838390350342, + 0.13944962620735168, + 0.11991359293460846, + 0.28113487362861633, + 0.28801897168159485, + 0.4426411986351013, + 0.47109293937683105, + 0.8467425107955933 + ], + "mean_feedback_forward_cosine": 0.2022549521766211, + "mean_feedback_forward_relative_error": 1.9356060435897426, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0030611385591328144, + 0.006984669715166092, + 0.015605964697897434, + 0.010937267914414406, + 0.02078421413898468, + 0.009977679699659348, + 0.03010556846857071, + 0.02181582897901535, + 0.06369832903146744, + 0.040703386068344116, + 0.09983091056346893, + 0.06209838390350342, + 0.13944962620735168, + 0.11991359293460846, + 0.28113487362861633, + 0.28801897168159485, + 0.4426411986351013, + 0.47109293937683105, + 0.8467425107955933 + ], + "teaching_negative_gradient_cosine": [ + 0.0030611385591328144, + 0.006984669715166092, + 0.015605964697897434, + 0.010937267914414406, + 0.02078421413898468, + 0.009977679699659348, + 0.03010556846857071, + 0.02181582897901535, + 0.06369832903146744, + 0.040703386068344116, + 0.09983091056346893, + 0.06209838390350342, + 0.13944962620735168, + 0.11991359293460846, + 0.28113487362861633, + 0.28801897168159485, + 0.4426411986351013, + 0.47109293937683105, + 0.8467425107955933 + ], + "wall_s": 0.10387682914733887 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.3056, + "eval_loss": 1.9553183349609375, + "eval_split": "validation", + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.0639842291302153 + }, + { + "epoch": 2, + "eval_accuracy": 0.3414, + "eval_loss": 1.8812863037109375, + "eval_split": "validation", + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.91367295850118 + }, + { + "epoch": 3, + "eval_accuracy": 0.3298, + "eval_loss": 1.8941572509765625, + "eval_split": "validation", + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 2.0059402748743693 + }, + { + "epoch": 4, + "eval_accuracy": 0.3868, + "eval_loss": 2.012137060546875, + "eval_split": "validation", + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 2.1139173667060005 + }, + { + "epoch": 5, + "eval_accuracy": 0.3448, + "eval_loss": 2.873762353515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 2.4378785884433323 + }, + { + "epoch": 6, + "eval_accuracy": 0.4028, + "eval_loss": 2.19697216796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 2.382185635460748 + }, + { + "epoch": 7, + "eval_accuracy": 0.3742, + "eval_loss": 2.617165966796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 2.240234103329976 + }, + { + "epoch": 8, + "eval_accuracy": 0.3884, + "eval_loss": 2.572106884765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 2.071946952946981 + }, + { + "epoch": 9, + "eval_accuracy": 0.4522, + "eval_loss": 1.9747636474609376, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 2.058614935133192 + }, + { + "epoch": 10, + "eval_accuracy": 0.432, + "eval_loss": 2.391416796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 1.9395815348519219 + }, + { + "epoch": 11, + "eval_accuracy": 0.3852, + "eval_loss": 2.6288126953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 1.9402141478644477 + }, + { + "epoch": 12, + "eval_accuracy": 0.419, + "eval_loss": 2.22531181640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 1.9182814904106988 + }, + { + "epoch": 13, + "eval_accuracy": 0.4612, + "eval_loss": 1.892072021484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 1.7968032083511352 + }, + { + "epoch": 14, + "eval_accuracy": 0.417, + "eval_loss": 2.081826513671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 1.8566136685053507 + }, + { + "epoch": 15, + "eval_accuracy": 0.438, + "eval_loss": 1.9989041015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 1.7629460686153835 + }, + { + "epoch": 16, + "eval_accuracy": 0.458, + "eval_loss": 2.0938974365234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 1.7258217601776122 + }, + { + "epoch": 17, + "eval_accuracy": 0.4368, + "eval_loss": 2.362848486328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 1.7666962187872992 + }, + { + "epoch": 18, + "eval_accuracy": 0.4846, + "eval_loss": 1.92572080078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 1.7164832483503554 + }, + { + "epoch": 19, + "eval_accuracy": 0.494, + "eval_loss": 1.8626768310546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 1.6503403984917535 + }, + { + "epoch": 20, + "eval_accuracy": 0.3992, + "eval_loss": 3.14110546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 1.7138124549018012 + }, + { + "epoch": 21, + "eval_accuracy": 0.4812, + "eval_loss": 1.944050341796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 1.6447507121192084 + }, + { + "epoch": 22, + "eval_accuracy": 0.4892, + "eval_loss": 1.911722021484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 1.6547758380042181 + }, + { + "epoch": 23, + "eval_accuracy": 0.485, + "eval_loss": 1.815741455078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 1.5776512994130452 + }, + { + "epoch": 24, + "eval_accuracy": 0.5166, + "eval_loss": 1.5387634521484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 1.61900724834866 + }, + { + "epoch": 25, + "eval_accuracy": 0.4492, + "eval_loss": 2.1026521484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 1.6446258316040039 + }, + { + "epoch": 26, + "eval_accuracy": 0.461, + "eval_loss": 1.787435693359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 1.635316059621175 + }, + { + "epoch": 27, + "eval_accuracy": 0.4612, + "eval_loss": 2.1551716796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 1.624604059685601 + }, + { + "epoch": 28, + "eval_accuracy": 0.4852, + "eval_loss": 1.75483876953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 1.6604180307176377 + }, + { + "epoch": 29, + "eval_accuracy": 0.5064, + "eval_loss": 1.5234459716796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 1.585603558646308 + }, + { + "epoch": 30, + "eval_accuracy": 0.4848, + "eval_loss": 1.8498150634765624, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 1.5916083756128947 + }, + { + "epoch": 31, + "eval_accuracy": 0.4718, + "eval_loss": 1.84035732421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 1.6653513647927178 + }, + { + "epoch": 32, + "eval_accuracy": 0.5348, + "eval_loss": 1.6685882568359376, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 1.6522833022647434 + }, + { + "epoch": 33, + "eval_accuracy": 0.463, + "eval_loss": 2.010286328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 1.581467076174418 + }, + { + "epoch": 34, + "eval_accuracy": 0.4372, + "eval_loss": 2.1312464111328127, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 1.6292456374486288 + }, + { + "epoch": 35, + "eval_accuracy": 0.5374, + "eval_loss": 1.577523291015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 1.5334896649254692 + }, + { + "epoch": 36, + "eval_accuracy": 0.4628, + "eval_loss": 1.90355205078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 1.504615032196045 + }, + { + "epoch": 37, + "eval_accuracy": 0.5238, + "eval_loss": 1.5645536376953124, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 1.596086120054457 + }, + { + "epoch": 38, + "eval_accuracy": 0.5198, + "eval_loss": 1.5882181884765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 1.5680132107628717 + }, + { + "epoch": 39, + "eval_accuracy": 0.5282, + "eval_loss": 1.6558640380859375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 1.5771280631595188 + }, + { + "epoch": 40, + "eval_accuracy": 0.5684, + "eval_loss": 1.37480546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 1.6025711276796129 + }, + { + "epoch": 41, + "eval_accuracy": 0.5168, + "eval_loss": 1.548530712890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 1.5799489162021214 + }, + { + "epoch": 42, + "eval_accuracy": 0.4688, + "eval_loss": 2.0833614990234377, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 1.5346269664764405 + }, + { + "epoch": 43, + "eval_accuracy": 0.5302, + "eval_loss": 1.5688693359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 1.5281175550248889 + }, + { + "epoch": 44, + "eval_accuracy": 0.483, + "eval_loss": 1.8670911376953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 1.4591134276707967 + }, + { + "epoch": 45, + "eval_accuracy": 0.5122, + "eval_loss": 1.7008560546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 1.4886928064134386 + }, + { + "epoch": 46, + "eval_accuracy": 0.5714, + "eval_loss": 1.379425, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 1.5013230222066243 + }, + { + "epoch": 47, + "eval_accuracy": 0.5548, + "eval_loss": 1.540955224609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 1.423521467505561 + }, + { + "epoch": 48, + "eval_accuracy": 0.5066, + "eval_loss": 1.8178400634765626, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 1.473559805340237 + }, + { + "epoch": 49, + "eval_accuracy": 0.5778, + "eval_loss": 1.350400732421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 1.512079693031311 + }, + { + "epoch": 50, + "eval_accuracy": 0.5722, + "eval_loss": 1.516371484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 1.461774546453688 + }, + { + "epoch": 51, + "eval_accuracy": 0.507, + "eval_loss": 1.701910693359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 1.464529783630371 + }, + { + "epoch": 52, + "eval_accuracy": 0.5038, + "eval_loss": 1.78495341796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 1.5189528352101644 + }, + { + "epoch": 53, + "eval_accuracy": 0.5186, + "eval_loss": 1.6093507080078124, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 1.4640200068155924 + }, + { + "epoch": 54, + "eval_accuracy": 0.52, + "eval_loss": 1.6729033447265624, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 1.4083326490402222 + }, + { + "epoch": 55, + "eval_accuracy": 0.5778, + "eval_loss": 1.3512170654296876, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 1.416956041378445 + }, + { + "epoch": 56, + "eval_accuracy": 0.5506, + "eval_loss": 1.6091921142578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 1.4798751659181384 + }, + { + "epoch": 57, + "eval_accuracy": 0.512, + "eval_loss": 1.6767433837890624, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 1.4291151117112901 + }, + { + "epoch": 58, + "eval_accuracy": 0.545, + "eval_loss": 1.5081556396484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 1.3918506565093993 + }, + { + "epoch": 59, + "eval_accuracy": 0.539, + "eval_loss": 1.5954337890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 1.5130316936280992 + }, + { + "epoch": 60, + "eval_accuracy": 0.5556, + "eval_loss": 1.492790185546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 1.47905190410614 + }, + { + "epoch": 61, + "eval_accuracy": 0.521, + "eval_loss": 1.7738572021484376, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 1.3852702054129706 + }, + { + "epoch": 62, + "eval_accuracy": 0.5432, + "eval_loss": 1.5759424560546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 1.4184113253699409 + }, + { + "epoch": 63, + "eval_accuracy": 0.5626, + "eval_loss": 1.61301494140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 1.5066941794501412 + }, + { + "epoch": 64, + "eval_accuracy": 0.5646, + "eval_loss": 1.675884912109375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 1.4340424686219957 + }, + { + "epoch": 65, + "eval_accuracy": 0.5372, + "eval_loss": 1.64473779296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 1.4120538077672322 + }, + { + "epoch": 66, + "eval_accuracy": 0.5084, + "eval_loss": 1.7685004150390624, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 1.4140199373457167 + }, + { + "epoch": 67, + "eval_accuracy": 0.5182, + "eval_loss": 1.8401543212890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 1.4402784141964382 + }, + { + "epoch": 68, + "eval_accuracy": 0.5674, + "eval_loss": 1.4143515380859375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 1.4850667947769165 + }, + { + "epoch": 69, + "eval_accuracy": 0.581, + "eval_loss": 1.3920767578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 1.3742551008436414 + }, + { + "epoch": 70, + "eval_accuracy": 0.526, + "eval_loss": 1.6470415283203126, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 1.439255021180047 + }, + { + "epoch": 71, + "eval_accuracy": 0.5476, + "eval_loss": 1.4234358154296876, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 1.402162924469842 + }, + { + "epoch": 72, + "eval_accuracy": 0.527, + "eval_loss": 1.6882696044921874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 1.3987669376797147 + }, + { + "epoch": 73, + "eval_accuracy": 0.5564, + "eval_loss": 1.527014453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 1.3347418962902493 + }, + { + "epoch": 74, + "eval_accuracy": 0.497, + "eval_loss": 1.6632053955078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 1.3455509513643054 + }, + { + "epoch": 75, + "eval_accuracy": 0.5554, + "eval_loss": 1.44293671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 1.3510505442301433 + }, + { + "epoch": 76, + "eval_accuracy": 0.562, + "eval_loss": 1.5632307861328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 1.4097723316192627 + }, + { + "epoch": 77, + "eval_accuracy": 0.4878, + "eval_loss": 1.86476484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 1.3837750133726332 + }, + { + "epoch": 78, + "eval_accuracy": 0.5332, + "eval_loss": 1.5847595947265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 1.396502939499749 + }, + { + "epoch": 79, + "eval_accuracy": 0.5112, + "eval_loss": 1.5564174560546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 1.3709289339065551 + }, + { + "epoch": 80, + "eval_accuracy": 0.5394, + "eval_loss": 1.9737873046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 1.4022963900883993 + }, + { + "epoch": 81, + "eval_accuracy": 0.5584, + "eval_loss": 1.5171572021484374, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 1.370726917775472 + }, + { + "epoch": 82, + "eval_accuracy": 0.5358, + "eval_loss": 1.4977206787109374, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 1.4500145688374837 + }, + { + "epoch": 83, + "eval_accuracy": 0.5152, + "eval_loss": 1.8416988037109374, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 1.4227493050893147 + }, + { + "epoch": 84, + "eval_accuracy": 0.5836, + "eval_loss": 1.352899951171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 1.3736794021606444 + }, + { + "epoch": 85, + "eval_accuracy": 0.5584, + "eval_loss": 1.6942635498046874, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 1.389241146045261 + }, + { + "epoch": 86, + "eval_accuracy": 0.5596, + "eval_loss": 1.4475465087890624, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 1.270623020532396 + }, + { + "epoch": 87, + "eval_accuracy": 0.5794, + "eval_loss": 1.3361758056640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 1.3661915603425767 + }, + { + "epoch": 88, + "eval_accuracy": 0.5566, + "eval_loss": 1.4134527587890624, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 1.3660098644044665 + }, + { + "epoch": 89, + "eval_accuracy": 0.53, + "eval_loss": 1.5531100341796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 1.2967509341133965 + }, + { + "epoch": 90, + "eval_accuracy": 0.51, + "eval_loss": 1.7390384521484374, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 1.3260442124684653 + }, + { + "epoch": 91, + "eval_accuracy": 0.5874, + "eval_loss": 1.3916345458984376, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 1.3153838235431248 + }, + { + "epoch": 92, + "eval_accuracy": 0.5636, + "eval_loss": 1.4337053955078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 1.2993551434834798 + }, + { + "epoch": 93, + "eval_accuracy": 0.533, + "eval_loss": 1.6667333984375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 1.3019694290796915 + }, + { + "epoch": 94, + "eval_accuracy": 0.5652, + "eval_loss": 1.4489898681640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 1.2822739417394002 + }, + { + "epoch": 95, + "eval_accuracy": 0.5668, + "eval_loss": 1.419910986328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 1.3006300278769598 + }, + { + "epoch": 96, + "eval_accuracy": 0.5394, + "eval_loss": 1.69527265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 1.238620239914788 + }, + { + "epoch": 97, + "eval_accuracy": 0.513, + "eval_loss": 1.7494765869140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 1.2671517435497708 + }, + { + "epoch": 98, + "eval_accuracy": 0.5838, + "eval_loss": 1.2681653076171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 1.3089203299416436 + }, + { + "epoch": 99, + "eval_accuracy": 0.5436, + "eval_loss": 1.52252724609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 1.3226820044411554 + }, + { + "epoch": 100, + "eval_accuracy": 0.6308, + "eval_loss": 1.1525076416015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 1.2924140570958456 + }, + { + "epoch": 101, + "eval_accuracy": 0.682, + "eval_loss": 0.922958056640625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.9534723613103231 + }, + { + "epoch": 102, + "eval_accuracy": 0.6796, + "eval_loss": 0.9390528076171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.9059994698312548 + }, + { + "epoch": 103, + "eval_accuracy": 0.686, + "eval_loss": 0.9127562133789062, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.8874733817100525 + }, + { + "epoch": 104, + "eval_accuracy": 0.6914, + "eval_loss": 0.89916376953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.8901613323529561 + }, + { + "epoch": 105, + "eval_accuracy": 0.6866, + "eval_loss": 0.9040415405273438, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.8719133009168837 + }, + { + "epoch": 106, + "eval_accuracy": 0.6934, + "eval_loss": 0.91411796875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.8583287696414523 + }, + { + "epoch": 107, + "eval_accuracy": 0.6924, + "eval_loss": 0.89068076171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.8588273622512818 + }, + { + "epoch": 108, + "eval_accuracy": 0.6922, + "eval_loss": 0.9007458984375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.8597999311447143 + }, + { + "epoch": 109, + "eval_accuracy": 0.7014, + "eval_loss": 0.8806451171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.8486230295923021 + }, + { + "epoch": 110, + "eval_accuracy": 0.689, + "eval_loss": 0.9200911865234375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.8488019719229805 + }, + { + "epoch": 111, + "eval_accuracy": 0.7036, + "eval_loss": 0.8662437866210938, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.8433535183800591 + }, + { + "epoch": 112, + "eval_accuracy": 0.6954, + "eval_loss": 0.8857200561523437, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.8497770226160685 + }, + { + "epoch": 113, + "eval_accuracy": 0.696, + "eval_loss": 0.8835077026367187, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.8537612689124213 + }, + { + "epoch": 114, + "eval_accuracy": 0.6884, + "eval_loss": 0.8951915893554687, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.8630593092494541 + }, + { + "epoch": 115, + "eval_accuracy": 0.692, + "eval_loss": 0.89095966796875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.8638030516730415 + }, + { + "epoch": 116, + "eval_accuracy": 0.7048, + "eval_loss": 0.877529052734375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.8609290717230903 + }, + { + "epoch": 117, + "eval_accuracy": 0.7066, + "eval_loss": 0.8617789794921875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.8482380716853671 + }, + { + "epoch": 118, + "eval_accuracy": 0.6902, + "eval_loss": 0.901466943359375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.8500684111277262 + }, + { + "epoch": 119, + "eval_accuracy": 0.6872, + "eval_loss": 0.9289511840820313, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.8877813004069858 + }, + { + "epoch": 120, + "eval_accuracy": 0.6874, + "eval_loss": 0.937879931640625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.9034233251783583 + }, + { + "epoch": 121, + "eval_accuracy": 0.6836, + "eval_loss": 0.9336000122070313, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.9166675170262655 + }, + { + "epoch": 122, + "eval_accuracy": 0.6874, + "eval_loss": 0.919823388671875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.8952297897338867 + }, + { + "epoch": 123, + "eval_accuracy": 0.6752, + "eval_loss": 0.9460063232421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.9035753014034695 + }, + { + "epoch": 124, + "eval_accuracy": 0.681, + "eval_loss": 0.939022802734375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.908453873623742 + }, + { + "epoch": 125, + "eval_accuracy": 0.6914, + "eval_loss": 0.9120480102539062, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.9134819952646891 + }, + { + "epoch": 126, + "eval_accuracy": 0.684, + "eval_loss": 0.9374011962890625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.906762312274509 + }, + { + "epoch": 127, + "eval_accuracy": 0.6816, + "eval_loss": 0.9479945434570313, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.900509298239814 + }, + { + "epoch": 128, + "eval_accuracy": 0.6754, + "eval_loss": 0.96343857421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.8953627472347684 + }, + { + "epoch": 129, + "eval_accuracy": 0.6742, + "eval_loss": 0.968261376953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.9279563033845689 + }, + { + "epoch": 130, + "eval_accuracy": 0.6846, + "eval_loss": 0.9416985961914063, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.9155617316775851 + }, + { + "epoch": 131, + "eval_accuracy": 0.6666, + "eval_loss": 0.9798123779296875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.9026238541709052 + }, + { + "epoch": 132, + "eval_accuracy": 0.6714, + "eval_loss": 0.96294951171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.9165382932874891 + }, + { + "epoch": 133, + "eval_accuracy": 0.678, + "eval_loss": 0.9479088256835938, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.8981040594206916 + }, + { + "epoch": 134, + "eval_accuracy": 0.6714, + "eval_loss": 0.9769468994140625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.9077048551559448 + }, + { + "epoch": 135, + "eval_accuracy": 0.6674, + "eval_loss": 0.966569775390625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.9450435054779053 + }, + { + "epoch": 136, + "eval_accuracy": 0.6658, + "eval_loss": 0.9645234619140625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.9295638001971774 + }, + { + "epoch": 137, + "eval_accuracy": 0.6238, + "eval_loss": 1.1051807861328125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.95655686923133 + }, + { + "epoch": 138, + "eval_accuracy": 0.642, + "eval_loss": 1.03306103515625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 1.0022516270955404 + }, + { + "epoch": 139, + "eval_accuracy": 0.6416, + "eval_loss": 1.0455357788085937, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.9911704215155708 + }, + { + "epoch": 140, + "eval_accuracy": 0.6474, + "eval_loss": 1.0605947875976562, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.9922565679762099 + }, + { + "epoch": 141, + "eval_accuracy": 0.6374, + "eval_loss": 1.07503681640625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 1.0034182527012294 + }, + { + "epoch": 142, + "eval_accuracy": 0.641, + "eval_loss": 1.061389990234375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 1.0350198850949606 + }, + { + "epoch": 143, + "eval_accuracy": 0.6272, + "eval_loss": 1.091273046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 1.015117134115431 + }, + { + "epoch": 144, + "eval_accuracy": 0.634, + "eval_loss": 1.0583884765625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 1.0244670013639663 + }, + { + "epoch": 145, + "eval_accuracy": 0.653, + "eval_loss": 1.0368972412109374, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 1.0196523662779067 + }, + { + "epoch": 146, + "eval_accuracy": 0.6564, + "eval_loss": 1.007610009765625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.981625235833062 + }, + { + "epoch": 147, + "eval_accuracy": 0.6548, + "eval_loss": 1.0109105224609376, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.9948846951166789 + }, + { + "epoch": 148, + "eval_accuracy": 0.6306, + "eval_loss": 1.0539881225585936, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 1.052033565235138 + }, + { + "epoch": 149, + "eval_accuracy": 0.61, + "eval_loss": 1.15952216796875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 1.0489593990325927 + }, + { + "epoch": 150, + "eval_accuracy": 0.6218, + "eval_loss": 1.085763623046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 1.1033223127788967 + }, + { + "epoch": 151, + "eval_accuracy": 0.654, + "eval_loss": 1.0092583862304687, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 1.0319644333945381 + }, + { + "epoch": 152, + "eval_accuracy": 0.676, + "eval_loss": 0.9723513427734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.994410558075375 + }, + { + "epoch": 153, + "eval_accuracy": 0.6766, + "eval_loss": 0.9589015747070313, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.9668529237747192 + }, + { + "epoch": 154, + "eval_accuracy": 0.6738, + "eval_loss": 0.95261396484375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.9483775505489773 + }, + { + "epoch": 155, + "eval_accuracy": 0.6798, + "eval_loss": 0.9446010498046875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.9422066597832573 + }, + { + "epoch": 156, + "eval_accuracy": 0.6814, + "eval_loss": 0.9320489990234375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.9290729566150241 + }, + { + "epoch": 157, + "eval_accuracy": 0.6882, + "eval_loss": 0.9252060791015625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.9260442474683126 + }, + { + "epoch": 158, + "eval_accuracy": 0.6916, + "eval_loss": 0.9229399536132813, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.9205480414920383 + }, + { + "epoch": 159, + "eval_accuracy": 0.6904, + "eval_loss": 0.9217549682617188, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.9238852113935683 + }, + { + "epoch": 160, + "eval_accuracy": 0.6848, + "eval_loss": 0.9292841796875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.9197275682343377 + }, + { + "epoch": 161, + "eval_accuracy": 0.6886, + "eval_loss": 0.9162351684570312, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.9102767890082465 + }, + { + "epoch": 162, + "eval_accuracy": 0.6914, + "eval_loss": 0.9114646118164063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.9038821694480048 + }, + { + "epoch": 163, + "eval_accuracy": 0.697, + "eval_loss": 0.9089584228515625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.9043405807918973 + }, + { + "epoch": 164, + "eval_accuracy": 0.6988, + "eval_loss": 0.9067180297851563, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.8994147229936388 + }, + { + "epoch": 165, + "eval_accuracy": 0.6932, + "eval_loss": 0.9099502319335937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.9026298767301771 + }, + { + "epoch": 166, + "eval_accuracy": 0.6942, + "eval_loss": 0.9057412231445312, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.9046581612269083 + }, + { + "epoch": 167, + "eval_accuracy": 0.6932, + "eval_loss": 0.906397509765625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.9021518977906969 + }, + { + "epoch": 168, + "eval_accuracy": 0.6908, + "eval_loss": 0.9119505981445313, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.9030990373081631 + }, + { + "epoch": 169, + "eval_accuracy": 0.6868, + "eval_loss": 0.9118242309570312, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.9032874836497836 + }, + { + "epoch": 170, + "eval_accuracy": 0.6822, + "eval_loss": 0.92456123046875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.9015067362467448 + }, + { + "epoch": 171, + "eval_accuracy": 0.686, + "eval_loss": 0.9098457397460937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.9019311337153116 + }, + { + "epoch": 172, + "eval_accuracy": 0.6856, + "eval_loss": 0.9247203979492188, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.9169406125810411 + }, + { + "epoch": 173, + "eval_accuracy": 0.6864, + "eval_loss": 0.9248978637695312, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.9140570906321208 + }, + { + "epoch": 174, + "eval_accuracy": 0.686, + "eval_loss": 0.916908447265625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.9104670696258544 + }, + { + "epoch": 175, + "eval_accuracy": 0.686, + "eval_loss": 0.92089794921875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.9013468429777357 + }, + { + "epoch": 176, + "eval_accuracy": 0.6868, + "eval_loss": 0.9163492797851562, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.9033487060334947 + }, + { + "epoch": 177, + "eval_accuracy": 0.6832, + "eval_loss": 0.9202944091796875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.9047088833385044 + }, + { + "epoch": 178, + "eval_accuracy": 0.6766, + "eval_loss": 0.9400230834960938, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.915128051270379 + }, + { + "epoch": 179, + "eval_accuracy": 0.68, + "eval_loss": 0.9347542114257813, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.9253638660642836 + }, + { + "epoch": 180, + "eval_accuracy": 0.6784, + "eval_loss": 0.9496093017578126, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.9585485564867655 + }, + { + "epoch": 181, + "eval_accuracy": 0.6754, + "eval_loss": 0.9441274047851562, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.9477499859068129 + }, + { + "epoch": 182, + "eval_accuracy": 0.6798, + "eval_loss": 0.93713427734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.9405778836568197 + }, + { + "epoch": 183, + "eval_accuracy": 0.6762, + "eval_loss": 0.9379406616210938, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.9201891781171163 + }, + { + "epoch": 184, + "eval_accuracy": 0.6798, + "eval_loss": 0.9241371215820312, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.9184683513959249 + }, + { + "epoch": 185, + "eval_accuracy": 0.681, + "eval_loss": 0.9289290649414063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.9075623473697239 + }, + { + "epoch": 186, + "eval_accuracy": 0.682, + "eval_loss": 0.9222794555664062, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.9138578772544861 + }, + { + "epoch": 187, + "eval_accuracy": 0.6778, + "eval_loss": 0.9411468017578125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.917494329823388 + }, + { + "epoch": 188, + "eval_accuracy": 0.6734, + "eval_loss": 0.952200244140625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.9213011764632331 + }, + { + "epoch": 189, + "eval_accuracy": 0.6812, + "eval_loss": 0.9310021484375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.9217213383992513 + }, + { + "epoch": 190, + "eval_accuracy": 0.682, + "eval_loss": 0.9291856201171875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.9160469234360589 + }, + { + "epoch": 191, + "eval_accuracy": 0.6774, + "eval_loss": 0.93428427734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.9122076215638055 + }, + { + "epoch": 192, + "eval_accuracy": 0.6754, + "eval_loss": 0.944618798828125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.9217920794275072 + }, + { + "epoch": 193, + "eval_accuracy": 0.6768, + "eval_loss": 0.9403953857421875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.9300884993659125 + }, + { + "epoch": 194, + "eval_accuracy": 0.6846, + "eval_loss": 0.9310226684570313, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.9326077370961507 + }, + { + "epoch": 195, + "eval_accuracy": 0.6812, + "eval_loss": 0.9314572143554688, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.9251852588123746 + }, + { + "epoch": 196, + "eval_accuracy": 0.6792, + "eval_loss": 0.9355191040039063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.924093881840176 + }, + { + "epoch": 197, + "eval_accuracy": 0.6796, + "eval_loss": 0.9381912963867187, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.9298513369878133 + }, + { + "epoch": 198, + "eval_accuracy": 0.6796, + "eval_loss": 0.9357055297851562, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.9319831230163574 + }, + { + "epoch": 199, + "eval_accuracy": 0.6888, + "eval_loss": 0.9251195556640625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.9235278500239055 + }, + { + "epoch": 200, + "eval_accuracy": 0.6854, + "eval_loss": 0.9467566040039063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.9284468433168199 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.6854, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.9467566040039063, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2009673728, + "peak_memory_reserved_bytes": 2550136832, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 21.07404398918152, + "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": 1361.8625671863556, + "train_wall_s": 1340.767660856247 + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 360978048000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 364959360000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 364959360000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 1090896768000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json.manifest.json new file mode 100644 index 0000000..4920cfc --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "hfa", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787034474.4473119, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 1365.4268357753754, + "epochs": 200, + "experiment_name": "resnet-r2-fa-d20-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "fa", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d20-lr0p1.json", + "output_exists": true, + "output_sha256": "4214a5ec034977d0fe8f927f9de040e282833008270204936eb81b419f16c399", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json new file mode 100644 index 0000000..a2f3e08 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json @@ -0,0 +1,2801 @@ +{ + "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": 461440, + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "hfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.03778243931010365, + "feedback_forward_cosine": [ + 0.13522903621196747, + 0.2389027625322342, + 0.1656235307455063, + 0.25669336318969727, + 0.08168920129537582, + 0.24313712120056152, + 0.03679082542657852, + 0.16495703160762787, + 0.043550122529268265, + 0.20206722617149353, + 0.052305079996585846, + 0.1600673943758011, + 0.04140310361981392, + 0.1889246553182602, + 0.029380766674876213, + 0.2227826863527298, + 0.028732571750879288, + 0.18993735313415527, + 0.032700490206480026, + 0.19198478758335114, + 0.031028691679239273, + 0.1495836079120636, + 0.017455775290727615, + 0.18182963132858276, + 0.03303111344575882, + 0.18500036001205444, + 0.003500960301607847, + 0.19491636753082275, + 0.02336471900343895, + 0.15348230302333832, + 0.7153934240341187 + ], + "feedback_forward_norm_ratio": [ + 1.6532303094863892, + 1.3528220653533936, + 1.8407398462295532, + 1.573699712753296, + 2.2097935676574707, + 1.9533603191375732, + 2.9062178134918213, + 2.332139253616333, + 2.6777613162994385, + 2.2939107418060303, + 2.361605405807495, + 1.6632351875305176, + 2.157731771469116, + 1.681369423866272, + 2.3610785007476807, + 1.9888919591903687, + 2.5523250102996826, + 2.211498737335205, + 2.42690110206604, + 2.11393141746521, + 2.266143560409546, + 1.3281869888305664, + 2.0636491775512695, + 1.2860081195831299, + 1.958283543586731, + 1.3220630884170532, + 1.8889795541763306, + 1.3131144046783447, + 2.109546184539795, + 1.3858412504196167, + 0.19881561398506165 + ], + "feedback_forward_relative_error": [ + 1.8127440214157104, + 1.477748990058899, + 1.9438579082489014, + 1.6335893869400024, + 2.349926710128784, + 1.9661505222320557, + 3.0384628772735596, + 2.3810644149780273, + 2.817298650741577, + 2.3097572326660156, + 2.5159752368927, + 1.7983025312423706, + 2.340327739715576, + 1.7865326404571533, + 2.5369174480438232, + 2.0173025131225586, + 2.7143495082855225, + 2.247361660003662, + 2.5944416522979736, + 2.158013343811035, + 2.448423147201538, + 1.5384182929992676, + 2.277411460876465, + 1.4785629510879517, + 2.169217586517334, + 1.5028926134109497, + 2.134248733520508, + 1.4874053001403809, + 2.3133537769317627, + 1.5796048641204834, + 0.8689447045326233 + ], + "innovation_negative_gradient_cosine": [ + 0.00606947299093008, + 0.009478073567152023, + 0.024084236472845078, + 0.022062642499804497, + 0.04974904656410217, + 0.037176258862018585, + 0.07845072448253632, + 0.028204817324876785, + 0.09064127504825592, + 0.03190784528851509, + 0.09537523239850998, + 0.06018902733922005, + 0.18250532448291779, + 0.06845758110284805, + 0.22662648558616638, + 0.09742648154497147, + 0.2728409171104431, + 0.07810325175523758, + 0.29824674129486084, + 0.08288314938545227, + 0.3066348731517792, + 0.09523385763168335, + 0.4438329339027405, + 0.10535746812820435, + 0.4930099844932556, + 0.13920406997203827, + 0.5292818546295166, + 0.2641158998012543, + 0.6173524260520935, + 0.31898194551467896, + 0.7836034893989563 + ], + "mean_feedback_forward_cosine": 0.1417885826930644, + "mean_feedback_forward_relative_error": 2.072213174835328, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.00606947299093008, + 0.009478073567152023, + 0.024084236472845078, + 0.022062642499804497, + 0.04974904656410217, + 0.037176258862018585, + 0.07845072448253632, + 0.028204817324876785, + 0.09064127504825592, + 0.03190784528851509, + 0.09537523239850998, + 0.06018902733922005, + 0.18250532448291779, + 0.06845758110284805, + 0.22662648558616638, + 0.09742648154497147, + 0.2728409171104431, + 0.07810325175523758, + 0.29824674129486084, + 0.08288314938545227, + 0.3066348731517792, + 0.09523385763168335, + 0.4438329339027405, + 0.10535746812820435, + 0.4930099844932556, + 0.13920406997203827, + 0.5292818546295166, + 0.2641158998012543, + 0.6173524260520935, + 0.31898194551467896, + 0.7836034893989563 + ], + "teaching_negative_gradient_cosine": [ + 0.00606947299093008, + 0.009478073567152023, + 0.024084236472845078, + 0.022062642499804497, + 0.04974904656410217, + 0.037176258862018585, + 0.07845072448253632, + 0.028204817324876785, + 0.09064127504825592, + 0.03190784528851509, + 0.09537523239850998, + 0.06018902733922005, + 0.18250532448291779, + 0.06845758110284805, + 0.22662648558616638, + 0.09742648154497147, + 0.2728409171104431, + 0.07810325175523758, + 0.29824674129486084, + 0.08288314938545227, + 0.3066348731517792, + 0.09523385763168335, + 0.4438329339027405, + 0.10535746812820435, + 0.4930099844932556, + 0.13920406997203827, + 0.5292818546295166, + 0.2641158998012543, + 0.6173524260520935, + 0.31898194551467896, + 0.7836034893989563 + ], + "wall_s": 0.11364865303039551 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.2956, + "eval_loss": 1.986936767578125, + "eval_split": "validation", + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.195509643088447 + }, + { + "epoch": 2, + "eval_accuracy": 0.2702, + "eval_loss": 2.53878330078125, + "eval_split": "validation", + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 2.3834243949042424 + }, + { + "epoch": 3, + "eval_accuracy": 0.2754, + "eval_loss": 3.973369384765625, + "eval_split": "validation", + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 3.8713371835496693 + }, + { + "epoch": 4, + "eval_accuracy": 0.272, + "eval_loss": 7.13771689453125, + "eval_split": "validation", + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 6.064967641448975 + }, + { + "epoch": 5, + "eval_accuracy": 0.2578, + "eval_loss": 9.420830078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 7.330231070963542 + }, + { + "epoch": 6, + "eval_accuracy": 0.3024, + "eval_loss": 7.5071580078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 7.459590669589573 + }, + { + "epoch": 7, + "eval_accuracy": 0.341, + "eval_loss": 9.0689892578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 6.910371139611138 + }, + { + "epoch": 8, + "eval_accuracy": 0.2986, + "eval_loss": 7.185187890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 7.848369245402019 + }, + { + "epoch": 9, + "eval_accuracy": 0.268, + "eval_loss": 7.8466748046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 7.287076099565294 + }, + { + "epoch": 10, + "eval_accuracy": 0.3114, + "eval_loss": 6.8664982421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 6.893469244893392 + }, + { + "epoch": 11, + "eval_accuracy": 0.2968, + "eval_loss": 6.61613349609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 7.038937747870551 + }, + { + "epoch": 12, + "eval_accuracy": 0.3484, + "eval_loss": 8.7687521484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 6.735957527669271 + }, + { + "epoch": 13, + "eval_accuracy": 0.3388, + "eval_loss": 6.116796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 6.188509490797255 + }, + { + "epoch": 14, + "eval_accuracy": 0.3352, + "eval_loss": 6.60065009765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 6.898669404178196 + }, + { + "epoch": 15, + "eval_accuracy": 0.3442, + "eval_loss": 7.7424158203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 6.403976938290066 + }, + { + "epoch": 16, + "eval_accuracy": 0.3778, + "eval_loss": 6.02613154296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 6.261937475925022 + }, + { + "epoch": 17, + "eval_accuracy": 0.3454, + "eval_loss": 6.16180947265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 5.652168246968587 + }, + { + "epoch": 18, + "eval_accuracy": 0.2904, + "eval_loss": 12.1865435546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 5.759846517774794 + }, + { + "epoch": 19, + "eval_accuracy": 0.3586, + "eval_loss": 6.29294453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 5.43130024939643 + }, + { + "epoch": 20, + "eval_accuracy": 0.3414, + "eval_loss": 6.084059375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 5.368133390469021 + }, + { + "epoch": 21, + "eval_accuracy": 0.3324, + "eval_loss": 7.98075751953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 5.172292945861816 + }, + { + "epoch": 22, + "eval_accuracy": 0.3692, + "eval_loss": 6.4959990234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 5.412291467454699 + }, + { + "epoch": 23, + "eval_accuracy": 0.362, + "eval_loss": 6.58297880859375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 5.337951386176215 + }, + { + "epoch": 24, + "eval_accuracy": 0.3838, + "eval_loss": 5.59174921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 5.547587069659763 + }, + { + "epoch": 25, + "eval_accuracy": 0.3284, + "eval_loss": 8.17561748046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 5.564922823164198 + }, + { + "epoch": 26, + "eval_accuracy": 0.4088, + "eval_loss": 5.76070908203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 5.24930276345147 + }, + { + "epoch": 27, + "eval_accuracy": 0.3016, + "eval_loss": 7.5848359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 5.235181749979655 + }, + { + "epoch": 28, + "eval_accuracy": 0.3864, + "eval_loss": 6.51845185546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 5.172612402937148 + }, + { + "epoch": 29, + "eval_accuracy": 0.3368, + "eval_loss": 7.98949931640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 5.318213782840305 + }, + { + "epoch": 30, + "eval_accuracy": 0.3612, + "eval_loss": 8.5869908203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 5.080642896185981 + }, + { + "epoch": 31, + "eval_accuracy": 0.3554, + "eval_loss": 6.6961919921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 5.069903329298231 + }, + { + "epoch": 32, + "eval_accuracy": 0.3928, + "eval_loss": 6.05183935546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 5.545271482086181 + }, + { + "epoch": 33, + "eval_accuracy": 0.3728, + "eval_loss": 7.6160564453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 5.059899168311225 + }, + { + "epoch": 34, + "eval_accuracy": 0.4254, + "eval_loss": 4.392253515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 4.735173885133531 + }, + { + "epoch": 35, + "eval_accuracy": 0.4154, + "eval_loss": 5.35446240234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 4.707660108100043 + }, + { + "epoch": 36, + "eval_accuracy": 0.4556, + "eval_loss": 5.477988671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 4.8679831588745115 + }, + { + "epoch": 37, + "eval_accuracy": 0.3956, + "eval_loss": 5.809548828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 5.061632610660129 + }, + { + "epoch": 38, + "eval_accuracy": 0.38, + "eval_loss": 8.34639951171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 5.17016610599094 + }, + { + "epoch": 39, + "eval_accuracy": 0.3596, + "eval_loss": 10.3157998046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 5.13461835738288 + }, + { + "epoch": 40, + "eval_accuracy": 0.4494, + "eval_loss": 3.57537568359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 4.814193962224325 + }, + { + "epoch": 41, + "eval_accuracy": 0.3972, + "eval_loss": 7.6974388671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 4.812456588575575 + }, + { + "epoch": 42, + "eval_accuracy": 0.4116, + "eval_loss": 5.53400478515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 4.879522632090251 + }, + { + "epoch": 43, + "eval_accuracy": 0.3586, + "eval_loss": 7.28328564453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 4.666603200615777 + }, + { + "epoch": 44, + "eval_accuracy": 0.4092, + "eval_loss": 5.62242421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 5.149397599199083 + }, + { + "epoch": 45, + "eval_accuracy": 0.4616, + "eval_loss": 4.175123388671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 4.626690245098538 + }, + { + "epoch": 46, + "eval_accuracy": 0.4692, + "eval_loss": 5.34783779296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 4.866097333908081 + }, + { + "epoch": 47, + "eval_accuracy": 0.4406, + "eval_loss": 5.89131806640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 4.398048905860053 + }, + { + "epoch": 48, + "eval_accuracy": 0.4358, + "eval_loss": 5.7748138671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 4.245615175713433 + }, + { + "epoch": 49, + "eval_accuracy": 0.5006, + "eval_loss": 4.4333083984375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 4.748879802788628 + }, + { + "epoch": 50, + "eval_accuracy": 0.4424, + "eval_loss": 4.04569853515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 4.1075999011993405 + }, + { + "epoch": 51, + "eval_accuracy": 0.4298, + "eval_loss": 4.8061736328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 4.082825281778971 + }, + { + "epoch": 52, + "eval_accuracy": 0.3572, + "eval_loss": 5.676712109375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 4.619352707672119 + }, + { + "epoch": 53, + "eval_accuracy": 0.423, + "eval_loss": 6.05435361328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 3.9728969636705185 + }, + { + "epoch": 54, + "eval_accuracy": 0.4374, + "eval_loss": 4.6374767578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 4.3532811665005156 + }, + { + "epoch": 55, + "eval_accuracy": 0.3908, + "eval_loss": 5.56046455078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 3.9575316127353246 + }, + { + "epoch": 56, + "eval_accuracy": 0.4264, + "eval_loss": 4.6221392578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 4.364873327128093 + }, + { + "epoch": 57, + "eval_accuracy": 0.4556, + "eval_loss": 3.86793564453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 3.9743894003974067 + }, + { + "epoch": 58, + "eval_accuracy": 0.401, + "eval_loss": 5.83616787109375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 4.274858414755927 + }, + { + "epoch": 59, + "eval_accuracy": 0.4062, + "eval_loss": 5.9770345703125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 4.168988529247708 + }, + { + "epoch": 60, + "eval_accuracy": 0.4162, + "eval_loss": 5.7100212890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 4.416859284888374 + }, + { + "epoch": 61, + "eval_accuracy": 0.4016, + "eval_loss": 4.9233658203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 4.03270875193278 + }, + { + "epoch": 62, + "eval_accuracy": 0.41, + "eval_loss": 7.848937890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 4.23637970928616 + }, + { + "epoch": 63, + "eval_accuracy": 0.4404, + "eval_loss": 8.1851318359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 4.32700122680664 + }, + { + "epoch": 64, + "eval_accuracy": 0.4304, + "eval_loss": 5.8247, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 4.107528818427192 + }, + { + "epoch": 65, + "eval_accuracy": 0.4472, + "eval_loss": 6.40511494140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 4.552944485388862 + }, + { + "epoch": 66, + "eval_accuracy": 0.395, + "eval_loss": 6.11930693359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 4.474366613896688 + }, + { + "epoch": 67, + "eval_accuracy": 0.4084, + "eval_loss": 6.2334013671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 4.986977877638075 + }, + { + "epoch": 68, + "eval_accuracy": 0.4232, + "eval_loss": 5.4462650390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 4.605135821109348 + }, + { + "epoch": 69, + "eval_accuracy": 0.3384, + "eval_loss": 6.6896943359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 4.689142846934001 + }, + { + "epoch": 70, + "eval_accuracy": 0.4314, + "eval_loss": 7.1887201171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 4.31898379787869 + }, + { + "epoch": 71, + "eval_accuracy": 0.4278, + "eval_loss": 7.44157431640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 4.600417254977756 + }, + { + "epoch": 72, + "eval_accuracy": 0.409, + "eval_loss": 6.05228955078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 4.883821003723145 + }, + { + "epoch": 73, + "eval_accuracy": 0.4002, + "eval_loss": 6.30699609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 4.3505779678344725 + }, + { + "epoch": 74, + "eval_accuracy": 0.4464, + "eval_loss": 6.32979833984375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 4.398072329033746 + }, + { + "epoch": 75, + "eval_accuracy": 0.4112, + "eval_loss": 6.6822724609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 4.89097486360338 + }, + { + "epoch": 76, + "eval_accuracy": 0.4596, + "eval_loss": 4.2010662109375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 4.577514958106147 + }, + { + "epoch": 77, + "eval_accuracy": 0.3782, + "eval_loss": 9.2190607421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 4.218928383212619 + }, + { + "epoch": 78, + "eval_accuracy": 0.4284, + "eval_loss": 9.0243501953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 4.1683951346503365 + }, + { + "epoch": 79, + "eval_accuracy": 0.3524, + "eval_loss": 8.21492509765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 4.540081050364177 + }, + { + "epoch": 80, + "eval_accuracy": 0.406, + "eval_loss": 6.62808154296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 4.4213002469380696 + }, + { + "epoch": 81, + "eval_accuracy": 0.4084, + "eval_loss": 7.6263900390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 4.814154179806179 + }, + { + "epoch": 82, + "eval_accuracy": 0.4404, + "eval_loss": 4.99802724609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 4.313245312457615 + }, + { + "epoch": 83, + "eval_accuracy": 0.4326, + "eval_loss": 4.130868115234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 4.179187333128187 + }, + { + "epoch": 84, + "eval_accuracy": 0.4598, + "eval_loss": 6.097021484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 4.335098880301581 + }, + { + "epoch": 85, + "eval_accuracy": 0.4, + "eval_loss": 6.01697724609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 4.538031046464708 + }, + { + "epoch": 86, + "eval_accuracy": 0.4672, + "eval_loss": 5.2787099609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 4.391405413013034 + }, + { + "epoch": 87, + "eval_accuracy": 0.328, + "eval_loss": 7.990158203125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 4.530714601389567 + }, + { + "epoch": 88, + "eval_accuracy": 0.4342, + "eval_loss": 5.50915888671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 4.405182353210449 + }, + { + "epoch": 89, + "eval_accuracy": 0.4862, + "eval_loss": 3.551749169921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 4.707141447533502 + }, + { + "epoch": 90, + "eval_accuracy": 0.4614, + "eval_loss": 5.64707607421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 4.449501358328925 + }, + { + "epoch": 91, + "eval_accuracy": 0.4286, + "eval_loss": 6.6657060546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 4.62092832438151 + }, + { + "epoch": 92, + "eval_accuracy": 0.391, + "eval_loss": 7.563044140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 4.743001677449544 + }, + { + "epoch": 93, + "eval_accuracy": 0.4466, + "eval_loss": 6.4824744140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 4.3673409123738605 + }, + { + "epoch": 94, + "eval_accuracy": 0.4352, + "eval_loss": 6.52478466796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 4.221223247782389 + }, + { + "epoch": 95, + "eval_accuracy": 0.4942, + "eval_loss": 3.797342041015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 4.0709133242713085 + }, + { + "epoch": 96, + "eval_accuracy": 0.3552, + "eval_loss": 11.0053232421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 4.388050371297201 + }, + { + "epoch": 97, + "eval_accuracy": 0.4682, + "eval_loss": 5.0339734375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 4.638358270899455 + }, + { + "epoch": 98, + "eval_accuracy": 0.4216, + "eval_loss": 7.56427587890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 4.665724361250136 + }, + { + "epoch": 99, + "eval_accuracy": 0.3896, + "eval_loss": 5.62162314453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 4.479719784079657 + }, + { + "epoch": 100, + "eval_accuracy": 0.4674, + "eval_loss": 6.46364716796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 5.066503650241428 + }, + { + "epoch": 101, + "eval_accuracy": 0.609, + "eval_loss": 1.586387109375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 1.721678271569146 + }, + { + "epoch": 102, + "eval_accuracy": 0.5992, + "eval_loss": 1.4983593505859376, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 1.3279080866071913 + }, + { + "epoch": 103, + "eval_accuracy": 0.6246, + "eval_loss": 1.256317919921875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 1.232182222451104 + }, + { + "epoch": 104, + "eval_accuracy": 0.5786, + "eval_loss": 1.4460140625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 1.1727672132703992 + }, + { + "epoch": 105, + "eval_accuracy": 0.6232, + "eval_loss": 1.2341310546875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 1.1651357496897379 + }, + { + "epoch": 106, + "eval_accuracy": 0.6248, + "eval_loss": 1.1990308349609375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 1.153296654340956 + }, + { + "epoch": 107, + "eval_accuracy": 0.627, + "eval_loss": 1.151527783203125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 1.1303095881144205 + }, + { + "epoch": 108, + "eval_accuracy": 0.6266, + "eval_loss": 1.139106787109375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 1.1244081782446966 + }, + { + "epoch": 109, + "eval_accuracy": 0.6224, + "eval_loss": 1.18290263671875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 1.121556561088562 + }, + { + "epoch": 110, + "eval_accuracy": 0.623, + "eval_loss": 1.1661594482421875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 1.1135010125690037 + }, + { + "epoch": 111, + "eval_accuracy": 0.6088, + "eval_loss": 1.201440478515625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 1.1076628268559774 + }, + { + "epoch": 112, + "eval_accuracy": 0.6354, + "eval_loss": 1.09391640625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 1.1044298093159994 + }, + { + "epoch": 113, + "eval_accuracy": 0.6282, + "eval_loss": 1.1468849365234375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 1.090382902208964 + }, + { + "epoch": 114, + "eval_accuracy": 0.6192, + "eval_loss": 1.1721736328125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 1.0959963248146904 + }, + { + "epoch": 115, + "eval_accuracy": 0.6326, + "eval_loss": 1.1188074951171876, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 1.0874598748736912 + }, + { + "epoch": 116, + "eval_accuracy": 0.6348, + "eval_loss": 1.119938037109375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 1.0830459158155654 + }, + { + "epoch": 117, + "eval_accuracy": 0.6334, + "eval_loss": 1.0968715087890626, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 1.0895018880844116 + }, + { + "epoch": 118, + "eval_accuracy": 0.6556, + "eval_loss": 1.06753017578125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 1.0812106000264485 + }, + { + "epoch": 119, + "eval_accuracy": 0.6272, + "eval_loss": 1.1438458740234374, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 1.0666006845792135 + }, + { + "epoch": 120, + "eval_accuracy": 0.6512, + "eval_loss": 1.0582649169921876, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 1.0607601169480219 + }, + { + "epoch": 121, + "eval_accuracy": 0.6484, + "eval_loss": 1.0663829833984375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 1.0521053105036418 + }, + { + "epoch": 122, + "eval_accuracy": 0.6476, + "eval_loss": 1.0945978271484376, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 1.049345637872484 + }, + { + "epoch": 123, + "eval_accuracy": 0.6316, + "eval_loss": 1.1688353271484375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 1.0518553288565742 + }, + { + "epoch": 124, + "eval_accuracy": 0.6596, + "eval_loss": 1.0690121826171874, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 1.050257617494795 + }, + { + "epoch": 125, + "eval_accuracy": 0.6264, + "eval_loss": 1.1545436279296875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 1.0591116312238906 + }, + { + "epoch": 126, + "eval_accuracy": 0.6508, + "eval_loss": 1.0498425415039063, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 1.0364673189057245 + }, + { + "epoch": 127, + "eval_accuracy": 0.6358, + "eval_loss": 1.1038102294921874, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 1.04405638639662 + }, + { + "epoch": 128, + "eval_accuracy": 0.6582, + "eval_loss": 1.0474168212890624, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 1.041512263361613 + }, + { + "epoch": 129, + "eval_accuracy": 0.6444, + "eval_loss": 1.0852489501953124, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 1.0383492436091104 + }, + { + "epoch": 130, + "eval_accuracy": 0.628, + "eval_loss": 1.138535400390625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 1.0526941585858662 + }, + { + "epoch": 131, + "eval_accuracy": 0.6438, + "eval_loss": 1.08569150390625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 1.0424592800352308 + }, + { + "epoch": 132, + "eval_accuracy": 0.6164, + "eval_loss": 1.1733404296875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 1.0678051187939115 + }, + { + "epoch": 133, + "eval_accuracy": 0.6428, + "eval_loss": 1.0911018798828125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 1.0511749137242634 + }, + { + "epoch": 134, + "eval_accuracy": 0.6284, + "eval_loss": 1.1672166748046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 1.0622696600808037 + }, + { + "epoch": 135, + "eval_accuracy": 0.6302, + "eval_loss": 1.107805029296875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 1.0528499618742202 + }, + { + "epoch": 136, + "eval_accuracy": 0.6398, + "eval_loss": 1.1062733154296875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 1.0609219336191813 + }, + { + "epoch": 137, + "eval_accuracy": 0.6424, + "eval_loss": 1.0821095703125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 1.0891843078931174 + }, + { + "epoch": 138, + "eval_accuracy": 0.6324, + "eval_loss": 1.0882267578125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 1.0656020960172017 + }, + { + "epoch": 139, + "eval_accuracy": 0.6506, + "eval_loss": 1.062216015625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 1.04979617284139 + }, + { + "epoch": 140, + "eval_accuracy": 0.6542, + "eval_loss": 1.0492350952148437, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 1.039383688206143 + }, + { + "epoch": 141, + "eval_accuracy": 0.634, + "eval_loss": 1.0878307861328125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 1.046560775396559 + }, + { + "epoch": 142, + "eval_accuracy": 0.6532, + "eval_loss": 1.0724321533203125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 1.0318034744050768 + }, + { + "epoch": 143, + "eval_accuracy": 0.6452, + "eval_loss": 1.06745087890625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 1.0459841170628865 + }, + { + "epoch": 144, + "eval_accuracy": 0.6494, + "eval_loss": 1.0546759521484375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 1.0411283879597981 + }, + { + "epoch": 145, + "eval_accuracy": 0.6406, + "eval_loss": 1.07638974609375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 1.03138457192315 + }, + { + "epoch": 146, + "eval_accuracy": 0.6496, + "eval_loss": 1.0701206787109374, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 1.021044852172004 + }, + { + "epoch": 147, + "eval_accuracy": 0.66, + "eval_loss": 1.0486619995117188, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 1.0230277407752144 + }, + { + "epoch": 148, + "eval_accuracy": 0.6444, + "eval_loss": 1.066232470703125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 1.0228486086951363 + }, + { + "epoch": 149, + "eval_accuracy": 0.6516, + "eval_loss": 1.0704355102539063, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 1.0319631608539157 + }, + { + "epoch": 150, + "eval_accuracy": 0.6548, + "eval_loss": 1.0434142211914061, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 1.0307153160836962 + }, + { + "epoch": 151, + "eval_accuracy": 0.6632, + "eval_loss": 1.0081596069335939, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.968408794424269 + }, + { + "epoch": 152, + "eval_accuracy": 0.6622, + "eval_loss": 1.0009671264648436, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.961383619011773 + }, + { + "epoch": 153, + "eval_accuracy": 0.6626, + "eval_loss": 1.0000935791015626, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.9587891780005561 + }, + { + "epoch": 154, + "eval_accuracy": 0.6666, + "eval_loss": 0.9947133178710937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.956661511707306 + }, + { + "epoch": 155, + "eval_accuracy": 0.667, + "eval_loss": 0.9916768676757812, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.9527008137491014 + }, + { + "epoch": 156, + "eval_accuracy": 0.6698, + "eval_loss": 0.9854952758789063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.9535861522886488 + }, + { + "epoch": 157, + "eval_accuracy": 0.6702, + "eval_loss": 0.9857482788085937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.9495243875291612 + }, + { + "epoch": 158, + "eval_accuracy": 0.6678, + "eval_loss": 0.9886399169921875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.9482226912604438 + }, + { + "epoch": 159, + "eval_accuracy": 0.6692, + "eval_loss": 0.9884629272460937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.9453291122012668 + }, + { + "epoch": 160, + "eval_accuracy": 0.6702, + "eval_loss": 0.98918701171875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.9501123986562093 + }, + { + "epoch": 161, + "eval_accuracy": 0.6724, + "eval_loss": 0.9819580810546875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.9426417893515693 + }, + { + "epoch": 162, + "eval_accuracy": 0.6688, + "eval_loss": 0.9828267333984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.9445630306455824 + }, + { + "epoch": 163, + "eval_accuracy": 0.6716, + "eval_loss": 0.9865187255859375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.9460579630851745 + }, + { + "epoch": 164, + "eval_accuracy": 0.672, + "eval_loss": 0.9791216430664063, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.9461332703696357 + }, + { + "epoch": 165, + "eval_accuracy": 0.671, + "eval_loss": 0.980871533203125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.9464517212549846 + }, + { + "epoch": 166, + "eval_accuracy": 0.6692, + "eval_loss": 0.9821738891601562, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.9537665562417772 + }, + { + "epoch": 167, + "eval_accuracy": 0.6728, + "eval_loss": 0.9786509887695313, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.9519137758678861 + }, + { + "epoch": 168, + "eval_accuracy": 0.6714, + "eval_loss": 0.9796623901367187, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.9474307224697537 + }, + { + "epoch": 169, + "eval_accuracy": 0.6732, + "eval_loss": 0.9656408813476562, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.9520944051954481 + }, + { + "epoch": 170, + "eval_accuracy": 0.6716, + "eval_loss": 0.9688844848632813, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.946247570154402 + }, + { + "epoch": 171, + "eval_accuracy": 0.6714, + "eval_loss": 0.9677995239257813, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.9489625805536905 + }, + { + "epoch": 172, + "eval_accuracy": 0.67, + "eval_loss": 0.9761918212890625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.9467603574222988 + }, + { + "epoch": 173, + "eval_accuracy": 0.6702, + "eval_loss": 0.968112744140625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.9482885099093119 + }, + { + "epoch": 174, + "eval_accuracy": 0.6734, + "eval_loss": 0.975332275390625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.9473493633694119 + }, + { + "epoch": 175, + "eval_accuracy": 0.6762, + "eval_loss": 0.9680783447265625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.9499148143344456 + }, + { + "epoch": 176, + "eval_accuracy": 0.6714, + "eval_loss": 0.9696656860351562, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.9587628659142389 + }, + { + "epoch": 177, + "eval_accuracy": 0.6732, + "eval_loss": 0.9686334350585938, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.955567098681132 + }, + { + "epoch": 178, + "eval_accuracy": 0.6734, + "eval_loss": 0.9767455810546875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.9594909250683255 + }, + { + "epoch": 179, + "eval_accuracy": 0.6736, + "eval_loss": 0.9675688354492188, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.9592544206619262 + }, + { + "epoch": 180, + "eval_accuracy": 0.674, + "eval_loss": 0.9692090087890625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.9600356777615018 + }, + { + "epoch": 181, + "eval_accuracy": 0.673, + "eval_loss": 0.9752307861328126, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.9585328283309936 + }, + { + "epoch": 182, + "eval_accuracy": 0.6732, + "eval_loss": 0.9749939697265625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.9576578861024645 + }, + { + "epoch": 183, + "eval_accuracy": 0.668, + "eval_loss": 0.9789309448242187, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.9648777356147766 + }, + { + "epoch": 184, + "eval_accuracy": 0.6728, + "eval_loss": 0.9707718383789062, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.9659501979933844 + }, + { + "epoch": 185, + "eval_accuracy": 0.6704, + "eval_loss": 0.97909375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.969595404349433 + }, + { + "epoch": 186, + "eval_accuracy": 0.6708, + "eval_loss": 0.9784969604492187, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.9665438864601983 + }, + { + "epoch": 187, + "eval_accuracy": 0.6696, + "eval_loss": 0.981677734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.9694981577661302 + }, + { + "epoch": 188, + "eval_accuracy": 0.6668, + "eval_loss": 0.9806268432617188, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.9674224305470784 + }, + { + "epoch": 189, + "eval_accuracy": 0.668, + "eval_loss": 0.97913740234375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.9666299751281738 + }, + { + "epoch": 190, + "eval_accuracy": 0.6686, + "eval_loss": 0.9715440063476563, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.9722860072347853 + }, + { + "epoch": 191, + "eval_accuracy": 0.6676, + "eval_loss": 0.9761428833007812, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.9745998671107822 + }, + { + "epoch": 192, + "eval_accuracy": 0.6666, + "eval_loss": 0.9851692260742188, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.9772617618454827 + }, + { + "epoch": 193, + "eval_accuracy": 0.6644, + "eval_loss": 0.9863299438476563, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.9811174005296496 + }, + { + "epoch": 194, + "eval_accuracy": 0.6638, + "eval_loss": 0.9889223022460938, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.9881035271220737 + }, + { + "epoch": 195, + "eval_accuracy": 0.6534, + "eval_loss": 1.0029588500976563, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.9892041141933865 + }, + { + "epoch": 196, + "eval_accuracy": 0.649, + "eval_loss": 1.0210272705078125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 1.0075765220854018 + }, + { + "epoch": 197, + "eval_accuracy": 0.6484, + "eval_loss": 1.0257604370117188, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 1.0129783128950332 + }, + { + "epoch": 198, + "eval_accuracy": 0.6512, + "eval_loss": 1.0173834716796875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 1.0185926798078748 + }, + { + "epoch": 199, + "eval_accuracy": 0.651, + "eval_loss": 1.0110072143554687, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 1.0170071312798394 + }, + { + "epoch": 200, + "eval_accuracy": 0.6548, + "eval_loss": 0.9980604614257812, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 1.0106280994839139 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.6548, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.9980604614257812, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2324148736, + "peak_memory_reserved_bytes": 3162505216, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 34.08719086647034, + "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": 2176.771654367447, + "train_wall_s": 2142.661477804184 + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 615782016000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 619763328000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 619763328000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 1855308672000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json.manifest.json new file mode 100644 index 0000000..d82f31b --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "hfa", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787029347.5506423, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 2180.355216741562, + "epochs": 200, + "experiment_name": "resnet-r2-fa-d32-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "fa", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d32-lr0p1.json", + "output_exists": true, + "output_sha256": "0a5f7017c2ab6fb991e98cf2b9c8e1438685f2575a2eaa41e80bdfdaa4162cf6", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json new file mode 100644 index 0000000..9eec779 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json @@ -0,0 +1,3065 @@ +{ + "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": 848512, + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "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": "hfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.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": 0, + "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": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "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": 0.08863549338032801, + "feedback_forward_cosine": [ + 0.0882125198841095, + 0.17185500264167786, + 0.10954485088586807, + 0.2577724754810333, + 0.0751032680273056, + 0.23121197521686554, + 0.030148033052682877, + 0.26019108295440674, + 0.050409719347953796, + 0.19008716940879822, + 0.0399165078997612, + 0.21744656562805176, + 0.0001835389994084835, + 0.1896965205669403, + 0.005662070121616125, + 0.20933006703853607, + 0.01125209592282772, + 0.1695132851600647, + 0.017937814816832542, + 0.1433870494365692, + 0.013061119243502617, + 0.1942385882139206, + 0.005128783639520407, + 0.2021099328994751, + -0.0006857686676084995, + 0.20494231581687927, + 0.0395386703312397, + 0.1583186686038971, + 0.016033835709095, + 0.18731288611888885, + 0.021174851804971695, + 0.16396485269069672, + 0.030186234042048454, + 0.1645135134458542, + 0.007649475708603859, + 0.15658310055732727, + 0.011635647155344486, + 0.1251221001148224, + -0.008313348516821861, + 0.15915539860725403, + -0.0014332830905914307, + 0.1680000275373459, + 0.011308528482913971, + 0.16522729396820068, + 0.006545661482959986, + 0.15889963507652283, + 0.006269207689911127, + 0.17684805393218994, + 0.006641895975917578, + 0.15410152077674866, + 0.01070297509431839, + 0.15132766962051392, + 0.00690346397459507, + 0.13190671801567078, + 0.6286851167678833 + ], + "feedback_forward_norm_ratio": [ + 1.6574243307113647, + 1.1986840963363647, + 1.6578441858291626, + 1.143052101135254, + 2.059192657470703, + 1.446895956993103, + 2.117647409439087, + 1.6426010131835938, + 2.1686482429504395, + 1.708986520767212, + 2.569164752960205, + 2.15853214263916, + 2.5361809730529785, + 1.9963648319244385, + 2.6983582973480225, + 2.1399641036987305, + 3.1610701084136963, + 2.68005633354187, + 2.4815921783447266, + 1.5921815633773804, + 2.099860191345215, + 1.5018749237060547, + 2.329667568206787, + 1.6975750923156738, + 2.184513807296753, + 1.6134381294250488, + 2.444791316986084, + 1.7782717943191528, + 2.475090503692627, + 1.9780645370483398, + 2.562943935394287, + 1.9800993204116821, + 2.487797498703003, + 1.9621620178222656, + 2.9988152980804443, + 2.466535806655884, + 2.4743497371673584, + 1.2880759239196777, + 1.9693315029144287, + 1.1085004806518555, + 1.9485834836959839, + 1.1041178703308105, + 1.955966830253601, + 1.1079226732254028, + 1.9907394647598267, + 1.2147018909454346, + 2.0718092918395996, + 1.319077968597412, + 2.0721075534820557, + 1.3644046783447266, + 2.153334856033325, + 1.4132341146469116, + 2.29006028175354, + 1.3792732954025269, + 0.112843818962574 + ], + "feedback_forward_relative_error": [ + 1.8586674928665161, + 1.4229700565338135, + 1.8398997783660889, + 1.3104478120803833, + 2.220578908920288, + 1.5570577383041382, + 2.3144640922546387, + 1.6862257719039917, + 2.3418779373168945, + 1.8085689544677734, + 2.7194676399230957, + 2.172678232192993, + 2.7260379791259766, + 2.0562262535095215, + 2.872382402420044, + 2.1641461849212646, + 3.3047280311584473, + 2.6970527172088623, + 2.6588101387023926, + 1.7545499801635742, + 2.3139920234680176, + 1.6346817016601562, + 2.5305044651031494, + 1.787615180015564, + 2.4031431674957275, + 1.7151849269866943, + 2.6045494079589844, + 1.8971514701843262, + 2.654562473297119, + 2.0424752235412598, + 2.73132586479187, + 2.06675124168396, + 2.6531007289886475, + 2.0504820346832275, + 3.1538889408111572, + 2.512242555618286, + 2.6579737663269043, + 1.5286616086959839, + 2.2160797119140625, + 1.369644284248352, + 2.1914751529693604, + 1.3594456911087036, + 2.1866798400878906, + 1.3643220663070679, + 2.2219321727752686, + 1.4454996585845947, + 2.2948672771453857, + 1.507784366607666, + 2.294799327850342, + 1.5623975992202759, + 2.36447811126709, + 1.6029685735702515, + 2.4925405979156494, + 1.5932745933532715, + 0.9331920146942139 + ], + "innovation_negative_gradient_cosine": [ + 0.008553816005587578, + 0.025900859385728836, + 0.0376245342195034, + 0.03193657472729683, + 0.09551645815372467, + 0.039293598383665085, + 0.13530397415161133, + 0.04041023924946785, + 0.158193439245224, + 0.035545770078897476, + 0.17993035912513733, + 0.04054297134280205, + 0.19581130146980286, + 0.04374796524643898, + 0.21109670400619507, + 0.04255548119544983, + 0.22164040803909302, + 0.05183442682027817, + 0.224349245429039, + 0.04830259829759598, + 0.26252907514572144, + 0.06813947856426239, + 0.2766053080558777, + 0.07549480348825455, + 0.29249435663223267, + 0.09466566145420074, + 0.3062906861305237, + 0.04542122036218643, + 0.31328457593917847, + 0.07723840326070786, + 0.3224060535430908, + 0.04863288253545761, + 0.327239990234375, + 0.04049643874168396, + 0.3324066400527954, + 0.08639751374721527, + 0.3271884620189667, + 0.06347719579935074, + 0.39523422718048096, + 0.04708479344844818, + 0.4143325090408325, + 0.06797029078006744, + 0.42466670274734497, + 0.08129915595054626, + 0.44150951504707336, + 0.07288248836994171, + 0.44816941022872925, + 0.15117616951465607, + 0.4542672038078308, + 0.14126504957675934, + 0.4803707003593445, + 0.19390016794204712, + 0.4851304888725281, + 0.22901073098182678, + 0.6826860904693604 + ], + "mean_feedback_forward_cosine": 0.11095394464209676, + "mean_feedback_forward_relative_error": 2.098663744059476, + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.008553816005587578, + 0.025900859385728836, + 0.0376245342195034, + 0.03193657472729683, + 0.09551645815372467, + 0.039293598383665085, + 0.13530397415161133, + 0.04041023924946785, + 0.158193439245224, + 0.035545770078897476, + 0.17993035912513733, + 0.04054297134280205, + 0.19581130146980286, + 0.04374796524643898, + 0.21109670400619507, + 0.04255548119544983, + 0.22164040803909302, + 0.05183442682027817, + 0.224349245429039, + 0.04830259829759598, + 0.26252907514572144, + 0.06813947856426239, + 0.2766053080558777, + 0.07549480348825455, + 0.29249435663223267, + 0.09466566145420074, + 0.3062906861305237, + 0.04542122036218643, + 0.31328457593917847, + 0.07723840326070786, + 0.3224060535430908, + 0.04863288253545761, + 0.327239990234375, + 0.04049643874168396, + 0.3324066400527954, + 0.08639751374721527, + 0.3271884620189667, + 0.06347719579935074, + 0.39523422718048096, + 0.04708479344844818, + 0.4143325090408325, + 0.06797029078006744, + 0.42466670274734497, + 0.08129915595054626, + 0.44150951504707336, + 0.07288248836994171, + 0.44816941022872925, + 0.15117616951465607, + 0.4542672038078308, + 0.14126504957675934, + 0.4803707003593445, + 0.19390016794204712, + 0.4851304888725281, + 0.22901073098182678, + 0.6826860904693604 + ], + "teaching_negative_gradient_cosine": [ + 0.008553816005587578, + 0.025900859385728836, + 0.0376245342195034, + 0.03193657472729683, + 0.09551645815372467, + 0.039293598383665085, + 0.13530397415161133, + 0.04041023924946785, + 0.158193439245224, + 0.035545770078897476, + 0.17993035912513733, + 0.04054297134280205, + 0.19581130146980286, + 0.04374796524643898, + 0.21109670400619507, + 0.04255548119544983, + 0.22164040803909302, + 0.05183442682027817, + 0.224349245429039, + 0.04830259829759598, + 0.26252907514572144, + 0.06813947856426239, + 0.2766053080558777, + 0.07549480348825455, + 0.29249435663223267, + 0.09466566145420074, + 0.3062906861305237, + 0.04542122036218643, + 0.31328457593917847, + 0.07723840326070786, + 0.3224060535430908, + 0.04863288253545761, + 0.327239990234375, + 0.04049643874168396, + 0.3324066400527954, + 0.08639751374721527, + 0.3271884620189667, + 0.06347719579935074, + 0.39523422718048096, + 0.04708479344844818, + 0.4143325090408325, + 0.06797029078006744, + 0.42466670274734497, + 0.08129915595054626, + 0.44150951504707336, + 0.07288248836994171, + 0.44816941022872925, + 0.15117616951465607, + 0.4542672038078308, + 0.14126504957675934, + 0.4803707003593445, + 0.19390016794204712, + 0.4851304888725281, + 0.22901073098182678, + 0.6826860904693604 + ], + "wall_s": 0.5787839889526367 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.2598, + "eval_loss": 2.63627958984375, + "eval_split": "validation", + "lr": 0.02, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.7094667250739204 + }, + { + "epoch": 2, + "eval_accuracy": 0.244, + "eval_loss": 8.75294052734375, + "eval_split": "validation", + "lr": 0.04, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 7.502461778089735 + }, + { + "epoch": 3, + "eval_accuracy": 0.248, + "eval_loss": 14.208425, + "eval_split": "validation", + "lr": 0.06000000000000001, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 13.557118966674805 + }, + { + "epoch": 4, + "eval_accuracy": 0.2646, + "eval_loss": 19.696252734375, + "eval_split": "validation", + "lr": 0.08, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 19.190646782769097 + }, + { + "epoch": 5, + "eval_accuracy": 0.2452, + "eval_loss": 33.2675609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 23.807205302937824 + }, + { + "epoch": 6, + "eval_accuracy": 0.2678, + "eval_loss": 40.026609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 24.055041174994575 + }, + { + "epoch": 7, + "eval_accuracy": 0.3196, + "eval_loss": 23.8587390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 24.712686389160154 + }, + { + "epoch": 8, + "eval_accuracy": 0.3278, + "eval_loss": 27.1260609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 23.086525019327798 + }, + { + "epoch": 9, + "eval_accuracy": 0.2696, + "eval_loss": 30.96389921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 24.295748993937174 + }, + { + "epoch": 10, + "eval_accuracy": 0.2384, + "eval_loss": 51.4358140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 21.023040753851998 + }, + { + "epoch": 11, + "eval_accuracy": 0.3404, + "eval_loss": 25.350265234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 23.475594851345488 + }, + { + "epoch": 12, + "eval_accuracy": 0.336, + "eval_loss": 22.1634453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 22.52266331515842 + }, + { + "epoch": 13, + "eval_accuracy": 0.3306, + "eval_loss": 23.82544453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 23.4229757297092 + }, + { + "epoch": 14, + "eval_accuracy": 0.2642, + "eval_loss": 34.5797359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 23.008342339409722 + }, + { + "epoch": 15, + "eval_accuracy": 0.375, + "eval_loss": 23.1807421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 20.60292657385932 + }, + { + "epoch": 16, + "eval_accuracy": 0.2904, + "eval_loss": 36.161196875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 19.139730751546224 + }, + { + "epoch": 17, + "eval_accuracy": 0.3454, + "eval_loss": 26.833031640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 20.002529979112413 + }, + { + "epoch": 18, + "eval_accuracy": 0.3586, + "eval_loss": 21.24528046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 18.727348812188044 + }, + { + "epoch": 19, + "eval_accuracy": 0.3632, + "eval_loss": 25.525848828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 17.44186926167806 + }, + { + "epoch": 20, + "eval_accuracy": 0.301, + "eval_loss": 36.5179703125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 17.90213321668837 + }, + { + "epoch": 21, + "eval_accuracy": 0.3958, + "eval_loss": 19.76304296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 17.890375726996528 + }, + { + "epoch": 22, + "eval_accuracy": 0.327, + "eval_loss": 28.7622765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 17.43689203559028 + }, + { + "epoch": 23, + "eval_accuracy": 0.2592, + "eval_loss": 41.68627578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 17.345420552571614 + }, + { + "epoch": 24, + "eval_accuracy": 0.4054, + "eval_loss": 21.936447265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 16.934118476019965 + }, + { + "epoch": 25, + "eval_accuracy": 0.2902, + "eval_loss": 30.93798828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 18.798614492119682 + }, + { + "epoch": 26, + "eval_accuracy": 0.3712, + "eval_loss": 29.504349609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 18.812271053059895 + }, + { + "epoch": 27, + "eval_accuracy": 0.342, + "eval_loss": 26.395713671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 18.473190761650933 + }, + { + "epoch": 28, + "eval_accuracy": 0.3854, + "eval_loss": 26.093215625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 18.005538069322373 + }, + { + "epoch": 29, + "eval_accuracy": 0.3012, + "eval_loss": 25.42871171875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 18.125954042222766 + }, + { + "epoch": 30, + "eval_accuracy": 0.4244, + "eval_loss": 19.527813671875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 18.05612528415256 + }, + { + "epoch": 31, + "eval_accuracy": 0.3614, + "eval_loss": 21.34248828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 19.15135492943658 + }, + { + "epoch": 32, + "eval_accuracy": 0.4132, + "eval_loss": 20.4143234375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 18.49009884847005 + }, + { + "epoch": 33, + "eval_accuracy": 0.4256, + "eval_loss": 16.4301474609375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 18.47117029859755 + }, + { + "epoch": 34, + "eval_accuracy": 0.4202, + "eval_loss": 15.4893919921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 17.859162979804147 + }, + { + "epoch": 35, + "eval_accuracy": 0.388, + "eval_loss": 16.226535546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 18.110402368503145 + }, + { + "epoch": 36, + "eval_accuracy": 0.4052, + "eval_loss": 18.623937890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 17.06648885972765 + }, + { + "epoch": 37, + "eval_accuracy": 0.3954, + "eval_loss": 18.58285390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 19.626207305230036 + }, + { + "epoch": 38, + "eval_accuracy": 0.424, + "eval_loss": 26.581407421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 18.639922721354168 + }, + { + "epoch": 39, + "eval_accuracy": 0.4192, + "eval_loss": 19.451832421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 18.782825440131294 + }, + { + "epoch": 40, + "eval_accuracy": 0.3804, + "eval_loss": 26.744866015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 17.683322509087457 + }, + { + "epoch": 41, + "eval_accuracy": 0.395, + "eval_loss": 34.78169375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 18.225805063544378 + }, + { + "epoch": 42, + "eval_accuracy": 0.412, + "eval_loss": 29.765325, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 17.984680138481988 + }, + { + "epoch": 43, + "eval_accuracy": 0.3786, + "eval_loss": 19.825704296875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 17.825573151312934 + }, + { + "epoch": 44, + "eval_accuracy": 0.3604, + "eval_loss": 26.8407375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 19.19038238796658 + }, + { + "epoch": 45, + "eval_accuracy": 0.4052, + "eval_loss": 12.7454291015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 16.198456757269966 + }, + { + "epoch": 46, + "eval_accuracy": 0.424, + "eval_loss": 15.499886328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 16.85351919962565 + }, + { + "epoch": 47, + "eval_accuracy": 0.4306, + "eval_loss": 16.8180568359375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 17.23167397986518 + }, + { + "epoch": 48, + "eval_accuracy": 0.4574, + "eval_loss": 15.12474765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 17.16782407752143 + }, + { + "epoch": 49, + "eval_accuracy": 0.3906, + "eval_loss": 33.472205859375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 17.989877756754556 + }, + { + "epoch": 50, + "eval_accuracy": 0.3944, + "eval_loss": 18.846742578125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 16.750614368184408 + }, + { + "epoch": 51, + "eval_accuracy": 0.3306, + "eval_loss": 20.342544921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 16.608609683566623 + }, + { + "epoch": 52, + "eval_accuracy": 0.4454, + "eval_loss": 21.5571515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 16.961722309366863 + }, + { + "epoch": 53, + "eval_accuracy": 0.3998, + "eval_loss": 23.065123828125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 16.919527493286132 + }, + { + "epoch": 54, + "eval_accuracy": 0.3482, + "eval_loss": 25.1215953125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 16.035512492031522 + }, + { + "epoch": 55, + "eval_accuracy": 0.469, + "eval_loss": 15.480585546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 17.075127189466688 + }, + { + "epoch": 56, + "eval_accuracy": 0.4464, + "eval_loss": 14.592352734375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 17.66614246198866 + }, + { + "epoch": 57, + "eval_accuracy": 0.4314, + "eval_loss": 22.738269140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 16.894754077487523 + }, + { + "epoch": 58, + "eval_accuracy": 0.3318, + "eval_loss": 45.84163984375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 19.05143542683919 + }, + { + "epoch": 59, + "eval_accuracy": 0.372, + "eval_loss": 31.423384375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 18.783138044230142 + }, + { + "epoch": 60, + "eval_accuracy": 0.418, + "eval_loss": 16.9460552734375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 16.77694915907118 + }, + { + "epoch": 61, + "eval_accuracy": 0.361, + "eval_loss": 22.063344140625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 16.14378400200738 + }, + { + "epoch": 62, + "eval_accuracy": 0.408, + "eval_loss": 24.965109765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 17.721834357367623 + }, + { + "epoch": 63, + "eval_accuracy": 0.4464, + "eval_loss": 20.872457421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 17.106755812411837 + }, + { + "epoch": 64, + "eval_accuracy": 0.351, + "eval_loss": 27.144359765625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 16.525339327663847 + }, + { + "epoch": 65, + "eval_accuracy": 0.401, + "eval_loss": 28.997582421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 17.898866891988117 + }, + { + "epoch": 66, + "eval_accuracy": 0.4388, + "eval_loss": 25.025003515625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 18.09744178805881 + }, + { + "epoch": 67, + "eval_accuracy": 0.4728, + "eval_loss": 16.6167294921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 17.755582858784994 + }, + { + "epoch": 68, + "eval_accuracy": 0.4108, + "eval_loss": 19.801496484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 17.263119559054903 + }, + { + "epoch": 69, + "eval_accuracy": 0.4906, + "eval_loss": 18.20902421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 17.33434205186632 + }, + { + "epoch": 70, + "eval_accuracy": 0.4292, + "eval_loss": 34.2252921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 18.17401082695855 + }, + { + "epoch": 71, + "eval_accuracy": 0.4798, + "eval_loss": 15.21786796875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 18.021223406304255 + }, + { + "epoch": 72, + "eval_accuracy": 0.4518, + "eval_loss": 22.791141015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 19.531677550930446 + }, + { + "epoch": 73, + "eval_accuracy": 0.368, + "eval_loss": 38.81344453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 18.620982016838923 + }, + { + "epoch": 74, + "eval_accuracy": 0.4228, + "eval_loss": 30.719596484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 20.337078319295248 + }, + { + "epoch": 75, + "eval_accuracy": 0.4418, + "eval_loss": 21.886297265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 20.1115397922092 + }, + { + "epoch": 76, + "eval_accuracy": 0.4868, + "eval_loss": 16.3975734375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 17.937315833706325 + }, + { + "epoch": 77, + "eval_accuracy": 0.325, + "eval_loss": 45.389496875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 21.48694476860894 + }, + { + "epoch": 78, + "eval_accuracy": 0.4012, + "eval_loss": 26.195025390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 22.31231401570638 + }, + { + "epoch": 79, + "eval_accuracy": 0.445, + "eval_loss": 27.56462890625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 21.119349666341147 + }, + { + "epoch": 80, + "eval_accuracy": 0.471, + "eval_loss": 17.123298046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 20.679742971123588 + }, + { + "epoch": 81, + "eval_accuracy": 0.4214, + "eval_loss": 27.093696484375, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 20.63988988172743 + }, + { + "epoch": 82, + "eval_accuracy": 0.4414, + "eval_loss": 26.012460546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 21.493989885457356 + }, + { + "epoch": 83, + "eval_accuracy": 0.3698, + "eval_loss": 30.0991921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 22.169163477918836 + }, + { + "epoch": 84, + "eval_accuracy": 0.3986, + "eval_loss": 31.00485078125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 23.3011970703125 + }, + { + "epoch": 85, + "eval_accuracy": 0.442, + "eval_loss": 23.574222265625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 21.34971837056478 + }, + { + "epoch": 86, + "eval_accuracy": 0.4296, + "eval_loss": 29.412936328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 21.91920473361545 + }, + { + "epoch": 87, + "eval_accuracy": 0.4386, + "eval_loss": 20.903475390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 22.060477025689018 + }, + { + "epoch": 88, + "eval_accuracy": 0.4506, + "eval_loss": 22.551448046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 21.711965223524306 + }, + { + "epoch": 89, + "eval_accuracy": 0.3526, + "eval_loss": 61.11854921875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 21.463106849839953 + }, + { + "epoch": 90, + "eval_accuracy": 0.4564, + "eval_loss": 23.186648046875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 24.00905051337348 + }, + { + "epoch": 91, + "eval_accuracy": 0.376, + "eval_loss": 60.64270546875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 21.28097244635688 + }, + { + "epoch": 92, + "eval_accuracy": 0.3868, + "eval_loss": 27.0673390625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 23.36898587782118 + }, + { + "epoch": 93, + "eval_accuracy": 0.4672, + "eval_loss": 22.039832421875, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 20.88351126539442 + }, + { + "epoch": 94, + "eval_accuracy": 0.3888, + "eval_loss": 25.698695703125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 21.979886994425456 + }, + { + "epoch": 95, + "eval_accuracy": 0.4346, + "eval_loss": 26.78786640625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 21.236860284084745 + }, + { + "epoch": 96, + "eval_accuracy": 0.3624, + "eval_loss": 36.5288453125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 21.10884918484158 + }, + { + "epoch": 97, + "eval_accuracy": 0.4434, + "eval_loss": 23.964403125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 20.604113126966688 + }, + { + "epoch": 98, + "eval_accuracy": 0.4674, + "eval_loss": 24.196286328125, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 19.82856385260688 + }, + { + "epoch": 99, + "eval_accuracy": 0.4354, + "eval_loss": 24.922025, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 19.128273210313584 + }, + { + "epoch": 100, + "eval_accuracy": 0.3788, + "eval_loss": 31.4631015625, + "eval_split": "validation", + "lr": 0.1, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 21.582442571343314 + }, + { + "epoch": 101, + "eval_accuracy": 0.5972, + "eval_loss": 4.836526171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 5.9259513893975155 + }, + { + "epoch": 102, + "eval_accuracy": 0.5624, + "eval_loss": 4.101729345703125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 3.619677195019192 + }, + { + "epoch": 103, + "eval_accuracy": 0.5986, + "eval_loss": 2.667718505859375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 2.880115439690484 + }, + { + "epoch": 104, + "eval_accuracy": 0.5958, + "eval_loss": 2.390951025390625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 2.4437281640370685 + }, + { + "epoch": 105, + "eval_accuracy": 0.574, + "eval_loss": 2.260222998046875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 2.226529400253296 + }, + { + "epoch": 106, + "eval_accuracy": 0.5504, + "eval_loss": 2.49472744140625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 2.185105883640713 + }, + { + "epoch": 107, + "eval_accuracy": 0.5266, + "eval_loss": 2.41015546875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 2.0920644403245716 + }, + { + "epoch": 108, + "eval_accuracy": 0.5886, + "eval_loss": 1.7911347412109375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 2.0197150400797526 + }, + { + "epoch": 109, + "eval_accuracy": 0.5726, + "eval_loss": 2.175161474609375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 1.8753828492058648 + }, + { + "epoch": 110, + "eval_accuracy": 0.58, + "eval_loss": 1.9253150146484375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 1.8506183098475137 + }, + { + "epoch": 111, + "eval_accuracy": 0.5912, + "eval_loss": 1.810534423828125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 1.7877488494873046 + }, + { + "epoch": 112, + "eval_accuracy": 0.5956, + "eval_loss": 1.6881039306640624, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 1.7687504001193577 + }, + { + "epoch": 113, + "eval_accuracy": 0.5764, + "eval_loss": 1.8380051513671876, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 1.7672821660783555 + }, + { + "epoch": 114, + "eval_accuracy": 0.5794, + "eval_loss": 1.6862110595703126, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 1.6414273686091105 + }, + { + "epoch": 115, + "eval_accuracy": 0.5752, + "eval_loss": 1.6944744384765624, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 1.6939214492797852 + }, + { + "epoch": 116, + "eval_accuracy": 0.5926, + "eval_loss": 1.582383349609375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 1.597629587088691 + }, + { + "epoch": 117, + "eval_accuracy": 0.5588, + "eval_loss": 1.856890869140625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 1.6295654713312784 + }, + { + "epoch": 118, + "eval_accuracy": 0.5834, + "eval_loss": 1.705884130859375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 1.5944301096810234 + }, + { + "epoch": 119, + "eval_accuracy": 0.608, + "eval_loss": 1.561981005859375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 1.5895581109788683 + }, + { + "epoch": 120, + "eval_accuracy": 0.5932, + "eval_loss": 1.6351948486328125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 1.5375917362001208 + }, + { + "epoch": 121, + "eval_accuracy": 0.5716, + "eval_loss": 1.6493072265625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 1.5845890295876397 + }, + { + "epoch": 122, + "eval_accuracy": 0.6058, + "eval_loss": 1.49820810546875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 1.5254730051252576 + }, + { + "epoch": 123, + "eval_accuracy": 0.5652, + "eval_loss": 1.77642001953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 1.5247097325219048 + }, + { + "epoch": 124, + "eval_accuracy": 0.6068, + "eval_loss": 1.376882470703125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 1.5499936188379924 + }, + { + "epoch": 125, + "eval_accuracy": 0.5676, + "eval_loss": 1.7324075927734375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 1.615310203382704 + }, + { + "epoch": 126, + "eval_accuracy": 0.5954, + "eval_loss": 1.5954062255859376, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 1.4982189151763916 + }, + { + "epoch": 127, + "eval_accuracy": 0.5846, + "eval_loss": 1.7751647705078124, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 1.5301205956353081 + }, + { + "epoch": 128, + "eval_accuracy": 0.5602, + "eval_loss": 1.6907564697265625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 1.4734539789411756 + }, + { + "epoch": 129, + "eval_accuracy": 0.6002, + "eval_loss": 1.40965439453125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 1.4641339381747775 + }, + { + "epoch": 130, + "eval_accuracy": 0.6192, + "eval_loss": 1.4004435791015626, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 1.5231939977010092 + }, + { + "epoch": 131, + "eval_accuracy": 0.5982, + "eval_loss": 1.45526376953125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 1.497420593600803 + }, + { + "epoch": 132, + "eval_accuracy": 0.5884, + "eval_loss": 1.550844140625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 1.469409120665656 + }, + { + "epoch": 133, + "eval_accuracy": 0.6256, + "eval_loss": 1.267538330078125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 1.4565586593204074 + }, + { + "epoch": 134, + "eval_accuracy": 0.5592, + "eval_loss": 1.77311513671875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 1.437351505109999 + }, + { + "epoch": 135, + "eval_accuracy": 0.5574, + "eval_loss": 1.940101904296875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 1.4218262961917454 + }, + { + "epoch": 136, + "eval_accuracy": 0.5774, + "eval_loss": 1.6147098388671874, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 1.411111260647244 + }, + { + "epoch": 137, + "eval_accuracy": 0.6128, + "eval_loss": 1.357390869140625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 1.5000552250120376 + }, + { + "epoch": 138, + "eval_accuracy": 0.5848, + "eval_loss": 1.632561083984375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 1.3685510100682576 + }, + { + "epoch": 139, + "eval_accuracy": 0.6142, + "eval_loss": 1.3133041015625, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 1.3762463880327014 + }, + { + "epoch": 140, + "eval_accuracy": 0.5754, + "eval_loss": 1.5640064453125, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 1.401551789580451 + }, + { + "epoch": 141, + "eval_accuracy": 0.594, + "eval_loss": 1.479112841796875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 1.4058031635072497 + }, + { + "epoch": 142, + "eval_accuracy": 0.5914, + "eval_loss": 1.5028365478515624, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 1.4034333087073432 + }, + { + "epoch": 143, + "eval_accuracy": 0.6102, + "eval_loss": 1.5048456298828126, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 1.3954313831117418 + }, + { + "epoch": 144, + "eval_accuracy": 0.5666, + "eval_loss": 1.6636828369140626, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 1.3667300954606798 + }, + { + "epoch": 145, + "eval_accuracy": 0.5612, + "eval_loss": 1.6101176025390624, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 1.404260525724623 + }, + { + "epoch": 146, + "eval_accuracy": 0.5556, + "eval_loss": 1.876030859375, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 1.4051593517515395 + }, + { + "epoch": 147, + "eval_accuracy": 0.5764, + "eval_loss": 1.6018065673828126, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 1.4870435859256321 + }, + { + "epoch": 148, + "eval_accuracy": 0.6056, + "eval_loss": 1.43818701171875, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 1.367290024079217 + }, + { + "epoch": 149, + "eval_accuracy": 0.6162, + "eval_loss": 1.3064309814453126, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 1.3197986438751221 + }, + { + "epoch": 150, + "eval_accuracy": 0.6296, + "eval_loss": 1.2498236572265624, + "eval_split": "validation", + "lr": 0.010000000000000002, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 1.3129838515387642 + }, + { + "epoch": 151, + "eval_accuracy": 0.6534, + "eval_loss": 1.07179833984375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 1.026788145626916 + }, + { + "epoch": 152, + "eval_accuracy": 0.6568, + "eval_loss": 1.0430185913085936, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 1.009344142680698 + }, + { + "epoch": 153, + "eval_accuracy": 0.658, + "eval_loss": 1.0307321044921876, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 1.0141865419387817 + }, + { + "epoch": 154, + "eval_accuracy": 0.659, + "eval_loss": 1.02932734375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 1.0022637588077121 + }, + { + "epoch": 155, + "eval_accuracy": 0.6604, + "eval_loss": 1.0242865844726563, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.9988587782012092 + }, + { + "epoch": 156, + "eval_accuracy": 0.6674, + "eval_loss": 1.0057396484375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.9961366279178195 + }, + { + "epoch": 157, + "eval_accuracy": 0.6548, + "eval_loss": 1.03570576171875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.9966530807283189 + }, + { + "epoch": 158, + "eval_accuracy": 0.6616, + "eval_loss": 1.0033466552734376, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 1.0007755590226914 + }, + { + "epoch": 159, + "eval_accuracy": 0.6624, + "eval_loss": 0.99225693359375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 1.0026236204147339 + }, + { + "epoch": 160, + "eval_accuracy": 0.661, + "eval_loss": 1.00777568359375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 1.0010402308146158 + }, + { + "epoch": 161, + "eval_accuracy": 0.6566, + "eval_loss": 1.0218382690429688, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.9932583589553833 + }, + { + "epoch": 162, + "eval_accuracy": 0.6546, + "eval_loss": 1.0098359619140624, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 1.0004163008266025 + }, + { + "epoch": 163, + "eval_accuracy": 0.662, + "eval_loss": 1.0110062622070313, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 1.0034730763753255 + }, + { + "epoch": 164, + "eval_accuracy": 0.6494, + "eval_loss": 1.0401352172851563, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 1.0073212409337362 + }, + { + "epoch": 165, + "eval_accuracy": 0.6474, + "eval_loss": 1.0578818725585937, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 1.0492681529257033 + }, + { + "epoch": 166, + "eval_accuracy": 0.6434, + "eval_loss": 1.0694183837890625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 1.0751917953279284 + }, + { + "epoch": 167, + "eval_accuracy": 0.6294, + "eval_loss": 1.1221494384765625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 1.0811405420939129 + }, + { + "epoch": 168, + "eval_accuracy": 0.6354, + "eval_loss": 1.0892340576171875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 1.0801018522050645 + }, + { + "epoch": 169, + "eval_accuracy": 0.6412, + "eval_loss": 1.0978382080078124, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 1.0860861628214518 + }, + { + "epoch": 170, + "eval_accuracy": 0.6346, + "eval_loss": 1.090638720703125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 1.0723167603810628 + }, + { + "epoch": 171, + "eval_accuracy": 0.6272, + "eval_loss": 1.1234599609375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 1.0920719518237645 + }, + { + "epoch": 172, + "eval_accuracy": 0.6284, + "eval_loss": 1.099410595703125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 1.0925795223236083 + }, + { + "epoch": 173, + "eval_accuracy": 0.63, + "eval_loss": 1.104658056640625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 1.087668931304084 + }, + { + "epoch": 174, + "eval_accuracy": 0.628, + "eval_loss": 1.1068023193359375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 1.094618272082011 + }, + { + "epoch": 175, + "eval_accuracy": 0.6318, + "eval_loss": 1.09778740234375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 1.0823923013687133 + }, + { + "epoch": 176, + "eval_accuracy": 0.6308, + "eval_loss": 1.10460810546875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 1.0776995142830743 + }, + { + "epoch": 177, + "eval_accuracy": 0.6418, + "eval_loss": 1.0786893798828125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 1.069732749218411 + }, + { + "epoch": 178, + "eval_accuracy": 0.6406, + "eval_loss": 1.0766607421875, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 1.0717433659447564 + }, + { + "epoch": 179, + "eval_accuracy": 0.635, + "eval_loss": 1.0887271728515624, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 1.0777871327506172 + }, + { + "epoch": 180, + "eval_accuracy": 0.6436, + "eval_loss": 1.062487890625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 1.070056160566542 + }, + { + "epoch": 181, + "eval_accuracy": 0.6448, + "eval_loss": 1.060628564453125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 1.0583967365688747 + }, + { + "epoch": 182, + "eval_accuracy": 0.6446, + "eval_loss": 1.058402197265625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 1.0691630625618829 + }, + { + "epoch": 183, + "eval_accuracy": 0.6366, + "eval_loss": 1.0809754150390625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 1.0663337293412951 + }, + { + "epoch": 184, + "eval_accuracy": 0.6452, + "eval_loss": 1.058856298828125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 1.0665179736455281 + }, + { + "epoch": 185, + "eval_accuracy": 0.6398, + "eval_loss": 1.06523369140625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 1.0648094150119358 + }, + { + "epoch": 186, + "eval_accuracy": 0.64, + "eval_loss": 1.060458837890625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 1.0582156455569798 + }, + { + "epoch": 187, + "eval_accuracy": 0.6538, + "eval_loss": 1.04734462890625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 1.0498132875124613 + }, + { + "epoch": 188, + "eval_accuracy": 0.6562, + "eval_loss": 1.0362562622070313, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 1.0539691292444864 + }, + { + "epoch": 189, + "eval_accuracy": 0.6546, + "eval_loss": 1.0296682861328126, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 1.050129378912184 + }, + { + "epoch": 190, + "eval_accuracy": 0.6484, + "eval_loss": 1.0466437377929687, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 1.051998159429762 + }, + { + "epoch": 191, + "eval_accuracy": 0.6538, + "eval_loss": 1.0468611083984376, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 1.0416840789159139 + }, + { + "epoch": 192, + "eval_accuracy": 0.6592, + "eval_loss": 1.0471241577148438, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 1.0441420639038086 + }, + { + "epoch": 193, + "eval_accuracy": 0.6478, + "eval_loss": 1.0665528564453124, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 1.048444183370802 + }, + { + "epoch": 194, + "eval_accuracy": 0.647, + "eval_loss": 1.0716345703125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 1.0648873981157938 + }, + { + "epoch": 195, + "eval_accuracy": 0.6476, + "eval_loss": 1.0518780395507812, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 1.0581021981451246 + }, + { + "epoch": 196, + "eval_accuracy": 0.647, + "eval_loss": 1.0528640258789062, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 1.0561888460053337 + }, + { + "epoch": 197, + "eval_accuracy": 0.6446, + "eval_loss": 1.0466599609375, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 1.0471556855837505 + }, + { + "epoch": 198, + "eval_accuracy": 0.653, + "eval_loss": 1.0403861328125, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 1.0446702560000949 + }, + { + "epoch": 199, + "eval_accuracy": 0.6442, + "eval_loss": 1.060380322265625, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 1.0528953741285536 + }, + { + "epoch": 200, + "eval_accuracy": 0.6436, + "eval_loss": 1.0704212158203126, + "eval_split": "validation", + "lr": 0.0010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 1.0543489791870118 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.6436, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 1.0704212158203126, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 3252628992, + "peak_memory_reserved_bytes": 4510973952, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 61.098384141922, + "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": 3827.3347294330597, + "train_wall_s": 3766.213252544403 + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1125389952000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 1129371264000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1129371264000000, + "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": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 3384132480000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json.manifest.json new file mode 100644 index 0000000..fa0f73d --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "hfa", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4" + ], + "completed_unix_time": 1787019593.461147, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 3838.2244839668274, + "epochs": 200, + "experiment_name": "resnet-r2-fa-d56-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "fa", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-fa-d56-lr0p1.json", + "output_exists": true, + "output_sha256": "db985851db69598966506f4067c2febdd411b2fa61b02d330263f6a9ed90a776", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json b/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json new file mode 100644 index 0000000..fbf60bd --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json @@ -0,0 +1,8265 @@ +{ + "architecture": { + "base_width": 16, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 269722, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 40, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "constant", + "max_steps": 0, + "method": "ff", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json", + "output_lr": 0.03, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.1174, + "evaluation_seconds": 1.5146634578704834, + "examples": 5000, + "finite": true, + "loss": null + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 5474375168, + "peak_memory_reserved_bytes": 5981077504, + "torch_version": "2.3.1+cu118" + }, + "layers": [ + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.4005121697319878, + "lr": 0.03, + "negative_goodness": 1.9601310234069824, + "pair_accuracy": 0.4739333333598243, + "positive_goodness": 1.960143262354533, + "runtime_seconds": 4.442354917526245 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.387007635582818, + "lr": 0.03, + "negative_goodness": 1.9987232278611924, + "pair_accuracy": 0.47797777780426876, + "positive_goodness": 1.9987444209204779, + "runtime_seconds": 4.020473480224609 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3869588102340697, + "lr": 0.03, + "negative_goodness": 1.9987902463701037, + "pair_accuracy": 0.48168888891537986, + "positive_goodness": 1.9988118118286133, + "runtime_seconds": 4.022093057632446 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3869192715115017, + "lr": 0.03, + "negative_goodness": 1.9989530067443848, + "pair_accuracy": 0.47273333336512247, + "positive_goodness": 1.9989705510033502, + "runtime_seconds": 4.014694690704346 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3867757701237997, + "lr": 0.03, + "negative_goodness": 1.999157364824083, + "pair_accuracy": 0.47580000001589456, + "positive_goodness": 1.9991776374816894, + "runtime_seconds": 4.016554832458496 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.386698933813307, + "lr": 0.03, + "negative_goodness": 1.9993364699469671, + "pair_accuracy": 0.4782444444656372, + "positive_goodness": 1.9993548776838515, + "runtime_seconds": 4.028818130493164 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3865385099834866, + "lr": 0.03, + "negative_goodness": 1.9995932447221545, + "pair_accuracy": 0.4743111111164093, + "positive_goodness": 1.9996111343807643, + "runtime_seconds": 4.022809743881226 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3864329044977823, + "lr": 0.03, + "negative_goodness": 1.999763582971361, + "pair_accuracy": 0.47788888888888886, + "positive_goodness": 1.9997798079808553, + "runtime_seconds": 4.012293815612793 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3863745658662585, + "lr": 0.03, + "negative_goodness": 1.9998732730017768, + "pair_accuracy": 0.48206666666666664, + "positive_goodness": 1.9998874375661215, + "runtime_seconds": 4.015212297439575 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.38633431869083, + "lr": 0.03, + "negative_goodness": 1.9999270596398249, + "pair_accuracy": 0.4774444444126553, + "positive_goodness": 1.9999369727240668, + "runtime_seconds": 4.020283460617065 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.3863177184846667, + "lr": 0.03, + "negative_goodness": 1.9999535449557835, + "pair_accuracy": 0.48113333332273694, + "positive_goodness": 1.9999625151316325, + "runtime_seconds": 4.0136635303497314 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3863131265004476, + "lr": 0.03, + "negative_goodness": 1.999965398449368, + "pair_accuracy": 0.479, + "positive_goodness": 1.9999739984088474, + "runtime_seconds": 4.013098478317261 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.3863072466108535, + "lr": 0.03, + "negative_goodness": 1.9999784149593778, + "pair_accuracy": 0.47993333331214055, + "positive_goodness": 1.9999848613527087, + "runtime_seconds": 3.9994678497314453 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3863020559734769, + "lr": 0.03, + "negative_goodness": 1.999984212176005, + "pair_accuracy": 0.4804888888888889, + "positive_goodness": 1.9999912085851033, + "runtime_seconds": 4.003154039382935 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.3862991156260172, + "lr": 0.03, + "negative_goodness": 1.9999898837831285, + "pair_accuracy": 0.47864444448682997, + "positive_goodness": 1.9999962279849581, + "runtime_seconds": 4.000916481018066 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3862972198486327, + "lr": 0.03, + "negative_goodness": 1.9999917033513388, + "pair_accuracy": 0.47744444443384804, + "positive_goodness": 1.999997399160597, + "runtime_seconds": 4.000702142715454 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3862957434124417, + "lr": 0.03, + "negative_goodness": 1.9999922891616821, + "pair_accuracy": 0.47704444445504085, + "positive_goodness": 1.9999996165593465, + "runtime_seconds": 4.0117340087890625 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3862954868740505, + "lr": 0.03, + "negative_goodness": 1.999992126803928, + "pair_accuracy": 0.47262222221162586, + "positive_goodness": 2.0000018541971842, + "runtime_seconds": 4.025058031082153 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.386294012811449, + "lr": 0.03, + "negative_goodness": 1.9999834719975789, + "pair_accuracy": 0.46595555555555557, + "positive_goodness": 2.000004315800137, + "runtime_seconds": 4.020641803741455 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3862845910602146, + "lr": 0.03, + "negative_goodness": 1.9999536404927571, + "pair_accuracy": 0.4767777777618832, + "positive_goodness": 2.000026096534729, + "runtime_seconds": 4.025409698486328 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3862585081736247, + "lr": 0.03, + "negative_goodness": 1.999894027794732, + "pair_accuracy": 0.48071111107932196, + "positive_goodness": 2.00008999065823, + "runtime_seconds": 4.013952255249023 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.3862183980941774, + "lr": 0.03, + "negative_goodness": 1.9998316767374675, + "pair_accuracy": 0.480444444460339, + "positive_goodness": 2.0001379673851862, + "runtime_seconds": 4.018738746643066 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3861571537653605, + "lr": 0.03, + "negative_goodness": 1.9997450352350872, + "pair_accuracy": 0.47771111108991837, + "positive_goodness": 2.0002439667595757, + "runtime_seconds": 4.02006196975708 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3861010393354627, + "lr": 0.03, + "negative_goodness": 1.9995863431294758, + "pair_accuracy": 0.4808888888994853, + "positive_goodness": 2.0003641815609403, + "runtime_seconds": 4.017390727996826 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3860183791266547, + "lr": 0.03, + "negative_goodness": 1.9994270960913765, + "pair_accuracy": 0.4767111110952165, + "positive_goodness": 2.0004117652045355, + "runtime_seconds": 4.0145299434661865 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3858748785018922, + "lr": 0.03, + "negative_goodness": 1.999183433702257, + "pair_accuracy": 0.47682222221692405, + "positive_goodness": 2.0007004488839044, + "runtime_seconds": 4.007335186004639 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3857943122016059, + "lr": 0.03, + "negative_goodness": 1.999058674536811, + "pair_accuracy": 0.4764666666560703, + "positive_goodness": 2.0009235049353706, + "runtime_seconds": 4.0228271484375 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3859377777311537, + "lr": 0.03, + "negative_goodness": 1.9993038423326281, + "pair_accuracy": 0.48022222221162586, + "positive_goodness": 2.000637299092611, + "runtime_seconds": 4.021588563919067 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3855776105456883, + "lr": 0.03, + "negative_goodness": 1.9986911282433404, + "pair_accuracy": 0.4796222222222222, + "positive_goodness": 2.0011943065219455, + "runtime_seconds": 4.012900114059448 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3854861862394545, + "lr": 0.03, + "negative_goodness": 1.998729441663954, + "pair_accuracy": 0.47946666669845583, + "positive_goodness": 2.001515883763631, + "runtime_seconds": 4.018191576004028 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3852819519466824, + "lr": 0.03, + "negative_goodness": 1.998212189102173, + "pair_accuracy": 0.4805555555661519, + "positive_goodness": 2.001762049272325, + "runtime_seconds": 4.014139890670776 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3854936304516263, + "lr": 0.03, + "negative_goodness": 1.998297024536133, + "pair_accuracy": 0.48102222224871316, + "positive_goodness": 2.0014716115103828, + "runtime_seconds": 4.005394697189331 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.385308371755812, + "lr": 0.03, + "negative_goodness": 1.9982786936865913, + "pair_accuracy": 0.4831555555873447, + "positive_goodness": 2.0018401529735987, + "runtime_seconds": 4.019372940063477 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3851973777982924, + "lr": 0.03, + "negative_goodness": 1.997974285613166, + "pair_accuracy": 0.48304444447623357, + "positive_goodness": 2.0018273219214544, + "runtime_seconds": 4.024045705795288 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.3853170914120145, + "lr": 0.03, + "negative_goodness": 1.9982329156239826, + "pair_accuracy": 0.4868444444762336, + "positive_goodness": 2.0019923520618015, + "runtime_seconds": 4.022265434265137 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.3852790722952948, + "lr": 0.03, + "negative_goodness": 1.9981024457719592, + "pair_accuracy": 0.48771111113230387, + "positive_goodness": 2.001845592329237, + "runtime_seconds": 4.022640943527222 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.38514535668691, + "lr": 0.03, + "negative_goodness": 1.9979354625066121, + "pair_accuracy": 0.4862888889418708, + "positive_goodness": 2.0020988483428956, + "runtime_seconds": 4.01956844329834 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.3853505014419556, + "lr": 0.03, + "negative_goodness": 1.9982058176676432, + "pair_accuracy": 0.48740000002119277, + "positive_goodness": 2.001756338289049, + "runtime_seconds": 4.021960020065308 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3851402035395304, + "lr": 0.03, + "negative_goodness": 1.9978458022223577, + "pair_accuracy": 0.4874222222275204, + "positive_goodness": 2.0021881000094943, + "runtime_seconds": 4.0125696659088135 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3851493269178603, + "lr": 0.03, + "negative_goodness": 1.9978735819498699, + "pair_accuracy": 0.4847333333492279, + "positive_goodness": 2.0020106267293296, + "runtime_seconds": 4.008484125137329 + } + ], + "layer": 0 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.3932117523829142, + "lr": 0.03, + "negative_goodness": 1.9616084063635932, + "pair_accuracy": 0.5247555555820466, + "positive_goodness": 1.974602836057875, + "runtime_seconds": 4.113959550857544 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3788628021876017, + "lr": 0.03, + "negative_goodness": 1.9895675208621555, + "pair_accuracy": 0.5424222222010294, + "positive_goodness": 2.014808629099528, + "runtime_seconds": 4.104449510574341 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3786113515641953, + "lr": 0.03, + "negative_goodness": 1.9886924804475572, + "pair_accuracy": 0.5455777777989705, + "positive_goodness": 2.015313993771871, + "runtime_seconds": 4.11234712600708 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.377780750550164, + "lr": 0.03, + "negative_goodness": 1.9879824965159099, + "pair_accuracy": 0.5495111111005148, + "positive_goodness": 2.017264287439982, + "runtime_seconds": 4.115158319473267 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3765279240290325, + "lr": 0.03, + "negative_goodness": 1.9865627773072985, + "pair_accuracy": 0.5553333333757189, + "positive_goodness": 2.019646032121446, + "runtime_seconds": 4.1120030879974365 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.376173061413235, + "lr": 0.03, + "negative_goodness": 1.9854636295954387, + "pair_accuracy": 0.5514000000317891, + "positive_goodness": 2.0205752878401015, + "runtime_seconds": 4.109644412994385 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.375299360148112, + "lr": 0.03, + "negative_goodness": 1.9841995730294122, + "pair_accuracy": 0.5567555555714502, + "positive_goodness": 2.0230227034038966, + "runtime_seconds": 4.113102436065674 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3759224034839206, + "lr": 0.03, + "negative_goodness": 1.985288997162713, + "pair_accuracy": 0.5568222222328186, + "positive_goodness": 2.021230537372165, + "runtime_seconds": 4.119059324264526 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3745707266277738, + "lr": 0.03, + "negative_goodness": 1.9830020762973362, + "pair_accuracy": 0.5587555555555556, + "positive_goodness": 2.024046246761746, + "runtime_seconds": 4.122097492218018 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3748952810499404, + "lr": 0.03, + "negative_goodness": 1.9827730545891655, + "pair_accuracy": 0.5614000000317891, + "positive_goodness": 2.023238257153829, + "runtime_seconds": 4.121074914932251 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.3740840565363566, + "lr": 0.03, + "negative_goodness": 1.9826951626459757, + "pair_accuracy": 0.5675333333757189, + "positive_goodness": 2.0253078161027696, + "runtime_seconds": 4.122215509414673 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3735402496761746, + "lr": 0.03, + "negative_goodness": 1.9814580365710788, + "pair_accuracy": 0.5683333333015442, + "positive_goodness": 2.0261573245578344, + "runtime_seconds": 4.118436336517334 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.3735931759940254, + "lr": 0.03, + "negative_goodness": 1.9812357808007135, + "pair_accuracy": 0.5697111111323039, + "positive_goodness": 2.0262974440256754, + "runtime_seconds": 4.121304035186768 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3727920609580146, + "lr": 0.03, + "negative_goodness": 1.9802444451861911, + "pair_accuracy": 0.5747333333757189, + "positive_goodness": 2.0279628845638697, + "runtime_seconds": 4.114097356796265 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.373754797384474, + "lr": 0.03, + "negative_goodness": 1.9805093935224745, + "pair_accuracy": 0.5681555555343628, + "positive_goodness": 2.025516858461168, + "runtime_seconds": 4.1124608516693115 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3733848461151124, + "lr": 0.03, + "negative_goodness": 1.9802470002916124, + "pair_accuracy": 0.5681777777989705, + "positive_goodness": 2.0270457220289444, + "runtime_seconds": 4.109096050262451 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3730305892096626, + "lr": 0.03, + "negative_goodness": 1.9798825311024983, + "pair_accuracy": 0.5712888889100817, + "positive_goodness": 2.0274167782677543, + "runtime_seconds": 4.112367630004883 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3722204663806492, + "lr": 0.03, + "negative_goodness": 1.9783378393809001, + "pair_accuracy": 0.5704666666984558, + "positive_goodness": 2.029221641243829, + "runtime_seconds": 4.114980220794678 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.3718132865269979, + "lr": 0.03, + "negative_goodness": 1.9782278279622396, + "pair_accuracy": 0.573044444433848, + "positive_goodness": 2.029831932682461, + "runtime_seconds": 4.117157459259033 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3723131368425157, + "lr": 0.03, + "negative_goodness": 1.977868480512831, + "pair_accuracy": 0.5662444444762336, + "positive_goodness": 2.0281764190673828, + "runtime_seconds": 4.110045671463013 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3716715061823528, + "lr": 0.03, + "negative_goodness": 1.977911436589559, + "pair_accuracy": 0.5648666666878595, + "positive_goodness": 2.0306532088597615, + "runtime_seconds": 4.111177921295166 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.3707589925130208, + "lr": 0.03, + "negative_goodness": 1.9765842014524673, + "pair_accuracy": 0.565288888894187, + "positive_goodness": 2.0322339486016165, + "runtime_seconds": 4.110898971557617 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3705475579367743, + "lr": 0.03, + "negative_goodness": 1.9760299474080403, + "pair_accuracy": 0.5623777777777778, + "positive_goodness": 2.032383829583062, + "runtime_seconds": 4.112079620361328 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3704794426176283, + "lr": 0.03, + "negative_goodness": 1.9754500739839342, + "pair_accuracy": 0.5654222222010294, + "positive_goodness": 2.0320675870259604, + "runtime_seconds": 4.110180377960205 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3701289758046469, + "lr": 0.03, + "negative_goodness": 1.9758330468071832, + "pair_accuracy": 0.5655777778201633, + "positive_goodness": 2.0334881067487927, + "runtime_seconds": 4.11220645904541 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3707050980673896, + "lr": 0.03, + "negative_goodness": 1.9754781870736016, + "pair_accuracy": 0.569222222243415, + "positive_goodness": 2.0323951642778186, + "runtime_seconds": 4.106773853302002 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3700828728569878, + "lr": 0.03, + "negative_goodness": 1.9751642570707533, + "pair_accuracy": 0.5675777777671814, + "positive_goodness": 2.032820465342204, + "runtime_seconds": 4.107302904129028 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3702864546669855, + "lr": 0.03, + "negative_goodness": 1.9752910245683457, + "pair_accuracy": 0.5660222222540113, + "positive_goodness": 2.0329235928429497, + "runtime_seconds": 4.10903525352478 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3692832588407728, + "lr": 0.03, + "negative_goodness": 1.974154246351454, + "pair_accuracy": 0.5685555555343628, + "positive_goodness": 2.034878907945421, + "runtime_seconds": 4.12065315246582 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3693706057442558, + "lr": 0.03, + "negative_goodness": 1.9733561725616455, + "pair_accuracy": 0.569355555597941, + "positive_goodness": 2.0345855024125843, + "runtime_seconds": 4.105545282363892 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.369382768503825, + "lr": 0.03, + "negative_goodness": 1.972923424741957, + "pair_accuracy": 0.571688888920678, + "positive_goodness": 2.034545042461819, + "runtime_seconds": 4.101029396057129 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3687460449006823, + "lr": 0.03, + "negative_goodness": 1.9730511844846936, + "pair_accuracy": 0.5806444444232517, + "positive_goodness": 2.0357176334381104, + "runtime_seconds": 4.110149383544922 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3699555436240303, + "lr": 0.03, + "negative_goodness": 1.9736906222449409, + "pair_accuracy": 0.5723111111534966, + "positive_goodness": 2.033873972998725, + "runtime_seconds": 4.110774993896484 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3682082608116997, + "lr": 0.03, + "negative_goodness": 1.972206425772773, + "pair_accuracy": 0.5771333333015441, + "positive_goodness": 2.037496131939358, + "runtime_seconds": 4.109734296798706 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.3686906690385607, + "lr": 0.03, + "negative_goodness": 1.9720414384629992, + "pair_accuracy": 0.5766666666348775, + "positive_goodness": 2.0361298236846923, + "runtime_seconds": 4.105909585952759 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.367443592071533, + "lr": 0.03, + "negative_goodness": 1.9704478031582302, + "pair_accuracy": 0.5784666667090522, + "positive_goodness": 2.0388355396906537, + "runtime_seconds": 4.10884690284729 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.369210840818617, + "lr": 0.03, + "negative_goodness": 1.9718139563030668, + "pair_accuracy": 0.5817111110793219, + "positive_goodness": 2.0347980559455023, + "runtime_seconds": 4.107010841369629 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.3679332441541883, + "lr": 0.03, + "negative_goodness": 1.9716569560792712, + "pair_accuracy": 0.5771111111111111, + "positive_goodness": 2.0374635754479304, + "runtime_seconds": 4.1074538230896 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3676768312030367, + "lr": 0.03, + "negative_goodness": 1.971307231648763, + "pair_accuracy": 0.5788666666560703, + "positive_goodness": 2.0386354187435574, + "runtime_seconds": 4.110522985458374 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3673629220750596, + "lr": 0.03, + "negative_goodness": 1.9699760090933907, + "pair_accuracy": 0.5747111111005148, + "positive_goodness": 2.038680723444621, + "runtime_seconds": 4.107245445251465 + } + ], + "layer": 1 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.377819001473321, + "lr": 0.03, + "negative_goodness": 1.988411927011278, + "pair_accuracy": 0.6025777777936724, + "positive_goodness": 2.0198782718870376, + "runtime_seconds": 4.189739227294922 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3689779887517293, + "lr": 0.03, + "negative_goodness": 1.977996727180481, + "pair_accuracy": 0.6368222222646077, + "positive_goodness": 2.0353663258446586, + "runtime_seconds": 4.187904357910156 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3653394732369317, + "lr": 0.03, + "negative_goodness": 1.9717015564176772, + "pair_accuracy": 0.6386222222222222, + "positive_goodness": 2.0421092027028402, + "runtime_seconds": 4.190337419509888 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3630898765775892, + "lr": 0.03, + "negative_goodness": 1.969029081132677, + "pair_accuracy": 0.6375333333545261, + "positive_goodness": 2.04675103225708, + "runtime_seconds": 4.1922993659973145 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3621609250174629, + "lr": 0.03, + "negative_goodness": 1.9674551836861505, + "pair_accuracy": 0.6351777777777777, + "positive_goodness": 2.0488910830603704, + "runtime_seconds": 4.192067861557007 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3584965751859877, + "lr": 0.03, + "negative_goodness": 1.962687554359436, + "pair_accuracy": 0.6484666666878595, + "positive_goodness": 2.0554449939727784, + "runtime_seconds": 4.188335657119751 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3579738311555651, + "lr": 0.03, + "negative_goodness": 1.9613158783170912, + "pair_accuracy": 0.6446222222222222, + "positive_goodness": 2.0575264681922065, + "runtime_seconds": 4.19608473777771 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3570324223836263, + "lr": 0.03, + "negative_goodness": 1.9603937348683675, + "pair_accuracy": 0.6486888889418708, + "positive_goodness": 2.0589015285491943, + "runtime_seconds": 4.192540407180786 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3565069114896986, + "lr": 0.03, + "negative_goodness": 1.9574748461405436, + "pair_accuracy": 0.6499555555237664, + "positive_goodness": 2.0602696283976236, + "runtime_seconds": 4.191425085067749 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3548604204389785, + "lr": 0.03, + "negative_goodness": 1.9565950325436063, + "pair_accuracy": 0.6549777777459886, + "positive_goodness": 2.0633816044277613, + "runtime_seconds": 4.1922101974487305 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.3576543834686279, + "lr": 0.03, + "negative_goodness": 1.9582678382025824, + "pair_accuracy": 0.6433777777671814, + "positive_goodness": 2.057178598573473, + "runtime_seconds": 4.190258026123047 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3543724989785089, + "lr": 0.03, + "negative_goodness": 1.9560766280492148, + "pair_accuracy": 0.6526222222063276, + "positive_goodness": 2.0639543571048313, + "runtime_seconds": 4.196247577667236 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.3541788685480753, + "lr": 0.03, + "negative_goodness": 1.9558083656522962, + "pair_accuracy": 0.6543333333015442, + "positive_goodness": 2.064950257788764, + "runtime_seconds": 4.195421934127808 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3527351928710938, + "lr": 0.03, + "negative_goodness": 1.9519619402355617, + "pair_accuracy": 0.6550888889153799, + "positive_goodness": 2.0670503174040054, + "runtime_seconds": 4.192782640457153 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.351364672279358, + "lr": 0.03, + "negative_goodness": 1.9520452878528172, + "pair_accuracy": 0.6645555556085374, + "positive_goodness": 2.0699025149027506, + "runtime_seconds": 4.191969633102417 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3504573486116198, + "lr": 0.03, + "negative_goodness": 1.9508601203494602, + "pair_accuracy": 0.6547555555343628, + "positive_goodness": 2.073028755739, + "runtime_seconds": 4.193140506744385 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3510946736865574, + "lr": 0.03, + "negative_goodness": 1.9499657888412476, + "pair_accuracy": 0.6579555555661519, + "positive_goodness": 2.0703362597147623, + "runtime_seconds": 4.193693399429321 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3515895300759209, + "lr": 0.03, + "negative_goodness": 1.951262267473009, + "pair_accuracy": 0.6517333333757188, + "positive_goodness": 2.0698351888868545, + "runtime_seconds": 4.201700925827026 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.348997885047065, + "lr": 0.03, + "negative_goodness": 1.9480472557915582, + "pair_accuracy": 0.6626666666454739, + "positive_goodness": 2.074948938645257, + "runtime_seconds": 4.194263696670532 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3493221042209202, + "lr": 0.03, + "negative_goodness": 1.9477416581683689, + "pair_accuracy": 0.6602888889312745, + "positive_goodness": 2.074237415568034, + "runtime_seconds": 4.191493511199951 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3491587013244628, + "lr": 0.03, + "negative_goodness": 1.9469362277560764, + "pair_accuracy": 0.6563555555449592, + "positive_goodness": 2.0752004085116917, + "runtime_seconds": 4.192288875579834 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.3484515547010634, + "lr": 0.03, + "negative_goodness": 1.9464157408396403, + "pair_accuracy": 0.6596444444656372, + "positive_goodness": 2.0758873582204185, + "runtime_seconds": 4.195840120315552 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3472602286868625, + "lr": 0.03, + "negative_goodness": 1.943950325287713, + "pair_accuracy": 0.6643555555343628, + "positive_goodness": 2.079246378941006, + "runtime_seconds": 4.19755220413208 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3445955679363675, + "lr": 0.03, + "negative_goodness": 1.9409946622636582, + "pair_accuracy": 0.6659999999894036, + "positive_goodness": 2.082460123782688, + "runtime_seconds": 4.191541910171509 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.347445178604126, + "lr": 0.03, + "negative_goodness": 1.944903024673462, + "pair_accuracy": 0.6602888888782925, + "positive_goodness": 2.078291101116604, + "runtime_seconds": 4.196112394332886 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3451454876581828, + "lr": 0.03, + "negative_goodness": 1.9426024759928386, + "pair_accuracy": 0.6640444444550408, + "positive_goodness": 2.0826487350463867, + "runtime_seconds": 4.196904182434082 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3476428396013047, + "lr": 0.03, + "negative_goodness": 1.9441010345458984, + "pair_accuracy": 0.6514444444656372, + "positive_goodness": 2.0779040483898585, + "runtime_seconds": 4.1969990730285645 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3460337111155192, + "lr": 0.03, + "negative_goodness": 1.9414667540656196, + "pair_accuracy": 0.6618666666878594, + "positive_goodness": 2.0802998406304254, + "runtime_seconds": 4.2013139724731445 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3455608212788899, + "lr": 0.03, + "negative_goodness": 1.9409561617745295, + "pair_accuracy": 0.6598888889100817, + "positive_goodness": 2.082043794208103, + "runtime_seconds": 4.191962003707886 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3452154147254096, + "lr": 0.03, + "negative_goodness": 1.9412786587609185, + "pair_accuracy": 0.6612000000423855, + "positive_goodness": 2.083108978568183, + "runtime_seconds": 4.188827753067017 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3440360547595553, + "lr": 0.03, + "negative_goodness": 1.9406251360575357, + "pair_accuracy": 0.6558444444550409, + "positive_goodness": 2.0848511718326144, + "runtime_seconds": 4.184606313705444 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3450186493767633, + "lr": 0.03, + "negative_goodness": 1.9395471615261501, + "pair_accuracy": 0.6597999999682108, + "positive_goodness": 2.083051565000746, + "runtime_seconds": 4.190711975097656 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3430828803804187, + "lr": 0.03, + "negative_goodness": 1.9397762009938557, + "pair_accuracy": 0.6647333333015442, + "positive_goodness": 2.086174620183309, + "runtime_seconds": 4.194082498550415 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3443509083429972, + "lr": 0.03, + "negative_goodness": 1.939459768401252, + "pair_accuracy": 0.6593333333651225, + "positive_goodness": 2.0850511862013077, + "runtime_seconds": 4.1875410079956055 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.343963647058275, + "lr": 0.03, + "negative_goodness": 1.938435262383355, + "pair_accuracy": 0.6605555555873447, + "positive_goodness": 2.0843620100657145, + "runtime_seconds": 4.1936259269714355 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.3425550727208455, + "lr": 0.03, + "negative_goodness": 1.9362288673612806, + "pair_accuracy": 0.6664444444762336, + "positive_goodness": 2.0886418140835232, + "runtime_seconds": 4.192058324813843 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3442666721132066, + "lr": 0.03, + "negative_goodness": 1.938911836412218, + "pair_accuracy": 0.6634222221904331, + "positive_goodness": 2.085127221340603, + "runtime_seconds": 4.2011988162994385 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.3421695910135905, + "lr": 0.03, + "negative_goodness": 1.938083782280816, + "pair_accuracy": 0.6660222222540114, + "positive_goodness": 2.0885892077551946, + "runtime_seconds": 4.189666986465454 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3424739128960503, + "lr": 0.03, + "negative_goodness": 1.9359384698232014, + "pair_accuracy": 0.6598222222646077, + "positive_goodness": 2.0879136048210993, + "runtime_seconds": 4.190378904342651 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3415609036551581, + "lr": 0.03, + "negative_goodness": 1.9366985896852282, + "pair_accuracy": 0.6621999999788072, + "positive_goodness": 2.0899602895524767, + "runtime_seconds": 4.189827919006348 + } + ], + "layer": 2 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.363385203827752, + "lr": 0.03, + "negative_goodness": 1.931928123558892, + "pair_accuracy": 0.6857333333863153, + "positive_goodness": 2.032686831580268, + "runtime_seconds": 4.109395265579224 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3369102628072103, + "lr": 0.03, + "negative_goodness": 1.94321404616038, + "pair_accuracy": 0.7363333333439297, + "positive_goodness": 2.0945876924726696, + "runtime_seconds": 4.1077492237091064 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3268187275568644, + "lr": 0.03, + "negative_goodness": 1.9339646657731797, + "pair_accuracy": 0.7559333333757189, + "positive_goodness": 2.1149529910193547, + "runtime_seconds": 4.110430717468262 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3245545294019911, + "lr": 0.03, + "negative_goodness": 1.9301436735788982, + "pair_accuracy": 0.7551111111217075, + "positive_goodness": 2.1197704235924615, + "runtime_seconds": 4.1098246574401855 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3178527632819281, + "lr": 0.03, + "negative_goodness": 1.9235243468814427, + "pair_accuracy": 0.7648888888888888, + "positive_goodness": 2.1320185907151963, + "runtime_seconds": 4.111273288726807 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.315113187874688, + "lr": 0.03, + "negative_goodness": 1.9183410339567397, + "pair_accuracy": 0.7679333333651225, + "positive_goodness": 2.1381145515441893, + "runtime_seconds": 4.114124059677124 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3114647230148315, + "lr": 0.03, + "negative_goodness": 1.9148493578169081, + "pair_accuracy": 0.7791555555237664, + "positive_goodness": 2.146503318532308, + "runtime_seconds": 4.1095290184021 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3104503438313801, + "lr": 0.03, + "negative_goodness": 1.9163819240993925, + "pair_accuracy": 0.7753777777459886, + "positive_goodness": 2.148409844631619, + "runtime_seconds": 4.1121180057525635 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3083513871298895, + "lr": 0.03, + "negative_goodness": 1.9117465037239922, + "pair_accuracy": 0.7786000000211928, + "positive_goodness": 2.153146329667833, + "runtime_seconds": 4.110498666763306 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3067161347495184, + "lr": 0.03, + "negative_goodness": 1.9115743683073256, + "pair_accuracy": 0.7828444444550409, + "positive_goodness": 2.1538376466963025, + "runtime_seconds": 4.123650550842285 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.305268999332852, + "lr": 0.03, + "negative_goodness": 1.9080196282280817, + "pair_accuracy": 0.7817333333333333, + "positive_goodness": 2.1598868948618573, + "runtime_seconds": 4.118238925933838 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3026913959291246, + "lr": 0.03, + "negative_goodness": 1.9037724354426067, + "pair_accuracy": 0.7814000000105964, + "positive_goodness": 2.164972158135308, + "runtime_seconds": 4.1203014850616455 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.2983160450617472, + "lr": 0.03, + "negative_goodness": 1.903342233742608, + "pair_accuracy": 0.789422222243415, + "positive_goodness": 2.1723839891645644, + "runtime_seconds": 4.110793828964233 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3024447118123372, + "lr": 0.03, + "negative_goodness": 1.9038895222345988, + "pair_accuracy": 0.785155555597941, + "positive_goodness": 2.1654021936204697, + "runtime_seconds": 4.112881660461426 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.300776331753201, + "lr": 0.03, + "negative_goodness": 1.904529643037584, + "pair_accuracy": 0.7814444444868299, + "positive_goodness": 2.1692047299702963, + "runtime_seconds": 4.1149022579193115 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.295808846007453, + "lr": 0.03, + "negative_goodness": 1.8953657155142891, + "pair_accuracy": 0.7936222222328186, + "positive_goodness": 2.1770795996348062, + "runtime_seconds": 4.110299825668335 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.2970855795330472, + "lr": 0.03, + "negative_goodness": 1.8987712918599446, + "pair_accuracy": 0.787111111164093, + "positive_goodness": 2.177000882636176, + "runtime_seconds": 4.1118690967559814 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.2918610087924534, + "lr": 0.03, + "negative_goodness": 1.8922594525231256, + "pair_accuracy": 0.7957333333015442, + "positive_goodness": 2.1850923847198485, + "runtime_seconds": 4.112123966217041 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.2919201947742038, + "lr": 0.03, + "negative_goodness": 1.8956940507888793, + "pair_accuracy": 0.7908000000211928, + "positive_goodness": 2.187077578353882, + "runtime_seconds": 4.1048362255096436 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.2930115173975627, + "lr": 0.03, + "negative_goodness": 1.8957206906636557, + "pair_accuracy": 0.7942666666772631, + "positive_goodness": 2.182431216812134, + "runtime_seconds": 4.117568731307983 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.2937716511832342, + "lr": 0.03, + "negative_goodness": 1.8946809824201796, + "pair_accuracy": 0.7905333333015442, + "positive_goodness": 2.185149971050686, + "runtime_seconds": 4.114433526992798 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.2919936807844374, + "lr": 0.03, + "negative_goodness": 1.8919096178266737, + "pair_accuracy": 0.7935555555449592, + "positive_goodness": 2.1863580347061156, + "runtime_seconds": 4.11887264251709 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.2913187218560114, + "lr": 0.03, + "negative_goodness": 1.8922362888124253, + "pair_accuracy": 0.7917555555237664, + "positive_goodness": 2.1884951011233857, + "runtime_seconds": 4.110252380371094 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.286971730486552, + "lr": 0.03, + "negative_goodness": 1.8853241403579712, + "pair_accuracy": 0.7964888888570998, + "positive_goodness": 2.195241275490655, + "runtime_seconds": 4.118106126785278 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.283953804037306, + "lr": 0.03, + "negative_goodness": 1.88493439678616, + "pair_accuracy": 0.8019999999788072, + "positive_goodness": 2.203698545413547, + "runtime_seconds": 4.117110252380371 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.2850521908442178, + "lr": 0.03, + "negative_goodness": 1.8889475898106893, + "pair_accuracy": 0.8008000000317892, + "positive_goodness": 2.2001524258083767, + "runtime_seconds": 4.116452217102051 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.2863224657058716, + "lr": 0.03, + "negative_goodness": 1.8848744878556993, + "pair_accuracy": 0.8008, + "positive_goodness": 2.198052465947469, + "runtime_seconds": 4.109140396118164 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.2888672659132216, + "lr": 0.03, + "negative_goodness": 1.8857604767693414, + "pair_accuracy": 0.7932888888782925, + "positive_goodness": 2.195436905585395, + "runtime_seconds": 4.109589099884033 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.287426065105862, + "lr": 0.03, + "negative_goodness": 1.8860534832848443, + "pair_accuracy": 0.796244444433848, + "positive_goodness": 2.1965638163672554, + "runtime_seconds": 4.114764928817749 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.2829054540634155, + "lr": 0.03, + "negative_goodness": 1.8822332061343723, + "pair_accuracy": 0.8018444444656372, + "positive_goodness": 2.2035463251749676, + "runtime_seconds": 4.117783546447754 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.2779868308173286, + "lr": 0.03, + "negative_goodness": 1.8805023892508612, + "pair_accuracy": 0.8104666666772631, + "positive_goodness": 2.2148855459425185, + "runtime_seconds": 4.109351873397827 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.2816419819513958, + "lr": 0.03, + "negative_goodness": 1.8800837342368233, + "pair_accuracy": 0.8048888888570997, + "positive_goodness": 2.2094416664123537, + "runtime_seconds": 4.113013744354248 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.2849041304270425, + "lr": 0.03, + "negative_goodness": 1.883195079506768, + "pair_accuracy": 0.8004222222752041, + "positive_goodness": 2.204313894059923, + "runtime_seconds": 4.10442852973938 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.2862440074920654, + "lr": 0.03, + "negative_goodness": 1.8819093544218275, + "pair_accuracy": 0.795333333354526, + "positive_goodness": 2.1977786841922335, + "runtime_seconds": 4.106237888336182 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.2795901706271702, + "lr": 0.03, + "negative_goodness": 1.878754656346639, + "pair_accuracy": 0.800511111164093, + "positive_goodness": 2.2136409760369196, + "runtime_seconds": 4.115696668624878 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.273770546468099, + "lr": 0.03, + "negative_goodness": 1.8739364440706041, + "pair_accuracy": 0.8139999999788072, + "positive_goodness": 2.2241989495595296, + "runtime_seconds": 4.120581150054932 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.2855378384484184, + "lr": 0.03, + "negative_goodness": 1.8780484038035075, + "pair_accuracy": 0.7950444444762336, + "positive_goodness": 2.200965281889174, + "runtime_seconds": 4.113895654678345 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.2787170732710096, + "lr": 0.03, + "negative_goodness": 1.8770143736097549, + "pair_accuracy": 0.8050222221904331, + "positive_goodness": 2.2143660593244765, + "runtime_seconds": 4.105283737182617 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.2779149461534287, + "lr": 0.03, + "negative_goodness": 1.8770735144085355, + "pair_accuracy": 0.8076444444232517, + "positive_goodness": 2.2181284451378716, + "runtime_seconds": 4.112576484680176 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.2746874359766642, + "lr": 0.03, + "negative_goodness": 1.874471704588996, + "pair_accuracy": 0.8097111110793219, + "positive_goodness": 2.2218546881781682, + "runtime_seconds": 4.11297082901001 + } + ], + "layer": 3 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.3157120634290906, + "lr": 0.03, + "negative_goodness": 1.9484853195614285, + "pair_accuracy": 0.7645777778201633, + "positive_goodness": 2.159882808007134, + "runtime_seconds": 4.204307317733765 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.288856547249688, + "lr": 0.03, + "negative_goodness": 1.922778161324395, + "pair_accuracy": 0.8473777778201633, + "positive_goodness": 2.2029851684570314, + "runtime_seconds": 4.196081161499023 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.2831481364780002, + "lr": 0.03, + "negative_goodness": 1.9161016929626464, + "pair_accuracy": 0.8459999999894037, + "positive_goodness": 2.2125621995713978, + "runtime_seconds": 4.201501846313477 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.2773577492396038, + "lr": 0.03, + "negative_goodness": 1.909256796052721, + "pair_accuracy": 0.8535333333439297, + "positive_goodness": 2.2219490837944877, + "runtime_seconds": 4.1992692947387695 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.281189894824558, + "lr": 0.03, + "negative_goodness": 1.9089860293918186, + "pair_accuracy": 0.8432444444974263, + "positive_goodness": 2.2147648111131457, + "runtime_seconds": 4.1973044872283936 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.267738566992018, + "lr": 0.03, + "negative_goodness": 1.9021486432393393, + "pair_accuracy": 0.862711111164093, + "positive_goodness": 2.2413983477274577, + "runtime_seconds": 4.20433235168457 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.2716917334874471, + "lr": 0.03, + "negative_goodness": 1.9015547044330172, + "pair_accuracy": 0.8528666667196486, + "positive_goodness": 2.232240942086114, + "runtime_seconds": 4.195948362350464 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.268565087721083, + "lr": 0.03, + "negative_goodness": 1.897976900906033, + "pair_accuracy": 0.8529777778307597, + "positive_goodness": 2.2403096967485214, + "runtime_seconds": 4.196468114852905 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.2722486602995131, + "lr": 0.03, + "negative_goodness": 1.8991807861116197, + "pair_accuracy": 0.8445555555661519, + "positive_goodness": 2.2317002367655436, + "runtime_seconds": 4.189431428909302 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.2715027571148343, + "lr": 0.03, + "negative_goodness": 1.8981195918824938, + "pair_accuracy": 0.8423777778307597, + "positive_goodness": 2.2346959274715847, + "runtime_seconds": 4.200538635253906 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.2652780989117092, + "lr": 0.03, + "negative_goodness": 1.8943852587382, + "pair_accuracy": 0.8531111111111112, + "positive_goodness": 2.247254833645291, + "runtime_seconds": 4.192700386047363 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.2609423194885254, + "lr": 0.03, + "negative_goodness": 1.8917788743125068, + "pair_accuracy": 0.8580888888888889, + "positive_goodness": 2.255651056628757, + "runtime_seconds": 4.193871736526489 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.2643611871931288, + "lr": 0.03, + "negative_goodness": 1.8916995449066163, + "pair_accuracy": 0.8551555555979411, + "positive_goodness": 2.2481832693311903, + "runtime_seconds": 4.203734874725342 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.2568806176291571, + "lr": 0.03, + "negative_goodness": 1.8886215185589261, + "pair_accuracy": 0.8588000000423855, + "positive_goodness": 2.263132833353678, + "runtime_seconds": 4.208460092544556 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.2610209760877822, + "lr": 0.03, + "negative_goodness": 1.8899302138646443, + "pair_accuracy": 0.8533111111005147, + "positive_goodness": 2.255463768386841, + "runtime_seconds": 4.208158493041992 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.2577741435580783, + "lr": 0.03, + "negative_goodness": 1.8883643886990018, + "pair_accuracy": 0.8593333333439297, + "positive_goodness": 2.2636377524058022, + "runtime_seconds": 4.199467182159424 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.2629735854466757, + "lr": 0.03, + "negative_goodness": 1.8870703446494208, + "pair_accuracy": 0.8501333333757188, + "positive_goodness": 2.2503213762495253, + "runtime_seconds": 4.200150012969971 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.2616565897623697, + "lr": 0.03, + "negative_goodness": 1.885458588112725, + "pair_accuracy": 0.8536222222328186, + "positive_goodness": 2.256426188108656, + "runtime_seconds": 4.1971845626831055 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.258948041703966, + "lr": 0.03, + "negative_goodness": 1.8856301119062635, + "pair_accuracy": 0.8553111111005147, + "positive_goodness": 2.2627197759840225, + "runtime_seconds": 4.201722621917725 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.2626366814931234, + "lr": 0.03, + "negative_goodness": 1.8864861684375338, + "pair_accuracy": 0.8458222222222223, + "positive_goodness": 2.251456928973728, + "runtime_seconds": 4.199927091598511 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.258236756600274, + "lr": 0.03, + "negative_goodness": 1.885933362197876, + "pair_accuracy": 0.8567555555661519, + "positive_goodness": 2.2624497179667156, + "runtime_seconds": 4.201143026351929 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.2537546156141492, + "lr": 0.03, + "negative_goodness": 1.8826512704849243, + "pair_accuracy": 0.8634222222222222, + "positive_goodness": 2.2695890622033015, + "runtime_seconds": 4.197016477584839 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.2545791104210748, + "lr": 0.03, + "negative_goodness": 1.8804515733718872, + "pair_accuracy": 0.8510888889418708, + "positive_goodness": 2.268062123362223, + "runtime_seconds": 4.2015540599823 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.2598558737860786, + "lr": 0.03, + "negative_goodness": 1.883889200952318, + "pair_accuracy": 0.8443555555661519, + "positive_goodness": 2.2604875445048016, + "runtime_seconds": 4.199657440185547 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.251000226465861, + "lr": 0.03, + "negative_goodness": 1.8796693729612561, + "pair_accuracy": 0.8598666666772631, + "positive_goodness": 2.277060289764404, + "runtime_seconds": 4.198919057846069 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.2538767469406127, + "lr": 0.03, + "negative_goodness": 1.8825780462900799, + "pair_accuracy": 0.8567555555767483, + "positive_goodness": 2.2724789172702367, + "runtime_seconds": 4.195985555648804 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.255681205982632, + "lr": 0.03, + "negative_goodness": 1.8782661268234253, + "pair_accuracy": 0.849644444433848, + "positive_goodness": 2.2660218763139515, + "runtime_seconds": 4.200608491897583 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.2569354396608141, + "lr": 0.03, + "negative_goodness": 1.881655883471171, + "pair_accuracy": 0.8555111110899184, + "positive_goodness": 2.265819185087416, + "runtime_seconds": 4.206806659698486 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.2572458042144776, + "lr": 0.03, + "negative_goodness": 1.8813825391557482, + "pair_accuracy": 0.8477777778201633, + "positive_goodness": 2.2641023755815293, + "runtime_seconds": 4.229756116867065 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.2528791308297051, + "lr": 0.03, + "negative_goodness": 1.8796792140748766, + "pair_accuracy": 0.855822222243415, + "positive_goodness": 2.2716078596750897, + "runtime_seconds": 4.198012590408325 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.2482066959169176, + "lr": 0.03, + "negative_goodness": 1.8749223246892293, + "pair_accuracy": 0.8649555555767483, + "positive_goodness": 2.283548356713189, + "runtime_seconds": 4.191958904266357 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.255266463194953, + "lr": 0.03, + "negative_goodness": 1.8821477475060358, + "pair_accuracy": 0.8520444444338481, + "positive_goodness": 2.2713998101764257, + "runtime_seconds": 4.196762800216675 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.2492809716542561, + "lr": 0.03, + "negative_goodness": 1.876965950287713, + "pair_accuracy": 0.8567111110899184, + "positive_goodness": 2.2810123310512966, + "runtime_seconds": 4.200186729431152 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.2522084690729778, + "lr": 0.03, + "negative_goodness": 1.8745041070302328, + "pair_accuracy": 0.8550666666560703, + "positive_goodness": 2.2744709118737116, + "runtime_seconds": 4.196854114532471 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.2531635791142781, + "lr": 0.03, + "negative_goodness": 1.8761790663825142, + "pair_accuracy": 0.8522444444974263, + "positive_goodness": 2.2732953005048966, + "runtime_seconds": 4.196591854095459 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.252931639374627, + "lr": 0.03, + "negative_goodness": 1.879946422068278, + "pair_accuracy": 0.8550666667090522, + "positive_goodness": 2.2732917787763807, + "runtime_seconds": 4.193106174468994 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.252428323788113, + "lr": 0.03, + "negative_goodness": 1.8761344367133246, + "pair_accuracy": 0.8489555555343627, + "positive_goodness": 2.2745753856658935, + "runtime_seconds": 4.191633701324463 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.2447429954952665, + "lr": 0.03, + "negative_goodness": 1.8715160071902806, + "pair_accuracy": 0.8606888888570997, + "positive_goodness": 2.2904863514794243, + "runtime_seconds": 4.1963582038879395 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.247154250441657, + "lr": 0.03, + "negative_goodness": 1.8738527304331463, + "pair_accuracy": 0.859177777756585, + "positive_goodness": 2.285267403793335, + "runtime_seconds": 4.185273885726929 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.254693643315633, + "lr": 0.03, + "negative_goodness": 1.878792155011495, + "pair_accuracy": 0.8467999999894036, + "positive_goodness": 2.2718647586186727, + "runtime_seconds": 4.191496133804321 + } + ], + "layer": 4 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2261495694054498, + "lr": 0.03, + "negative_goodness": 1.846273912217882, + "pair_accuracy": 0.8636666667196485, + "positive_goodness": 2.3070808848487006, + "runtime_seconds": 4.106675386428833 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.1712898673481411, + "lr": 0.03, + "negative_goodness": 1.8397306237326727, + "pair_accuracy": 0.8874222222116258, + "positive_goodness": 2.438155224609375, + "runtime_seconds": 4.117546796798706 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.1548335404925876, + "lr": 0.03, + "negative_goodness": 1.824768670739068, + "pair_accuracy": 0.8901333333757189, + "positive_goodness": 2.476646268251207, + "runtime_seconds": 4.115573167800903 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.150726117346022, + "lr": 0.03, + "negative_goodness": 1.8177449900097318, + "pair_accuracy": 0.8958000000317892, + "positive_goodness": 2.4877128871070013, + "runtime_seconds": 4.110372066497803 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.13550559586419, + "lr": 0.03, + "negative_goodness": 1.8021509422302247, + "pair_accuracy": 0.9006, + "positive_goodness": 2.5281707704755996, + "runtime_seconds": 4.108009576797485 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.1413805837101407, + "lr": 0.03, + "negative_goodness": 1.8074235565609402, + "pair_accuracy": 0.8972222222222223, + "positive_goodness": 2.5102762009090847, + "runtime_seconds": 4.108757257461548 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.1273996325598823, + "lr": 0.03, + "negative_goodness": 1.7942975905524359, + "pair_accuracy": 0.8991555555237664, + "positive_goodness": 2.548309704123603, + "runtime_seconds": 4.10643196105957 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.1150318355242412, + "lr": 0.03, + "negative_goodness": 1.7895531732135348, + "pair_accuracy": 0.9064222222116258, + "positive_goodness": 2.575343133629693, + "runtime_seconds": 4.1074371337890625 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.1278533650928073, + "lr": 0.03, + "negative_goodness": 1.7907552389568753, + "pair_accuracy": 0.8966000000317892, + "positive_goodness": 2.546345408121745, + "runtime_seconds": 4.112310171127319 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.119874164422353, + "lr": 0.03, + "negative_goodness": 1.7889745242860582, + "pair_accuracy": 0.9062444444444444, + "positive_goodness": 2.5609932209014894, + "runtime_seconds": 4.1086225509643555 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.1267456806288825, + "lr": 0.03, + "negative_goodness": 1.787269320869446, + "pair_accuracy": 0.890511111164093, + "positive_goodness": 2.551787612279256, + "runtime_seconds": 4.104816675186157 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.1227829819573296, + "lr": 0.03, + "negative_goodness": 1.7890382353040908, + "pair_accuracy": 0.8938666666878594, + "positive_goodness": 2.557402565765381, + "runtime_seconds": 4.106895923614502 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.1133449703322515, + "lr": 0.03, + "negative_goodness": 1.7809584218555026, + "pair_accuracy": 0.9034222222222222, + "positive_goodness": 2.5775802551693388, + "runtime_seconds": 4.117905139923096 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.1116062195248073, + "lr": 0.03, + "negative_goodness": 1.776624480480618, + "pair_accuracy": 0.9046000000105964, + "positive_goodness": 2.5840108837975397, + "runtime_seconds": 4.107118368148804 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.104581275452508, + "lr": 0.03, + "negative_goodness": 1.7667369323306614, + "pair_accuracy": 0.9027333333121406, + "positive_goodness": 2.604363182406955, + "runtime_seconds": 4.101730585098267 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.0890382770114475, + "lr": 0.03, + "negative_goodness": 1.7625344626108805, + "pair_accuracy": 0.9158666666507721, + "positive_goodness": 2.6373711786058216, + "runtime_seconds": 4.107868194580078 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.0999188919067382, + "lr": 0.03, + "negative_goodness": 1.7646269097646077, + "pair_accuracy": 0.9048222222275204, + "positive_goodness": 2.608219020292494, + "runtime_seconds": 4.10097336769104 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.1012301017973158, + "lr": 0.03, + "negative_goodness": 1.7666165583292643, + "pair_accuracy": 0.9074444444444445, + "positive_goodness": 2.6064119266933865, + "runtime_seconds": 4.104971170425415 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.1002144324620564, + "lr": 0.03, + "negative_goodness": 1.7622497907426622, + "pair_accuracy": 0.9065111111111112, + "positive_goodness": 2.6119467107984753, + "runtime_seconds": 4.113715171813965 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.0859359534687467, + "lr": 0.03, + "negative_goodness": 1.7537220824771458, + "pair_accuracy": 0.9099111111005147, + "positive_goodness": 2.6461770452287463, + "runtime_seconds": 4.10743260383606 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.0961598378075494, + "lr": 0.03, + "negative_goodness": 1.7621391343010797, + "pair_accuracy": 0.9046222221904331, + "positive_goodness": 2.623654704793294, + "runtime_seconds": 4.109234809875488 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.1111296663496228, + "lr": 0.03, + "negative_goodness": 1.7718900956471761, + "pair_accuracy": 0.8995777777777778, + "positive_goodness": 2.582934309556749, + "runtime_seconds": 4.10819673538208 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.1079474945492214, + "lr": 0.03, + "negative_goodness": 1.7705527676264445, + "pair_accuracy": 0.9076444444126553, + "positive_goodness": 2.5943449639638265, + "runtime_seconds": 4.102893114089966 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.1167044221242268, + "lr": 0.03, + "negative_goodness": 1.773430486339993, + "pair_accuracy": 0.894977777756585, + "positive_goodness": 2.577734714041816, + "runtime_seconds": 4.106491565704346 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.09606968460083, + "lr": 0.03, + "negative_goodness": 1.7651921961890327, + "pair_accuracy": 0.9110000000529819, + "positive_goodness": 2.622268991088867, + "runtime_seconds": 4.111208915710449 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.1023774824778239, + "lr": 0.03, + "negative_goodness": 1.7662188081317478, + "pair_accuracy": 0.9075333333015442, + "positive_goodness": 2.602567477501763, + "runtime_seconds": 4.108267545700073 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.1002922629462348, + "lr": 0.03, + "negative_goodness": 1.7583807843738133, + "pair_accuracy": 0.8983555555449592, + "positive_goodness": 2.6132256741841635, + "runtime_seconds": 4.110975742340088 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.0839439553578694, + "lr": 0.03, + "negative_goodness": 1.7537993270874024, + "pair_accuracy": 0.9174666666666667, + "positive_goodness": 2.649188068177965, + "runtime_seconds": 4.1149396896362305 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.10768420914544, + "lr": 0.03, + "negative_goodness": 1.762215388361613, + "pair_accuracy": 0.8966222222646077, + "positive_goodness": 2.597895335430569, + "runtime_seconds": 4.108246326446533 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.106371989271376, + "lr": 0.03, + "negative_goodness": 1.7670591165966458, + "pair_accuracy": 0.8955111110899183, + "positive_goodness": 2.600708002344767, + "runtime_seconds": 4.100904941558838 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.0930078121609157, + "lr": 0.03, + "negative_goodness": 1.755316619152493, + "pair_accuracy": 0.9030888888782925, + "positive_goodness": 2.627472414991591, + "runtime_seconds": 4.101433038711548 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.0807727708816528, + "lr": 0.03, + "negative_goodness": 1.747190545739068, + "pair_accuracy": 0.9093333333121406, + "positive_goodness": 2.657519578721788, + "runtime_seconds": 4.1081531047821045 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.086436436780294, + "lr": 0.03, + "negative_goodness": 1.749000685988532, + "pair_accuracy": 0.9083999999682109, + "positive_goodness": 2.6419464111751982, + "runtime_seconds": 4.118070602416992 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.0928397960450915, + "lr": 0.03, + "negative_goodness": 1.7556655973646376, + "pair_accuracy": 0.903822222243415, + "positive_goodness": 2.629086883884006, + "runtime_seconds": 4.112451076507568 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.0838440064960055, + "lr": 0.03, + "negative_goodness": 1.747723941675822, + "pair_accuracy": 0.9109999999788072, + "positive_goodness": 2.652326032723321, + "runtime_seconds": 4.115649700164795 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.0802116849687364, + "lr": 0.03, + "negative_goodness": 1.7458540690316093, + "pair_accuracy": 0.9100222222328186, + "positive_goodness": 2.6584667794969348, + "runtime_seconds": 4.117010593414307 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.0919653782950507, + "lr": 0.03, + "negative_goodness": 1.7514284866756864, + "pair_accuracy": 0.9058888889418708, + "positive_goodness": 2.633768251885308, + "runtime_seconds": 4.11143159866333 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.080985579999288, + "lr": 0.03, + "negative_goodness": 1.742019308895535, + "pair_accuracy": 0.9063777777777777, + "positive_goodness": 2.65904129647149, + "runtime_seconds": 4.102772951126099 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.1102697087817721, + "lr": 0.03, + "negative_goodness": 1.7628061603122287, + "pair_accuracy": 0.8910444444762335, + "positive_goodness": 2.5946157149420843, + "runtime_seconds": 4.106356143951416 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.0745362322701348, + "lr": 0.03, + "negative_goodness": 1.739738830312093, + "pair_accuracy": 0.9123777777459886, + "positive_goodness": 2.6703601700676813, + "runtime_seconds": 4.105858564376831 + } + ], + "layer": 5 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2282417302873399, + "lr": 0.03, + "negative_goodness": 1.9139733407550388, + "pair_accuracy": 0.838422222190433, + "positive_goodness": 2.3548981219821505, + "runtime_seconds": 4.183380365371704 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.1528942929797703, + "lr": 0.03, + "negative_goodness": 1.8545381876627605, + "pair_accuracy": 0.9038666666666667, + "positive_goodness": 2.489146173010932, + "runtime_seconds": 4.186413049697876 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.1441439374923705, + "lr": 0.03, + "negative_goodness": 1.8414307845433553, + "pair_accuracy": 0.9059111111640931, + "positive_goodness": 2.51084839714898, + "runtime_seconds": 4.182114362716675 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.1254109309090508, + "lr": 0.03, + "negative_goodness": 1.8262093807644315, + "pair_accuracy": 0.9185777777777778, + "positive_goodness": 2.5477708988613554, + "runtime_seconds": 4.181799411773682 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.1301531390508015, + "lr": 0.03, + "negative_goodness": 1.827271707810296, + "pair_accuracy": 0.9052888889418708, + "positive_goodness": 2.546695994652642, + "runtime_seconds": 4.187775611877441 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.120451905632019, + "lr": 0.03, + "negative_goodness": 1.8158629444970025, + "pair_accuracy": 0.9123111111111111, + "positive_goodness": 2.5558012626224094, + "runtime_seconds": 4.189437389373779 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.1356065869861178, + "lr": 0.03, + "negative_goodness": 1.824257005458408, + "pair_accuracy": 0.9037777777883742, + "positive_goodness": 2.5264420747968885, + "runtime_seconds": 4.190229892730713 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.127683812099033, + "lr": 0.03, + "negative_goodness": 1.8262822903315226, + "pair_accuracy": 0.9100666667196485, + "positive_goodness": 2.5570729092491997, + "runtime_seconds": 4.185404300689697 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.1134229776382447, + "lr": 0.03, + "negative_goodness": 1.8045417392518786, + "pair_accuracy": 0.9136000000529819, + "positive_goodness": 2.5714300395117866, + "runtime_seconds": 4.188624858856201 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.115643950843811, + "lr": 0.03, + "negative_goodness": 1.808674778535631, + "pair_accuracy": 0.9150444444444444, + "positive_goodness": 2.5773485818227133, + "runtime_seconds": 4.18666934967041 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.1108136850145127, + "lr": 0.03, + "negative_goodness": 1.8010588664160834, + "pair_accuracy": 0.9103333333121406, + "positive_goodness": 2.5858695802900526, + "runtime_seconds": 4.193897724151611 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.10547662908766, + "lr": 0.03, + "negative_goodness": 1.8018836910247802, + "pair_accuracy": 0.9129555555979411, + "positive_goodness": 2.599337922244602, + "runtime_seconds": 4.186630725860596 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.101928511153327, + "lr": 0.03, + "negative_goodness": 1.7928600411732991, + "pair_accuracy": 0.9142888888888889, + "positive_goodness": 2.6065821189456515, + "runtime_seconds": 4.187521934509277 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.102749447886149, + "lr": 0.03, + "negative_goodness": 1.7906048328399657, + "pair_accuracy": 0.9124888888782925, + "positive_goodness": 2.605717048687405, + "runtime_seconds": 4.1851372718811035 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.1206658038669162, + "lr": 0.03, + "negative_goodness": 1.8006284732394748, + "pair_accuracy": 0.9025555555237664, + "positive_goodness": 2.5556880288441977, + "runtime_seconds": 4.18665623664856 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.1069506627188788, + "lr": 0.03, + "negative_goodness": 1.7928888461642796, + "pair_accuracy": 0.9088888888676961, + "positive_goodness": 2.5999124927944606, + "runtime_seconds": 4.185266733169556 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.0970086763275995, + "lr": 0.03, + "negative_goodness": 1.7887196825027465, + "pair_accuracy": 0.9157999999788072, + "positive_goodness": 2.614145469792684, + "runtime_seconds": 4.192104816436768 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.1046667453765868, + "lr": 0.03, + "negative_goodness": 1.7854950257195368, + "pair_accuracy": 0.9040888888782925, + "positive_goodness": 2.59957791756524, + "runtime_seconds": 4.182960033416748 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.0964878037876553, + "lr": 0.03, + "negative_goodness": 1.7866339332580568, + "pair_accuracy": 0.9166444444338481, + "positive_goodness": 2.617560800382826, + "runtime_seconds": 4.182856321334839 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.104281988588969, + "lr": 0.03, + "negative_goodness": 1.7853485932032267, + "pair_accuracy": 0.9070000000317892, + "positive_goodness": 2.6027050721910263, + "runtime_seconds": 4.18218994140625 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.093880547205607, + "lr": 0.03, + "negative_goodness": 1.7818320153978136, + "pair_accuracy": 0.9113555555449592, + "positive_goodness": 2.624588359790378, + "runtime_seconds": 4.188371658325195 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.1089344637976752, + "lr": 0.03, + "negative_goodness": 1.7902253051757813, + "pair_accuracy": 0.9037777778201633, + "positive_goodness": 2.5970583329518635, + "runtime_seconds": 4.194866180419922 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.0973363797505697, + "lr": 0.03, + "negative_goodness": 1.7793532372368706, + "pair_accuracy": 0.9114666667090522, + "positive_goodness": 2.6173748268551296, + "runtime_seconds": 4.182745695114136 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.083446769481235, + "lr": 0.03, + "negative_goodness": 1.7700866875330608, + "pair_accuracy": 0.9105777777883741, + "positive_goodness": 2.6520619473351372, + "runtime_seconds": 4.200013637542725 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.0895815067397223, + "lr": 0.03, + "negative_goodness": 1.78178900964525, + "pair_accuracy": 0.9148666666666667, + "positive_goodness": 2.633288595453898, + "runtime_seconds": 4.198870420455933 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.0899646204630533, + "lr": 0.03, + "negative_goodness": 1.7726043417400783, + "pair_accuracy": 0.908755555566152, + "positive_goodness": 2.6313519655015734, + "runtime_seconds": 4.189499616622925 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.0940151066886055, + "lr": 0.03, + "negative_goodness": 1.7806745727327136, + "pair_accuracy": 0.9119555555767483, + "positive_goodness": 2.6332524874369305, + "runtime_seconds": 4.1830267906188965 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.0918055002212523, + "lr": 0.03, + "negative_goodness": 1.7750634720908272, + "pair_accuracy": 0.9102888888782925, + "positive_goodness": 2.626335217497084, + "runtime_seconds": 4.1856865882873535 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.0909241199705335, + "lr": 0.03, + "negative_goodness": 1.7725908629099527, + "pair_accuracy": 0.9091999999894036, + "positive_goodness": 2.641830966398451, + "runtime_seconds": 4.18153190612793 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.092774199485779, + "lr": 0.03, + "negative_goodness": 1.7789551886876425, + "pair_accuracy": 0.9198888888994853, + "positive_goodness": 2.631766346232096, + "runtime_seconds": 4.17955207824707 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.08910296459198, + "lr": 0.03, + "negative_goodness": 1.7688072917726305, + "pair_accuracy": 0.906911111079322, + "positive_goodness": 2.636232932366265, + "runtime_seconds": 4.182530403137207 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.0668831248283386, + "lr": 0.03, + "negative_goodness": 1.7581169848548042, + "pair_accuracy": 0.9193111111005148, + "positive_goodness": 2.68321595954895, + "runtime_seconds": 4.194822072982788 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.0772013047112359, + "lr": 0.03, + "negative_goodness": 1.7653497921413845, + "pair_accuracy": 0.917666666677263, + "positive_goodness": 2.670431421110365, + "runtime_seconds": 4.18279767036438 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.0893278295728896, + "lr": 0.03, + "negative_goodness": 1.7705945396847196, + "pair_accuracy": 0.9088666666454739, + "positive_goodness": 2.643431201722887, + "runtime_seconds": 4.187745094299316 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.0939264350043403, + "lr": 0.03, + "negative_goodness": 1.7715770559946695, + "pair_accuracy": 0.9066666666984559, + "positive_goodness": 2.624885014682346, + "runtime_seconds": 4.188612222671509 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.1082750990761652, + "lr": 0.03, + "negative_goodness": 1.7769605804231432, + "pair_accuracy": 0.8950222222328186, + "positive_goodness": 2.601079196972317, + "runtime_seconds": 4.187227725982666 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.0996608392503526, + "lr": 0.03, + "negative_goodness": 1.7816196267445883, + "pair_accuracy": 0.9034444444868299, + "positive_goodness": 2.620765762286716, + "runtime_seconds": 4.184179306030273 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.0955652079688178, + "lr": 0.03, + "negative_goodness": 1.7699977742089166, + "pair_accuracy": 0.9041555555555556, + "positive_goodness": 2.621982020441691, + "runtime_seconds": 4.185506343841553 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.0766044742584229, + "lr": 0.03, + "negative_goodness": 1.7584998134613037, + "pair_accuracy": 0.9079777777671814, + "positive_goodness": 2.663644162326389, + "runtime_seconds": 4.1832146644592285 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.0892708026674058, + "lr": 0.03, + "negative_goodness": 1.76780527282291, + "pair_accuracy": 0.9054444444868299, + "positive_goodness": 2.643021335644192, + "runtime_seconds": 4.185601472854614 + } + ], + "layer": 6 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1253311761008369, + "lr": 0.03, + "negative_goodness": 1.6575051356845432, + "pair_accuracy": 0.8460666666560703, + "positive_goodness": 2.557292859141032, + "runtime_seconds": 3.891817331314087 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.0763676758024427, + "lr": 0.03, + "negative_goodness": 1.6518669692569308, + "pair_accuracy": 0.8623555555555555, + "positive_goodness": 2.6672118925730386, + "runtime_seconds": 3.896362066268921 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.0728448446273804, + "lr": 0.03, + "negative_goodness": 1.6546233544773525, + "pair_accuracy": 0.8680888888570998, + "positive_goodness": 2.6859911384158663, + "runtime_seconds": 3.8920247554779053 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0365681575987074, + "lr": 0.03, + "negative_goodness": 1.6315087844000922, + "pair_accuracy": 0.8820222222381168, + "positive_goodness": 2.769213263744778, + "runtime_seconds": 3.8668699264526367 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.0445627415868972, + "lr": 0.03, + "negative_goodness": 1.629291656451755, + "pair_accuracy": 0.8768222222116259, + "positive_goodness": 2.7458765820821127, + "runtime_seconds": 3.879272699356079 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.0211790038638644, + "lr": 0.03, + "negative_goodness": 1.6018875737508138, + "pair_accuracy": 0.8818444444762336, + "positive_goodness": 2.8127167895846896, + "runtime_seconds": 3.890521287918091 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.029574119038052, + "lr": 0.03, + "negative_goodness": 1.6114481822543674, + "pair_accuracy": 0.8778000000211927, + "positive_goodness": 2.7895690158843993, + "runtime_seconds": 3.8903748989105225 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0264175943374634, + "lr": 0.03, + "negative_goodness": 1.6093906185574, + "pair_accuracy": 0.8832222222646078, + "positive_goodness": 2.7925160817040338, + "runtime_seconds": 3.890129327774048 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.006596037949456, + "lr": 0.03, + "negative_goodness": 1.5959457813898723, + "pair_accuracy": 0.8902666666454739, + "positive_goodness": 2.8440994382222495, + "runtime_seconds": 3.862400770187378 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.0075055130640667, + "lr": 0.03, + "negative_goodness": 1.5890174460517035, + "pair_accuracy": 0.8848888888994852, + "positive_goodness": 2.8418201305389403, + "runtime_seconds": 3.891815185546875 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.9913885815090603, + "lr": 0.03, + "negative_goodness": 1.5767886158625284, + "pair_accuracy": 0.8924666666984558, + "positive_goodness": 2.886002661683824, + "runtime_seconds": 3.8908305168151855 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.9671852130148146, + "lr": 0.03, + "negative_goodness": 1.5379837447802225, + "pair_accuracy": 0.892977777756585, + "positive_goodness": 2.9561121763865152, + "runtime_seconds": 3.90191912651062 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.9981403942320082, + "lr": 0.03, + "negative_goodness": 1.5739102121565076, + "pair_accuracy": 0.8828000000211927, + "positive_goodness": 2.879414331817627, + "runtime_seconds": 3.8799173831939697 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.9865688554763794, + "lr": 0.03, + "negative_goodness": 1.5627147153218588, + "pair_accuracy": 0.8913111111217075, + "positive_goodness": 2.898682731543647, + "runtime_seconds": 3.8865065574645996 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.9876086483637492, + "lr": 0.03, + "negative_goodness": 1.5663233857472738, + "pair_accuracy": 0.8917333333015441, + "positive_goodness": 2.900077297973633, + "runtime_seconds": 3.879426956176758 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.9722018902036879, + "lr": 0.03, + "negative_goodness": 1.5425182858361137, + "pair_accuracy": 0.8934888888994853, + "positive_goodness": 2.951553590053982, + "runtime_seconds": 3.8823554515838623 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.9493757187631395, + "lr": 0.03, + "negative_goodness": 1.519185211859809, + "pair_accuracy": 0.900533333322737, + "positive_goodness": 3.002834382587009, + "runtime_seconds": 3.8933327198028564 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.9704292489581637, + "lr": 0.03, + "negative_goodness": 1.5397263514412773, + "pair_accuracy": 0.8953777777777778, + "positive_goodness": 2.9483693684895833, + "runtime_seconds": 3.8986411094665527 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.9569258130603366, + "lr": 0.03, + "negative_goodness": 1.5266976169162327, + "pair_accuracy": 0.8955999999788072, + "positive_goodness": 2.9991216736687556, + "runtime_seconds": 3.8721351623535156 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.9736129059897529, + "lr": 0.03, + "negative_goodness": 1.5432440814759996, + "pair_accuracy": 0.8900666666666667, + "positive_goodness": 2.9432943698883056, + "runtime_seconds": 3.8891103267669678 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.9708328408771091, + "lr": 0.03, + "negative_goodness": 1.538618601820204, + "pair_accuracy": 0.889688888920678, + "positive_goodness": 2.962089084752401, + "runtime_seconds": 3.8976728916168213 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.9948084070735508, + "lr": 0.03, + "negative_goodness": 1.5637291308720906, + "pair_accuracy": 0.8846666666454739, + "positive_goodness": 2.9047905408647323, + "runtime_seconds": 3.9057857990264893 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.971773373858134, + "lr": 0.03, + "negative_goodness": 1.5433490192201402, + "pair_accuracy": 0.8976888889312744, + "positive_goodness": 2.956057413185967, + "runtime_seconds": 3.878641128540039 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.9662275307549371, + "lr": 0.03, + "negative_goodness": 1.5293324873182508, + "pair_accuracy": 0.8927333333015441, + "positive_goodness": 2.982958706198798, + "runtime_seconds": 3.892656087875366 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.9604807125939263, + "lr": 0.03, + "negative_goodness": 1.5226243591308595, + "pair_accuracy": 0.8936000000317892, + "positive_goodness": 2.984456341764662, + "runtime_seconds": 3.8817129135131836 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9586343354649014, + "lr": 0.03, + "negative_goodness": 1.5128548272662692, + "pair_accuracy": 0.8897999999682109, + "positive_goodness": 3.0081344315422904, + "runtime_seconds": 4.018235683441162 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.9640578417248196, + "lr": 0.03, + "negative_goodness": 1.5206767811245387, + "pair_accuracy": 0.8883333333651224, + "positive_goodness": 2.9983419269561766, + "runtime_seconds": 4.060157299041748 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9197704285091824, + "lr": 0.03, + "negative_goodness": 1.4732675714492798, + "pair_accuracy": 0.9063555555343628, + "positive_goodness": 3.1108201388888888, + "runtime_seconds": 4.026797771453857 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.9398780798594157, + "lr": 0.03, + "negative_goodness": 1.4960400100072224, + "pair_accuracy": 0.8978000000317892, + "positive_goodness": 3.057522085020277, + "runtime_seconds": 3.886897563934326 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9511253603299459, + "lr": 0.03, + "negative_goodness": 1.502223103989495, + "pair_accuracy": 0.8925333333333333, + "positive_goodness": 3.03213802579244, + "runtime_seconds": 3.8795042037963867 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.9603229959487916, + "lr": 0.03, + "negative_goodness": 1.517600659942627, + "pair_accuracy": 0.8949333333863152, + "positive_goodness": 3.017794240866767, + "runtime_seconds": 3.8384146690368652 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.950588881556193, + "lr": 0.03, + "negative_goodness": 1.4990133396148682, + "pair_accuracy": 0.8959555556085375, + "positive_goodness": 3.031329246902466, + "runtime_seconds": 3.8312416076660156 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9335020172542996, + "lr": 0.03, + "negative_goodness": 1.47775428276062, + "pair_accuracy": 0.8924222222434149, + "positive_goodness": 3.081496426773071, + "runtime_seconds": 3.824446678161621 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.944277149242825, + "lr": 0.03, + "negative_goodness": 1.4969571764204237, + "pair_accuracy": 0.8948888888676961, + "positive_goodness": 3.057461026679145, + "runtime_seconds": 3.8228912353515625 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9630947551515368, + "lr": 0.03, + "negative_goodness": 1.5212136888715957, + "pair_accuracy": 0.8920666666560703, + "positive_goodness": 2.9990695167965358, + "runtime_seconds": 3.8245339393615723 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9452382517602709, + "lr": 0.03, + "negative_goodness": 1.4966110318342845, + "pair_accuracy": 0.8946, + "positive_goodness": 3.050751921335856, + "runtime_seconds": 3.8248226642608643 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9562018103970422, + "lr": 0.03, + "negative_goodness": 1.5105079015043046, + "pair_accuracy": 0.8909777777777778, + "positive_goodness": 3.019462661149767, + "runtime_seconds": 3.827463150024414 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.907993979771932, + "lr": 0.03, + "negative_goodness": 1.456880996746487, + "pair_accuracy": 0.9066222222646078, + "positive_goodness": 3.1567500873989527, + "runtime_seconds": 3.827312469482422 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.9444151743994819, + "lr": 0.03, + "negative_goodness": 1.495000817404853, + "pair_accuracy": 0.8988222222540113, + "positive_goodness": 3.0617941925472683, + "runtime_seconds": 3.8533265590667725 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.970937155342102, + "lr": 0.03, + "negative_goodness": 1.5101799403932359, + "pair_accuracy": 0.8834444444444445, + "positive_goodness": 2.9957221263885496, + "runtime_seconds": 3.8618123531341553 + } + ], + "layer": 7 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1843314747916327, + "lr": 0.03, + "negative_goodness": 1.8096785085466174, + "pair_accuracy": 0.8119333333439297, + "positive_goodness": 2.4477177166408963, + "runtime_seconds": 3.907862663269043 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.0908262498007881, + "lr": 0.03, + "negative_goodness": 1.7145724468655057, + "pair_accuracy": 0.8672000000105964, + "positive_goodness": 2.639638025156657, + "runtime_seconds": 3.8882288932800293 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.0643919554392496, + "lr": 0.03, + "negative_goodness": 1.668543790075514, + "pair_accuracy": 0.8724888889418708, + "positive_goodness": 2.7071398744371202, + "runtime_seconds": 3.8655688762664795 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0630058233790927, + "lr": 0.03, + "negative_goodness": 1.6633555345111424, + "pair_accuracy": 0.8711333333333333, + "positive_goodness": 2.720318223190308, + "runtime_seconds": 3.8666837215423584 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.0618469442155627, + "lr": 0.03, + "negative_goodness": 1.6594931891123454, + "pair_accuracy": 0.8728444444974264, + "positive_goodness": 2.7255787406073675, + "runtime_seconds": 3.855844497680664 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.0495743374294706, + "lr": 0.03, + "negative_goodness": 1.6361259392844305, + "pair_accuracy": 0.8753333333651224, + "positive_goodness": 2.7702808492024737, + "runtime_seconds": 3.87961483001709 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.0467650150299073, + "lr": 0.03, + "negative_goodness": 1.6289938568115234, + "pair_accuracy": 0.8731111110899183, + "positive_goodness": 2.7646711762746174, + "runtime_seconds": 3.8812477588653564 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0531198591974047, + "lr": 0.03, + "negative_goodness": 1.6303391473770141, + "pair_accuracy": 0.8746444444126553, + "positive_goodness": 2.768021809471978, + "runtime_seconds": 3.901270866394043 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.0322821862750584, + "lr": 0.03, + "negative_goodness": 1.6214960707982382, + "pair_accuracy": 0.8846666667196486, + "positive_goodness": 2.805942648654514, + "runtime_seconds": 3.924118757247925 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.0120673136605156, + "lr": 0.03, + "negative_goodness": 1.5919928324805366, + "pair_accuracy": 0.8879111111111111, + "positive_goodness": 2.8543076766120064, + "runtime_seconds": 3.912341356277466 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.0134641191800435, + "lr": 0.03, + "negative_goodness": 1.5882788895289104, + "pair_accuracy": 0.8836, + "positive_goodness": 2.8654654865264892, + "runtime_seconds": 4.003767013549805 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.0213920141008166, + "lr": 0.03, + "negative_goodness": 1.6006039105945162, + "pair_accuracy": 0.883533333322737, + "positive_goodness": 2.8506203154246013, + "runtime_seconds": 3.9204907417297363 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.0042789044274225, + "lr": 0.03, + "negative_goodness": 1.585235983551873, + "pair_accuracy": 0.8835333333015442, + "positive_goodness": 2.889981962500678, + "runtime_seconds": 3.9103479385375977 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.0242807126363118, + "lr": 0.03, + "negative_goodness": 1.5924005566279094, + "pair_accuracy": 0.8827777777777778, + "positive_goodness": 2.8300068995157877, + "runtime_seconds": 3.9340481758117676 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.0146125266498989, + "lr": 0.03, + "negative_goodness": 1.589675334654914, + "pair_accuracy": 0.8811111111111111, + "positive_goodness": 2.872640680355496, + "runtime_seconds": 3.8742222785949707 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.02594399383333, + "lr": 0.03, + "negative_goodness": 1.586355850007799, + "pair_accuracy": 0.8740444444444444, + "positive_goodness": 2.8491206364949546, + "runtime_seconds": 3.909990072250366 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.0040110695732964, + "lr": 0.03, + "negative_goodness": 1.5713011403401693, + "pair_accuracy": 0.889177777756585, + "positive_goodness": 2.899435606765747, + "runtime_seconds": 3.8839914798736572 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.0131056555324132, + "lr": 0.03, + "negative_goodness": 1.574580996937222, + "pair_accuracy": 0.8814444444974263, + "positive_goodness": 2.8823492155710855, + "runtime_seconds": 3.8894004821777344 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.0000256185849508, + "lr": 0.03, + "negative_goodness": 1.5647295669131809, + "pair_accuracy": 0.8848, + "positive_goodness": 2.919537128194173, + "runtime_seconds": 4.051071405410767 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.0019603678597344, + "lr": 0.03, + "negative_goodness": 1.5666641600926716, + "pair_accuracy": 0.8837555556085375, + "positive_goodness": 2.9175658261617023, + "runtime_seconds": 4.1140007972717285 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.0088668394936455, + "lr": 0.03, + "negative_goodness": 1.567333915922377, + "pair_accuracy": 0.8832888889100816, + "positive_goodness": 2.8925293108410304, + "runtime_seconds": 4.086546421051025 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.0147882745954726, + "lr": 0.03, + "negative_goodness": 1.5675391729142931, + "pair_accuracy": 0.8811333333121406, + "positive_goodness": 2.8851874069213865, + "runtime_seconds": 3.9882283210754395 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.9775476300557454, + "lr": 0.03, + "negative_goodness": 1.5453736410564847, + "pair_accuracy": 0.8896000000105964, + "positive_goodness": 2.9565597864786786, + "runtime_seconds": 4.124258279800415 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.0086930328263177, + "lr": 0.03, + "negative_goodness": 1.5672792819764878, + "pair_accuracy": 0.8848888888570997, + "positive_goodness": 2.9004936158921986, + "runtime_seconds": 4.13614296913147 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.004484528933631, + "lr": 0.03, + "negative_goodness": 1.5652869699902006, + "pair_accuracy": 0.8776444444444444, + "positive_goodness": 2.8996909873962404, + "runtime_seconds": 3.9356095790863037 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9996444772614373, + "lr": 0.03, + "negative_goodness": 1.5542908275816176, + "pair_accuracy": 0.8870888888888889, + "positive_goodness": 2.9176612499660917, + "runtime_seconds": 3.8748137950897217 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.99581383389367, + "lr": 0.03, + "negative_goodness": 1.554695945739746, + "pair_accuracy": 0.8837777778307597, + "positive_goodness": 2.932477813212077, + "runtime_seconds": 4.096490859985352 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9769430025524564, + "lr": 0.03, + "negative_goodness": 1.54367327770657, + "pair_accuracy": 0.8932888889418708, + "positive_goodness": 2.967214637798733, + "runtime_seconds": 4.005157470703125 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.9958069711261326, + "lr": 0.03, + "negative_goodness": 1.5501911715401544, + "pair_accuracy": 0.8857333333439297, + "positive_goodness": 2.9392935397254094, + "runtime_seconds": 4.02176308631897 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9945689824210273, + "lr": 0.03, + "negative_goodness": 1.549483028878106, + "pair_accuracy": 0.8896444444444445, + "positive_goodness": 2.9278550802230834, + "runtime_seconds": 3.8879575729370117 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.0012718841340806, + "lr": 0.03, + "negative_goodness": 1.5568728165096706, + "pair_accuracy": 0.8787111111323038, + "positive_goodness": 2.9311372423807778, + "runtime_seconds": 3.9071977138519287 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.9884383135371738, + "lr": 0.03, + "negative_goodness": 1.5362172483020358, + "pair_accuracy": 0.8891111111005148, + "positive_goodness": 2.9436491321987575, + "runtime_seconds": 3.878896474838257 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9966020227644179, + "lr": 0.03, + "negative_goodness": 1.5529784621344673, + "pair_accuracy": 0.8856444444762336, + "positive_goodness": 2.942101957702637, + "runtime_seconds": 3.882232666015625 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.0121936600367227, + "lr": 0.03, + "negative_goodness": 1.5539330653296577, + "pair_accuracy": 0.879, + "positive_goodness": 2.8834142702314587, + "runtime_seconds": 3.9066760540008545 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9778893494076198, + "lr": 0.03, + "negative_goodness": 1.5371667425791422, + "pair_accuracy": 0.8920444444126553, + "positive_goodness": 2.9968213414933946, + "runtime_seconds": 3.908698320388794 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9934703800413344, + "lr": 0.03, + "negative_goodness": 1.5379041339238484, + "pair_accuracy": 0.8794444444444445, + "positive_goodness": 2.9410683835347493, + "runtime_seconds": 3.920137405395508 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9869493277973599, + "lr": 0.03, + "negative_goodness": 1.539103998629252, + "pair_accuracy": 0.8869777778201633, + "positive_goodness": 2.9599283844417994, + "runtime_seconds": 3.9171736240386963 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.983934252527025, + "lr": 0.03, + "negative_goodness": 1.5345206484688654, + "pair_accuracy": 0.884066666645474, + "positive_goodness": 2.978272841644287, + "runtime_seconds": 3.8899288177490234 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.977417014143202, + "lr": 0.03, + "negative_goodness": 1.527415295622084, + "pair_accuracy": 0.892, + "positive_goodness": 2.9887521514468722, + "runtime_seconds": 3.863647222518921 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.0022231745613945, + "lr": 0.03, + "negative_goodness": 1.5487360752741495, + "pair_accuracy": 0.8798, + "positive_goodness": 2.929055310100979, + "runtime_seconds": 3.8910257816314697 + } + ], + "layer": 8 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0358832864337497, + "lr": 0.03, + "negative_goodness": 1.6675561219321358, + "pair_accuracy": 0.8819555555873447, + "positive_goodness": 2.7537703990936278, + "runtime_seconds": 3.8370935916900635 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.9891615707291497, + "lr": 0.03, + "negative_goodness": 1.6530320317586262, + "pair_accuracy": 0.8782222222010294, + "positive_goodness": 2.8915867245144313, + "runtime_seconds": 3.83945894241333 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8878141170077853, + "lr": 0.03, + "negative_goodness": 1.5606034367879231, + "pair_accuracy": 0.9192888888888889, + "positive_goodness": 3.1682924705929225, + "runtime_seconds": 3.837035894393921 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.9117577192730374, + "lr": 0.03, + "negative_goodness": 1.5643655072106255, + "pair_accuracy": 0.9013777777724796, + "positive_goodness": 3.146012950621711, + "runtime_seconds": 3.910130262374878 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.9163189773877461, + "lr": 0.03, + "negative_goodness": 1.5582847270329794, + "pair_accuracy": 0.9064888888888889, + "positive_goodness": 3.129040741475423, + "runtime_seconds": 3.9021759033203125 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.9082701563305325, + "lr": 0.03, + "negative_goodness": 1.545086894162496, + "pair_accuracy": 0.8994000000211928, + "positive_goodness": 3.1755112706926134, + "runtime_seconds": 3.9099040031433105 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.9188763835112254, + "lr": 0.03, + "negative_goodness": 1.5447528271357218, + "pair_accuracy": 0.8950222222222223, + "positive_goodness": 3.1430123339335125, + "runtime_seconds": 3.9067559242248535 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.9009803359561497, + "lr": 0.03, + "negative_goodness": 1.5375380429797703, + "pair_accuracy": 0.8955777777777778, + "positive_goodness": 3.210897341240777, + "runtime_seconds": 3.9030160903930664 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.9296967094845242, + "lr": 0.03, + "negative_goodness": 1.558184096230401, + "pair_accuracy": 0.8940444444762335, + "positive_goodness": 3.1142229231940375, + "runtime_seconds": 3.893656015396118 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.9302207252502441, + "lr": 0.03, + "negative_goodness": 1.5469240951326158, + "pair_accuracy": 0.8794222222328186, + "positive_goodness": 3.1327387749989826, + "runtime_seconds": 3.8969104290008545 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.9186549046622382, + "lr": 0.03, + "negative_goodness": 1.5371480662027994, + "pair_accuracy": 0.8908222222169241, + "positive_goodness": 3.157744024954902, + "runtime_seconds": 3.904487133026123 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.9055257738961113, + "lr": 0.03, + "negative_goodness": 1.5495680402119953, + "pair_accuracy": 0.9050444444974264, + "positive_goodness": 3.2144851754506427, + "runtime_seconds": 3.903032064437866 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.9011442115677728, + "lr": 0.03, + "negative_goodness": 1.5247069632424248, + "pair_accuracy": 0.9033999999682109, + "positive_goodness": 3.2015889662424724, + "runtime_seconds": 3.897266387939453 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.884363050450219, + "lr": 0.03, + "negative_goodness": 1.5073396715164185, + "pair_accuracy": 0.9005777777671814, + "positive_goodness": 3.254066605207655, + "runtime_seconds": 3.904639482498169 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.8884530794249641, + "lr": 0.03, + "negative_goodness": 1.5119021244684856, + "pair_accuracy": 0.9037111111111111, + "positive_goodness": 3.2621248037550186, + "runtime_seconds": 3.903503179550171 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.9187686367140876, + "lr": 0.03, + "negative_goodness": 1.5215731049643622, + "pair_accuracy": 0.8779333333333333, + "positive_goodness": 3.1877019118414984, + "runtime_seconds": 3.907388210296631 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.8834266141679552, + "lr": 0.03, + "negative_goodness": 1.5007580158869425, + "pair_accuracy": 0.9014000000211928, + "positive_goodness": 3.2883937250349256, + "runtime_seconds": 3.8962643146514893 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.9097687305238512, + "lr": 0.03, + "negative_goodness": 1.5162624812232124, + "pair_accuracy": 0.8775333333651225, + "positive_goodness": 3.227445184495714, + "runtime_seconds": 3.90580153465271 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.8739692093743219, + "lr": 0.03, + "negative_goodness": 1.4910369480980767, + "pair_accuracy": 0.9014, + "positive_goodness": 3.3236650600857205, + "runtime_seconds": 3.906627893447876 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.8951612075911628, + "lr": 0.03, + "negative_goodness": 1.5078423842324151, + "pair_accuracy": 0.8975555555555556, + "positive_goodness": 3.27828961452908, + "runtime_seconds": 3.9129366874694824 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.9092739360809327, + "lr": 0.03, + "negative_goodness": 1.5301395987616644, + "pair_accuracy": 0.8998444444974264, + "positive_goodness": 3.1839873522016737, + "runtime_seconds": 3.918229818344116 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.873044596391254, + "lr": 0.03, + "negative_goodness": 1.4895911621411642, + "pair_accuracy": 0.9045777777777778, + "positive_goodness": 3.346990058135986, + "runtime_seconds": 3.906986951828003 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.8575657847086589, + "lr": 0.03, + "negative_goodness": 1.4663995742162068, + "pair_accuracy": 0.9096444444444445, + "positive_goodness": 3.391477436404758, + "runtime_seconds": 3.908240795135498 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.8869966570748223, + "lr": 0.03, + "negative_goodness": 1.495113319545322, + "pair_accuracy": 0.8983555555237664, + "positive_goodness": 3.297498961639404, + "runtime_seconds": 3.9032323360443115 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.8865034440676371, + "lr": 0.03, + "negative_goodness": 1.490049732568529, + "pair_accuracy": 0.8895777778095669, + "positive_goodness": 3.3070731699201796, + "runtime_seconds": 3.9077041149139404 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.8749743803977966, + "lr": 0.03, + "negative_goodness": 1.4780747811635335, + "pair_accuracy": 0.8995555555343628, + "positive_goodness": 3.3651324629889596, + "runtime_seconds": 3.908973217010498 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.8773001370800866, + "lr": 0.03, + "negative_goodness": 1.4838626975165472, + "pair_accuracy": 0.9038444444444445, + "positive_goodness": 3.318234583706326, + "runtime_seconds": 3.90739369392395 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.877095800866021, + "lr": 0.03, + "negative_goodness": 1.4652223713556924, + "pair_accuracy": 0.8914444444550408, + "positive_goodness": 3.348999482006497, + "runtime_seconds": 3.9011974334716797 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.8842959954897562, + "lr": 0.03, + "negative_goodness": 1.4718309762742785, + "pair_accuracy": 0.8925777778307596, + "positive_goodness": 3.3168835601382787, + "runtime_seconds": 3.9106037616729736 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.8843337440278795, + "lr": 0.03, + "negative_goodness": 1.492759492195977, + "pair_accuracy": 0.892955555597941, + "positive_goodness": 3.3245956931644014, + "runtime_seconds": 3.9105334281921387 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.8821527151955498, + "lr": 0.03, + "negative_goodness": 1.4774398944642808, + "pair_accuracy": 0.8911999999788073, + "positive_goodness": 3.3469902198367647, + "runtime_seconds": 3.9100475311279297 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.8507096779081557, + "lr": 0.03, + "negative_goodness": 1.4567865567737155, + "pair_accuracy": 0.9117111111005147, + "positive_goodness": 3.4275777196248374, + "runtime_seconds": 3.8992552757263184 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.8544254676394992, + "lr": 0.03, + "negative_goodness": 1.457909365272522, + "pair_accuracy": 0.9106444444444445, + "positive_goodness": 3.415316786532932, + "runtime_seconds": 3.9088897705078125 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.8812329434818692, + "lr": 0.03, + "negative_goodness": 1.4804432313495213, + "pair_accuracy": 0.8978888888782925, + "positive_goodness": 3.3376074785868326, + "runtime_seconds": 3.868914842605591 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.8678501529693603, + "lr": 0.03, + "negative_goodness": 1.4742986324628193, + "pair_accuracy": 0.9004444444444445, + "positive_goodness": 3.3758217328389484, + "runtime_seconds": 3.8653645515441895 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.8640371051364475, + "lr": 0.03, + "negative_goodness": 1.450820818625556, + "pair_accuracy": 0.9100444444444444, + "positive_goodness": 3.40370659061008, + "runtime_seconds": 3.872363567352295 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.8628585001839532, + "lr": 0.03, + "negative_goodness": 1.4698487561543783, + "pair_accuracy": 0.9032222222328186, + "positive_goodness": 3.4137300084431965, + "runtime_seconds": 3.845562696456909 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.8846077808168199, + "lr": 0.03, + "negative_goodness": 1.4818413024054633, + "pair_accuracy": 0.8956222222010295, + "positive_goodness": 3.3225142946455213, + "runtime_seconds": 3.863171339035034 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.8155634877734714, + "lr": 0.03, + "negative_goodness": 1.4268885684543187, + "pair_accuracy": 0.9235555556085374, + "positive_goodness": 3.550938623046875, + "runtime_seconds": 3.847903251647949 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.8900680318832398, + "lr": 0.03, + "negative_goodness": 1.4901371713638305, + "pair_accuracy": 0.8961333333121406, + "positive_goodness": 3.3048774739159477, + "runtime_seconds": 3.904465913772583 + } + ], + "layer": 9 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0558351916684046, + "lr": 0.03, + "negative_goodness": 1.6067137994554308, + "pair_accuracy": 0.849644444433848, + "positive_goodness": 2.9186081733703615, + "runtime_seconds": 3.915478467941284 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.9601502426147461, + "lr": 0.03, + "negative_goodness": 1.4416416626824273, + "pair_accuracy": 0.8825333333439297, + "positive_goodness": 3.0841339188045924, + "runtime_seconds": 3.9097015857696533 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.9383914781358507, + "lr": 0.03, + "negative_goodness": 1.421867149840461, + "pair_accuracy": 0.8879333333015442, + "positive_goodness": 3.128555516857571, + "runtime_seconds": 3.909867525100708 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.9118359811358981, + "lr": 0.03, + "negative_goodness": 1.4038405808554755, + "pair_accuracy": 0.8943777777459886, + "positive_goodness": 3.1915371131896975, + "runtime_seconds": 3.8653600215911865 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.9123529336505466, + "lr": 0.03, + "negative_goodness": 1.4165537573708429, + "pair_accuracy": 0.8956444444762336, + "positive_goodness": 3.2077453693389892, + "runtime_seconds": 3.9085640907287598 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.8871742343266805, + "lr": 0.03, + "negative_goodness": 1.3885399338616264, + "pair_accuracy": 0.89604444448683, + "positive_goodness": 3.2800258140563967, + "runtime_seconds": 3.907447338104248 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.9020539134343465, + "lr": 0.03, + "negative_goodness": 1.409446581946479, + "pair_accuracy": 0.888466666677263, + "positive_goodness": 3.2381191890292698, + "runtime_seconds": 3.9050984382629395 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.8911553741455078, + "lr": 0.03, + "negative_goodness": 1.3994558487786186, + "pair_accuracy": 0.8967999999788072, + "positive_goodness": 3.2583565584818523, + "runtime_seconds": 3.9226937294006348 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.8563773856480916, + "lr": 0.03, + "negative_goodness": 1.3645244947857327, + "pair_accuracy": 0.9061111111217075, + "positive_goodness": 3.390390911738078, + "runtime_seconds": 3.9016947746276855 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8591050770123799, + "lr": 0.03, + "negative_goodness": 1.357219104854266, + "pair_accuracy": 0.9074444444444445, + "positive_goodness": 3.3817755713568793, + "runtime_seconds": 3.8427021503448486 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.8486003678851658, + "lr": 0.03, + "negative_goodness": 1.3438505852275424, + "pair_accuracy": 0.9098222222222222, + "positive_goodness": 3.4300622630649142, + "runtime_seconds": 3.844301700592041 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.8622047729598151, + "lr": 0.03, + "negative_goodness": 1.3653751761436463, + "pair_accuracy": 0.9028222222222222, + "positive_goodness": 3.4188227526346844, + "runtime_seconds": 3.8534493446350098 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.8448330414242214, + "lr": 0.03, + "negative_goodness": 1.3307753030141194, + "pair_accuracy": 0.9062444444656372, + "positive_goodness": 3.4596204611036514, + "runtime_seconds": 3.853543758392334 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.8678845011499193, + "lr": 0.03, + "negative_goodness": 1.3660157205581664, + "pair_accuracy": 0.8988222222222222, + "positive_goodness": 3.3959336793687607, + "runtime_seconds": 3.852619171142578 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.8356462349467807, + "lr": 0.03, + "negative_goodness": 1.3266758021672567, + "pair_accuracy": 0.9064222222222222, + "positive_goodness": 3.5070715271843804, + "runtime_seconds": 3.850132465362549 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.8201515030172136, + "lr": 0.03, + "negative_goodness": 1.3000073939429388, + "pair_accuracy": 0.9118, + "positive_goodness": 3.5903649731953937, + "runtime_seconds": 3.8485333919525146 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.8472950864579942, + "lr": 0.03, + "negative_goodness": 1.335951808643341, + "pair_accuracy": 0.9033333333333333, + "positive_goodness": 3.4772789139641658, + "runtime_seconds": 3.844280242919922 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.8416874306360881, + "lr": 0.03, + "negative_goodness": 1.334253598382738, + "pair_accuracy": 0.9083111111429003, + "positive_goodness": 3.504263954035441, + "runtime_seconds": 3.848423957824707 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.845147163910336, + "lr": 0.03, + "negative_goodness": 1.3505227263397641, + "pair_accuracy": 0.9065333333333333, + "positive_goodness": 3.474437249077691, + "runtime_seconds": 3.8709206581115723 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.8251677840338812, + "lr": 0.03, + "negative_goodness": 1.3000160573323567, + "pair_accuracy": 0.9118666666454739, + "positive_goodness": 3.563138605456882, + "runtime_seconds": 3.9167888164520264 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.8264957674794727, + "lr": 0.03, + "negative_goodness": 1.3109123407072492, + "pair_accuracy": 0.9071333333333333, + "positive_goodness": 3.5757242730034724, + "runtime_seconds": 3.9021761417388916 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.8075486160066393, + "lr": 0.03, + "negative_goodness": 1.2898654060151842, + "pair_accuracy": 0.9191555556085375, + "positive_goodness": 3.644220988845825, + "runtime_seconds": 3.903717279434204 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.8173870400746663, + "lr": 0.03, + "negative_goodness": 1.2887617046356201, + "pair_accuracy": 0.9076666666454739, + "positive_goodness": 3.6470405620151096, + "runtime_seconds": 3.9132630825042725 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.8551283103095161, + "lr": 0.03, + "negative_goodness": 1.3388442260954114, + "pair_accuracy": 0.8978222222540113, + "positive_goodness": 3.5073866847144233, + "runtime_seconds": 3.8511555194854736 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.8107140611224705, + "lr": 0.03, + "negative_goodness": 1.279279676331414, + "pair_accuracy": 0.9116222222010295, + "positive_goodness": 3.655449427837796, + "runtime_seconds": 3.8524422645568848 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.8283477087550694, + "lr": 0.03, + "negative_goodness": 1.3093871060901219, + "pair_accuracy": 0.909733333322737, + "positive_goodness": 3.6376003587086996, + "runtime_seconds": 3.8554325103759766 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.8208261838277181, + "lr": 0.03, + "negative_goodness": 1.3111751277711656, + "pair_accuracy": 0.9124666666984558, + "positive_goodness": 3.592458839458889, + "runtime_seconds": 3.855693817138672 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.8106666393703884, + "lr": 0.03, + "negative_goodness": 1.2786455172220865, + "pair_accuracy": 0.9120666667090522, + "positive_goodness": 3.6696208088345, + "runtime_seconds": 3.8595430850982666 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.8037709504445394, + "lr": 0.03, + "negative_goodness": 1.2663682245678372, + "pair_accuracy": 0.9128222221904331, + "positive_goodness": 3.6842671020507813, + "runtime_seconds": 3.8626623153686523 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7958856056743198, + "lr": 0.03, + "negative_goodness": 1.2591524255540636, + "pair_accuracy": 0.9156444444444445, + "positive_goodness": 3.7226974939982096, + "runtime_seconds": 3.854762315750122 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.7791351869556639, + "lr": 0.03, + "negative_goodness": 1.234616523162524, + "pair_accuracy": 0.9174, + "positive_goodness": 3.877089628346761, + "runtime_seconds": 3.8604252338409424 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.815367404323154, + "lr": 0.03, + "negative_goodness": 1.2798442450205485, + "pair_accuracy": 0.9144888888570998, + "positive_goodness": 3.654650759209527, + "runtime_seconds": 3.913480520248413 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.7859028269873725, + "lr": 0.03, + "negative_goodness": 1.2477321193059285, + "pair_accuracy": 0.9167333333545261, + "positive_goodness": 3.755795608859592, + "runtime_seconds": 3.8648133277893066 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.8295221572875977, + "lr": 0.03, + "negative_goodness": 1.2960095588260228, + "pair_accuracy": 0.9003333333121406, + "positive_goodness": 3.604596774843004, + "runtime_seconds": 3.8969411849975586 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.8073719693289862, + "lr": 0.03, + "negative_goodness": 1.2717388545778063, + "pair_accuracy": 0.9144666667196486, + "positive_goodness": 3.695022182591756, + "runtime_seconds": 3.888864517211914 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.8209946473942863, + "lr": 0.03, + "negative_goodness": 1.2879728712293836, + "pair_accuracy": 0.9033555555555556, + "positive_goodness": 3.656694803788927, + "runtime_seconds": 3.878284215927124 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.8024388017018637, + "lr": 0.03, + "negative_goodness": 1.2641130877600777, + "pair_accuracy": 0.9175111111429003, + "positive_goodness": 3.702834570058187, + "runtime_seconds": 3.9240264892578125 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.7886253163284725, + "lr": 0.03, + "negative_goodness": 1.248301322301229, + "pair_accuracy": 0.9202, + "positive_goodness": 3.7417381613837346, + "runtime_seconds": 3.912605047225952 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.7849113559828864, + "lr": 0.03, + "negative_goodness": 1.2513682071738772, + "pair_accuracy": 0.9186444444444445, + "positive_goodness": 3.763354557715522, + "runtime_seconds": 3.9177615642547607 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.786431483274036, + "lr": 0.03, + "negative_goodness": 1.2500878911866082, + "pair_accuracy": 0.9154666666666667, + "positive_goodness": 3.7885238456726076, + "runtime_seconds": 3.9139344692230225 + } + ], + "layer": 10 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9468249512778388, + "lr": 0.03, + "negative_goodness": 1.3204057310316297, + "pair_accuracy": 0.863777777783076, + "positive_goodness": 3.0895350177764893, + "runtime_seconds": 3.9005627632141113 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8487998251067268, + "lr": 0.03, + "negative_goodness": 1.2718467063268026, + "pair_accuracy": 0.9001999999788073, + "positive_goodness": 3.39873157365587, + "runtime_seconds": 3.8964779376983643 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.854365719127655, + "lr": 0.03, + "negative_goodness": 1.2608610014809503, + "pair_accuracy": 0.8967555555449592, + "positive_goodness": 3.38607457347446, + "runtime_seconds": 3.8429393768310547 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7880270669725206, + "lr": 0.03, + "negative_goodness": 1.1702091563330757, + "pair_accuracy": 0.9146000000529819, + "positive_goodness": 3.6401710270351835, + "runtime_seconds": 3.877209186553955 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8010227835973104, + "lr": 0.03, + "negative_goodness": 1.1939093811459012, + "pair_accuracy": 0.9088222222752042, + "positive_goodness": 3.609933455106947, + "runtime_seconds": 3.838693141937256 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7712847319920858, + "lr": 0.03, + "negative_goodness": 1.1547626070022583, + "pair_accuracy": 0.9135555555449592, + "positive_goodness": 3.7263993840535483, + "runtime_seconds": 3.8337302207946777 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7530221975061628, + "lr": 0.03, + "negative_goodness": 1.1332294219070012, + "pair_accuracy": 0.9189555555555555, + "positive_goodness": 3.816787183041043, + "runtime_seconds": 3.9034388065338135 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7625816127035353, + "lr": 0.03, + "negative_goodness": 1.1389508597691853, + "pair_accuracy": 0.9139111111111111, + "positive_goodness": 3.8304185469733345, + "runtime_seconds": 3.8991498947143555 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7548678985171848, + "lr": 0.03, + "negative_goodness": 1.1211944930818345, + "pair_accuracy": 0.9162888888888889, + "positive_goodness": 3.856129520458645, + "runtime_seconds": 3.9053709506988525 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7666283572329415, + "lr": 0.03, + "negative_goodness": 1.1345050811502668, + "pair_accuracy": 0.9106, + "positive_goodness": 3.8186040117899576, + "runtime_seconds": 3.89774489402771 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7755829943233066, + "lr": 0.03, + "negative_goodness": 1.1560776280826992, + "pair_accuracy": 0.9156444444444445, + "positive_goodness": 3.804604751671685, + "runtime_seconds": 3.9063711166381836 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7591350213898552, + "lr": 0.03, + "negative_goodness": 1.1263152501000298, + "pair_accuracy": 0.9170666666878594, + "positive_goodness": 3.8892634816487632, + "runtime_seconds": 3.9011969566345215 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7373749463982052, + "lr": 0.03, + "negative_goodness": 1.0926800289339489, + "pair_accuracy": 0.9170666666666667, + "positive_goodness": 4.00246992263794, + "runtime_seconds": 3.895674228668213 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.781813658036126, + "lr": 0.03, + "negative_goodness": 1.155914085706075, + "pair_accuracy": 0.911577777756585, + "positive_goodness": 3.8111412300957572, + "runtime_seconds": 3.8969686031341553 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7285303099685245, + "lr": 0.03, + "negative_goodness": 1.087854132408566, + "pair_accuracy": 0.9219777777777778, + "positive_goodness": 4.027747608862983, + "runtime_seconds": 3.8957676887512207 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7384331241342756, + "lr": 0.03, + "negative_goodness": 1.1174515972773233, + "pair_accuracy": 0.9209777777777778, + "positive_goodness": 3.9917766626146105, + "runtime_seconds": 3.898087739944458 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7153649346987406, + "lr": 0.03, + "negative_goodness": 1.0875280453364053, + "pair_accuracy": 0.92404444448683, + "positive_goodness": 4.1265461230383975, + "runtime_seconds": 3.8917388916015625 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.725791405889723, + "lr": 0.03, + "negative_goodness": 1.095050508732266, + "pair_accuracy": 0.9222000000211927, + "positive_goodness": 4.053754569329156, + "runtime_seconds": 3.8963663578033447 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6751564452701144, + "lr": 0.03, + "negative_goodness": 1.0129327408896551, + "pair_accuracy": 0.9333555555343628, + "positive_goodness": 4.3327192353990345, + "runtime_seconds": 3.9045045375823975 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7049860657162137, + "lr": 0.03, + "negative_goodness": 1.0386129658805, + "pair_accuracy": 0.9245333333439297, + "positive_goodness": 4.2470057428571915, + "runtime_seconds": 3.8986971378326416 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6562255241605971, + "lr": 0.03, + "negative_goodness": 0.9907547793282403, + "pair_accuracy": 0.9356444444656372, + "positive_goodness": 4.441044807179769, + "runtime_seconds": 3.8853600025177 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6918718002425299, + "lr": 0.03, + "negative_goodness": 1.0240726950327554, + "pair_accuracy": 0.9295111111005148, + "positive_goodness": 4.268153716108534, + "runtime_seconds": 3.909026861190796 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7078117112265693, + "lr": 0.03, + "negative_goodness": 1.0638484600279066, + "pair_accuracy": 0.9281999999788072, + "positive_goodness": 4.252589094712999, + "runtime_seconds": 3.8973166942596436 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7089377976523505, + "lr": 0.03, + "negative_goodness": 1.0535264881134032, + "pair_accuracy": 0.9250444444762336, + "positive_goodness": 4.218918970870972, + "runtime_seconds": 3.8400402069091797 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.7071906132592095, + "lr": 0.03, + "negative_goodness": 1.047954317771064, + "pair_accuracy": 0.9195111110793219, + "positive_goodness": 4.262293440416125, + "runtime_seconds": 3.8457603454589844 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7116877640088399, + "lr": 0.03, + "negative_goodness": 1.0522976061291165, + "pair_accuracy": 0.9203333333015442, + "positive_goodness": 4.242380603790283, + "runtime_seconds": 3.842684030532837 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7251088712586297, + "lr": 0.03, + "negative_goodness": 1.0552146358913845, + "pair_accuracy": 0.9188000000529819, + "positive_goodness": 4.1465361424764, + "runtime_seconds": 3.8365907669067383 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6870019319693248, + "lr": 0.03, + "negative_goodness": 1.0192479966693455, + "pair_accuracy": 0.928, + "positive_goodness": 4.348951516977946, + "runtime_seconds": 3.8479909896850586 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.715013603401184, + "lr": 0.03, + "negative_goodness": 1.0573060506714715, + "pair_accuracy": 0.9198444444656372, + "positive_goodness": 4.278624683761596, + "runtime_seconds": 3.8979618549346924 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7120110624127918, + "lr": 0.03, + "negative_goodness": 1.0707091020849016, + "pair_accuracy": 0.9245555555555556, + "positive_goodness": 4.16941495997111, + "runtime_seconds": 3.9019105434417725 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6906147976769341, + "lr": 0.03, + "negative_goodness": 1.0338624197218154, + "pair_accuracy": 0.9305555555343628, + "positive_goodness": 4.3780581447177465, + "runtime_seconds": 3.833395481109619 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7117188526418474, + "lr": 0.03, + "negative_goodness": 1.0395845964749655, + "pair_accuracy": 0.9166444444444445, + "positive_goodness": 4.327581964111328, + "runtime_seconds": 3.8332207202911377 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6844747766123878, + "lr": 0.03, + "negative_goodness": 1.0150699672487047, + "pair_accuracy": 0.9281111110899184, + "positive_goodness": 4.411062888336182, + "runtime_seconds": 3.8316667079925537 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6922942256556617, + "lr": 0.03, + "negative_goodness": 1.0258845878601075, + "pair_accuracy": 0.9252444444444444, + "positive_goodness": 4.365875509728326, + "runtime_seconds": 3.843471050262451 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6570231073644426, + "lr": 0.03, + "negative_goodness": 0.9738702452977498, + "pair_accuracy": 0.9317777777777778, + "positive_goodness": 4.58300728386773, + "runtime_seconds": 3.8823084831237793 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6567871012793647, + "lr": 0.03, + "negative_goodness": 0.977057303958469, + "pair_accuracy": 0.936733333322737, + "positive_goodness": 4.575164974255032, + "runtime_seconds": 3.910874605178833 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6961285183323754, + "lr": 0.03, + "negative_goodness": 1.0176973945299785, + "pair_accuracy": 0.9249999999894036, + "positive_goodness": 4.405254905954997, + "runtime_seconds": 3.907802104949951 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6981639866669973, + "lr": 0.03, + "negative_goodness": 1.0126278137683868, + "pair_accuracy": 0.9239111111111111, + "positive_goodness": 4.418038633812799, + "runtime_seconds": 3.900681495666504 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6729135822402106, + "lr": 0.03, + "negative_goodness": 0.9919594741715325, + "pair_accuracy": 0.9310222222222222, + "positive_goodness": 4.5101987101236976, + "runtime_seconds": 3.9013195037841797 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6710425524605645, + "lr": 0.03, + "negative_goodness": 0.9931388754314846, + "pair_accuracy": 0.930977777756585, + "positive_goodness": 4.57458455051846, + "runtime_seconds": 3.9087913036346436 + } + ], + "layer": 11 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.3185271839353774, + "lr": 0.03, + "negative_goodness": 1.887508733092414, + "pair_accuracy": 0.7703555555555556, + "positive_goodness": 2.946332304679023, + "runtime_seconds": 3.9298603534698486 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.0538476404296027, + "lr": 0.03, + "negative_goodness": 1.5676008949279785, + "pair_accuracy": 0.8383555555979411, + "positive_goodness": 3.155625390709771, + "runtime_seconds": 3.9115865230560303 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.9585096132066515, + "lr": 0.03, + "negative_goodness": 1.437667080031501, + "pair_accuracy": 0.8750888888570998, + "positive_goodness": 3.2918078805711533, + "runtime_seconds": 3.887042760848999 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8724830559412639, + "lr": 0.03, + "negative_goodness": 1.3157048323313396, + "pair_accuracy": 0.9029333333439297, + "positive_goodness": 3.426985045030382, + "runtime_seconds": 3.9251480102539062 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8774001080407037, + "lr": 0.03, + "negative_goodness": 1.3067769260618423, + "pair_accuracy": 0.8980000000423856, + "positive_goodness": 3.4186257697635227, + "runtime_seconds": 3.9104325771331787 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.8864061154047648, + "lr": 0.03, + "negative_goodness": 1.3154028912968105, + "pair_accuracy": 0.900911111079322, + "positive_goodness": 3.321017227300008, + "runtime_seconds": 3.936570644378662 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.834978289964464, + "lr": 0.03, + "negative_goodness": 1.2430314351823595, + "pair_accuracy": 0.9102444444762335, + "positive_goodness": 3.496619998847114, + "runtime_seconds": 3.9334516525268555 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.8177921928193834, + "lr": 0.03, + "negative_goodness": 1.2044719324111939, + "pair_accuracy": 0.9155555555767483, + "positive_goodness": 3.5216175231933593, + "runtime_seconds": 3.845991611480713 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7978795686086019, + "lr": 0.03, + "negative_goodness": 1.180627177450392, + "pair_accuracy": 0.9162666666984558, + "positive_goodness": 3.5885706604851615, + "runtime_seconds": 3.8512234687805176 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8082160500632392, + "lr": 0.03, + "negative_goodness": 1.1830865351570976, + "pair_accuracy": 0.9140222222222222, + "positive_goodness": 3.555669383494059, + "runtime_seconds": 3.85798978805542 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7956344126595392, + "lr": 0.03, + "negative_goodness": 1.1694402980380587, + "pair_accuracy": 0.917555555566152, + "positive_goodness": 3.635924807484945, + "runtime_seconds": 3.8609530925750732 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7968834280437893, + "lr": 0.03, + "negative_goodness": 1.186173814227846, + "pair_accuracy": 0.9196444444338481, + "positive_goodness": 3.5892775128682453, + "runtime_seconds": 3.890718460083008 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7943821539137098, + "lr": 0.03, + "negative_goodness": 1.1597710718366834, + "pair_accuracy": 0.9185777777777778, + "positive_goodness": 3.622129463068644, + "runtime_seconds": 3.9260261058807373 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7961443520122105, + "lr": 0.03, + "negative_goodness": 1.1603307704607646, + "pair_accuracy": 0.9179333333757188, + "positive_goodness": 3.6188617546081545, + "runtime_seconds": 3.9222264289855957 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7437063765843709, + "lr": 0.03, + "negative_goodness": 1.1043735557556151, + "pair_accuracy": 0.9294666666666667, + "positive_goodness": 3.784000438944499, + "runtime_seconds": 3.9175736904144287 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7791582089953952, + "lr": 0.03, + "negative_goodness": 1.1289865065892537, + "pair_accuracy": 0.9187555555979411, + "positive_goodness": 3.7111513886769614, + "runtime_seconds": 3.880596160888672 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7622104647742377, + "lr": 0.03, + "negative_goodness": 1.1050140443589953, + "pair_accuracy": 0.9213111110899184, + "positive_goodness": 3.7612183052910697, + "runtime_seconds": 3.861327648162842 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7940157075511085, + "lr": 0.03, + "negative_goodness": 1.1450878811306424, + "pair_accuracy": 0.9135111110793219, + "positive_goodness": 3.6287584171719023, + "runtime_seconds": 3.8529882431030273 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7525702519522773, + "lr": 0.03, + "negative_goodness": 1.097413838789198, + "pair_accuracy": 0.9259555556085375, + "positive_goodness": 3.792832419204712, + "runtime_seconds": 3.851757287979126 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7390556241883172, + "lr": 0.03, + "negative_goodness": 1.086934466489156, + "pair_accuracy": 0.9296666666560703, + "positive_goodness": 3.8372004843817815, + "runtime_seconds": 3.8560264110565186 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7477235282262167, + "lr": 0.03, + "negative_goodness": 1.0905819366878933, + "pair_accuracy": 0.9251777777671814, + "positive_goodness": 3.812375106302897, + "runtime_seconds": 3.894591808319092 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7522649048752255, + "lr": 0.03, + "negative_goodness": 1.1055291634559632, + "pair_accuracy": 0.9259777777671814, + "positive_goodness": 3.7835853234185115, + "runtime_seconds": 3.8674988746643066 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7339181278864543, + "lr": 0.03, + "negative_goodness": 1.0716687260945639, + "pair_accuracy": 0.9285333333015442, + "positive_goodness": 3.9176464138454863, + "runtime_seconds": 3.8688933849334717 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7607215806007386, + "lr": 0.03, + "negative_goodness": 1.1035220854335361, + "pair_accuracy": 0.9231111111005147, + "positive_goodness": 3.7789207315233018, + "runtime_seconds": 3.8720293045043945 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.7213977890014649, + "lr": 0.03, + "negative_goodness": 1.0410647917429605, + "pair_accuracy": 0.9305111111217075, + "positive_goodness": 3.9565952503204347, + "runtime_seconds": 3.872450590133667 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7464911210378011, + "lr": 0.03, + "negative_goodness": 1.0664513245264688, + "pair_accuracy": 0.9239555556085375, + "positive_goodness": 3.8778968577914767, + "runtime_seconds": 3.869954824447632 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7114085529751247, + "lr": 0.03, + "negative_goodness": 1.0316443641026816, + "pair_accuracy": 0.9300666666560703, + "positive_goodness": 3.986421192720201, + "runtime_seconds": 3.881387710571289 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.731734022670322, + "lr": 0.03, + "negative_goodness": 1.0467169155544704, + "pair_accuracy": 0.9265777777989705, + "positive_goodness": 3.8731126875135633, + "runtime_seconds": 3.852855682373047 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.718660465865665, + "lr": 0.03, + "negative_goodness": 1.0528238191922505, + "pair_accuracy": 0.9299999999788072, + "positive_goodness": 3.9243148715549045, + "runtime_seconds": 3.9162094593048096 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7376200407769945, + "lr": 0.03, + "negative_goodness": 1.075409499973721, + "pair_accuracy": 0.9242, + "positive_goodness": 3.8837724106258817, + "runtime_seconds": 3.9272773265838623 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.7357056013319228, + "lr": 0.03, + "negative_goodness": 1.0630966669082642, + "pair_accuracy": 0.9235777778307597, + "positive_goodness": 3.894642550616794, + "runtime_seconds": 3.9176580905914307 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7394641547467974, + "lr": 0.03, + "negative_goodness": 1.0713817902988858, + "pair_accuracy": 0.925133333322737, + "positive_goodness": 3.877066040971544, + "runtime_seconds": 3.925891160964966 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.7395559886084663, + "lr": 0.03, + "negative_goodness": 1.0492063057051764, + "pair_accuracy": 0.9221999999788072, + "positive_goodness": 3.911743345896403, + "runtime_seconds": 3.917706251144409 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.7363708227157593, + "lr": 0.03, + "negative_goodness": 1.0597754657745362, + "pair_accuracy": 0.9264222222222223, + "positive_goodness": 3.9006665195041235, + "runtime_seconds": 3.9313409328460693 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.752273914972941, + "lr": 0.03, + "negative_goodness": 1.0814482376734416, + "pair_accuracy": 0.9230444444974264, + "positive_goodness": 3.8442159672207303, + "runtime_seconds": 3.9416258335113525 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.7138455394956801, + "lr": 0.03, + "negative_goodness": 1.0092146866692437, + "pair_accuracy": 0.9274000000423855, + "positive_goodness": 4.0174319735633, + "runtime_seconds": 3.922100067138672 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.7154065210766263, + "lr": 0.03, + "negative_goodness": 1.0389859244876438, + "pair_accuracy": 0.9323555555873447, + "positive_goodness": 4.0220463450961645, + "runtime_seconds": 3.928891181945801 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6992003760231866, + "lr": 0.03, + "negative_goodness": 0.9963172398991055, + "pair_accuracy": 0.933577777756585, + "positive_goodness": 4.098868013381958, + "runtime_seconds": 3.9161112308502197 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.7195847440719605, + "lr": 0.03, + "negative_goodness": 1.037339400683509, + "pair_accuracy": 0.9262888888570997, + "positive_goodness": 4.004261943266127, + "runtime_seconds": 3.918886423110962 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.7497493489477369, + "lr": 0.03, + "negative_goodness": 1.0539330623626708, + "pair_accuracy": 0.9207777778201633, + "positive_goodness": 3.9021214069366454, + "runtime_seconds": 3.928858995437622 + } + ], + "layer": 12 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9897457589467367, + "lr": 0.03, + "negative_goodness": 1.3422176013628642, + "pair_accuracy": 0.8501555555449591, + "positive_goodness": 2.9998260765923392, + "runtime_seconds": 3.926103115081787 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.941633687930637, + "lr": 0.03, + "negative_goodness": 1.3315876194636027, + "pair_accuracy": 0.8691999999894037, + "positive_goodness": 3.152401608403524, + "runtime_seconds": 3.899630546569824 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.9204310355504354, + "lr": 0.03, + "negative_goodness": 1.298874842262268, + "pair_accuracy": 0.8750666666348775, + "positive_goodness": 3.219549514304267, + "runtime_seconds": 3.902782440185547 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8869137210951911, + "lr": 0.03, + "negative_goodness": 1.26926410235299, + "pair_accuracy": 0.8853999999788073, + "positive_goodness": 3.300233348634508, + "runtime_seconds": 3.8953773975372314 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8810357227325439, + "lr": 0.03, + "negative_goodness": 1.2697638557010227, + "pair_accuracy": 0.8888888888994853, + "positive_goodness": 3.3525470156987507, + "runtime_seconds": 3.9022879600524902 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.8823558606041803, + "lr": 0.03, + "negative_goodness": 1.2548078119913737, + "pair_accuracy": 0.8866222222222222, + "positive_goodness": 3.351088674672445, + "runtime_seconds": 3.9149863719940186 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.8628695985158285, + "lr": 0.03, + "negative_goodness": 1.2392077796088325, + "pair_accuracy": 0.8926666666984558, + "positive_goodness": 3.4001668996598986, + "runtime_seconds": 3.9063034057617188 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.8743702362060547, + "lr": 0.03, + "negative_goodness": 1.2445717261844211, + "pair_accuracy": 0.8858222222010295, + "positive_goodness": 3.3836656655629476, + "runtime_seconds": 3.9007363319396973 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.8505621334075928, + "lr": 0.03, + "negative_goodness": 1.219017782825894, + "pair_accuracy": 0.8937111111005147, + "positive_goodness": 3.4841313712649877, + "runtime_seconds": 3.893037796020508 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8643822475539313, + "lr": 0.03, + "negative_goodness": 1.2311093984391954, + "pair_accuracy": 0.8920222222646077, + "positive_goodness": 3.426124864112006, + "runtime_seconds": 3.893289089202881 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.8484905865351359, + "lr": 0.03, + "negative_goodness": 1.211715002632141, + "pair_accuracy": 0.8937333333015441, + "positive_goodness": 3.4991822474161784, + "runtime_seconds": 3.925253391265869 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.8382850954744551, + "lr": 0.03, + "negative_goodness": 1.1946074604988097, + "pair_accuracy": 0.8964444444444445, + "positive_goodness": 3.5412085612826876, + "runtime_seconds": 3.9113609790802 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.8589689746220907, + "lr": 0.03, + "negative_goodness": 1.2189074726528593, + "pair_accuracy": 0.8918222222222222, + "positive_goodness": 3.460806776682536, + "runtime_seconds": 3.909353733062744 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.8382921874576145, + "lr": 0.03, + "negative_goodness": 1.194120587433709, + "pair_accuracy": 0.8962444444868299, + "positive_goodness": 3.5494751014709474, + "runtime_seconds": 3.9132654666900635 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.8373078263706631, + "lr": 0.03, + "negative_goodness": 1.189800752984153, + "pair_accuracy": 0.8975555555979411, + "positive_goodness": 3.5562288522932266, + "runtime_seconds": 3.9149625301361084 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.8138099120563931, + "lr": 0.03, + "negative_goodness": 1.1630562759823269, + "pair_accuracy": 0.903022222190433, + "positive_goodness": 3.651068496025933, + "runtime_seconds": 3.9146087169647217 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.8208305387285021, + "lr": 0.03, + "negative_goodness": 1.1565215217060514, + "pair_accuracy": 0.9010666667196485, + "positive_goodness": 3.6449738268958196, + "runtime_seconds": 3.9044671058654785 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.8327840298334758, + "lr": 0.03, + "negative_goodness": 1.181743448193868, + "pair_accuracy": 0.8977111111534967, + "positive_goodness": 3.581564500596788, + "runtime_seconds": 3.87178111076355 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.8467262108590867, + "lr": 0.03, + "negative_goodness": 1.1977639463980994, + "pair_accuracy": 0.8923111111111112, + "positive_goodness": 3.5630890951368546, + "runtime_seconds": 3.8162147998809814 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.8348355386310153, + "lr": 0.03, + "negative_goodness": 1.1784591431087919, + "pair_accuracy": 0.8952888888676961, + "positive_goodness": 3.5852722641415067, + "runtime_seconds": 3.821647882461548 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.8180928344514635, + "lr": 0.03, + "negative_goodness": 1.1677586513519287, + "pair_accuracy": 0.8998666666984558, + "positive_goodness": 3.654529478242662, + "runtime_seconds": 3.9178028106689453 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.8001247010654874, + "lr": 0.03, + "negative_goodness": 1.129678199450175, + "pair_accuracy": 0.9043555555555556, + "positive_goodness": 3.730193137359619, + "runtime_seconds": 3.9250919818878174 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.83402632194625, + "lr": 0.03, + "negative_goodness": 1.1668323285738627, + "pair_accuracy": 0.8942888888782925, + "positive_goodness": 3.6007708541022407, + "runtime_seconds": 3.910881996154785 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7959150106006199, + "lr": 0.03, + "negative_goodness": 1.1224587226867675, + "pair_accuracy": 0.9039777777459886, + "positive_goodness": 3.7748881131066216, + "runtime_seconds": 3.8910136222839355 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.8257771210776435, + "lr": 0.03, + "negative_goodness": 1.1688468457698822, + "pair_accuracy": 0.8971111111005148, + "positive_goodness": 3.6351160851372613, + "runtime_seconds": 3.804781675338745 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7914416019545661, + "lr": 0.03, + "negative_goodness": 1.1017765376620823, + "pair_accuracy": 0.9060000000529819, + "positive_goodness": 3.7985594579060873, + "runtime_seconds": 3.8129026889801025 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.814687057240804, + "lr": 0.03, + "negative_goodness": 1.1498000959184436, + "pair_accuracy": 0.9018444444603391, + "positive_goodness": 3.686912654368083, + "runtime_seconds": 3.8221235275268555 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.8145993664953444, + "lr": 0.03, + "negative_goodness": 1.1493916024631925, + "pair_accuracy": 0.9037555555237664, + "positive_goodness": 3.6813575778961183, + "runtime_seconds": 3.8154408931732178 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.8077748132175869, + "lr": 0.03, + "negative_goodness": 1.1383171142896016, + "pair_accuracy": 0.9051333333863152, + "positive_goodness": 3.7131362295362687, + "runtime_seconds": 3.81921124458313 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7890516102049085, + "lr": 0.03, + "negative_goodness": 1.1225600020938449, + "pair_accuracy": 0.9120666666666667, + "positive_goodness": 3.811327413898044, + "runtime_seconds": 3.8269670009613037 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.8040497264650133, + "lr": 0.03, + "negative_goodness": 1.1267237514919706, + "pair_accuracy": 0.9041333333651225, + "positive_goodness": 3.74125780359904, + "runtime_seconds": 3.8397576808929443 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7980873599264356, + "lr": 0.03, + "negative_goodness": 1.1179847688462998, + "pair_accuracy": 0.9049777778307597, + "positive_goodness": 3.7891956053415936, + "runtime_seconds": 3.8322391510009766 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.7983331745783488, + "lr": 0.03, + "negative_goodness": 1.1104851909001667, + "pair_accuracy": 0.9047111111217075, + "positive_goodness": 3.7770513941870796, + "runtime_seconds": 3.836911916732788 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.7962364349047343, + "lr": 0.03, + "negative_goodness": 1.1184681528197395, + "pair_accuracy": 0.9070888888570997, + "positive_goodness": 3.7629022566901313, + "runtime_seconds": 3.823380470275879 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.7819444798151652, + "lr": 0.03, + "negative_goodness": 1.0936477658377752, + "pair_accuracy": 0.9085777777989705, + "positive_goodness": 3.9046398670196534, + "runtime_seconds": 3.8161845207214355 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.8161905946731567, + "lr": 0.03, + "negative_goodness": 1.1309946965535482, + "pair_accuracy": 0.8963333333439297, + "positive_goodness": 3.723821687698364, + "runtime_seconds": 3.8143906593322754 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.7990322351031833, + "lr": 0.03, + "negative_goodness": 1.104172683821784, + "pair_accuracy": 0.9037333333333334, + "positive_goodness": 3.8091714139726425, + "runtime_seconds": 3.8179116249084473 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.8323696867624919, + "lr": 0.03, + "negative_goodness": 1.1567761015997993, + "pair_accuracy": 0.8956222222328186, + "positive_goodness": 3.659682942920261, + "runtime_seconds": 3.812333583831787 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.8112501290109423, + "lr": 0.03, + "negative_goodness": 1.133375339741177, + "pair_accuracy": 0.9000000000529819, + "positive_goodness": 3.727873260243734, + "runtime_seconds": 3.8229198455810547 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.7952498845418294, + "lr": 0.03, + "negative_goodness": 1.1138768318600125, + "pair_accuracy": 0.9064444444550408, + "positive_goodness": 3.814595316441854, + "runtime_seconds": 3.8077547550201416 + } + ], + "layer": 13 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9927435837109884, + "lr": 0.03, + "negative_goodness": 1.371428248977661, + "pair_accuracy": 0.8501333333651224, + "positive_goodness": 3.180533321762085, + "runtime_seconds": 3.8475613594055176 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8972017687320709, + "lr": 0.03, + "negative_goodness": 1.2478669930352104, + "pair_accuracy": 0.8862888888888889, + "positive_goodness": 3.4034515160454646, + "runtime_seconds": 3.8471381664276123 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.83355728225708, + "lr": 0.03, + "negative_goodness": 1.1898207000308567, + "pair_accuracy": 0.9050888888676961, + "positive_goodness": 3.567763674884372, + "runtime_seconds": 3.86142897605896 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8317280115445455, + "lr": 0.03, + "negative_goodness": 1.170232958624098, + "pair_accuracy": 0.8990222222010295, + "positive_goodness": 3.5746266167110865, + "runtime_seconds": 3.8573617935180664 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7972440318531461, + "lr": 0.03, + "negative_goodness": 1.139726331975725, + "pair_accuracy": 0.911044444433848, + "positive_goodness": 3.6475575857374403, + "runtime_seconds": 3.8659579753875732 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7975238952530755, + "lr": 0.03, + "negative_goodness": 1.1403054419729444, + "pair_accuracy": 0.9085555555449591, + "positive_goodness": 3.6817897068447536, + "runtime_seconds": 3.865429639816284 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.755075655301412, + "lr": 0.03, + "negative_goodness": 1.094656585714552, + "pair_accuracy": 0.9218444444126553, + "positive_goodness": 3.845978601582845, + "runtime_seconds": 3.864448308944702 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7316005357954237, + "lr": 0.03, + "negative_goodness": 1.0845576380199857, + "pair_accuracy": 0.9312444444232517, + "positive_goodness": 3.890253694788615, + "runtime_seconds": 3.8736095428466797 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7511945588694678, + "lr": 0.03, + "negative_goodness": 1.1034989507039388, + "pair_accuracy": 0.9195111111111111, + "positive_goodness": 3.798990623219808, + "runtime_seconds": 3.873521089553833 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7429910631921556, + "lr": 0.03, + "negative_goodness": 1.0841457685258653, + "pair_accuracy": 0.9202222222116259, + "positive_goodness": 3.8820672032674155, + "runtime_seconds": 3.8549673557281494 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7543365297847324, + "lr": 0.03, + "negative_goodness": 1.1100962123129103, + "pair_accuracy": 0.9203555556085374, + "positive_goodness": 3.803879145558675, + "runtime_seconds": 3.8532605171203613 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7291592262479994, + "lr": 0.03, + "negative_goodness": 1.08297777200275, + "pair_accuracy": 0.9221111111429002, + "positive_goodness": 3.92532549273173, + "runtime_seconds": 3.8545966148376465 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.69681714764701, + "lr": 0.03, + "negative_goodness": 1.0417558953391182, + "pair_accuracy": 0.9319999999788072, + "positive_goodness": 4.092233706071641, + "runtime_seconds": 3.861940383911133 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7102607422298856, + "lr": 0.03, + "negative_goodness": 1.0484091600524055, + "pair_accuracy": 0.9279555555555555, + "positive_goodness": 3.9972046440124513, + "runtime_seconds": 4.024444103240967 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7117048489040798, + "lr": 0.03, + "negative_goodness": 1.060182642724779, + "pair_accuracy": 0.9296000000105964, + "positive_goodness": 3.9884732605404323, + "runtime_seconds": 4.101162672042847 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6871303003523085, + "lr": 0.03, + "negative_goodness": 1.0202624196370442, + "pair_accuracy": 0.9323555555661519, + "positive_goodness": 4.1345319319831, + "runtime_seconds": 4.092319965362549 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6819168406062656, + "lr": 0.03, + "negative_goodness": 1.0196824308819241, + "pair_accuracy": 0.9364000000529819, + "positive_goodness": 4.115802656597562, + "runtime_seconds": 4.097151279449463 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6917262025409274, + "lr": 0.03, + "negative_goodness": 1.0199751360151503, + "pair_accuracy": 0.9304222222540114, + "positive_goodness": 4.073841888978746, + "runtime_seconds": 4.088978052139282 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6883058398564657, + "lr": 0.03, + "negative_goodness": 1.0209365759319728, + "pair_accuracy": 0.9309555555979411, + "positive_goodness": 4.111548299556309, + "runtime_seconds": 3.913853406906128 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.696681451649136, + "lr": 0.03, + "negative_goodness": 1.021330736372206, + "pair_accuracy": 0.9268888889312744, + "positive_goodness": 4.101939597829183, + "runtime_seconds": 4.093047142028809 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6933414397875468, + "lr": 0.03, + "negative_goodness": 1.0383107833226521, + "pair_accuracy": 0.9320222221904331, + "positive_goodness": 4.112108587392171, + "runtime_seconds": 4.072264194488525 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6650079068289863, + "lr": 0.03, + "negative_goodness": 0.9933130378087361, + "pair_accuracy": 0.9379999999788072, + "positive_goodness": 4.169857550811767, + "runtime_seconds": 3.919973850250244 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6846415030691358, + "lr": 0.03, + "negative_goodness": 1.0140913085513645, + "pair_accuracy": 0.9311333333121405, + "positive_goodness": 4.137001554361979, + "runtime_seconds": 3.868976354598999 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6584406207826402, + "lr": 0.03, + "negative_goodness": 0.9772191049257914, + "pair_accuracy": 0.9373777777671813, + "positive_goodness": 4.230124160512289, + "runtime_seconds": 3.964521646499634 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6743088566462199, + "lr": 0.03, + "negative_goodness": 1.002647195010715, + "pair_accuracy": 0.93384444448683, + "positive_goodness": 4.192681758541531, + "runtime_seconds": 4.096384048461914 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6550116949505276, + "lr": 0.03, + "negative_goodness": 0.9566675271881951, + "pair_accuracy": 0.9388222222222222, + "positive_goodness": 4.2652373776753745, + "runtime_seconds": 4.04522705078125 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6542717175801595, + "lr": 0.03, + "negative_goodness": 0.9637270623948839, + "pair_accuracy": 0.9356222222222222, + "positive_goodness": 4.28298183063931, + "runtime_seconds": 3.917681932449341 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6674591303083631, + "lr": 0.03, + "negative_goodness": 0.985511946911282, + "pair_accuracy": 0.935733333322737, + "positive_goodness": 4.2471505895402695, + "runtime_seconds": 3.9382097721099854 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.648469323152966, + "lr": 0.03, + "negative_goodness": 0.9452383399248123, + "pair_accuracy": 0.937, + "positive_goodness": 4.350823145294189, + "runtime_seconds": 4.013874292373657 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6465691491656833, + "lr": 0.03, + "negative_goodness": 0.9511275424639384, + "pair_accuracy": 0.9390666666454739, + "positive_goodness": 4.326571492597792, + "runtime_seconds": 4.078093767166138 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6329634171168009, + "lr": 0.03, + "negative_goodness": 0.9330124511082967, + "pair_accuracy": 0.9445333333227369, + "positive_goodness": 4.433930954530504, + "runtime_seconds": 4.0952982902526855 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6323839157104493, + "lr": 0.03, + "negative_goodness": 0.9197256020651924, + "pair_accuracy": 0.9417555555343627, + "positive_goodness": 4.447618661499024, + "runtime_seconds": 4.093653917312622 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.653815532901552, + "lr": 0.03, + "negative_goodness": 0.9478567444059584, + "pair_accuracy": 0.9345555555449592, + "positive_goodness": 4.338125744374593, + "runtime_seconds": 4.0948405265808105 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.627855408583747, + "lr": 0.03, + "negative_goodness": 0.9294390690273708, + "pair_accuracy": 0.9454888889100816, + "positive_goodness": 4.431742012532552, + "runtime_seconds": 4.058185815811157 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6444618001937866, + "lr": 0.03, + "negative_goodness": 0.9484672688802084, + "pair_accuracy": 0.9396444444126553, + "positive_goodness": 4.355242827267117, + "runtime_seconds": 4.096928834915161 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.637737031671736, + "lr": 0.03, + "negative_goodness": 0.9296492751227485, + "pair_accuracy": 0.9429777778201633, + "positive_goodness": 4.411582122378879, + "runtime_seconds": 4.100308895111084 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6394215944290161, + "lr": 0.03, + "negative_goodness": 0.9395501088248359, + "pair_accuracy": 0.939777777756585, + "positive_goodness": 4.435339725579156, + "runtime_seconds": 4.093689680099487 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.660016583442688, + "lr": 0.03, + "negative_goodness": 0.943231183327569, + "pair_accuracy": 0.9350666666348775, + "positive_goodness": 4.319819572533501, + "runtime_seconds": 4.012077808380127 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6339202422671848, + "lr": 0.03, + "negative_goodness": 0.918854190381368, + "pair_accuracy": 0.9398222222116258, + "positive_goodness": 4.4606144256591795, + "runtime_seconds": 3.8652520179748535 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.656699076880349, + "lr": 0.03, + "negative_goodness": 0.9560864576816559, + "pair_accuracy": 0.9348, + "positive_goodness": 4.325829438527425, + "runtime_seconds": 3.8723371028900146 + } + ], + "layer": 14 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.5874056291156345, + "lr": 0.03, + "negative_goodness": 0.8470072862625122, + "pair_accuracy": 0.9333555555767483, + "positive_goodness": 4.389488594267103, + "runtime_seconds": 4.048981666564941 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.47105628654691906, + "lr": 0.03, + "negative_goodness": 0.7114742046515147, + "pair_accuracy": 0.9556888888888889, + "positive_goodness": 4.912495910135905, + "runtime_seconds": 3.9583826065063477 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.4308567408985562, + "lr": 0.03, + "negative_goodness": 0.6499376036961874, + "pair_accuracy": 0.9612888888782926, + "positive_goodness": 5.2190226319207085, + "runtime_seconds": 4.097906827926636 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.4745926030211978, + "lr": 0.03, + "negative_goodness": 0.6863426258616977, + "pair_accuracy": 0.9478888888782925, + "positive_goodness": 4.957175006866455, + "runtime_seconds": 4.098802328109741 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.4426095139662425, + "lr": 0.03, + "negative_goodness": 0.6423796742863125, + "pair_accuracy": 0.9527333333227369, + "positive_goodness": 5.181111384582519, + "runtime_seconds": 4.080105543136597 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.4406747067186568, + "lr": 0.03, + "negative_goodness": 0.6447717011610667, + "pair_accuracy": 0.96, + "positive_goodness": 5.298106381310357, + "runtime_seconds": 4.069923639297485 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.4207660053888957, + "lr": 0.03, + "negative_goodness": 0.6116653859668307, + "pair_accuracy": 0.9629333333545261, + "positive_goodness": 5.373381487909953, + "runtime_seconds": 4.062923431396484 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.43222096531126236, + "lr": 0.03, + "negative_goodness": 0.64458302090168, + "pair_accuracy": 0.9613333333333334, + "positive_goodness": 5.2987389734056265, + "runtime_seconds": 3.9239490032196045 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.3955785537560781, + "lr": 0.03, + "negative_goodness": 0.5884488096343147, + "pair_accuracy": 0.9707777777777777, + "positive_goodness": 5.6218479604085285, + "runtime_seconds": 3.8564014434814453 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.4290270928356383, + "lr": 0.03, + "negative_goodness": 0.6101725010050668, + "pair_accuracy": 0.9582222222222222, + "positive_goodness": 5.475240735032823, + "runtime_seconds": 3.8505990505218506 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.41946038685639697, + "lr": 0.03, + "negative_goodness": 0.6068836446656121, + "pair_accuracy": 0.9627111111111111, + "positive_goodness": 5.5583525326199, + "runtime_seconds": 3.8543567657470703 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.4444979146321615, + "lr": 0.03, + "negative_goodness": 0.6325329741795858, + "pair_accuracy": 0.9586444444444444, + "positive_goodness": 5.364362528297637, + "runtime_seconds": 3.858032464981079 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.44069798616833156, + "lr": 0.03, + "negative_goodness": 0.627452710893419, + "pair_accuracy": 0.953111111164093, + "positive_goodness": 5.302849096510146, + "runtime_seconds": 3.8605687618255615 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.41172359619140625, + "lr": 0.03, + "negative_goodness": 0.5818674978892009, + "pair_accuracy": 0.960933333322737, + "positive_goodness": 5.59815097410414, + "runtime_seconds": 3.856013774871826 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.44305318718486364, + "lr": 0.03, + "negative_goodness": 0.6154837809960048, + "pair_accuracy": 0.953, + "positive_goodness": 5.389977985890706, + "runtime_seconds": 3.8575384616851807 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.44391995067596435, + "lr": 0.03, + "negative_goodness": 0.6160402979956733, + "pair_accuracy": 0.950577777756585, + "positive_goodness": 5.370386554718017, + "runtime_seconds": 3.8563365936279297 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.4118602144771152, + "lr": 0.03, + "negative_goodness": 0.5980614718172286, + "pair_accuracy": 0.9630888888888889, + "positive_goodness": 5.6768775515238445, + "runtime_seconds": 3.8624608516693115 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.38235890719625687, + "lr": 0.03, + "negative_goodness": 0.5542257090568542, + "pair_accuracy": 0.9673555555555555, + "positive_goodness": 5.740799926842583, + "runtime_seconds": 3.85457444190979 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.42380321776602004, + "lr": 0.03, + "negative_goodness": 0.6037785151375664, + "pair_accuracy": 0.9600888888676962, + "positive_goodness": 5.565293125576443, + "runtime_seconds": 3.85642147064209 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.43805616834428573, + "lr": 0.03, + "negative_goodness": 0.6248121809562047, + "pair_accuracy": 0.9541555555555555, + "positive_goodness": 5.5280421120537655, + "runtime_seconds": 3.852433919906616 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.43026385887463886, + "lr": 0.03, + "negative_goodness": 0.6134875072214339, + "pair_accuracy": 0.9572, + "positive_goodness": 5.4865948466830785, + "runtime_seconds": 3.8529438972473145 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.42612292109065586, + "lr": 0.03, + "negative_goodness": 0.6118213372018603, + "pair_accuracy": 0.9569777777777778, + "positive_goodness": 5.547145421176486, + "runtime_seconds": 3.9022300243377686 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.44833453663720024, + "lr": 0.03, + "negative_goodness": 0.6359109555986192, + "pair_accuracy": 0.9488222222116258, + "positive_goodness": 5.363537948735555, + "runtime_seconds": 3.910219430923462 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.39738104627132415, + "lr": 0.03, + "negative_goodness": 0.5696806381834878, + "pair_accuracy": 0.9637111111111111, + "positive_goodness": 5.685835601552327, + "runtime_seconds": 3.865561008453369 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.3963717205842336, + "lr": 0.03, + "negative_goodness": 0.5697817426045736, + "pair_accuracy": 0.9676666666666667, + "positive_goodness": 5.820200190056695, + "runtime_seconds": 4.025519609451294 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.4633304613961114, + "lr": 0.03, + "negative_goodness": 0.6604052895651923, + "pair_accuracy": 0.9484666667196485, + "positive_goodness": 5.350510113016765, + "runtime_seconds": 3.936511278152466 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.4176096972200606, + "lr": 0.03, + "negative_goodness": 0.6100497753461201, + "pair_accuracy": 0.9613111111111111, + "positive_goodness": 5.535516904703776, + "runtime_seconds": 3.838644027709961 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.3967975211090512, + "lr": 0.03, + "negative_goodness": 0.5798405318419139, + "pair_accuracy": 0.9669777777777778, + "positive_goodness": 5.747317598639595, + "runtime_seconds": 4.039176940917969 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.3850191614892748, + "lr": 0.03, + "negative_goodness": 0.5582752846399943, + "pair_accuracy": 0.967511111079322, + "positive_goodness": 5.854789434729682, + "runtime_seconds": 4.087815523147583 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.38856920256614685, + "lr": 0.03, + "negative_goodness": 0.561785100279914, + "pair_accuracy": 0.9690666666666666, + "positive_goodness": 5.767497074635823, + "runtime_seconds": 3.9395267963409424 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.372534764152103, + "lr": 0.03, + "negative_goodness": 0.5292157534705269, + "pair_accuracy": 0.9677555555555556, + "positive_goodness": 5.966233510038588, + "runtime_seconds": 3.8350164890289307 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.3893827283196979, + "lr": 0.03, + "negative_goodness": 0.5555531898710463, + "pair_accuracy": 0.9670666666666666, + "positive_goodness": 5.761491807132297, + "runtime_seconds": 3.8568344116210938 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.4788663980854882, + "lr": 0.03, + "negative_goodness": 0.662757611656189, + "pair_accuracy": 0.9472222222116259, + "positive_goodness": 5.3611576752556696, + "runtime_seconds": 4.094604015350342 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.39045151273674433, + "lr": 0.03, + "negative_goodness": 0.5571704278045231, + "pair_accuracy": 0.965, + "positive_goodness": 5.748908765920003, + "runtime_seconds": 4.021821737289429 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.37656840449439155, + "lr": 0.03, + "negative_goodness": 0.5376582881715563, + "pair_accuracy": 0.9679333333015442, + "positive_goodness": 5.959503935496012, + "runtime_seconds": 4.106711149215698 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.4395785480181376, + "lr": 0.03, + "negative_goodness": 0.5974242279820972, + "pair_accuracy": 0.9535333333333333, + "positive_goodness": 5.49539454744127, + "runtime_seconds": 4.100239276885986 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.41272952580981787, + "lr": 0.03, + "negative_goodness": 0.573578044933743, + "pair_accuracy": 0.9560444444762336, + "positive_goodness": 5.7954757748921715, + "runtime_seconds": 4.061889886856079 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.38825500514772204, + "lr": 0.03, + "negative_goodness": 0.5436893267525567, + "pair_accuracy": 0.9621111111111111, + "positive_goodness": 5.86184908879598, + "runtime_seconds": 3.9957008361816406 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.4095791998253928, + "lr": 0.03, + "negative_goodness": 0.5780717928144666, + "pair_accuracy": 0.9560888888888889, + "positive_goodness": 5.6972717988755965, + "runtime_seconds": 3.857776403427124 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.3962808748351203, + "lr": 0.03, + "negative_goodness": 0.5595591606775919, + "pair_accuracy": 0.9651777777777778, + "positive_goodness": 5.816897706010606, + "runtime_seconds": 3.996206760406494 + } + ], + "layer": 15 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1037286151885985, + "lr": 0.03, + "negative_goodness": 1.735691058391995, + "pair_accuracy": 0.830711111108462, + "positive_goodness": 3.837062057601081, + "runtime_seconds": 4.126862287521362 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8588871418529087, + "lr": 0.03, + "negative_goodness": 1.2944788153860305, + "pair_accuracy": 0.8987555555237664, + "positive_goodness": 4.293171892632379, + "runtime_seconds": 4.0779149532318115 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7315956226295895, + "lr": 0.03, + "negative_goodness": 1.0853145355754428, + "pair_accuracy": 0.9222, + "positive_goodness": 4.647820052761502, + "runtime_seconds": 4.1251561641693115 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6619982407357957, + "lr": 0.03, + "negative_goodness": 1.1070927302678426, + "pair_accuracy": 0.9313999999894036, + "positive_goodness": 4.594846694352892, + "runtime_seconds": 3.9373412132263184 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.5750638111856249, + "lr": 0.03, + "negative_goodness": 0.9377299954732259, + "pair_accuracy": 0.9423555555449592, + "positive_goodness": 4.921641605292426, + "runtime_seconds": 3.9327919483184814 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.5742392288049062, + "lr": 0.03, + "negative_goodness": 0.9272008467886182, + "pair_accuracy": 0.9345333333333333, + "positive_goodness": 4.8462381005181205, + "runtime_seconds": 3.933500289916992 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.5599243617110782, + "lr": 0.03, + "negative_goodness": 0.9053771364635892, + "pair_accuracy": 0.9406222222222222, + "positive_goodness": 4.970169921535915, + "runtime_seconds": 3.9310765266418457 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.5326668416500091, + "lr": 0.03, + "negative_goodness": 0.8437669569068484, + "pair_accuracy": 0.9474444444444444, + "positive_goodness": 5.013439182281494, + "runtime_seconds": 3.9390313625335693 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.5460394308990902, + "lr": 0.03, + "negative_goodness": 0.845555419254303, + "pair_accuracy": 0.9432666666666667, + "positive_goodness": 5.065125306955974, + "runtime_seconds": 3.93769907951355 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5330875303321414, + "lr": 0.03, + "negative_goodness": 0.8124970892853207, + "pair_accuracy": 0.9441777777777778, + "positive_goodness": 5.122416002824571, + "runtime_seconds": 3.8959507942199707 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5309866003513336, + "lr": 0.03, + "negative_goodness": 0.8298380652639601, + "pair_accuracy": 0.9489333333333333, + "positive_goodness": 5.096726090833876, + "runtime_seconds": 3.859955310821533 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5383232957628038, + "lr": 0.03, + "negative_goodness": 0.8181714871088663, + "pair_accuracy": 0.9383777777459886, + "positive_goodness": 5.112697885682848, + "runtime_seconds": 3.8623080253601074 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5479773717562357, + "lr": 0.03, + "negative_goodness": 0.838069969092475, + "pair_accuracy": 0.9364888888888889, + "positive_goodness": 4.972529047648112, + "runtime_seconds": 3.9010822772979736 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.520786135398017, + "lr": 0.03, + "negative_goodness": 0.8044685966915555, + "pair_accuracy": 0.9486444444391463, + "positive_goodness": 5.160231547461616, + "runtime_seconds": 3.9002559185028076 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5520766305393643, + "lr": 0.03, + "negative_goodness": 0.8407056345621745, + "pair_accuracy": 0.9341333333863152, + "positive_goodness": 4.9466104638841415, + "runtime_seconds": 3.8662993907928467 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5354720631493463, + "lr": 0.03, + "negative_goodness": 0.8058247510062324, + "pair_accuracy": 0.9435999999788073, + "positive_goodness": 5.106605230161879, + "runtime_seconds": 3.93524432182312 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5473982710520426, + "lr": 0.03, + "negative_goodness": 0.8237621835072835, + "pair_accuracy": 0.9367111111270057, + "positive_goodness": 5.025638469780816, + "runtime_seconds": 3.938058376312256 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5240521452373929, + "lr": 0.03, + "negative_goodness": 0.808945811843872, + "pair_accuracy": 0.9465333333121406, + "positive_goodness": 5.13514719420539, + "runtime_seconds": 3.930907964706421 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5530481625874837, + "lr": 0.03, + "negative_goodness": 0.8190036996629503, + "pair_accuracy": 0.9357555555608538, + "positive_goodness": 5.0518151269488865, + "runtime_seconds": 3.923275947570801 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5178424617555406, + "lr": 0.03, + "negative_goodness": 0.7922904553731283, + "pair_accuracy": 0.9450444444444445, + "positive_goodness": 5.126710022629632, + "runtime_seconds": 3.9474802017211914 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.512259197420544, + "lr": 0.03, + "negative_goodness": 0.7750590584701962, + "pair_accuracy": 0.9474444444444444, + "positive_goodness": 5.2578440849304195, + "runtime_seconds": 3.860387086868286 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5031339485115475, + "lr": 0.03, + "negative_goodness": 0.7453542940563626, + "pair_accuracy": 0.9477777777777778, + "positive_goodness": 5.316610199483236, + "runtime_seconds": 3.865333318710327 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5425699098110199, + "lr": 0.03, + "negative_goodness": 0.7986878082010481, + "pair_accuracy": 0.9396666666666667, + "positive_goodness": 5.159570078870985, + "runtime_seconds": 3.8740484714508057 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.49105331598917645, + "lr": 0.03, + "negative_goodness": 0.7497807261361016, + "pair_accuracy": 0.9568, + "positive_goodness": 5.375977751837836, + "runtime_seconds": 3.9087319374084473 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5327753486421373, + "lr": 0.03, + "negative_goodness": 0.8034203391181098, + "pair_accuracy": 0.946533333349228, + "positive_goodness": 5.237360865359836, + "runtime_seconds": 3.9415297508239746 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.4939642111990187, + "lr": 0.03, + "negative_goodness": 0.7426050375620524, + "pair_accuracy": 0.9537999999682109, + "positive_goodness": 5.30258372569614, + "runtime_seconds": 3.9194276332855225 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5131330797195435, + "lr": 0.03, + "negative_goodness": 0.7747595157835219, + "pair_accuracy": 0.9469333333651225, + "positive_goodness": 5.355814484786987, + "runtime_seconds": 3.9268736839294434 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5239260049608019, + "lr": 0.03, + "negative_goodness": 0.7787311667972141, + "pair_accuracy": 0.9446666666613684, + "positive_goodness": 5.188008808432685, + "runtime_seconds": 3.917374610900879 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5136728825251261, + "lr": 0.03, + "negative_goodness": 0.7285850036833021, + "pair_accuracy": 0.9436000000211927, + "positive_goodness": 5.2877294560750325, + "runtime_seconds": 3.928753137588501 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5313088116063012, + "lr": 0.03, + "negative_goodness": 0.7762623659769694, + "pair_accuracy": 0.9395333333333333, + "positive_goodness": 5.097755367702908, + "runtime_seconds": 3.9291746616363525 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5308934855381647, + "lr": 0.03, + "negative_goodness": 0.7775476137744056, + "pair_accuracy": 0.9412888888888888, + "positive_goodness": 5.331867823791504, + "runtime_seconds": 3.926022529602051 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.4894051421377394, + "lr": 0.03, + "negative_goodness": 0.7350317971759373, + "pair_accuracy": 0.9523333333015442, + "positive_goodness": 5.574683961613973, + "runtime_seconds": 3.93035626411438 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5046816167460547, + "lr": 0.03, + "negative_goodness": 0.7302403922186957, + "pair_accuracy": 0.9442, + "positive_goodness": 5.390593143208822, + "runtime_seconds": 3.937390089035034 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.49669114627838135, + "lr": 0.03, + "negative_goodness": 0.7316012315326267, + "pair_accuracy": 0.9505999999894036, + "positive_goodness": 5.454772280565898, + "runtime_seconds": 3.9452366828918457 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.48576984630160863, + "lr": 0.03, + "negative_goodness": 0.72541312851376, + "pair_accuracy": 0.9542444444603391, + "positive_goodness": 5.593988674439324, + "runtime_seconds": 3.9366493225097656 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5151901540968153, + "lr": 0.03, + "negative_goodness": 0.7566550712691413, + "pair_accuracy": 0.9448222222222222, + "positive_goodness": 5.431130543687608, + "runtime_seconds": 3.946904182434082 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.4867814237515132, + "lr": 0.03, + "negative_goodness": 0.7182685162173377, + "pair_accuracy": 0.9577555555555556, + "positive_goodness": 5.579440804799398, + "runtime_seconds": 3.8974452018737793 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.54305230398443, + "lr": 0.03, + "negative_goodness": 0.8053101126511891, + "pair_accuracy": 0.9395111111111111, + "positive_goodness": 5.255170621405708, + "runtime_seconds": 3.8682634830474854 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.4659505391862657, + "lr": 0.03, + "negative_goodness": 0.678499389690823, + "pair_accuracy": 0.9601555555767484, + "positive_goodness": 5.73503406706916, + "runtime_seconds": 3.8704633712768555 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.48317568202548555, + "lr": 0.03, + "negative_goodness": 0.7059040532430013, + "pair_accuracy": 0.9540888888888889, + "positive_goodness": 5.684920119306776, + "runtime_seconds": 3.8675014972686768 + } + ], + "layer": 16 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.7586615669674344, + "lr": 0.03, + "negative_goodness": 1.0916804257922703, + "pair_accuracy": 0.9087777778095669, + "positive_goodness": 4.250027406671312, + "runtime_seconds": 3.8413751125335693 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.677990246603224, + "lr": 0.03, + "negative_goodness": 1.038208615451389, + "pair_accuracy": 0.9208888888676962, + "positive_goodness": 4.640136895031399, + "runtime_seconds": 3.924159288406372 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.665008257039388, + "lr": 0.03, + "negative_goodness": 1.0324396991729736, + "pair_accuracy": 0.9220666666719649, + "positive_goodness": 4.753602556800843, + "runtime_seconds": 3.911895275115967 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6081227573606703, + "lr": 0.03, + "negative_goodness": 0.9626450615353055, + "pair_accuracy": 0.9412, + "positive_goodness": 4.908655993991428, + "runtime_seconds": 3.9062507152557373 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.5654408628198836, + "lr": 0.03, + "negative_goodness": 0.892465837044186, + "pair_accuracy": 0.9483555555555555, + "positive_goodness": 5.364968763054741, + "runtime_seconds": 3.9162437915802 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.5766828758504655, + "lr": 0.03, + "negative_goodness": 0.9025845632129246, + "pair_accuracy": 0.9469555555555556, + "positive_goodness": 5.153242272186279, + "runtime_seconds": 3.9091036319732666 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.5997738396750556, + "lr": 0.03, + "negative_goodness": 0.927967771297031, + "pair_accuracy": 0.9350444444444445, + "positive_goodness": 5.092716813871595, + "runtime_seconds": 3.9133336544036865 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.575873211812973, + "lr": 0.03, + "negative_goodness": 0.8906434065288967, + "pair_accuracy": 0.9438888888888889, + "positive_goodness": 5.097391577233209, + "runtime_seconds": 3.907472610473633 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.5493676128387451, + "lr": 0.03, + "negative_goodness": 0.8714729200363159, + "pair_accuracy": 0.9527333333757189, + "positive_goodness": 5.279890227296617, + "runtime_seconds": 3.912637710571289 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5550840728759766, + "lr": 0.03, + "negative_goodness": 0.8720231152852377, + "pair_accuracy": 0.9490888888888889, + "positive_goodness": 5.18629994925393, + "runtime_seconds": 3.8823142051696777 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5674689796129863, + "lr": 0.03, + "negative_goodness": 0.8695816284391615, + "pair_accuracy": 0.9381777777459887, + "positive_goodness": 5.120570919715034, + "runtime_seconds": 3.916275978088379 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5543431355423397, + "lr": 0.03, + "negative_goodness": 0.8568709933280945, + "pair_accuracy": 0.9465333333227369, + "positive_goodness": 5.3013614087422685, + "runtime_seconds": 3.915222644805908 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5346418526649475, + "lr": 0.03, + "negative_goodness": 0.8228881141079797, + "pair_accuracy": 0.9497333333333333, + "positive_goodness": 5.474570681762695, + "runtime_seconds": 3.9182536602020264 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5305073735766941, + "lr": 0.03, + "negative_goodness": 0.8096861326641507, + "pair_accuracy": 0.9500888889312744, + "positive_goodness": 5.7211670981089275, + "runtime_seconds": 3.919935703277588 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5643990787929959, + "lr": 0.03, + "negative_goodness": 0.8612753474341498, + "pair_accuracy": 0.9458, + "positive_goodness": 5.213083960639106, + "runtime_seconds": 3.9164929389953613 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5615045519193014, + "lr": 0.03, + "negative_goodness": 0.8571677096048991, + "pair_accuracy": 0.9465111111217075, + "positive_goodness": 5.204213348558214, + "runtime_seconds": 3.9158213138580322 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5322039067983627, + "lr": 0.03, + "negative_goodness": 0.8142418759610918, + "pair_accuracy": 0.9532666666666667, + "positive_goodness": 5.363461328464084, + "runtime_seconds": 3.9106087684631348 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5342219806777107, + "lr": 0.03, + "negative_goodness": 0.8096399970610937, + "pair_accuracy": 0.9488222222222222, + "positive_goodness": 5.36584740041097, + "runtime_seconds": 3.9353952407836914 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5398867576069302, + "lr": 0.03, + "negative_goodness": 0.8226270603285896, + "pair_accuracy": 0.9465333333333333, + "positive_goodness": 5.349246341196696, + "runtime_seconds": 3.9077253341674805 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5554774314244588, + "lr": 0.03, + "negative_goodness": 0.8395211765713162, + "pair_accuracy": 0.9442888889100817, + "positive_goodness": 5.4120756100124785, + "runtime_seconds": 3.9449379444122314 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5330274446275499, + "lr": 0.03, + "negative_goodness": 0.8118214014265273, + "pair_accuracy": 0.95, + "positive_goodness": 5.495063440450033, + "runtime_seconds": 3.920408248901367 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5230589733335707, + "lr": 0.03, + "negative_goodness": 0.7914677900526259, + "pair_accuracy": 0.9504888889153799, + "positive_goodness": 5.567852670563592, + "runtime_seconds": 3.9231085777282715 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5355902297284868, + "lr": 0.03, + "negative_goodness": 0.7926053251584371, + "pair_accuracy": 0.9438444444232517, + "positive_goodness": 5.483232170274523, + "runtime_seconds": 3.936486005783081 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5063716541290283, + "lr": 0.03, + "negative_goodness": 0.7821036146640777, + "pair_accuracy": 0.9559111111111112, + "positive_goodness": 5.5131356434292265, + "runtime_seconds": 3.9241185188293457 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5512631209638383, + "lr": 0.03, + "negative_goodness": 0.8129799520227644, + "pair_accuracy": 0.9451777777777778, + "positive_goodness": 5.345480949910482, + "runtime_seconds": 3.856553792953491 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5341362471845414, + "lr": 0.03, + "negative_goodness": 0.7949520338800219, + "pair_accuracy": 0.9437555555555556, + "positive_goodness": 5.522969711303711, + "runtime_seconds": 3.8554799556732178 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5331730677975549, + "lr": 0.03, + "negative_goodness": 0.7955434938483769, + "pair_accuracy": 0.9489777777777778, + "positive_goodness": 5.387353615654839, + "runtime_seconds": 3.847607135772705 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5493709773328569, + "lr": 0.03, + "negative_goodness": 0.7981132944160038, + "pair_accuracy": 0.9417111111111112, + "positive_goodness": 5.459244980282254, + "runtime_seconds": 3.8501980304718018 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5543026714033551, + "lr": 0.03, + "negative_goodness": 0.8178763434622023, + "pair_accuracy": 0.9414666666666667, + "positive_goodness": 5.316324805619981, + "runtime_seconds": 3.8513782024383545 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.4939218893369039, + "lr": 0.03, + "negative_goodness": 0.7453480491214328, + "pair_accuracy": 0.9584222222222222, + "positive_goodness": 5.7769214179144965, + "runtime_seconds": 3.865675926208496 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5152683940463596, + "lr": 0.03, + "negative_goodness": 0.7800449671851264, + "pair_accuracy": 0.9518666666560703, + "positive_goodness": 5.806410187106662, + "runtime_seconds": 3.8533811569213867 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5127227973514134, + "lr": 0.03, + "negative_goodness": 0.7604451662805345, + "pair_accuracy": 0.9573333333333334, + "positive_goodness": 5.582389385732015, + "runtime_seconds": 3.853175640106201 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5290071175893147, + "lr": 0.03, + "negative_goodness": 0.7957968071407742, + "pair_accuracy": 0.9529777778307597, + "positive_goodness": 5.592521416812473, + "runtime_seconds": 3.8508284091949463 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.4685826285203298, + "lr": 0.03, + "negative_goodness": 0.7164890910360548, + "pair_accuracy": 0.9639333333333333, + "positive_goodness": 5.908779762437608, + "runtime_seconds": 3.914210319519043 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5018731364674038, + "lr": 0.03, + "negative_goodness": 0.7547937106238471, + "pair_accuracy": 0.9528222222116258, + "positive_goodness": 5.617706099361843, + "runtime_seconds": 3.85937237739563 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.47497297270033095, + "lr": 0.03, + "negative_goodness": 0.720595511118571, + "pair_accuracy": 0.9626000000211927, + "positive_goodness": 6.028655215072632, + "runtime_seconds": 3.9092907905578613 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5243999756601122, + "lr": 0.03, + "negative_goodness": 0.7624058069970873, + "pair_accuracy": 0.9500888888888889, + "positive_goodness": 5.697151492140028, + "runtime_seconds": 3.921797752380371 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.521622302447425, + "lr": 0.03, + "negative_goodness": 0.7687800934261746, + "pair_accuracy": 0.9544666667196485, + "positive_goodness": 5.54342650718689, + "runtime_seconds": 3.848965883255005 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.4913499216609531, + "lr": 0.03, + "negative_goodness": 0.7406809664620293, + "pair_accuracy": 0.9587777777671814, + "positive_goodness": 5.926230840471056, + "runtime_seconds": 3.8797240257263184 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5429708968692356, + "lr": 0.03, + "negative_goodness": 0.7853209460788303, + "pair_accuracy": 0.9461555555555555, + "positive_goodness": 5.379881164635552, + "runtime_seconds": 3.8645825386047363 + } + ], + "layer": 17 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8259686123847961, + "lr": 0.03, + "negative_goodness": 1.3389674243927001, + "pair_accuracy": 0.9036222222010295, + "positive_goodness": 4.806562017440796, + "runtime_seconds": 3.8636341094970703 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.5788217274241977, + "lr": 0.03, + "negative_goodness": 0.9419700681262546, + "pair_accuracy": 0.9484222222222223, + "positive_goodness": 5.49278571150038, + "runtime_seconds": 3.8657214641571045 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.5452739104800755, + "lr": 0.03, + "negative_goodness": 0.8748219245910644, + "pair_accuracy": 0.9482444444444444, + "positive_goodness": 5.398229738532172, + "runtime_seconds": 3.854307174682617 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.45720737796094685, + "lr": 0.03, + "negative_goodness": 0.7167868900378546, + "pair_accuracy": 0.9614222222116259, + "positive_goodness": 5.8161153249104816, + "runtime_seconds": 3.8811235427856445 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.469736750263638, + "lr": 0.03, + "negative_goodness": 0.7384170273621877, + "pair_accuracy": 0.9646, + "positive_goodness": 5.531330157894558, + "runtime_seconds": 3.881183624267578 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.41790379668341743, + "lr": 0.03, + "negative_goodness": 0.6347421670542823, + "pair_accuracy": 0.9651777777777778, + "positive_goodness": 6.030887801615397, + "runtime_seconds": 4.097397327423096 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.4006347717497084, + "lr": 0.03, + "negative_goodness": 0.6212350602679783, + "pair_accuracy": 0.9732222222752042, + "positive_goodness": 6.3339286349402535, + "runtime_seconds": 3.9917373657226562 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.46259186188644835, + "lr": 0.03, + "negative_goodness": 0.6732803741746478, + "pair_accuracy": 0.9578, + "positive_goodness": 5.527439904954698, + "runtime_seconds": 3.919882297515869 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.36039457421567705, + "lr": 0.03, + "negative_goodness": 0.5354044926855299, + "pair_accuracy": 0.9762, + "positive_goodness": 6.4684110660976835, + "runtime_seconds": 3.93808913230896 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.4191694047821893, + "lr": 0.03, + "negative_goodness": 0.614789413494534, + "pair_accuracy": 0.9667555555979411, + "positive_goodness": 5.857750079939101, + "runtime_seconds": 3.918116331100464 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.37569413325256773, + "lr": 0.03, + "negative_goodness": 0.5344191539525985, + "pair_accuracy": 0.9682222222222222, + "positive_goodness": 6.470014559766981, + "runtime_seconds": 3.913426637649536 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.3965109835863113, + "lr": 0.03, + "negative_goodness": 0.5750921640978919, + "pair_accuracy": 0.9684888888888888, + "positive_goodness": 6.197246535746257, + "runtime_seconds": 3.8500475883483887 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.3876774400419659, + "lr": 0.03, + "negative_goodness": 0.547981151898702, + "pair_accuracy": 0.9671777777777778, + "positive_goodness": 6.139969396633572, + "runtime_seconds": 3.8462307453155518 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.3792311737007565, + "lr": 0.03, + "negative_goodness": 0.5350332353671392, + "pair_accuracy": 0.9701333333333333, + "positive_goodness": 6.3542265944586855, + "runtime_seconds": 3.8593735694885254 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.3618579012711843, + "lr": 0.03, + "negative_goodness": 0.5199745007316271, + "pair_accuracy": 0.9734888888888888, + "positive_goodness": 6.281533173116048, + "runtime_seconds": 3.870372772216797 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.39634108844333227, + "lr": 0.03, + "negative_goodness": 0.5631719219578637, + "pair_accuracy": 0.9643111111111111, + "positive_goodness": 6.278186319393582, + "runtime_seconds": 3.8494772911071777 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.3803172600640191, + "lr": 0.03, + "negative_goodness": 0.5487459373447631, + "pair_accuracy": 0.9673777777777778, + "positive_goodness": 6.219796577623155, + "runtime_seconds": 3.8615305423736572 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.4019638591342502, + "lr": 0.03, + "negative_goodness": 0.5593759052700467, + "pair_accuracy": 0.9621555555555555, + "positive_goodness": 6.256822263675266, + "runtime_seconds": 3.8488924503326416 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.3416882675250371, + "lr": 0.03, + "negative_goodness": 0.49287176938586763, + "pair_accuracy": 0.9773333333333334, + "positive_goodness": 6.576029263729519, + "runtime_seconds": 3.8501155376434326 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.3620601107623842, + "lr": 0.03, + "negative_goodness": 0.5124692486286163, + "pair_accuracy": 0.9753777777777778, + "positive_goodness": 6.3990465777926975, + "runtime_seconds": 3.8538014888763428 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.4049675173494551, + "lr": 0.03, + "negative_goodness": 0.5472224340756734, + "pair_accuracy": 0.9586888888888889, + "positive_goodness": 6.37498036066691, + "runtime_seconds": 3.8620553016662598 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.33891054705778756, + "lr": 0.03, + "negative_goodness": 0.4881928093380398, + "pair_accuracy": 0.9782, + "positive_goodness": 6.537544328308106, + "runtime_seconds": 3.8455708026885986 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.3837569868829515, + "lr": 0.03, + "negative_goodness": 0.5320657742924161, + "pair_accuracy": 0.9676666666666667, + "positive_goodness": 6.2128113115946455, + "runtime_seconds": 3.8544373512268066 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.32475028533935546, + "lr": 0.03, + "negative_goodness": 0.4454149994638231, + "pair_accuracy": 0.9772888888888889, + "positive_goodness": 6.929710346137153, + "runtime_seconds": 3.8468000888824463 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.3527931089639664, + "lr": 0.03, + "negative_goodness": 0.4730946593635612, + "pair_accuracy": 0.9728, + "positive_goodness": 6.8493069038391114, + "runtime_seconds": 3.8451690673828125 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.34312050483756595, + "lr": 0.03, + "negative_goodness": 0.4535034031060007, + "pair_accuracy": 0.9716, + "positive_goodness": 6.763877388848199, + "runtime_seconds": 3.8402135372161865 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.3597784379296833, + "lr": 0.03, + "negative_goodness": 0.4927623445802265, + "pair_accuracy": 0.9721333333333333, + "positive_goodness": 6.5638175576951765, + "runtime_seconds": 3.8595786094665527 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.3394632726510366, + "lr": 0.03, + "negative_goodness": 0.4670485525290171, + "pair_accuracy": 0.9775333333333334, + "positive_goodness": 6.526536104668511, + "runtime_seconds": 3.925543785095215 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.3669456438673867, + "lr": 0.03, + "negative_goodness": 0.4731154606872135, + "pair_accuracy": 0.9623333333333334, + "positive_goodness": 6.5778701544019915, + "runtime_seconds": 3.931903123855591 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.3297052407688565, + "lr": 0.03, + "negative_goodness": 0.43899190334214105, + "pair_accuracy": 0.9747555555449592, + "positive_goodness": 7.103966812939114, + "runtime_seconds": 3.933610677719116 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.3071729453722636, + "lr": 0.03, + "negative_goodness": 0.4208682724263933, + "pair_accuracy": 0.9811555555555556, + "positive_goodness": 6.976453070661757, + "runtime_seconds": 3.9413397312164307 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.35308734004232617, + "lr": 0.03, + "negative_goodness": 0.4886600203990936, + "pair_accuracy": 0.9735111111005147, + "positive_goodness": 6.51723375278049, + "runtime_seconds": 3.9345340728759766 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.34387368368572657, + "lr": 0.03, + "negative_goodness": 0.43323115503523085, + "pair_accuracy": 0.9716, + "positive_goodness": 6.770022542995877, + "runtime_seconds": 3.921149969100952 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.38037150200737846, + "lr": 0.03, + "negative_goodness": 0.5160685161060757, + "pair_accuracy": 0.9654222222222222, + "positive_goodness": 6.391655477227105, + "runtime_seconds": 3.9093334674835205 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.31865622033013236, + "lr": 0.03, + "negative_goodness": 0.43610168860091103, + "pair_accuracy": 0.9794222222222222, + "positive_goodness": 6.734662050543891, + "runtime_seconds": 3.9346818923950195 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.3391467007901933, + "lr": 0.03, + "negative_goodness": 0.4630040694342719, + "pair_accuracy": 0.9745555555555555, + "positive_goodness": 6.817466590457492, + "runtime_seconds": 3.9160642623901367 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.331196120707194, + "lr": 0.03, + "negative_goodness": 0.4256816998852624, + "pair_accuracy": 0.9733777777777778, + "positive_goodness": 6.937445837815603, + "runtime_seconds": 3.880622625350952 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.3578230931599935, + "lr": 0.03, + "negative_goodness": 0.4748589723057217, + "pair_accuracy": 0.9688222222328186, + "positive_goodness": 6.682817717255486, + "runtime_seconds": 3.8480892181396484 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.28852497049702536, + "lr": 0.03, + "negative_goodness": 0.3886095251321793, + "pair_accuracy": 0.9874888888888889, + "positive_goodness": 7.2759385794745555, + "runtime_seconds": 3.8858482837677 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.3394897342628903, + "lr": 0.03, + "negative_goodness": 0.4562199644751019, + "pair_accuracy": 0.9749555555555556, + "positive_goodness": 6.458949321492513, + "runtime_seconds": 3.9331071376800537 + } + ], + "layer": 18 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8378556284162734, + "lr": 0.03, + "negative_goodness": 0.8190931993140115, + "pair_accuracy": 0.9032444444444444, + "positive_goodness": 4.4816375066280365, + "runtime_seconds": 3.6656405925750732 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.793786353026496, + "lr": 0.03, + "negative_goodness": 0.6602042104509142, + "pair_accuracy": 0.926400000052982, + "positive_goodness": 5.252249863974253, + "runtime_seconds": 3.638366937637329 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7150230112287733, + "lr": 0.03, + "negative_goodness": 0.6579606157302856, + "pair_accuracy": 0.9446444444762336, + "positive_goodness": 5.449562658172184, + "runtime_seconds": 3.688756227493286 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7272881580776639, + "lr": 0.03, + "negative_goodness": 0.6310853884696961, + "pair_accuracy": 0.9392444444126553, + "positive_goodness": 5.403566014586555, + "runtime_seconds": 3.660064697265625 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8136597938987944, + "lr": 0.03, + "negative_goodness": 0.8834357947243584, + "pair_accuracy": 0.9400222222222222, + "positive_goodness": 4.934488987308078, + "runtime_seconds": 3.666287660598755 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.792785745271047, + "lr": 0.03, + "negative_goodness": 0.796174953142802, + "pair_accuracy": 0.943, + "positive_goodness": 4.682592287784153, + "runtime_seconds": 3.6586015224456787 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7878554633590911, + "lr": 0.03, + "negative_goodness": 0.8333511306232876, + "pair_accuracy": 0.9417111111111112, + "positive_goodness": 5.0035317302280005, + "runtime_seconds": 3.693408966064453 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7611826844321357, + "lr": 0.03, + "negative_goodness": 0.797366199795405, + "pair_accuracy": 0.9456666666560702, + "positive_goodness": 4.750197190602621, + "runtime_seconds": 3.6431591510772705 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.710203143035041, + "lr": 0.03, + "negative_goodness": 0.7506467061678569, + "pair_accuracy": 0.9620222222116258, + "positive_goodness": 5.208370551088121, + "runtime_seconds": 3.662313461303711 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8188559944788615, + "lr": 0.03, + "negative_goodness": 0.8794479369693332, + "pair_accuracy": 0.9532222222222222, + "positive_goodness": 4.602978541819255, + "runtime_seconds": 3.670264959335327 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7283904992262522, + "lr": 0.03, + "negative_goodness": 0.7886248685306972, + "pair_accuracy": 0.9553111111111111, + "positive_goodness": 5.024405931684706, + "runtime_seconds": 3.623258352279663 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7273276751677196, + "lr": 0.03, + "negative_goodness": 0.7504454702642229, + "pair_accuracy": 0.9398222222222222, + "positive_goodness": 5.26897135603163, + "runtime_seconds": 3.6219589710235596 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7431851822535197, + "lr": 0.03, + "negative_goodness": 0.817262694782681, + "pair_accuracy": 0.9430888888994853, + "positive_goodness": 4.927849314382341, + "runtime_seconds": 3.6218836307525635 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7888229042927424, + "lr": 0.03, + "negative_goodness": 0.8557820429590013, + "pair_accuracy": 0.9523111111111111, + "positive_goodness": 4.649183097966512, + "runtime_seconds": 3.622542142868042 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.790370363468594, + "lr": 0.03, + "negative_goodness": 0.8172142630471123, + "pair_accuracy": 0.9441777778307597, + "positive_goodness": 4.7578598287794325, + "runtime_seconds": 3.6189162731170654 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7213186428281996, + "lr": 0.03, + "negative_goodness": 0.8018780158148872, + "pair_accuracy": 0.9576888888888889, + "positive_goodness": 4.893322066201104, + "runtime_seconds": 3.643550395965576 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.719649231531885, + "lr": 0.03, + "negative_goodness": 0.7912019467406802, + "pair_accuracy": 0.9655111111111111, + "positive_goodness": 5.176326021364, + "runtime_seconds": 3.6200449466705322 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.762235623354382, + "lr": 0.03, + "negative_goodness": 0.7961893403265211, + "pair_accuracy": 0.9496888888888889, + "positive_goodness": 4.963705967288547, + "runtime_seconds": 3.623379707336426 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7888209527810415, + "lr": 0.03, + "negative_goodness": 0.940478202925788, + "pair_accuracy": 0.9644222222222222, + "positive_goodness": 4.398103315819634, + "runtime_seconds": 3.625978708267212 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6804959466298421, + "lr": 0.03, + "negative_goodness": 0.8154352258258396, + "pair_accuracy": 0.9699555555608538, + "positive_goodness": 5.2733856780582, + "runtime_seconds": 3.6139540672302246 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7537416289859348, + "lr": 0.03, + "negative_goodness": 0.8231160748852624, + "pair_accuracy": 0.9586666666878594, + "positive_goodness": 4.74653024031851, + "runtime_seconds": 3.617305040359497 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7601175495412614, + "lr": 0.03, + "negative_goodness": 0.8482662359343635, + "pair_accuracy": 0.9593555555555555, + "positive_goodness": 5.0593434859805635, + "runtime_seconds": 3.620288610458374 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7552843845208486, + "lr": 0.03, + "negative_goodness": 0.8298653539551629, + "pair_accuracy": 0.9546222222222223, + "positive_goodness": 4.979660187827216, + "runtime_seconds": 3.6558327674865723 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7132218266805013, + "lr": 0.03, + "negative_goodness": 0.7899800141970317, + "pair_accuracy": 0.9603555555343628, + "positive_goodness": 5.05491555112203, + "runtime_seconds": 3.66025447845459 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6736454852210151, + "lr": 0.03, + "negative_goodness": 0.7284934794690874, + "pair_accuracy": 0.9603111111323038, + "positive_goodness": 5.179482285955217, + "runtime_seconds": 3.6590731143951416 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7802763431125217, + "lr": 0.03, + "negative_goodness": 0.8739584117412568, + "pair_accuracy": 0.9496666666666667, + "positive_goodness": 4.604382725000382, + "runtime_seconds": 3.6600799560546875 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7418865736219618, + "lr": 0.03, + "negative_goodness": 0.8144318610509237, + "pair_accuracy": 0.9569777777671814, + "positive_goodness": 4.733769334030152, + "runtime_seconds": 3.663022041320801 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.720155341498057, + "lr": 0.03, + "negative_goodness": 0.8167131082746718, + "pair_accuracy": 0.9581111111111111, + "positive_goodness": 4.770802198876275, + "runtime_seconds": 3.6647558212280273 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.7766031409396066, + "lr": 0.03, + "negative_goodness": 0.8580984510845608, + "pair_accuracy": 0.9486222222222223, + "positive_goodness": 4.6893890176985, + "runtime_seconds": 3.65832781791687 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.8182663829273648, + "lr": 0.03, + "negative_goodness": 0.9199759199460348, + "pair_accuracy": 0.9460444444444445, + "positive_goodness": 4.3285252565171985, + "runtime_seconds": 3.779449462890625 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.7592799105008443, + "lr": 0.03, + "negative_goodness": 0.852420579984453, + "pair_accuracy": 0.9513111110899184, + "positive_goodness": 4.933446850967408, + "runtime_seconds": 3.670698404312134 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.718013380908966, + "lr": 0.03, + "negative_goodness": 0.8205504271719191, + "pair_accuracy": 0.9663555556085375, + "positive_goodness": 4.834306415685018, + "runtime_seconds": 3.658837080001831 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6914673953586155, + "lr": 0.03, + "negative_goodness": 0.7858992307821909, + "pair_accuracy": 0.9560888888994853, + "positive_goodness": 5.414765327805943, + "runtime_seconds": 3.657630681991577 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.7121127066877153, + "lr": 0.03, + "negative_goodness": 0.7977258121066624, + "pair_accuracy": 0.9614888888888888, + "positive_goodness": 4.991760984590319, + "runtime_seconds": 3.6261799335479736 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.7520923722293642, + "lr": 0.03, + "negative_goodness": 0.8486520180649227, + "pair_accuracy": 0.9546666666666667, + "positive_goodness": 4.7654911400265165, + "runtime_seconds": 3.6199333667755127 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.7217549567752414, + "lr": 0.03, + "negative_goodness": 0.8545575268745422, + "pair_accuracy": 0.9610444444523917, + "positive_goodness": 4.983532989623812, + "runtime_seconds": 3.61509370803833 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.7317499878406525, + "lr": 0.03, + "negative_goodness": 0.821341920264562, + "pair_accuracy": 0.9602222222222222, + "positive_goodness": 4.659991877237956, + "runtime_seconds": 3.615964889526367 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.7587645548264186, + "lr": 0.03, + "negative_goodness": 0.8634413420465258, + "pair_accuracy": 0.9634222222222222, + "positive_goodness": 4.76607735824585, + "runtime_seconds": 3.6209447383880615 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.7140099500762092, + "lr": 0.03, + "negative_goodness": 0.8132642613516914, + "pair_accuracy": 0.9644222222222222, + "positive_goodness": 4.859307693311903, + "runtime_seconds": 3.6259891986846924 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.7358407176097234, + "lr": 0.03, + "negative_goodness": 0.842765372882949, + "pair_accuracy": 0.9631333333333333, + "positive_goodness": 4.857692943572998, + "runtime_seconds": 3.628889322280884 + } + ], + "layer": 19 + } + ], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 3180.144710302353, + "work": { + "candidate_label_evaluation_presentations": 50000, + "completed_global_epochs": 800, + "feedforward_example_passes": 72000000, + "forward_macs_per_example": 40551040, + "forward_parameter_count": 269722, + "local_target_backward_example_evaluations": 72000000, + "local_vjp_example_evaluations": 0, + "logical_task_loss_queries": 0, + "num_trainable_layers": 20, + "ordinary_training_examples": 36000000, + "ordinary_validation_examples": 5000, + "relaxation_example_passes": 0, + "training_example_presentations": 72000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json.manifest.json new file mode 100644 index 0000000..11a25f7 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "ff", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "40", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.03", + "--output_lr", + "0.03", + "--lr_schedule", + "constant", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--ff_threshold", + "2.0", + "--ff_score_from_layer", + "1" + ], + "completed_unix_time": 1787002781.8964257, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 3187.272476911545, + "epochs": 40, + "experiment_name": "resnet-r2-ff-d20-lr0p03", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "constant", + "method": "ff", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d20-lr0p03.json", + "output_exists": true, + "output_sha256": "e6cde27ffd50512b43afb18b6ddccdf1a49b2185001bf6740c96ec84ee7a89ec", + "rate": 0.03, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json b/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json new file mode 100644 index 0000000..05436e7 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json @@ -0,0 +1,13125 @@ +{ + "architecture": { + "base_width": 16, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 464154, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 40, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "constant", + "max_steps": 0, + "method": "ff", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json", + "output_lr": 0.03, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.1228, + "evaluation_seconds": 2.436674118041992, + "examples": 5000, + "finite": true, + "loss": null + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 8251766272, + "peak_memory_reserved_bytes": 8801746944, + "torch_version": "2.3.1+cu118" + }, + "layers": [ + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.4005121707492405, + "lr": 0.03, + "negative_goodness": 1.9601310303582085, + "pair_accuracy": 0.4738666666560703, + "positive_goodness": 1.9601432648976644, + "runtime_seconds": 7.918581247329712 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3870076318528917, + "lr": 0.03, + "negative_goodness": 1.9987232434590658, + "pair_accuracy": 0.4773555555820465, + "positive_goodness": 1.998744419225057, + "runtime_seconds": 5.874989986419678 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3869588137732611, + "lr": 0.03, + "negative_goodness": 1.9987902426401774, + "pair_accuracy": 0.48162222224871315, + "positive_goodness": 1.9988118033515083, + "runtime_seconds": 5.86069130897522 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3869192726771036, + "lr": 0.03, + "negative_goodness": 1.9989530081007216, + "pair_accuracy": 0.47257777780956695, + "positive_goodness": 1.9989705465952556, + "runtime_seconds": 5.860997200012207 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3867757738537259, + "lr": 0.03, + "negative_goodness": 1.9991573658413357, + "pair_accuracy": 0.475644444460339, + "positive_goodness": 1.999177642610338, + "runtime_seconds": 5.855912923812866 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3866989333258735, + "lr": 0.03, + "negative_goodness": 1.9993364679124621, + "pair_accuracy": 0.477755555539661, + "positive_goodness": 1.9993548810746935, + "runtime_seconds": 5.8581647872924805 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3865385059144761, + "lr": 0.03, + "negative_goodness": 1.9995932420094809, + "pair_accuracy": 0.4741111111164093, + "positive_goodness": 1.9996111384497748, + "runtime_seconds": 5.860776901245117 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.386432902124193, + "lr": 0.03, + "negative_goodness": 1.9997635799196032, + "pair_accuracy": 0.47753333332273695, + "positive_goodness": 1.9997798184500801, + "runtime_seconds": 5.863062858581543 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3863745677524142, + "lr": 0.03, + "negative_goodness": 1.99987327164544, + "pair_accuracy": 0.48231111111111113, + "positive_goodness": 1.9998874392615424, + "runtime_seconds": 5.861711025238037 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3863343165079753, + "lr": 0.03, + "negative_goodness": 1.999927060317993, + "pair_accuracy": 0.4777555555237664, + "positive_goodness": 1.9999369669596354, + "runtime_seconds": 5.872966766357422 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.386317720519172, + "lr": 0.03, + "negative_goodness": 1.9999535439385308, + "pair_accuracy": 0.48062222224871315, + "positive_goodness": 1.999962510723538, + "runtime_seconds": 5.875638246536255 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3863131285349528, + "lr": 0.03, + "negative_goodness": 1.9999653892940945, + "pair_accuracy": 0.47864444444444443, + "positive_goodness": 1.9999740058686999, + "runtime_seconds": 5.877273082733154 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.386307242880927, + "lr": 0.03, + "negative_goodness": 1.9999784241570366, + "pair_accuracy": 0.48011111108991833, + "positive_goodness": 1.9999848674562242, + "runtime_seconds": 5.877172946929932 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3863020617379083, + "lr": 0.03, + "negative_goodness": 1.9999842087851631, + "pair_accuracy": 0.4804222222222222, + "positive_goodness": 1.9999912102805244, + "runtime_seconds": 5.881123781204224 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.386299115286933, + "lr": 0.03, + "negative_goodness": 1.9999898784425523, + "pair_accuracy": 0.4792444444762336, + "positive_goodness": 1.999996218193902, + "runtime_seconds": 5.882385730743408 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3862972230911255, + "lr": 0.03, + "negative_goodness": 1.9999917033513388, + "pair_accuracy": 0.4775777777671814, + "positive_goodness": 1.9999973849190607, + "runtime_seconds": 5.885278940200806 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.386295739004347, + "lr": 0.03, + "negative_goodness": 1.9999922837363349, + "pair_accuracy": 0.47706666667726305, + "positive_goodness": 1.9999996148639254, + "runtime_seconds": 5.885771036148071 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3862954873614841, + "lr": 0.03, + "negative_goodness": 1.9999921279059516, + "pair_accuracy": 0.47224444446033903, + "positive_goodness": 2.0000018464830185, + "runtime_seconds": 5.887912750244141 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.386294015375773, + "lr": 0.03, + "negative_goodness": 1.9999834595150419, + "pair_accuracy": 0.46633333332273696, + "positive_goodness": 2.000004312769572, + "runtime_seconds": 5.885218143463135 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3862846208996242, + "lr": 0.03, + "negative_goodness": 1.9999540635638766, + "pair_accuracy": 0.4766888888729943, + "positive_goodness": 2.0000266984091866, + "runtime_seconds": 5.886136770248413 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3862511160532633, + "lr": 0.03, + "negative_goodness": 1.9998888256920708, + "pair_accuracy": 0.48159999996821085, + "positive_goodness": 2.000104170015123, + "runtime_seconds": 5.881669521331787 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.3862158044815063, + "lr": 0.03, + "negative_goodness": 1.9998076720343696, + "pair_accuracy": 0.47960000002649095, + "positive_goodness": 2.000140270190769, + "runtime_seconds": 5.882010221481323 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3861503973431057, + "lr": 0.03, + "negative_goodness": 1.9997307079103257, + "pair_accuracy": 0.47773333332273693, + "positive_goodness": 2.0002485545688207, + "runtime_seconds": 5.884494304656982 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3860986304601033, + "lr": 0.03, + "negative_goodness": 1.999570548502604, + "pair_accuracy": 0.48088888891008164, + "positive_goodness": 2.0003663606855606, + "runtime_seconds": 5.8838465213775635 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3860176994959514, + "lr": 0.03, + "negative_goodness": 1.9994311214023166, + "pair_accuracy": 0.4760666666507721, + "positive_goodness": 2.0004171453264026, + "runtime_seconds": 5.883695363998413 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3858704584757486, + "lr": 0.03, + "negative_goodness": 1.9991913842307196, + "pair_accuracy": 0.4773111111164093, + "positive_goodness": 2.0007020690282187, + "runtime_seconds": 5.884255886077881 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3857602746115791, + "lr": 0.03, + "negative_goodness": 1.9989943292829726, + "pair_accuracy": 0.47764444442325166, + "positive_goodness": 2.0009385062323677, + "runtime_seconds": 5.881941556930542 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3858194495307075, + "lr": 0.03, + "negative_goodness": 1.9991373915354411, + "pair_accuracy": 0.47948888891537983, + "positive_goodness": 2.0008429104275174, + "runtime_seconds": 5.881671905517578 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.38555989718967, + "lr": 0.03, + "negative_goodness": 1.9986262475967407, + "pair_accuracy": 0.4800222222328186, + "positive_goodness": 2.0012139696757, + "runtime_seconds": 5.884804964065552 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3854964096069335, + "lr": 0.03, + "negative_goodness": 1.9987245820151436, + "pair_accuracy": 0.48113333338631525, + "positive_goodness": 2.0014742348564996, + "runtime_seconds": 5.883755922317505 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.385248271285163, + "lr": 0.03, + "negative_goodness": 1.9981655681186252, + "pair_accuracy": 0.48097777778837414, + "positive_goodness": 2.001850898785061, + "runtime_seconds": 5.889410018920898 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.385392057991028, + "lr": 0.03, + "negative_goodness": 1.9981381525039672, + "pair_accuracy": 0.4823777778042687, + "positive_goodness": 2.0016636032528345, + "runtime_seconds": 5.888308048248291 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3852959673351712, + "lr": 0.03, + "negative_goodness": 1.9982130109786986, + "pair_accuracy": 0.4832444444762336, + "positive_goodness": 2.0018869299146864, + "runtime_seconds": 5.891618490219116 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3851991245693631, + "lr": 0.03, + "negative_goodness": 1.9979276790618896, + "pair_accuracy": 0.4842444444762336, + "positive_goodness": 2.0018074752807618, + "runtime_seconds": 5.891409873962402 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.3853057106865776, + "lr": 0.03, + "negative_goodness": 1.9982167551676433, + "pair_accuracy": 0.48620000003178915, + "positive_goodness": 2.0020106845643784, + "runtime_seconds": 5.892041206359863 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.385270743921068, + "lr": 0.03, + "negative_goodness": 1.9980789685143365, + "pair_accuracy": 0.48755555557674835, + "positive_goodness": 2.0018536292182074, + "runtime_seconds": 5.893585443496704 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3851377257876927, + "lr": 0.03, + "negative_goodness": 1.9979134016248914, + "pair_accuracy": 0.4860444444126553, + "positive_goodness": 2.0021071075015597, + "runtime_seconds": 5.891669034957886 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.3853445621914333, + "lr": 0.03, + "negative_goodness": 1.9981865581088596, + "pair_accuracy": 0.4877111111217075, + "positive_goodness": 2.0017612493303085, + "runtime_seconds": 5.898106336593628 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3851350643793743, + "lr": 0.03, + "negative_goodness": 1.997820346154107, + "pair_accuracy": 0.4867999999947018, + "positive_goodness": 2.002197341156006, + "runtime_seconds": 5.894902944564819 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3851453823513455, + "lr": 0.03, + "negative_goodness": 1.9978969047122532, + "pair_accuracy": 0.4845333333492279, + "positive_goodness": 2.0020384525722927, + "runtime_seconds": 5.892822742462158 + } + ], + "layer": 0 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.3931700608783297, + "lr": 0.03, + "negative_goodness": 1.9614571956210667, + "pair_accuracy": 0.5248444444444444, + "positive_goodness": 1.9745125068240696, + "runtime_seconds": 6.00260853767395 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3787266453637017, + "lr": 0.03, + "negative_goodness": 1.989453277905782, + "pair_accuracy": 0.5433777777777777, + "positive_goodness": 2.014946661482917, + "runtime_seconds": 5.984441041946411 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.378524174838596, + "lr": 0.03, + "negative_goodness": 1.9885209112591213, + "pair_accuracy": 0.5478666666984559, + "positive_goodness": 2.0155353230582342, + "runtime_seconds": 5.986544847488403 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3775977412541707, + "lr": 0.03, + "negative_goodness": 1.9877562760882908, + "pair_accuracy": 0.5504666666931576, + "positive_goodness": 2.0176366748809813, + "runtime_seconds": 5.980956315994263 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.376350413746304, + "lr": 0.03, + "negative_goodness": 1.986241997549269, + "pair_accuracy": 0.5562888889206781, + "positive_goodness": 2.0199487218644885, + "runtime_seconds": 5.982185363769531 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3759631296581691, + "lr": 0.03, + "negative_goodness": 1.9850278862635296, + "pair_accuracy": 0.5534888889100816, + "positive_goodness": 2.0210267003801134, + "runtime_seconds": 5.980340957641602 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3751652083502874, + "lr": 0.03, + "negative_goodness": 1.9837275034798516, + "pair_accuracy": 0.5558666666931577, + "positive_goodness": 2.0232700168397693, + "runtime_seconds": 5.981940031051636 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3758167764875624, + "lr": 0.03, + "negative_goodness": 1.9848760172102187, + "pair_accuracy": 0.5562888888994852, + "positive_goodness": 2.021539633644952, + "runtime_seconds": 5.985486030578613 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3744675419489543, + "lr": 0.03, + "negative_goodness": 1.9826782192018297, + "pair_accuracy": 0.5586444444126553, + "positive_goodness": 2.0242542195214166, + "runtime_seconds": 5.981043100357056 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.374949085489909, + "lr": 0.03, + "negative_goodness": 1.9826056987550524, + "pair_accuracy": 0.5610222222646077, + "positive_goodness": 2.023200198364258, + "runtime_seconds": 5.98160982131958 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.3740675876193575, + "lr": 0.03, + "negative_goodness": 1.9827120547824435, + "pair_accuracy": 0.567711111164093, + "positive_goodness": 2.0254339715745715, + "runtime_seconds": 5.9803690910339355 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3736978960461086, + "lr": 0.03, + "negative_goodness": 1.981502300092909, + "pair_accuracy": 0.5674888889206781, + "positive_goodness": 2.025813584009806, + "runtime_seconds": 5.980071544647217 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.3736572660446167, + "lr": 0.03, + "negative_goodness": 1.9813354211383396, + "pair_accuracy": 0.5705111111217075, + "positive_goodness": 2.0261692496829564, + "runtime_seconds": 5.981170892715454 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3728231609132555, + "lr": 0.03, + "negative_goodness": 1.9803242225858901, + "pair_accuracy": 0.5755333333651225, + "positive_goodness": 2.0277427441914875, + "runtime_seconds": 5.978522300720215 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.373797341410319, + "lr": 0.03, + "negative_goodness": 1.9806811288197834, + "pair_accuracy": 0.568044444433848, + "positive_goodness": 2.025475881322225, + "runtime_seconds": 5.983184099197388 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3733719135920206, + "lr": 0.03, + "negative_goodness": 1.9803072258843315, + "pair_accuracy": 0.5688666666666666, + "positive_goodness": 2.027118152533637, + "runtime_seconds": 5.981003284454346 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.372928720050388, + "lr": 0.03, + "negative_goodness": 1.9799542486826578, + "pair_accuracy": 0.5754888888888889, + "positive_goodness": 2.0275567568884956, + "runtime_seconds": 5.978998422622681 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3721297845840454, + "lr": 0.03, + "negative_goodness": 1.9784513053470187, + "pair_accuracy": 0.5755555555555556, + "positive_goodness": 2.029322801208496, + "runtime_seconds": 5.980375528335571 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.3716845730463665, + "lr": 0.03, + "negative_goodness": 1.9782871550665961, + "pair_accuracy": 0.5791999999894036, + "positive_goodness": 2.0300268153296575, + "runtime_seconds": 5.980324983596802 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3721484000311956, + "lr": 0.03, + "negative_goodness": 1.9778074828677707, + "pair_accuracy": 0.5736888888676961, + "positive_goodness": 2.028543465508355, + "runtime_seconds": 5.9810755252838135 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3716066225687662, + "lr": 0.03, + "negative_goodness": 1.977788461028205, + "pair_accuracy": 0.5746444444444444, + "positive_goodness": 2.030718470658196, + "runtime_seconds": 5.980867147445679 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.3705314564387003, + "lr": 0.03, + "negative_goodness": 1.97657086192237, + "pair_accuracy": 0.5748444444338481, + "positive_goodness": 2.032610629865858, + "runtime_seconds": 5.982075452804565 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3702781814787124, + "lr": 0.03, + "negative_goodness": 1.9758473294999865, + "pair_accuracy": 0.5718444444444445, + "positive_goodness": 2.032681765153673, + "runtime_seconds": 5.986770391464233 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3701435327741835, + "lr": 0.03, + "negative_goodness": 1.9752755856619941, + "pair_accuracy": 0.5749999999682108, + "positive_goodness": 2.0326180130428737, + "runtime_seconds": 5.981854677200317 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3697423634211223, + "lr": 0.03, + "negative_goodness": 1.9753647101932101, + "pair_accuracy": 0.5738666667090522, + "positive_goodness": 2.0340875378078884, + "runtime_seconds": 5.984050989151001 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3701823308520846, + "lr": 0.03, + "negative_goodness": 1.9749206221898397, + "pair_accuracy": 0.5801777778095669, + "positive_goodness": 2.0332959334903293, + "runtime_seconds": 5.983195781707764 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3694166609446208, + "lr": 0.03, + "negative_goodness": 1.974562843577067, + "pair_accuracy": 0.5795777777671814, + "positive_goodness": 2.033927973980374, + "runtime_seconds": 5.980337381362915 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3696382489734227, + "lr": 0.03, + "negative_goodness": 1.9743131326887342, + "pair_accuracy": 0.577555555597941, + "positive_goodness": 2.0340135498894587, + "runtime_seconds": 5.979592561721802 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3688199961132475, + "lr": 0.03, + "negative_goodness": 1.9735227844662135, + "pair_accuracy": 0.5801555555237664, + "positive_goodness": 2.035634282981025, + "runtime_seconds": 5.97942042350769 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3686969990836249, + "lr": 0.03, + "negative_goodness": 1.972615485848321, + "pair_accuracy": 0.5782888889418708, + "positive_goodness": 2.0357347773657906, + "runtime_seconds": 5.980018615722656 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3687058734681872, + "lr": 0.03, + "negative_goodness": 1.9721841576046413, + "pair_accuracy": 0.5804000000423856, + "positive_goodness": 2.035713754611545, + "runtime_seconds": 5.9822611808776855 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3682003465016683, + "lr": 0.03, + "negative_goodness": 1.9719386617448595, + "pair_accuracy": 0.5854666666560703, + "positive_goodness": 2.036688850593567, + "runtime_seconds": 5.977877855300903 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3692989912245008, + "lr": 0.03, + "negative_goodness": 1.9726413374582927, + "pair_accuracy": 0.5781999999682108, + "positive_goodness": 2.0350098616706, + "runtime_seconds": 5.9875288009643555 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.367435570356581, + "lr": 0.03, + "negative_goodness": 1.970960520807902, + "pair_accuracy": 0.5836888888782925, + "positive_goodness": 2.0386133412255183, + "runtime_seconds": 5.978238582611084 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.367826961517334, + "lr": 0.03, + "negative_goodness": 1.9709638761732313, + "pair_accuracy": 0.5834222222752041, + "positive_goodness": 2.037747360399034, + "runtime_seconds": 5.981059312820435 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.3666473015679252, + "lr": 0.03, + "negative_goodness": 1.9695034294340346, + "pair_accuracy": 0.582422222190433, + "positive_goodness": 2.040335519727071, + "runtime_seconds": 5.980744123458862 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3683875297758314, + "lr": 0.03, + "negative_goodness": 1.9706050326453315, + "pair_accuracy": 0.5854222222328186, + "positive_goodness": 2.0359173857371013, + "runtime_seconds": 5.983423233032227 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.367209183353848, + "lr": 0.03, + "negative_goodness": 1.9703588547812567, + "pair_accuracy": 0.581111111164093, + "positive_goodness": 2.0389629511091445, + "runtime_seconds": 5.989146709442139 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3670525961558024, + "lr": 0.03, + "negative_goodness": 1.9700104254404704, + "pair_accuracy": 0.5829999999682108, + "positive_goodness": 2.039833468712701, + "runtime_seconds": 5.9869232177734375 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3665830942577786, + "lr": 0.03, + "negative_goodness": 1.9684993154737684, + "pair_accuracy": 0.580244444433848, + "positive_goodness": 2.039880865987142, + "runtime_seconds": 5.98329758644104 + } + ], + "layer": 1 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.3778934198379515, + "lr": 0.03, + "negative_goodness": 1.9883329700469972, + "pair_accuracy": 0.6015555555767483, + "positive_goodness": 2.020156074439155, + "runtime_seconds": 6.0666563510894775 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3692912198384604, + "lr": 0.03, + "negative_goodness": 1.977916879526774, + "pair_accuracy": 0.6224666666348775, + "positive_goodness": 2.0346438342200384, + "runtime_seconds": 6.0654613971710205 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3656193161646526, + "lr": 0.03, + "negative_goodness": 1.972359447945489, + "pair_accuracy": 0.631377777756585, + "positive_goodness": 2.0417931645287406, + "runtime_seconds": 6.066296577453613 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3636689510981241, + "lr": 0.03, + "negative_goodness": 1.969901711569892, + "pair_accuracy": 0.63304444448683, + "positive_goodness": 2.0459449424319796, + "runtime_seconds": 6.062884330749512 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3624872768190173, + "lr": 0.03, + "negative_goodness": 1.968433802625868, + "pair_accuracy": 0.6327777777459886, + "positive_goodness": 2.048124256324768, + "runtime_seconds": 6.06137752532959 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3585968855539958, + "lr": 0.03, + "negative_goodness": 1.9645866083145143, + "pair_accuracy": 0.6498666666560703, + "positive_goodness": 2.055677861404419, + "runtime_seconds": 6.064768075942993 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3579612738927205, + "lr": 0.03, + "negative_goodness": 1.9625601061715021, + "pair_accuracy": 0.6409555555237664, + "positive_goodness": 2.0577649586995443, + "runtime_seconds": 6.060542106628418 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.357287814288669, + "lr": 0.03, + "negative_goodness": 1.9620870353910658, + "pair_accuracy": 0.6463333333227369, + "positive_goodness": 2.058215067715115, + "runtime_seconds": 6.060739040374756 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.356436611535814, + "lr": 0.03, + "negative_goodness": 1.9598341552522447, + "pair_accuracy": 0.6462666666878595, + "positive_goodness": 2.060737878290812, + "runtime_seconds": 6.060312509536743 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3552057264116075, + "lr": 0.03, + "negative_goodness": 1.9594891665140788, + "pair_accuracy": 0.6499777778307597, + "positive_goodness": 2.0628951744503445, + "runtime_seconds": 6.059125661849976 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.3578014404084948, + "lr": 0.03, + "negative_goodness": 1.9598490172068277, + "pair_accuracy": 0.6387111111270057, + "positive_goodness": 2.057129967159695, + "runtime_seconds": 6.061015367507935 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.354744299613105, + "lr": 0.03, + "negative_goodness": 1.9582753913031683, + "pair_accuracy": 0.6472666666613685, + "positive_goodness": 2.0633156706492106, + "runtime_seconds": 6.0633440017700195 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.354120409880744, + "lr": 0.03, + "negative_goodness": 1.9572144768185085, + "pair_accuracy": 0.6486000000423855, + "positive_goodness": 2.0649203235202367, + "runtime_seconds": 6.06241512298584 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3528062055163914, + "lr": 0.03, + "negative_goodness": 1.9545223477045695, + "pair_accuracy": 0.6516444444550408, + "positive_goodness": 2.0674305085288154, + "runtime_seconds": 6.06486439704895 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.351523060268826, + "lr": 0.03, + "negative_goodness": 1.9547262173334758, + "pair_accuracy": 0.6563999999682109, + "positive_goodness": 2.069753249189589, + "runtime_seconds": 6.063844680786133 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3505450826856824, + "lr": 0.03, + "negative_goodness": 1.952416684044732, + "pair_accuracy": 0.6524222222540114, + "positive_goodness": 2.0723450057135686, + "runtime_seconds": 6.063648223876953 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3510271078533596, + "lr": 0.03, + "negative_goodness": 1.9524554694493612, + "pair_accuracy": 0.6523555555343628, + "positive_goodness": 2.071400096469455, + "runtime_seconds": 6.064286708831787 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3515247582541572, + "lr": 0.03, + "negative_goodness": 1.9529400699191624, + "pair_accuracy": 0.6467777777459887, + "positive_goodness": 2.0700243264516196, + "runtime_seconds": 6.065088510513306 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.3492372905095418, + "lr": 0.03, + "negative_goodness": 1.950092873403761, + "pair_accuracy": 0.6549777777565851, + "positive_goodness": 2.0744359222835964, + "runtime_seconds": 6.062382459640503 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.349206513404846, + "lr": 0.03, + "negative_goodness": 1.9500429633670382, + "pair_accuracy": 0.6557999999682108, + "positive_goodness": 2.075095561811659, + "runtime_seconds": 6.062449216842651 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.349302830229865, + "lr": 0.03, + "negative_goodness": 1.9498704238891602, + "pair_accuracy": 0.6529555555820465, + "positive_goodness": 2.074691123750475, + "runtime_seconds": 6.061194896697998 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.3483241590499877, + "lr": 0.03, + "negative_goodness": 1.9486380299038357, + "pair_accuracy": 0.6539777778307597, + "positive_goodness": 2.0761523702409534, + "runtime_seconds": 6.062600374221802 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.347282488293118, + "lr": 0.03, + "negative_goodness": 1.9468271743562486, + "pair_accuracy": 0.6572222222222223, + "positive_goodness": 2.0789933425267537, + "runtime_seconds": 6.062525749206543 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3448558597352769, + "lr": 0.03, + "negative_goodness": 1.9445306475109525, + "pair_accuracy": 0.6564888888994853, + "positive_goodness": 2.0825855370415582, + "runtime_seconds": 6.061218976974487 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3473351562288072, + "lr": 0.03, + "negative_goodness": 1.9469996535407172, + "pair_accuracy": 0.6567111111534967, + "positive_goodness": 2.0786677921295165, + "runtime_seconds": 6.062284469604492 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.345635716650221, + "lr": 0.03, + "negative_goodness": 1.9449720490349665, + "pair_accuracy": 0.6616000000105964, + "positive_goodness": 2.0821274549272326, + "runtime_seconds": 6.062480688095093 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3469362598207262, + "lr": 0.03, + "negative_goodness": 1.9468740383995904, + "pair_accuracy": 0.6513111111323039, + "positive_goodness": 2.079769244066874, + "runtime_seconds": 6.065801620483398 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.346161561203003, + "lr": 0.03, + "negative_goodness": 1.9432483650631376, + "pair_accuracy": 0.6587333333651225, + "positive_goodness": 2.0803432929992676, + "runtime_seconds": 6.06447172164917 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3451417684343125, + "lr": 0.03, + "negative_goodness": 1.9428280779732598, + "pair_accuracy": 0.6592222222752041, + "positive_goodness": 2.083250764041477, + "runtime_seconds": 6.06339693069458 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3448662763383654, + "lr": 0.03, + "negative_goodness": 1.9437873603396945, + "pair_accuracy": 0.6600666666878594, + "positive_goodness": 2.084092647128635, + "runtime_seconds": 6.066731691360474 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3446798132578532, + "lr": 0.03, + "negative_goodness": 1.9434373878690931, + "pair_accuracy": 0.6546444444444445, + "positive_goodness": 2.083723704910278, + "runtime_seconds": 6.064776182174683 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3458749353832669, + "lr": 0.03, + "negative_goodness": 1.9426654205322265, + "pair_accuracy": 0.6564888888570998, + "positive_goodness": 2.081389229838053, + "runtime_seconds": 6.0665411949157715 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3433646739112006, + "lr": 0.03, + "negative_goodness": 1.9427647684521145, + "pair_accuracy": 0.661555555566152, + "positive_goodness": 2.086705190404256, + "runtime_seconds": 6.0651984214782715 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3445697229173448, + "lr": 0.03, + "negative_goodness": 1.942930364926656, + "pair_accuracy": 0.6604444444232517, + "positive_goodness": 2.083599764463637, + "runtime_seconds": 6.061059236526489 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.343607105869717, + "lr": 0.03, + "negative_goodness": 1.9416292188220554, + "pair_accuracy": 0.663222222190433, + "positive_goodness": 2.0854547542148167, + "runtime_seconds": 6.0617454051971436 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.3419898032930162, + "lr": 0.03, + "negative_goodness": 1.9397027740478516, + "pair_accuracy": 0.6669111111217075, + "positive_goodness": 2.0894151977539064, + "runtime_seconds": 6.0608134269714355 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3434102685716418, + "lr": 0.03, + "negative_goodness": 1.940571518474155, + "pair_accuracy": 0.6646888888676962, + "positive_goodness": 2.086845216263665, + "runtime_seconds": 6.059274673461914 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.3426040760040283, + "lr": 0.03, + "negative_goodness": 1.940912329864502, + "pair_accuracy": 0.6653777777989706, + "positive_goodness": 2.0880502849578857, + "runtime_seconds": 6.062561750411987 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3420939029905532, + "lr": 0.03, + "negative_goodness": 1.9387054932276409, + "pair_accuracy": 0.6625111111429003, + "positive_goodness": 2.0892556584676107, + "runtime_seconds": 6.059795379638672 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3407587654537625, + "lr": 0.03, + "negative_goodness": 1.9381091831843058, + "pair_accuracy": 0.6653777777671814, + "positive_goodness": 2.091323936568366, + "runtime_seconds": 6.061957120895386 + } + ], + "layer": 2 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.3694264541625976, + "lr": 0.03, + "negative_goodness": 1.9315426686180963, + "pair_accuracy": 0.6433999999894037, + "positive_goodness": 2.022064396116469, + "runtime_seconds": 5.979110479354858 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3466723141564263, + "lr": 0.03, + "negative_goodness": 1.9455105621761746, + "pair_accuracy": 0.6989777778307597, + "positive_goodness": 2.078012590408325, + "runtime_seconds": 5.979700088500977 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3409328441196018, + "lr": 0.03, + "negative_goodness": 1.941532026248508, + "pair_accuracy": 0.7171777778201633, + "positive_goodness": 2.0903180645836725, + "runtime_seconds": 5.981481552124023 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3384243650012546, + "lr": 0.03, + "negative_goodness": 1.9390470175001355, + "pair_accuracy": 0.7205333333863152, + "positive_goodness": 2.094773719829983, + "runtime_seconds": 5.980740547180176 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3348754941092598, + "lr": 0.03, + "negative_goodness": 1.9352488935258654, + "pair_accuracy": 0.7259333333439297, + "positive_goodness": 2.1019654900021023, + "runtime_seconds": 5.983296155929565 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3311071152793037, + "lr": 0.03, + "negative_goodness": 1.9325834632661607, + "pair_accuracy": 0.7347777778095669, + "positive_goodness": 2.1095293689727783, + "runtime_seconds": 5.983576774597168 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3298779280132718, + "lr": 0.03, + "negative_goodness": 1.9299019046147665, + "pair_accuracy": 0.7346222222222222, + "positive_goodness": 2.1126911966111925, + "runtime_seconds": 5.98113226890564 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3293858040067885, + "lr": 0.03, + "negative_goodness": 1.9291187158372667, + "pair_accuracy": 0.7287555555343628, + "positive_goodness": 2.1144703827752007, + "runtime_seconds": 5.984424829483032 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.328323615031772, + "lr": 0.03, + "negative_goodness": 1.928339974255032, + "pair_accuracy": 0.7338888888782925, + "positive_goodness": 2.116397580167982, + "runtime_seconds": 5.984027624130249 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3282192443211873, + "lr": 0.03, + "negative_goodness": 1.9294531637191772, + "pair_accuracy": 0.7326000000317892, + "positive_goodness": 2.1155966016981336, + "runtime_seconds": 5.984655141830444 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.3259414041519164, + "lr": 0.03, + "negative_goodness": 1.9258427941216363, + "pair_accuracy": 0.7369999999788073, + "positive_goodness": 2.1211818587409126, + "runtime_seconds": 5.985104560852051 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3231838992436726, + "lr": 0.03, + "negative_goodness": 1.9216014865239461, + "pair_accuracy": 0.7374222222752042, + "positive_goodness": 2.1265720929675633, + "runtime_seconds": 5.983922719955444 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.3217691022660998, + "lr": 0.03, + "negative_goodness": 1.9225371470981174, + "pair_accuracy": 0.7440888888994853, + "positive_goodness": 2.1299233885023328, + "runtime_seconds": 5.986240386962891 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3258742632124159, + "lr": 0.03, + "negative_goodness": 1.922986926375495, + "pair_accuracy": 0.7358222222116259, + "positive_goodness": 2.122034566116333, + "runtime_seconds": 5.98368763923645 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.32442424820794, + "lr": 0.03, + "negative_goodness": 1.9242325100368924, + "pair_accuracy": 0.7310000000423855, + "positive_goodness": 2.125202389526367, + "runtime_seconds": 5.983877658843994 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3205884724299113, + "lr": 0.03, + "negative_goodness": 1.9184309008068507, + "pair_accuracy": 0.7436666666878594, + "positive_goodness": 2.1308239685906303, + "runtime_seconds": 5.982682943344116 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3207264395183986, + "lr": 0.03, + "negative_goodness": 1.921188094054328, + "pair_accuracy": 0.7407555555449592, + "positive_goodness": 2.1331847368028427, + "runtime_seconds": 5.980348825454712 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3180421180937025, + "lr": 0.03, + "negative_goodness": 1.9147770264731514, + "pair_accuracy": 0.7467999999894036, + "positive_goodness": 2.137587122175429, + "runtime_seconds": 5.980019569396973 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.318621134376526, + "lr": 0.03, + "negative_goodness": 1.9175735144297281, + "pair_accuracy": 0.7434888889312744, + "positive_goodness": 2.136951607852512, + "runtime_seconds": 5.979837656021118 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.319987634891934, + "lr": 0.03, + "negative_goodness": 1.919391003312005, + "pair_accuracy": 0.737488888920678, + "positive_goodness": 2.1344727956559923, + "runtime_seconds": 5.978689670562744 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3180361602147421, + "lr": 0.03, + "negative_goodness": 1.9148269819895427, + "pair_accuracy": 0.7378444444974264, + "positive_goodness": 2.137727655749851, + "runtime_seconds": 5.9797515869140625 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.317678520096673, + "lr": 0.03, + "negative_goodness": 1.9167088392893474, + "pair_accuracy": 0.7423777777671814, + "positive_goodness": 2.138864570511712, + "runtime_seconds": 5.980733633041382 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3178222114139133, + "lr": 0.03, + "negative_goodness": 1.9149838419596354, + "pair_accuracy": 0.7369777777459886, + "positive_goodness": 2.1378890107896593, + "runtime_seconds": 5.981074333190918 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3148595873302884, + "lr": 0.03, + "negative_goodness": 1.9132602583779228, + "pair_accuracy": 0.7428444444550408, + "positive_goodness": 2.1442959336598713, + "runtime_seconds": 5.98176646232605 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3135741770426432, + "lr": 0.03, + "negative_goodness": 1.910923408847385, + "pair_accuracy": 0.7475555555449592, + "positive_goodness": 2.1477724839104546, + "runtime_seconds": 5.979963302612305 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3160499358707005, + "lr": 0.03, + "negative_goodness": 1.913453252029419, + "pair_accuracy": 0.7455111111217075, + "positive_goodness": 2.1418466620551215, + "runtime_seconds": 5.983236074447632 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.316577659204271, + "lr": 0.03, + "negative_goodness": 1.914289217482673, + "pair_accuracy": 0.7425333333651225, + "positive_goodness": 2.140870028177897, + "runtime_seconds": 5.980891227722168 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.316730691189236, + "lr": 0.03, + "negative_goodness": 1.9139939239290025, + "pair_accuracy": 0.7385333333121405, + "positive_goodness": 2.1415105464087594, + "runtime_seconds": 5.982870101928711 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.315346255959405, + "lr": 0.03, + "negative_goodness": 1.9110439316219754, + "pair_accuracy": 0.7405777778307597, + "positive_goodness": 2.1448754462771946, + "runtime_seconds": 5.985513210296631 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3137559794955784, + "lr": 0.03, + "negative_goodness": 1.9102261783387926, + "pair_accuracy": 0.7443111111005147, + "positive_goodness": 2.145622188652886, + "runtime_seconds": 5.9830687046051025 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3125492536756727, + "lr": 0.03, + "negative_goodness": 1.9118073127746582, + "pair_accuracy": 0.7490666667090522, + "positive_goodness": 2.150018466462029, + "runtime_seconds": 5.982649326324463 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3122342109892104, + "lr": 0.03, + "negative_goodness": 1.9098376069810656, + "pair_accuracy": 0.7460888889153798, + "positive_goodness": 2.1514166417863634, + "runtime_seconds": 5.984927415847778 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3144113831626045, + "lr": 0.03, + "negative_goodness": 1.9110068290286595, + "pair_accuracy": 0.7438222222063277, + "positive_goodness": 2.1460519907633464, + "runtime_seconds": 5.98341965675354 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.314858454577128, + "lr": 0.03, + "negative_goodness": 1.90949225669437, + "pair_accuracy": 0.7395555555979411, + "positive_goodness": 2.144171138381958, + "runtime_seconds": 5.981204271316528 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.3123024072647094, + "lr": 0.03, + "negative_goodness": 1.9082330724928114, + "pair_accuracy": 0.7400000000105964, + "positive_goodness": 2.1511556092580157, + "runtime_seconds": 5.980629920959473 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.3084842159907024, + "lr": 0.03, + "negative_goodness": 1.9054335157394409, + "pair_accuracy": 0.7525333333651225, + "positive_goodness": 2.1573879462771948, + "runtime_seconds": 5.980962038040161 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3150889170328777, + "lr": 0.03, + "negative_goodness": 1.909355968983968, + "pair_accuracy": 0.7383333333651225, + "positive_goodness": 2.1458897659725613, + "runtime_seconds": 5.981602191925049 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.311930605676439, + "lr": 0.03, + "negative_goodness": 1.9078128904130724, + "pair_accuracy": 0.7456444444232517, + "positive_goodness": 2.1513866523742675, + "runtime_seconds": 5.980937719345093 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3105542700449626, + "lr": 0.03, + "negative_goodness": 1.9074810123231676, + "pair_accuracy": 0.7495555555343628, + "positive_goodness": 2.154290027152167, + "runtime_seconds": 5.979893445968628 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3091086486180623, + "lr": 0.03, + "negative_goodness": 1.9047713115480212, + "pair_accuracy": 0.748933333354526, + "positive_goodness": 2.1578838062286376, + "runtime_seconds": 5.9804463386535645 + } + ], + "layer": 3 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2790996762169733, + "lr": 0.03, + "negative_goodness": 1.9320242347717285, + "pair_accuracy": 0.8267333333015442, + "positive_goodness": 2.236565159988403, + "runtime_seconds": 6.059277534484863 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.2372933910157946, + "lr": 0.03, + "negative_goodness": 1.894013934940762, + "pair_accuracy": 0.8933777777671814, + "positive_goodness": 2.312923978551229, + "runtime_seconds": 6.061666250228882 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.2213343842400446, + "lr": 0.03, + "negative_goodness": 1.8759458523644341, + "pair_accuracy": 0.8886222221904331, + "positive_goodness": 2.3443591727786592, + "runtime_seconds": 6.0614330768585205 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.2150324143515694, + "lr": 0.03, + "negative_goodness": 1.871982069545322, + "pair_accuracy": 0.896177777756585, + "positive_goodness": 2.357221519385444, + "runtime_seconds": 6.059678554534912 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.2227705931557549, + "lr": 0.03, + "negative_goodness": 1.8720421888987224, + "pair_accuracy": 0.8831111110899184, + "positive_goodness": 2.3392312397850885, + "runtime_seconds": 6.066678524017334 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.2007182056427002, + "lr": 0.03, + "negative_goodness": 1.858509766176012, + "pair_accuracy": 0.9067777777989705, + "positive_goodness": 2.3868888347201875, + "runtime_seconds": 6.0650200843811035 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.2047750562773811, + "lr": 0.03, + "negative_goodness": 1.854804564306471, + "pair_accuracy": 0.8879111110899184, + "positive_goodness": 2.375061264801025, + "runtime_seconds": 6.063778877258301 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.1958627138561673, + "lr": 0.03, + "negative_goodness": 1.851102579498291, + "pair_accuracy": 0.9002888889312745, + "positive_goodness": 2.398383729934692, + "runtime_seconds": 6.0660295486450195 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.1987770407146878, + "lr": 0.03, + "negative_goodness": 1.8513284273359512, + "pair_accuracy": 0.8969999999682109, + "positive_goodness": 2.3890313185797796, + "runtime_seconds": 6.065452337265015 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.2013778301874796, + "lr": 0.03, + "negative_goodness": 1.8514497347513834, + "pair_accuracy": 0.8865333333333333, + "positive_goodness": 2.3866108526017933, + "runtime_seconds": 6.065209150314331 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.1969052589416505, + "lr": 0.03, + "negative_goodness": 1.849414856423272, + "pair_accuracy": 0.8961333333333333, + "positive_goodness": 2.3950613297780357, + "runtime_seconds": 6.061498165130615 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.1845088395224677, + "lr": 0.03, + "negative_goodness": 1.8398546699735854, + "pair_accuracy": 0.9020888888888889, + "positive_goodness": 2.4231562284257677, + "runtime_seconds": 6.061634302139282 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.1931991404003568, + "lr": 0.03, + "negative_goodness": 1.8435046942181057, + "pair_accuracy": 0.895377777756585, + "positive_goodness": 2.405434050454034, + "runtime_seconds": 6.064945697784424 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.185102932548523, + "lr": 0.03, + "negative_goodness": 1.8360718784120347, + "pair_accuracy": 0.8936888889312744, + "positive_goodness": 2.4219772477467854, + "runtime_seconds": 6.061174631118774 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.191577965227763, + "lr": 0.03, + "negative_goodness": 1.843565831459893, + "pair_accuracy": 0.8907555555555555, + "positive_goodness": 2.41326848487854, + "runtime_seconds": 6.0632617473602295 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.1865951836480035, + "lr": 0.03, + "negative_goodness": 1.8362281233257718, + "pair_accuracy": 0.9004888889418708, + "positive_goodness": 2.4201734641604955, + "runtime_seconds": 6.063931226730347 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.1986465691672432, + "lr": 0.03, + "negative_goodness": 1.845456612332662, + "pair_accuracy": 0.884822222190433, + "positive_goodness": 2.3923546896192764, + "runtime_seconds": 6.061584711074829 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.1824353651046753, + "lr": 0.03, + "negative_goodness": 1.8325393890168933, + "pair_accuracy": 0.8981333333121406, + "positive_goodness": 2.429313689761692, + "runtime_seconds": 6.06391167640686 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.1852495226330226, + "lr": 0.03, + "negative_goodness": 1.8337557284673054, + "pair_accuracy": 0.8977555555237664, + "positive_goodness": 2.4290720226287843, + "runtime_seconds": 6.062767505645752 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.1913312978320652, + "lr": 0.03, + "negative_goodness": 1.8350722527821859, + "pair_accuracy": 0.888511111079322, + "positive_goodness": 2.4110671156989203, + "runtime_seconds": 6.063162803649902 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.1842537308798895, + "lr": 0.03, + "negative_goodness": 1.8360518987443712, + "pair_accuracy": 0.8960000000529819, + "positive_goodness": 2.431038141547309, + "runtime_seconds": 6.067229509353638 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.1778004178577, + "lr": 0.03, + "negative_goodness": 1.8273498172336156, + "pair_accuracy": 0.9048222222752041, + "positive_goodness": 2.439615209197998, + "runtime_seconds": 6.0656349658966064 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.1845065527174208, + "lr": 0.03, + "negative_goodness": 1.8306520388921101, + "pair_accuracy": 0.8894888888888889, + "positive_goodness": 2.426795822313097, + "runtime_seconds": 6.074563503265381 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.1820990962982179, + "lr": 0.03, + "negative_goodness": 1.8291268906911213, + "pair_accuracy": 0.8931777777671814, + "positive_goodness": 2.433557123904758, + "runtime_seconds": 6.065053224563599 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.1774473353915744, + "lr": 0.03, + "negative_goodness": 1.8281886732313368, + "pair_accuracy": 0.8973999999894037, + "positive_goodness": 2.442788556543986, + "runtime_seconds": 6.064608335494995 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.1813383837381999, + "lr": 0.03, + "negative_goodness": 1.8314955468707614, + "pair_accuracy": 0.8986888889418708, + "positive_goodness": 2.4338408478418985, + "runtime_seconds": 6.063029050827026 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.1860559944788616, + "lr": 0.03, + "negative_goodness": 1.8318558712853326, + "pair_accuracy": 0.8873333333333333, + "positive_goodness": 2.4253181638929577, + "runtime_seconds": 6.060927629470825 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.184213071484036, + "lr": 0.03, + "negative_goodness": 1.8328081658469306, + "pair_accuracy": 0.894555555566152, + "positive_goodness": 2.42784982486301, + "runtime_seconds": 6.063875913619995 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.185441330909729, + "lr": 0.03, + "negative_goodness": 1.8311142799589368, + "pair_accuracy": 0.8920888888570997, + "positive_goodness": 2.4240012288835313, + "runtime_seconds": 6.062417268753052 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.1763895898183188, + "lr": 0.03, + "negative_goodness": 1.8293777327219645, + "pair_accuracy": 0.9011999999788073, + "positive_goodness": 2.445229783460829, + "runtime_seconds": 6.062203884124756 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.1767015352037218, + "lr": 0.03, + "negative_goodness": 1.8244261776394315, + "pair_accuracy": 0.9043999999788073, + "positive_goodness": 2.445290773179796, + "runtime_seconds": 6.061763286590576 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.1886791415108575, + "lr": 0.03, + "negative_goodness": 1.836533439678616, + "pair_accuracy": 0.8852444444338481, + "positive_goodness": 2.424700284279717, + "runtime_seconds": 6.061740398406982 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.1747555481804741, + "lr": 0.03, + "negative_goodness": 1.8218524259355333, + "pair_accuracy": 0.9007555555555555, + "positive_goodness": 2.4449608136071097, + "runtime_seconds": 6.062439441680908 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.1793666855706109, + "lr": 0.03, + "negative_goodness": 1.8263251133177016, + "pair_accuracy": 0.8933555555555556, + "positive_goodness": 2.439794603220622, + "runtime_seconds": 6.060402154922485 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.182122101656596, + "lr": 0.03, + "negative_goodness": 1.828336991373698, + "pair_accuracy": 0.8952666666454739, + "positive_goodness": 2.4364880695343016, + "runtime_seconds": 6.0615761280059814 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.189055893919203, + "lr": 0.03, + "negative_goodness": 1.8320734663221572, + "pair_accuracy": 0.8870888888782925, + "positive_goodness": 2.4191097621070012, + "runtime_seconds": 6.070591926574707 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.1795562043613859, + "lr": 0.03, + "negative_goodness": 1.8225096825705633, + "pair_accuracy": 0.8887555555343628, + "positive_goodness": 2.441470533498128, + "runtime_seconds": 6.071572542190552 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.176062773683336, + "lr": 0.03, + "negative_goodness": 1.8214178182390002, + "pair_accuracy": 0.8963555555555556, + "positive_goodness": 2.4484717436472576, + "runtime_seconds": 6.07112455368042 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.1734984332932366, + "lr": 0.03, + "negative_goodness": 1.8226626316070558, + "pair_accuracy": 0.8947999999894036, + "positive_goodness": 2.4538106683095298, + "runtime_seconds": 6.069512367248535 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.193426409403483, + "lr": 0.03, + "negative_goodness": 1.8323564805136787, + "pair_accuracy": 0.8805999999682108, + "positive_goodness": 2.4151936037275528, + "runtime_seconds": 6.063399314880371 + } + ], + "layer": 4 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.206510053677029, + "lr": 0.03, + "negative_goodness": 1.7792172163221571, + "pair_accuracy": 0.849644444433848, + "positive_goodness": 2.3613981796688504, + "runtime_seconds": 5.9840404987335205 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.1180472915437487, + "lr": 0.03, + "negative_goodness": 1.7537119994057548, + "pair_accuracy": 0.9066222222222222, + "positive_goodness": 2.5668498105367026, + "runtime_seconds": 5.9834020137786865 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.0761442077636718, + "lr": 0.03, + "negative_goodness": 1.7192611765119765, + "pair_accuracy": 0.9183111111111111, + "positive_goodness": 2.6737394202338325, + "runtime_seconds": 5.984705686569214 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0767657746209038, + "lr": 0.03, + "negative_goodness": 1.7165096852196589, + "pair_accuracy": 0.9234444444550408, + "positive_goodness": 2.6744019661373564, + "runtime_seconds": 5.984201669692993 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.0511816959804958, + "lr": 0.03, + "negative_goodness": 1.6945178568522135, + "pair_accuracy": 0.9325555555979411, + "positive_goodness": 2.739649661509196, + "runtime_seconds": 5.9865686893463135 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.0598680033153958, + "lr": 0.03, + "negative_goodness": 1.7017980873319838, + "pair_accuracy": 0.9239999999788072, + "positive_goodness": 2.715990206866794, + "runtime_seconds": 5.9873270988464355 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.0364842529296876, + "lr": 0.03, + "negative_goodness": 1.6706852739122178, + "pair_accuracy": 0.9279555555343628, + "positive_goodness": 2.7817265841166177, + "runtime_seconds": 5.985645294189453 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0157055593066746, + "lr": 0.03, + "negative_goodness": 1.6606598917007447, + "pair_accuracy": 0.9399777777883742, + "positive_goodness": 2.8447105996449786, + "runtime_seconds": 5.9859161376953125 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.022450678369734, + "lr": 0.03, + "negative_goodness": 1.658062926250034, + "pair_accuracy": 0.9371555555343628, + "positive_goodness": 2.817877022467719, + "runtime_seconds": 5.986541509628296 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.0199218417697482, + "lr": 0.03, + "negative_goodness": 1.6600687336815727, + "pair_accuracy": 0.9450222222010295, + "positive_goodness": 2.8277223790062798, + "runtime_seconds": 5.986910820007324 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.0293013444582622, + "lr": 0.03, + "negative_goodness": 1.6601685360166762, + "pair_accuracy": 0.927977777756585, + "positive_goodness": 2.8063861876593696, + "runtime_seconds": 5.984683513641357 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.997588968149821, + "lr": 0.03, + "negative_goodness": 1.6347474789513483, + "pair_accuracy": 0.942044444433848, + "positive_goodness": 2.903607641559177, + "runtime_seconds": 5.986644268035889 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.9766871611383227, + "lr": 0.03, + "negative_goodness": 1.612153264511956, + "pair_accuracy": 0.9464, + "positive_goodness": 2.945292927593655, + "runtime_seconds": 5.985624313354492 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.9830098033905029, + "lr": 0.03, + "negative_goodness": 1.6152430071512858, + "pair_accuracy": 0.9434888889206781, + "positive_goodness": 2.9363582916259765, + "runtime_seconds": 5.9841718673706055 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.9902497747209337, + "lr": 0.03, + "negative_goodness": 1.621737363327874, + "pair_accuracy": 0.9442444444550409, + "positive_goodness": 2.9309814730750188, + "runtime_seconds": 5.987169981002808 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.9778145734892951, + "lr": 0.03, + "negative_goodness": 1.6086522109349568, + "pair_accuracy": 0.9457111110899183, + "positive_goodness": 2.9578793236202663, + "runtime_seconds": 5.9855968952178955 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.976281268649631, + "lr": 0.03, + "negative_goodness": 1.6040572290208606, + "pair_accuracy": 0.9406666666454739, + "positive_goodness": 2.977479224650065, + "runtime_seconds": 5.984076499938965 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.9736170813454522, + "lr": 0.03, + "negative_goodness": 1.5982097276475695, + "pair_accuracy": 0.9449555555343628, + "positive_goodness": 2.9742093078189424, + "runtime_seconds": 5.985750198364258 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.9931732725037469, + "lr": 0.03, + "negative_goodness": 1.6185183919906616, + "pair_accuracy": 0.9413555555237664, + "positive_goodness": 2.923843538453844, + "runtime_seconds": 5.986278533935547 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.9578904231283399, + "lr": 0.03, + "negative_goodness": 1.5847804288652207, + "pair_accuracy": 0.9502444444444444, + "positive_goodness": 3.0183473300086128, + "runtime_seconds": 5.983867883682251 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.9709372259775797, + "lr": 0.03, + "negative_goodness": 1.5950287723541259, + "pair_accuracy": 0.9445333333863153, + "positive_goodness": 2.995971722708808, + "runtime_seconds": 5.98766016960144 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.9745934847831726, + "lr": 0.03, + "negative_goodness": 1.592425140147739, + "pair_accuracy": 0.9373555555449592, + "positive_goodness": 2.98869778298272, + "runtime_seconds": 5.989475965499878 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.9722689702033996, + "lr": 0.03, + "negative_goodness": 1.5920691163169014, + "pair_accuracy": 0.9372222222010295, + "positive_goodness": 2.9851054923587377, + "runtime_seconds": 5.993381977081299 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.9841771196577284, + "lr": 0.03, + "negative_goodness": 1.600395458624098, + "pair_accuracy": 0.9337999999788072, + "positive_goodness": 2.9726584322187635, + "runtime_seconds": 5.983206272125244 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.952827390437656, + "lr": 0.03, + "negative_goodness": 1.5768287056393093, + "pair_accuracy": 0.9476444444974264, + "positive_goodness": 3.0435272327423095, + "runtime_seconds": 5.987974166870117 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9687997609880236, + "lr": 0.03, + "negative_goodness": 1.585767903751797, + "pair_accuracy": 0.9429333333121406, + "positive_goodness": 3.0114864821539986, + "runtime_seconds": 5.98924708366394 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.9706575076421102, + "lr": 0.03, + "negative_goodness": 1.5900301490995619, + "pair_accuracy": 0.9430222222222222, + "positive_goodness": 3.009912428410848, + "runtime_seconds": 5.988724946975708 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9777957746399774, + "lr": 0.03, + "negative_goodness": 1.5941182352383931, + "pair_accuracy": 0.9354, + "positive_goodness": 2.99299986199273, + "runtime_seconds": 5.994703769683838 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.9718369442621867, + "lr": 0.03, + "negative_goodness": 1.5880136258655124, + "pair_accuracy": 0.9396666666454739, + "positive_goodness": 2.998744737709893, + "runtime_seconds": 5.984786748886108 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9926770297580295, + "lr": 0.03, + "negative_goodness": 1.599764779747857, + "pair_accuracy": 0.9272000000105964, + "positive_goodness": 2.9585340937720406, + "runtime_seconds": 5.988449573516846 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.9576410236358642, + "lr": 0.03, + "negative_goodness": 1.5798669538285997, + "pair_accuracy": 0.9470888888782925, + "positive_goodness": 3.0436539843241373, + "runtime_seconds": 5.985822916030884 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.9512689460542467, + "lr": 0.03, + "negative_goodness": 1.5691325120078192, + "pair_accuracy": 0.9425777777989706, + "positive_goodness": 3.057830171839396, + "runtime_seconds": 5.985580682754517 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9545161602232192, + "lr": 0.03, + "negative_goodness": 1.5719103706359863, + "pair_accuracy": 0.9422444444444444, + "positive_goodness": 3.0525292229970296, + "runtime_seconds": 5.995856285095215 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.9701163140720791, + "lr": 0.03, + "negative_goodness": 1.5868076204511854, + "pair_accuracy": 0.9429555555343628, + "positive_goodness": 3.0039700173272026, + "runtime_seconds": 5.986920118331909 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9578141057332357, + "lr": 0.03, + "negative_goodness": 1.5693687444474962, + "pair_accuracy": 0.9405333333333333, + "positive_goodness": 3.0643607595655653, + "runtime_seconds": 5.98850154876709 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9406514992713928, + "lr": 0.03, + "negative_goodness": 1.5536981528176201, + "pair_accuracy": 0.9433555555237664, + "positive_goodness": 3.086789667214288, + "runtime_seconds": 5.987348556518555 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9521825499110752, + "lr": 0.03, + "negative_goodness": 1.5643756360583836, + "pair_accuracy": 0.9530666666666666, + "positive_goodness": 3.0750645878685847, + "runtime_seconds": 5.986860513687134 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.9526853174421522, + "lr": 0.03, + "negative_goodness": 1.561588542620341, + "pair_accuracy": 0.9451111110899183, + "positive_goodness": 3.0681092639923095, + "runtime_seconds": 5.99743390083313 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.9581019984563192, + "lr": 0.03, + "negative_goodness": 1.5657325627644856, + "pair_accuracy": 0.943200000026491, + "positive_goodness": 3.0598972651163736, + "runtime_seconds": 5.988001823425293 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.9551496694140964, + "lr": 0.03, + "negative_goodness": 1.5619863805770875, + "pair_accuracy": 0.9408000000105964, + "positive_goodness": 3.0749873738182916, + "runtime_seconds": 5.988463878631592 + } + ], + "layer": 5 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2069369535022312, + "lr": 0.03, + "negative_goodness": 1.878007424185011, + "pair_accuracy": 0.8386444444232517, + "positive_goodness": 2.4071879845089383, + "runtime_seconds": 6.072202920913696 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.0915344909032185, + "lr": 0.03, + "negative_goodness": 1.7735054194556341, + "pair_accuracy": 0.9285333333757189, + "positive_goodness": 2.6250424639383954, + "runtime_seconds": 6.072357654571533 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.0732878299607171, + "lr": 0.03, + "negative_goodness": 1.7514820478015476, + "pair_accuracy": 0.9346222222222222, + "positive_goodness": 2.6683965268029106, + "runtime_seconds": 6.078364372253418 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.05770603129069, + "lr": 0.03, + "negative_goodness": 1.7357088331858317, + "pair_accuracy": 0.9372222222116259, + "positive_goodness": 2.6996505723317465, + "runtime_seconds": 6.065775394439697 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.0473989458296034, + "lr": 0.03, + "negative_goodness": 1.7237313971201578, + "pair_accuracy": 0.9394888888782925, + "positive_goodness": 2.7289328232659233, + "runtime_seconds": 6.0661234855651855 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.037606486331092, + "lr": 0.03, + "negative_goodness": 1.7133918589909871, + "pair_accuracy": 0.9358444444338481, + "positive_goodness": 2.7471106541103785, + "runtime_seconds": 6.07236647605896 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.0584386195924547, + "lr": 0.03, + "negative_goodness": 1.7242147430207995, + "pair_accuracy": 0.9277333333015442, + "positive_goodness": 2.7116743752797445, + "runtime_seconds": 6.070155382156372 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0457144526481628, + "lr": 0.03, + "negative_goodness": 1.7180665937423707, + "pair_accuracy": 0.9348222222116258, + "positive_goodness": 2.738128447002835, + "runtime_seconds": 6.076332092285156 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.0255863619486492, + "lr": 0.03, + "negative_goodness": 1.699936665725708, + "pair_accuracy": 0.9464444444974264, + "positive_goodness": 2.7851904121398925, + "runtime_seconds": 6.070559024810791 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.0242784050835503, + "lr": 0.03, + "negative_goodness": 1.69669497856564, + "pair_accuracy": 0.9442000000105963, + "positive_goodness": 2.789823032675849, + "runtime_seconds": 6.068063974380493 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.015664781750573, + "lr": 0.03, + "negative_goodness": 1.686445381779141, + "pair_accuracy": 0.9453333333121405, + "positive_goodness": 2.8125445082770453, + "runtime_seconds": 6.071450471878052 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.0195351774957444, + "lr": 0.03, + "negative_goodness": 1.6885072526507907, + "pair_accuracy": 0.9416444444338481, + "positive_goodness": 2.810015179655287, + "runtime_seconds": 6.071309566497803 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.00817935095893, + "lr": 0.03, + "negative_goodness": 1.6779471594492594, + "pair_accuracy": 0.9409111111111111, + "positive_goodness": 2.8320267719268797, + "runtime_seconds": 6.071361541748047 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.007525092909071, + "lr": 0.03, + "negative_goodness": 1.6830598696814643, + "pair_accuracy": 0.9468444444709354, + "positive_goodness": 2.838360868390401, + "runtime_seconds": 6.073633909225464 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.0145651638348898, + "lr": 0.03, + "negative_goodness": 1.6810208875656127, + "pair_accuracy": 0.9421777777671814, + "positive_goodness": 2.818259706285265, + "runtime_seconds": 6.072213888168335 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.0037081009229025, + "lr": 0.03, + "negative_goodness": 1.6738623853471544, + "pair_accuracy": 0.9443999999682109, + "positive_goodness": 2.8474527298397487, + "runtime_seconds": 6.07149338722229 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.0021930273479887, + "lr": 0.03, + "negative_goodness": 1.670481028535631, + "pair_accuracy": 0.9456444444338481, + "positive_goodness": 2.8536205819447837, + "runtime_seconds": 6.073810338973999 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.007950345993042, + "lr": 0.03, + "negative_goodness": 1.6785386692470974, + "pair_accuracy": 0.9394888888570997, + "positive_goodness": 2.839594729593065, + "runtime_seconds": 6.065096616744995 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.0088157450887891, + "lr": 0.03, + "negative_goodness": 1.6750930501514012, + "pair_accuracy": 0.9404888888782925, + "positive_goodness": 2.8309512388441296, + "runtime_seconds": 6.073385238647461 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.9921604364183214, + "lr": 0.03, + "negative_goodness": 1.6598668162027994, + "pair_accuracy": 0.9460444444232516, + "positive_goodness": 2.8794004062652587, + "runtime_seconds": 6.070342302322388 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.9971431394471063, + "lr": 0.03, + "negative_goodness": 1.6693712210125393, + "pair_accuracy": 0.9492222222116259, + "positive_goodness": 2.863982703781128, + "runtime_seconds": 6.074413061141968 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.0021582485304938, + "lr": 0.03, + "negative_goodness": 1.6662490332285562, + "pair_accuracy": 0.9422666666666667, + "positive_goodness": 2.859551836437649, + "runtime_seconds": 6.073232173919678 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.9976600789176093, + "lr": 0.03, + "negative_goodness": 1.6613353423436483, + "pair_accuracy": 0.9471111111005147, + "positive_goodness": 2.8707634510040285, + "runtime_seconds": 6.070436239242554 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.9903258057700263, + "lr": 0.03, + "negative_goodness": 1.653003014416165, + "pair_accuracy": 0.9471777777459887, + "positive_goodness": 2.893606533135308, + "runtime_seconds": 6.071918487548828 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.009785226143731, + "lr": 0.03, + "negative_goodness": 1.6677443305333455, + "pair_accuracy": 0.9366444444444444, + "positive_goodness": 2.8462919802771673, + "runtime_seconds": 6.073101997375488 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9885845294528537, + "lr": 0.03, + "negative_goodness": 1.650520522371928, + "pair_accuracy": 0.9457777778201633, + "positive_goodness": 2.8906780707677204, + "runtime_seconds": 6.071959495544434 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.9841681716071234, + "lr": 0.03, + "negative_goodness": 1.647103882514106, + "pair_accuracy": 0.950422222190433, + "positive_goodness": 2.915805278438992, + "runtime_seconds": 6.072985887527466 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9921259487893846, + "lr": 0.03, + "negative_goodness": 1.653497701687283, + "pair_accuracy": 0.9445777777777777, + "positive_goodness": 2.887381821187337, + "runtime_seconds": 6.073259115219116 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.9848030892690023, + "lr": 0.03, + "negative_goodness": 1.64258929977417, + "pair_accuracy": 0.9477333333757189, + "positive_goodness": 2.913162755330404, + "runtime_seconds": 6.072667598724365 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9919881761762831, + "lr": 0.03, + "negative_goodness": 1.6550657899008856, + "pair_accuracy": 0.946066666677263, + "positive_goodness": 2.893144809850057, + "runtime_seconds": 6.07134747505188 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.9999136259502834, + "lr": 0.03, + "negative_goodness": 1.652131803618537, + "pair_accuracy": 0.9444888888676961, + "positive_goodness": 2.8744963713328042, + "runtime_seconds": 6.070787191390991 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.960671629386478, + "lr": 0.03, + "negative_goodness": 1.6201310202704535, + "pair_accuracy": 0.9545333333227369, + "positive_goodness": 2.966533075290256, + "runtime_seconds": 6.0723631381988525 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9847816694471572, + "lr": 0.03, + "negative_goodness": 1.6451870479159885, + "pair_accuracy": 0.9456444444232517, + "positive_goodness": 2.914908781094021, + "runtime_seconds": 6.070354223251343 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.9970335137155321, + "lr": 0.03, + "negative_goodness": 1.6522415831247965, + "pair_accuracy": 0.9473111111005147, + "positive_goodness": 2.8823232448154026, + "runtime_seconds": 6.071749448776245 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9961873066584269, + "lr": 0.03, + "negative_goodness": 1.6535688753975761, + "pair_accuracy": 0.9442444444444444, + "positive_goodness": 2.8842692738427056, + "runtime_seconds": 6.0727386474609375 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.0037440880033706, + "lr": 0.03, + "negative_goodness": 1.657117398791843, + "pair_accuracy": 0.9377999999788073, + "positive_goodness": 2.8660266600290933, + "runtime_seconds": 6.07116174697876 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.0063784518877665, + "lr": 0.03, + "negative_goodness": 1.6601487005021838, + "pair_accuracy": 0.9383333333015442, + "positive_goodness": 2.8592266990237767, + "runtime_seconds": 6.073184967041016 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.9901651751836141, + "lr": 0.03, + "negative_goodness": 1.648253130531311, + "pair_accuracy": 0.9438222222222222, + "positive_goodness": 2.9074770587497287, + "runtime_seconds": 6.073035001754761 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.967422561094496, + "lr": 0.03, + "negative_goodness": 1.6187495983759561, + "pair_accuracy": 0.9461777777777778, + "positive_goodness": 2.967290100309584, + "runtime_seconds": 6.073230981826782 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.9753550059000651, + "lr": 0.03, + "negative_goodness": 1.6301390432781644, + "pair_accuracy": 0.9489777777671814, + "positive_goodness": 2.9507186506059435, + "runtime_seconds": 6.074027061462402 + } + ], + "layer": 6 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1047713654624092, + "lr": 0.03, + "negative_goodness": 1.6705051990509032, + "pair_accuracy": 0.8816222222116259, + "positive_goodness": 2.602321488486396, + "runtime_seconds": 5.986935615539551 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.016976987944709, + "lr": 0.03, + "negative_goodness": 1.631722716331482, + "pair_accuracy": 0.9128222222010295, + "positive_goodness": 2.832969377178616, + "runtime_seconds": 5.988238573074341 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.9977135389010111, + "lr": 0.03, + "negative_goodness": 1.6069021512561374, + "pair_accuracy": 0.9187999999788072, + "positive_goodness": 2.886049237569173, + "runtime_seconds": 5.984768867492676 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.9943932462268406, + "lr": 0.03, + "negative_goodness": 1.5976091989305283, + "pair_accuracy": 0.9140000000105963, + "positive_goodness": 2.9023216100904676, + "runtime_seconds": 5.984936714172363 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.9874814539379544, + "lr": 0.03, + "negative_goodness": 1.5974239786571927, + "pair_accuracy": 0.9172444444550408, + "positive_goodness": 2.909827044169108, + "runtime_seconds": 5.989271879196167 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.9741817536989847, + "lr": 0.03, + "negative_goodness": 1.5764076160642835, + "pair_accuracy": 0.924911111164093, + "positive_goodness": 2.9777524338616264, + "runtime_seconds": 5.985894441604614 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.9711338120778402, + "lr": 0.03, + "negative_goodness": 1.5747888959248861, + "pair_accuracy": 0.9204444444762336, + "positive_goodness": 2.969945328860813, + "runtime_seconds": 5.987898826599121 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.9701211907068888, + "lr": 0.03, + "negative_goodness": 1.5662317122353449, + "pair_accuracy": 0.9218222222222222, + "positive_goodness": 2.981061100302802, + "runtime_seconds": 5.990257978439331 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.9521167525185479, + "lr": 0.03, + "negative_goodness": 1.561932893562317, + "pair_accuracy": 0.9317999999788072, + "positive_goodness": 3.0328010031806096, + "runtime_seconds": 5.987656116485596 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.9671788024478488, + "lr": 0.03, + "negative_goodness": 1.5661966579225328, + "pair_accuracy": 0.9207333333333333, + "positive_goodness": 2.983630551486545, + "runtime_seconds": 5.987100124359131 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.9431731047736274, + "lr": 0.03, + "negative_goodness": 1.5414130928675334, + "pair_accuracy": 0.9303777777459886, + "positive_goodness": 3.0535691778818768, + "runtime_seconds": 5.988179683685303 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.9127565748214722, + "lr": 0.03, + "negative_goodness": 1.5096803912268744, + "pair_accuracy": 0.9335777778307597, + "positive_goodness": 3.137108304553562, + "runtime_seconds": 5.988799810409546 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.9532928494983249, + "lr": 0.03, + "negative_goodness": 1.5432824754503038, + "pair_accuracy": 0.9244222222222223, + "positive_goodness": 3.055563843366835, + "runtime_seconds": 5.988455772399902 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.9423422786924575, + "lr": 0.03, + "negative_goodness": 1.5319547372394138, + "pair_accuracy": 0.9252444444550408, + "positive_goodness": 3.057128486336602, + "runtime_seconds": 5.987005233764648 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.9383319293340047, + "lr": 0.03, + "negative_goodness": 1.527830424118042, + "pair_accuracy": 0.9273999999788073, + "positive_goodness": 3.0833778854793974, + "runtime_seconds": 5.98850417137146 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.9378395057466296, + "lr": 0.03, + "negative_goodness": 1.5310434899012249, + "pair_accuracy": 0.9292222222646077, + "positive_goodness": 3.073428839577569, + "runtime_seconds": 5.988821268081665 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.9205313366783989, + "lr": 0.03, + "negative_goodness": 1.5096513600879244, + "pair_accuracy": 0.9328444444126553, + "positive_goodness": 3.128542161432902, + "runtime_seconds": 5.991384744644165 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.9310264198144277, + "lr": 0.03, + "negative_goodness": 1.5177855179892645, + "pair_accuracy": 0.9304888888888889, + "positive_goodness": 3.09437004614936, + "runtime_seconds": 5.9875733852386475 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.9294039735158285, + "lr": 0.03, + "negative_goodness": 1.5174929947747124, + "pair_accuracy": 0.9257555555767483, + "positive_goodness": 3.1104327276441786, + "runtime_seconds": 5.982976913452148 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.9491744428793589, + "lr": 0.03, + "negative_goodness": 1.5340221051745944, + "pair_accuracy": 0.9218888888888889, + "positive_goodness": 3.049798480139838, + "runtime_seconds": 5.982738018035889 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.9352030433019002, + "lr": 0.03, + "negative_goodness": 1.517447774442037, + "pair_accuracy": 0.9263777777989706, + "positive_goodness": 3.105410221057468, + "runtime_seconds": 5.987628698348999 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.9320375057432386, + "lr": 0.03, + "negative_goodness": 1.5153394739574857, + "pair_accuracy": 0.926222222190433, + "positive_goodness": 3.095835459560818, + "runtime_seconds": 5.985572338104248 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.9311841476864285, + "lr": 0.03, + "negative_goodness": 1.5237145315382217, + "pair_accuracy": 0.9341999999894036, + "positive_goodness": 3.1003103837331136, + "runtime_seconds": 5.9819793701171875 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.912362519370185, + "lr": 0.03, + "negative_goodness": 1.5012026485443115, + "pair_accuracy": 0.9338222222434149, + "positive_goodness": 3.174440860536363, + "runtime_seconds": 5.983969688415527 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.9105262146843804, + "lr": 0.03, + "negative_goodness": 1.4983820819642808, + "pair_accuracy": 0.9330222222222222, + "positive_goodness": 3.150548135842217, + "runtime_seconds": 5.984005689620972 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9031359250704447, + "lr": 0.03, + "negative_goodness": 1.4875914692348904, + "pair_accuracy": 0.9338666666560703, + "positive_goodness": 3.1700603759765626, + "runtime_seconds": 5.986890554428101 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.9099008700264825, + "lr": 0.03, + "negative_goodness": 1.4936886887020535, + "pair_accuracy": 0.9379111111534967, + "positive_goodness": 3.2135098967658147, + "runtime_seconds": 5.985763072967529 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9165781401210361, + "lr": 0.03, + "negative_goodness": 1.503474468697442, + "pair_accuracy": 0.9362444444232517, + "positive_goodness": 3.1406457023620606, + "runtime_seconds": 5.9881272315979 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.9089858049392701, + "lr": 0.03, + "negative_goodness": 1.4858922930823433, + "pair_accuracy": 0.9309777777777778, + "positive_goodness": 3.1716213034735787, + "runtime_seconds": 5.988808631896973 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9295066186269124, + "lr": 0.03, + "negative_goodness": 1.5103753222677443, + "pair_accuracy": 0.9228222222116258, + "positive_goodness": 3.115150204891629, + "runtime_seconds": 5.984308958053589 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.9192026178783841, + "lr": 0.03, + "negative_goodness": 1.5064749988132053, + "pair_accuracy": 0.9323333333121405, + "positive_goodness": 3.1332605503929987, + "runtime_seconds": 5.985523700714111 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.9167141386244032, + "lr": 0.03, + "negative_goodness": 1.5056812538994684, + "pair_accuracy": 0.9327555555555556, + "positive_goodness": 3.148329245461358, + "runtime_seconds": 5.987178802490234 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9223440536075168, + "lr": 0.03, + "negative_goodness": 1.494900909529792, + "pair_accuracy": 0.9310666666560703, + "positive_goodness": 3.1124453901926676, + "runtime_seconds": 5.987573623657227 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.9166980059093899, + "lr": 0.03, + "negative_goodness": 1.5045458637873332, + "pair_accuracy": 0.9308222222222222, + "positive_goodness": 3.1543955130683052, + "runtime_seconds": 5.987436532974243 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9482819736268785, + "lr": 0.03, + "negative_goodness": 1.5264000133726332, + "pair_accuracy": 0.9218222221904331, + "positive_goodness": 3.0546964040544298, + "runtime_seconds": 5.987787246704102 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9267946164131164, + "lr": 0.03, + "negative_goodness": 1.5136960118823581, + "pair_accuracy": 0.9272666666560703, + "positive_goodness": 3.112067362128364, + "runtime_seconds": 5.986567735671997 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9114419485727946, + "lr": 0.03, + "negative_goodness": 1.4833301217185126, + "pair_accuracy": 0.9347777777989705, + "positive_goodness": 3.1738905006196765, + "runtime_seconds": 5.9867517948150635 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.9168089244630602, + "lr": 0.03, + "negative_goodness": 1.4999325443267821, + "pair_accuracy": 0.9342888889418708, + "positive_goodness": 3.162919767464532, + "runtime_seconds": 5.986276149749756 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.902870134947035, + "lr": 0.03, + "negative_goodness": 1.4803310693105063, + "pair_accuracy": 0.93404444445769, + "positive_goodness": 3.200960938051012, + "runtime_seconds": 5.985558271408081 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.912474150774214, + "lr": 0.03, + "negative_goodness": 1.482125117641025, + "pair_accuracy": 0.934355555597941, + "positive_goodness": 3.1629071749369304, + "runtime_seconds": 5.984673023223877 + } + ], + "layer": 7 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1020045685450237, + "lr": 0.03, + "negative_goodness": 1.7628332803514268, + "pair_accuracy": 0.860822222243415, + "positive_goodness": 2.6932270421346027, + "runtime_seconds": 6.0700225830078125 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.9492320774820115, + "lr": 0.03, + "negative_goodness": 1.6018734668307835, + "pair_accuracy": 0.9216000000423855, + "positive_goodness": 2.9815992197248673, + "runtime_seconds": 6.070222854614258 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.928488591893514, + "lr": 0.03, + "negative_goodness": 1.5785449941847058, + "pair_accuracy": 0.926844444460339, + "positive_goodness": 3.0363014639960393, + "runtime_seconds": 6.068427562713623 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8991748455789355, + "lr": 0.03, + "negative_goodness": 1.527305171087053, + "pair_accuracy": 0.9369111110899183, + "positive_goodness": 3.1129736905415855, + "runtime_seconds": 6.06777548789978 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.9281481657875909, + "lr": 0.03, + "negative_goodness": 1.560334613249037, + "pair_accuracy": 0.9268444444232516, + "positive_goodness": 3.05851059328715, + "runtime_seconds": 6.069596767425537 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.9159989105118646, + "lr": 0.03, + "negative_goodness": 1.536494003868103, + "pair_accuracy": 0.9262000000529819, + "positive_goodness": 3.0850818045722113, + "runtime_seconds": 6.06955623626709 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.9055507817798191, + "lr": 0.03, + "negative_goodness": 1.534139963552687, + "pair_accuracy": 0.9230666666666667, + "positive_goodness": 3.1073875303056506, + "runtime_seconds": 6.070082902908325 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.9077624881108602, + "lr": 0.03, + "negative_goodness": 1.517895080142551, + "pair_accuracy": 0.922155555597941, + "positive_goodness": 3.1214220897250704, + "runtime_seconds": 6.068735122680664 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.8704601972897847, + "lr": 0.03, + "negative_goodness": 1.4886600719239977, + "pair_accuracy": 0.9326, + "positive_goodness": 3.202496231842041, + "runtime_seconds": 6.069361925125122 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8567272817187839, + "lr": 0.03, + "negative_goodness": 1.4702416769451565, + "pair_accuracy": 0.9394000000529819, + "positive_goodness": 3.266849716737535, + "runtime_seconds": 6.069270610809326 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.8713401335610284, + "lr": 0.03, + "negative_goodness": 1.4847298002031115, + "pair_accuracy": 0.9323777778201633, + "positive_goodness": 3.2272797509087456, + "runtime_seconds": 6.06751561164856 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.8798764757050408, + "lr": 0.03, + "negative_goodness": 1.493644551255968, + "pair_accuracy": 0.9339333333333333, + "positive_goodness": 3.2029139891730414, + "runtime_seconds": 6.071386814117432 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.8679597430441115, + "lr": 0.03, + "negative_goodness": 1.4752227288352118, + "pair_accuracy": 0.9328666666666666, + "positive_goodness": 3.222412074110243, + "runtime_seconds": 6.070573091506958 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.8723996558083428, + "lr": 0.03, + "negative_goodness": 1.4763035846074422, + "pair_accuracy": 0.9324444444232517, + "positive_goodness": 3.228742024951511, + "runtime_seconds": 6.0685715675354 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.8628667241626315, + "lr": 0.03, + "negative_goodness": 1.4541002182218763, + "pair_accuracy": 0.9271777777671814, + "positive_goodness": 3.2528708149380154, + "runtime_seconds": 6.069413423538208 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.8518355544408163, + "lr": 0.03, + "negative_goodness": 1.4543006882773506, + "pair_accuracy": 0.9396666666560702, + "positive_goodness": 3.2889412854088675, + "runtime_seconds": 6.072787523269653 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.8351003908051385, + "lr": 0.03, + "negative_goodness": 1.427501477792528, + "pair_accuracy": 0.9333333333333333, + "positive_goodness": 3.3562607432047527, + "runtime_seconds": 6.076222658157349 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.8590668502489726, + "lr": 0.03, + "negative_goodness": 1.4539884293662177, + "pair_accuracy": 0.9288222222116258, + "positive_goodness": 3.2642128629048663, + "runtime_seconds": 6.075667381286621 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.8774957138273451, + "lr": 0.03, + "negative_goodness": 1.4652082245084974, + "pair_accuracy": 0.9267555555449591, + "positive_goodness": 3.241902503204346, + "runtime_seconds": 6.074560642242432 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.8579738805241055, + "lr": 0.03, + "negative_goodness": 1.464208017486996, + "pair_accuracy": 0.9350222222222222, + "positive_goodness": 3.2834549722459583, + "runtime_seconds": 6.074399948120117 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.8398716508441502, + "lr": 0.03, + "negative_goodness": 1.4244015479193792, + "pair_accuracy": 0.9332666666666667, + "positive_goodness": 3.307374367014567, + "runtime_seconds": 6.070643901824951 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.8522136627939012, + "lr": 0.03, + "negative_goodness": 1.4459857014867994, + "pair_accuracy": 0.927577777756585, + "positive_goodness": 3.304133714675903, + "runtime_seconds": 6.079803943634033 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.8438733010821873, + "lr": 0.03, + "negative_goodness": 1.4317530353546142, + "pair_accuracy": 0.9376444444338481, + "positive_goodness": 3.3201043831719295, + "runtime_seconds": 6.076130390167236 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.8629701731152004, + "lr": 0.03, + "negative_goodness": 1.4571535093943278, + "pair_accuracy": 0.9346222222222222, + "positive_goodness": 3.2761405941433375, + "runtime_seconds": 6.0749053955078125 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.8338736295700073, + "lr": 0.03, + "negative_goodness": 1.4187097135119968, + "pair_accuracy": 0.9316444444338481, + "positive_goodness": 3.3564652384016247, + "runtime_seconds": 6.077089786529541 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.8451450922648112, + "lr": 0.03, + "negative_goodness": 1.4319051986058553, + "pair_accuracy": 0.9278444444550408, + "positive_goodness": 3.323463198216756, + "runtime_seconds": 6.073301553726196 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.842925273068746, + "lr": 0.03, + "negative_goodness": 1.427319827885098, + "pair_accuracy": 0.9334888888782925, + "positive_goodness": 3.334396543841892, + "runtime_seconds": 6.073834419250488 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.830776940536499, + "lr": 0.03, + "negative_goodness": 1.4206801030688816, + "pair_accuracy": 0.9405333333863153, + "positive_goodness": 3.3535934253692625, + "runtime_seconds": 6.072809934616089 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.831003140449524, + "lr": 0.03, + "negative_goodness": 1.4146732489267986, + "pair_accuracy": 0.9350222221904331, + "positive_goodness": 3.3705124121771917, + "runtime_seconds": 6.07473611831665 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.840055184533861, + "lr": 0.03, + "negative_goodness": 1.42825186314053, + "pair_accuracy": 0.9355111110793219, + "positive_goodness": 3.330011825603909, + "runtime_seconds": 6.076419115066528 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.867908125866784, + "lr": 0.03, + "negative_goodness": 1.4557019358740912, + "pair_accuracy": 0.9274222222222223, + "positive_goodness": 3.2820441030714247, + "runtime_seconds": 6.071070671081543 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.8521756847699483, + "lr": 0.03, + "negative_goodness": 1.426819183921814, + "pair_accuracy": 0.9251777777671814, + "positive_goodness": 3.3017540806240504, + "runtime_seconds": 6.0731096267700195 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.8371311879581875, + "lr": 0.03, + "negative_goodness": 1.4196140237066481, + "pair_accuracy": 0.9340666666560703, + "positive_goodness": 3.3609882208506265, + "runtime_seconds": 6.0734703540802 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.8540305017471314, + "lr": 0.03, + "negative_goodness": 1.4412937421162924, + "pair_accuracy": 0.9335111111111111, + "positive_goodness": 3.2844327359729344, + "runtime_seconds": 6.073189735412598 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.8291359105004205, + "lr": 0.03, + "negative_goodness": 1.407533947414822, + "pair_accuracy": 0.9316444444444445, + "positive_goodness": 3.3817940876430934, + "runtime_seconds": 6.075547695159912 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.8605761903868782, + "lr": 0.03, + "negative_goodness": 1.4368469270070394, + "pair_accuracy": 0.9255111110899183, + "positive_goodness": 3.3015544641282824, + "runtime_seconds": 6.071113109588623 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.853195690366957, + "lr": 0.03, + "negative_goodness": 1.4270244799296061, + "pair_accuracy": 0.9235777777777778, + "positive_goodness": 3.3364019151475692, + "runtime_seconds": 6.0725390911102295 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.862823566034105, + "lr": 0.03, + "negative_goodness": 1.4427758745829264, + "pair_accuracy": 0.924777777756585, + "positive_goodness": 3.2939404490152997, + "runtime_seconds": 6.075156211853027 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.8144843714184231, + "lr": 0.03, + "negative_goodness": 1.396576434135437, + "pair_accuracy": 0.9436222222222223, + "positive_goodness": 3.4234986339992948, + "runtime_seconds": 6.07093358039856 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.8494788181304932, + "lr": 0.03, + "negative_goodness": 1.4337654742770725, + "pair_accuracy": 0.9385999999682109, + "positive_goodness": 3.3083818290286593, + "runtime_seconds": 6.071460723876953 + } + ], + "layer": 8 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9252230487611559, + "lr": 0.03, + "negative_goodness": 1.5421253234439427, + "pair_accuracy": 0.924111111079322, + "positive_goodness": 3.075616000450982, + "runtime_seconds": 5.993659257888794 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.9073852990468343, + "lr": 0.03, + "negative_goodness": 1.5477063755459255, + "pair_accuracy": 0.9245333333333333, + "positive_goodness": 3.160488071568807, + "runtime_seconds": 5.991990804672241 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8500285108778212, + "lr": 0.03, + "negative_goodness": 1.4833457489861381, + "pair_accuracy": 0.9310222222116259, + "positive_goodness": 3.3122185432010225, + "runtime_seconds": 5.991520166397095 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8836794876098633, + "lr": 0.03, + "negative_goodness": 1.512553148566352, + "pair_accuracy": 0.9184222222328186, + "positive_goodness": 3.2444915763431124, + "runtime_seconds": 5.991036653518677 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8751606356514825, + "lr": 0.03, + "negative_goodness": 1.5018275901794433, + "pair_accuracy": 0.9186666666454739, + "positive_goodness": 3.2898421436309815, + "runtime_seconds": 5.9918413162231445 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.8525611279169718, + "lr": 0.03, + "negative_goodness": 1.4780015188005236, + "pair_accuracy": 0.9288666666348775, + "positive_goodness": 3.3319861938052706, + "runtime_seconds": 5.995177984237671 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.8581229745864868, + "lr": 0.03, + "negative_goodness": 1.4991639843622844, + "pair_accuracy": 0.9244888888676961, + "positive_goodness": 3.3266509593115914, + "runtime_seconds": 5.991699934005737 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.8601771071222093, + "lr": 0.03, + "negative_goodness": 1.4904380973921882, + "pair_accuracy": 0.9225555555555556, + "positive_goodness": 3.344894269095527, + "runtime_seconds": 5.9884114265441895 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.8787720206684536, + "lr": 0.03, + "negative_goodness": 1.5144748461617363, + "pair_accuracy": 0.9178222222222222, + "positive_goodness": 3.282415651957194, + "runtime_seconds": 5.988785743713379 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8621221910476684, + "lr": 0.03, + "negative_goodness": 1.496292439227634, + "pair_accuracy": 0.920711111164093, + "positive_goodness": 3.341027947828505, + "runtime_seconds": 5.991317987442017 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.8538880255169339, + "lr": 0.03, + "negative_goodness": 1.4821043611314562, + "pair_accuracy": 0.9229333333121406, + "positive_goodness": 3.339417123921712, + "runtime_seconds": 5.993986129760742 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.8513538282606337, + "lr": 0.03, + "negative_goodness": 1.4880363442314997, + "pair_accuracy": 0.9264888889418708, + "positive_goodness": 3.3747141927931046, + "runtime_seconds": 5.994272470474243 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.810726534181171, + "lr": 0.03, + "negative_goodness": 1.4388472917662727, + "pair_accuracy": 0.9357333333333333, + "positive_goodness": 3.490012640974257, + "runtime_seconds": 5.99190616607666 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.8506154332796733, + "lr": 0.03, + "negative_goodness": 1.4724765898386638, + "pair_accuracy": 0.9207999999788072, + "positive_goodness": 3.410022364171346, + "runtime_seconds": 5.993544101715088 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.8452483251041836, + "lr": 0.03, + "negative_goodness": 1.469147465218438, + "pair_accuracy": 0.9261111111111111, + "positive_goodness": 3.402167128499349, + "runtime_seconds": 5.9945526123046875 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.840870250638326, + "lr": 0.03, + "negative_goodness": 1.4466584872775607, + "pair_accuracy": 0.9131333333545261, + "positive_goodness": 3.42511480140686, + "runtime_seconds": 5.992929458618164 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.8440263682047526, + "lr": 0.03, + "negative_goodness": 1.468047823439704, + "pair_accuracy": 0.9296444444232517, + "positive_goodness": 3.429710871887207, + "runtime_seconds": 5.991706848144531 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.8532044361538357, + "lr": 0.03, + "negative_goodness": 1.4663492472966513, + "pair_accuracy": 0.9211555555343628, + "positive_goodness": 3.4028311371273463, + "runtime_seconds": 5.991961479187012 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.8142939900928073, + "lr": 0.03, + "negative_goodness": 1.4085986281394958, + "pair_accuracy": 0.9283111111111111, + "positive_goodness": 3.5179487695482043, + "runtime_seconds": 5.992000102996826 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.8506964598973592, + "lr": 0.03, + "negative_goodness": 1.457425954691569, + "pair_accuracy": 0.9185111111005148, + "positive_goodness": 3.4234212147182888, + "runtime_seconds": 5.991859436035156 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.8415451057222154, + "lr": 0.03, + "negative_goodness": 1.4496082900153267, + "pair_accuracy": 0.9143111111376021, + "positive_goodness": 3.4359027148776584, + "runtime_seconds": 5.9919350147247314 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.8064213075637817, + "lr": 0.03, + "negative_goodness": 1.4204449804517958, + "pair_accuracy": 0.9303999999894036, + "positive_goodness": 3.5313755450778537, + "runtime_seconds": 5.993561029434204 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.8217925706863404, + "lr": 0.03, + "negative_goodness": 1.4196162192662556, + "pair_accuracy": 0.9242444444444444, + "positive_goodness": 3.521615335676405, + "runtime_seconds": 5.992141008377075 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.8220098312166002, + "lr": 0.03, + "negative_goodness": 1.4260145701726277, + "pair_accuracy": 0.9327555555555556, + "positive_goodness": 3.5006757882436115, + "runtime_seconds": 5.994147300720215 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.8845799657185872, + "lr": 0.03, + "negative_goodness": 1.5023331307729084, + "pair_accuracy": 0.9096222222646078, + "positive_goodness": 3.3200522799597847, + "runtime_seconds": 5.992006778717041 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.8182631401962704, + "lr": 0.03, + "negative_goodness": 1.4238248562494913, + "pair_accuracy": 0.9265777777777777, + "positive_goodness": 3.5200738868713377, + "runtime_seconds": 5.9933552742004395 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.8469810448434618, + "lr": 0.03, + "negative_goodness": 1.4528388444052802, + "pair_accuracy": 0.9164888888888889, + "positive_goodness": 3.446702257071601, + "runtime_seconds": 5.994479417800903 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.825912569363912, + "lr": 0.03, + "negative_goodness": 1.4275735221862793, + "pair_accuracy": 0.9200444444550409, + "positive_goodness": 3.501613513056437, + "runtime_seconds": 5.993117094039917 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.8213581595844692, + "lr": 0.03, + "negative_goodness": 1.4158523527993097, + "pair_accuracy": 0.9203777777777777, + "positive_goodness": 3.5014595917595757, + "runtime_seconds": 5.9922194480896 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.853421005736457, + "lr": 0.03, + "negative_goodness": 1.457511041853163, + "pair_accuracy": 0.9109333333545261, + "positive_goodness": 3.418493969768948, + "runtime_seconds": 5.994363307952881 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.8551224330478244, + "lr": 0.03, + "negative_goodness": 1.4567466590033638, + "pair_accuracy": 0.919155555597941, + "positive_goodness": 3.4092905641767715, + "runtime_seconds": 5.994402170181274 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.8220669362280104, + "lr": 0.03, + "negative_goodness": 1.420109809091356, + "pair_accuracy": 0.9267777777671814, + "positive_goodness": 3.5304730926089816, + "runtime_seconds": 5.993524551391602 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.8064127449247572, + "lr": 0.03, + "negative_goodness": 1.3960862805260552, + "pair_accuracy": 0.9265777777989705, + "positive_goodness": 3.6148565970102946, + "runtime_seconds": 5.991242408752441 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.8289019759760963, + "lr": 0.03, + "negative_goodness": 1.4184717717170716, + "pair_accuracy": 0.9184888888888889, + "positive_goodness": 3.473149604203966, + "runtime_seconds": 5.994807958602905 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.8406077211221059, + "lr": 0.03, + "negative_goodness": 1.4353287262068855, + "pair_accuracy": 0.9166222222222222, + "positive_goodness": 3.455549937099881, + "runtime_seconds": 5.994017601013184 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.8346524012989468, + "lr": 0.03, + "negative_goodness": 1.4306789912117852, + "pair_accuracy": 0.9174, + "positive_goodness": 3.483180759599474, + "runtime_seconds": 5.9919891357421875 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.834660606320699, + "lr": 0.03, + "negative_goodness": 1.4361849598566692, + "pair_accuracy": 0.9221555555343628, + "positive_goodness": 3.4552013250139026, + "runtime_seconds": 5.996401786804199 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.8137459458669026, + "lr": 0.03, + "negative_goodness": 1.415790456199646, + "pair_accuracy": 0.9202222222752041, + "positive_goodness": 3.521031880187988, + "runtime_seconds": 5.992907762527466 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.8331851607216729, + "lr": 0.03, + "negative_goodness": 1.4343686966790092, + "pair_accuracy": 0.919155555597941, + "positive_goodness": 3.4923780093722874, + "runtime_seconds": 5.9922871589660645 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.8334452665964762, + "lr": 0.03, + "negative_goodness": 1.4312845500098335, + "pair_accuracy": 0.9156222222222222, + "positive_goodness": 3.4830199651082356, + "runtime_seconds": 5.994171619415283 + } + ], + "layer": 9 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9890905321862963, + "lr": 0.03, + "negative_goodness": 1.6019476457383897, + "pair_accuracy": 0.8981111111111111, + "positive_goodness": 3.06274335776435, + "runtime_seconds": 6.075602293014526 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8732810376326243, + "lr": 0.03, + "negative_goodness": 1.4952655436197917, + "pair_accuracy": 0.9353555555555556, + "positive_goodness": 3.2324363563537597, + "runtime_seconds": 6.076925039291382 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8021117896185981, + "lr": 0.03, + "negative_goodness": 1.4031559561835394, + "pair_accuracy": 0.9512222222752041, + "positive_goodness": 3.416219871139526, + "runtime_seconds": 6.0743279457092285 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7697314586639404, + "lr": 0.03, + "negative_goodness": 1.3596199298858642, + "pair_accuracy": 0.9461555556085375, + "positive_goodness": 3.498901305347019, + "runtime_seconds": 6.075823068618774 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7754493910895454, + "lr": 0.03, + "negative_goodness": 1.3627476545969646, + "pair_accuracy": 0.9438222222222222, + "positive_goodness": 3.5047267249213325, + "runtime_seconds": 6.077818393707275 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7605254633373685, + "lr": 0.03, + "negative_goodness": 1.3680135048972235, + "pair_accuracy": 0.9575111111323039, + "positive_goodness": 3.5149522807227243, + "runtime_seconds": 6.076742649078369 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7420539406458537, + "lr": 0.03, + "negative_goodness": 1.3453533321380615, + "pair_accuracy": 0.9590666666799121, + "positive_goodness": 3.5789984555138483, + "runtime_seconds": 6.074727773666382 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7543877152548896, + "lr": 0.03, + "negative_goodness": 1.3381143490685357, + "pair_accuracy": 0.9532666666878594, + "positive_goodness": 3.541262004470825, + "runtime_seconds": 6.076810359954834 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7517229100227356, + "lr": 0.03, + "negative_goodness": 1.3319505395677356, + "pair_accuracy": 0.9439333333121406, + "positive_goodness": 3.547979349263509, + "runtime_seconds": 6.075562238693237 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.722237421242396, + "lr": 0.03, + "negative_goodness": 1.294380122894711, + "pair_accuracy": 0.9547777777777777, + "positive_goodness": 3.6500768900553386, + "runtime_seconds": 6.076650142669678 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7114437907377879, + "lr": 0.03, + "negative_goodness": 1.2715443696551854, + "pair_accuracy": 0.9566, + "positive_goodness": 3.700647439066569, + "runtime_seconds": 6.073492765426636 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7271749235365126, + "lr": 0.03, + "negative_goodness": 1.2997770890129936, + "pair_accuracy": 0.9482222222222222, + "positive_goodness": 3.6760531901041666, + "runtime_seconds": 6.075574636459351 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7016696636517843, + "lr": 0.03, + "negative_goodness": 1.2515966574774848, + "pair_accuracy": 0.9516444444338481, + "positive_goodness": 3.7375212923261856, + "runtime_seconds": 6.075350284576416 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7316251453081767, + "lr": 0.03, + "negative_goodness": 1.2944209300147163, + "pair_accuracy": 0.9527111111111111, + "positive_goodness": 3.6671283017052545, + "runtime_seconds": 6.075261116027832 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7189945008595785, + "lr": 0.03, + "negative_goodness": 1.2603947084214953, + "pair_accuracy": 0.9452888888782925, + "positive_goodness": 3.7435136762830945, + "runtime_seconds": 6.075253248214722 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7211734177483453, + "lr": 0.03, + "negative_goodness": 1.2699761622746786, + "pair_accuracy": 0.9528666666560703, + "positive_goodness": 3.69564802992079, + "runtime_seconds": 6.075754165649414 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7195421195560031, + "lr": 0.03, + "negative_goodness": 1.2662246457841662, + "pair_accuracy": 0.9455777777777777, + "positive_goodness": 3.70419994269477, + "runtime_seconds": 6.073704242706299 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7021967890739441, + "lr": 0.03, + "negative_goodness": 1.2382457555452984, + "pair_accuracy": 0.9529111111111112, + "positive_goodness": 3.781830401187473, + "runtime_seconds": 6.076978445053101 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6951017627398173, + "lr": 0.03, + "negative_goodness": 1.238087816174825, + "pair_accuracy": 0.9564000000317892, + "positive_goodness": 3.7843771968841553, + "runtime_seconds": 6.0731728076934814 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6938328893025716, + "lr": 0.03, + "negative_goodness": 1.2245993598514133, + "pair_accuracy": 0.9514222222222222, + "positive_goodness": 3.8336377605862086, + "runtime_seconds": 6.075738430023193 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7005297884464264, + "lr": 0.03, + "negative_goodness": 1.241312129804823, + "pair_accuracy": 0.9589555555555556, + "positive_goodness": 3.7787800181070965, + "runtime_seconds": 6.076793909072876 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6731307301097446, + "lr": 0.03, + "negative_goodness": 1.209531447982788, + "pair_accuracy": 0.9629333333333333, + "positive_goodness": 3.879496694691976, + "runtime_seconds": 6.07799220085144 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6967803355958727, + "lr": 0.03, + "negative_goodness": 1.2294261678907605, + "pair_accuracy": 0.9448444444974263, + "positive_goodness": 3.837099274741279, + "runtime_seconds": 6.0753631591796875 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7166289329369863, + "lr": 0.03, + "negative_goodness": 1.2516094962437947, + "pair_accuracy": 0.9456222222222223, + "positive_goodness": 3.729293309275309, + "runtime_seconds": 6.077150821685791 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.7139879804399278, + "lr": 0.03, + "negative_goodness": 1.2608631392584906, + "pair_accuracy": 0.9470222222540113, + "positive_goodness": 3.7619093352423776, + "runtime_seconds": 6.075390100479126 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7027071105533176, + "lr": 0.03, + "negative_goodness": 1.228983147875468, + "pair_accuracy": 0.9532222222222222, + "positive_goodness": 3.788905825551351, + "runtime_seconds": 6.077399492263794 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6742443373150295, + "lr": 0.03, + "negative_goodness": 1.2154707177268134, + "pair_accuracy": 0.9524, + "positive_goodness": 3.887509992896186, + "runtime_seconds": 6.07465934753418 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.696591824192471, + "lr": 0.03, + "negative_goodness": 1.2291403232574463, + "pair_accuracy": 0.9531333333121406, + "positive_goodness": 3.796242000113593, + "runtime_seconds": 6.071532249450684 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.651862729687161, + "lr": 0.03, + "negative_goodness": 1.1858953115675184, + "pair_accuracy": 0.9648888888570998, + "positive_goodness": 3.948794576348199, + "runtime_seconds": 6.0720202922821045 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6908868989573584, + "lr": 0.03, + "negative_goodness": 1.2194946820365058, + "pair_accuracy": 0.9532222222222222, + "positive_goodness": 3.839168114980062, + "runtime_seconds": 6.071333885192871 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6799391881095038, + "lr": 0.03, + "negative_goodness": 1.200992816374037, + "pair_accuracy": 0.9590222222222222, + "positive_goodness": 3.90738802702162, + "runtime_seconds": 6.079413414001465 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6810854959593878, + "lr": 0.03, + "negative_goodness": 1.2065426541222466, + "pair_accuracy": 0.9543111111111111, + "positive_goodness": 3.838090062544081, + "runtime_seconds": 6.071967840194702 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6678847272766961, + "lr": 0.03, + "negative_goodness": 1.1898766979641384, + "pair_accuracy": 0.9489111110899183, + "positive_goodness": 3.922656563398573, + "runtime_seconds": 6.071629047393799 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6778730167918735, + "lr": 0.03, + "negative_goodness": 1.1906973229726157, + "pair_accuracy": 0.951711111164093, + "positive_goodness": 3.888378163316515, + "runtime_seconds": 6.07435941696167 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6682064533127678, + "lr": 0.03, + "negative_goodness": 1.1731624277962578, + "pair_accuracy": 0.9533999999682109, + "positive_goodness": 3.946334639570448, + "runtime_seconds": 6.073904037475586 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6858460974799262, + "lr": 0.03, + "negative_goodness": 1.210319314596388, + "pair_accuracy": 0.950511111079322, + "positive_goodness": 3.8593234970516628, + "runtime_seconds": 6.082851409912109 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6486436084323459, + "lr": 0.03, + "negative_goodness": 1.1645399390856426, + "pair_accuracy": 0.9626666666666667, + "positive_goodness": 4.037427917438083, + "runtime_seconds": 6.070228099822998 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6591485054704878, + "lr": 0.03, + "negative_goodness": 1.163366610262129, + "pair_accuracy": 0.9580222222222222, + "positive_goodness": 3.9707039728800457, + "runtime_seconds": 6.071400880813599 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6562998566521538, + "lr": 0.03, + "negative_goodness": 1.1686889727804395, + "pair_accuracy": 0.9605333333333334, + "positive_goodness": 3.964452907477485, + "runtime_seconds": 6.069666624069214 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6598564396328396, + "lr": 0.03, + "negative_goodness": 1.174902680375841, + "pair_accuracy": 0.9587111111005148, + "positive_goodness": 3.971752312045627, + "runtime_seconds": 6.0701823234558105 + } + ], + "layer": 10 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9106370119730631, + "lr": 0.03, + "negative_goodness": 1.399916353225708, + "pair_accuracy": 0.8843555555714502, + "positive_goodness": 3.161377728483412, + "runtime_seconds": 5.764100074768066 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8187113421334161, + "lr": 0.03, + "negative_goodness": 1.3280047248840332, + "pair_accuracy": 0.9122666666825613, + "positive_goodness": 3.4527837754567465, + "runtime_seconds": 5.702158451080322 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8447445643319024, + "lr": 0.03, + "negative_goodness": 1.3336001342349582, + "pair_accuracy": 0.9059333333757189, + "positive_goodness": 3.4219687914530437, + "runtime_seconds": 5.744431495666504 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8117092312494913, + "lr": 0.03, + "negative_goodness": 1.317666572740343, + "pair_accuracy": 0.9178222222222222, + "positive_goodness": 3.490631975724962, + "runtime_seconds": 5.719973564147949 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8014655340512593, + "lr": 0.03, + "negative_goodness": 1.2954574620564778, + "pair_accuracy": 0.9161111111111111, + "positive_goodness": 3.543264004558987, + "runtime_seconds": 5.7465980052948 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7901647670004103, + "lr": 0.03, + "negative_goodness": 1.252397782198588, + "pair_accuracy": 0.9165777777777778, + "positive_goodness": 3.620525307973226, + "runtime_seconds": 5.747977018356323 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.762624009153578, + "lr": 0.03, + "negative_goodness": 1.237898624168502, + "pair_accuracy": 0.9271777777777778, + "positive_goodness": 3.6974511996799047, + "runtime_seconds": 5.750403165817261 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7796003316508399, + "lr": 0.03, + "negative_goodness": 1.253549135957824, + "pair_accuracy": 0.9237333333333333, + "positive_goodness": 3.6653653501722547, + "runtime_seconds": 5.74618935585022 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7871165150112576, + "lr": 0.03, + "negative_goodness": 1.2588836609098646, + "pair_accuracy": 0.9162888888888889, + "positive_goodness": 3.644911286968655, + "runtime_seconds": 5.748783826828003 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8019548162460327, + "lr": 0.03, + "negative_goodness": 1.2926832546340095, + "pair_accuracy": 0.9182666666454739, + "positive_goodness": 3.57136676097446, + "runtime_seconds": 5.726208209991455 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7906611551178826, + "lr": 0.03, + "negative_goodness": 1.2701714059140947, + "pair_accuracy": 0.9206222222222222, + "positive_goodness": 3.636381914477878, + "runtime_seconds": 5.755251407623291 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7865540919833713, + "lr": 0.03, + "negative_goodness": 1.2720130904091729, + "pair_accuracy": 0.9226444444656372, + "positive_goodness": 3.634465662299262, + "runtime_seconds": 5.7532453536987305 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7696931166436937, + "lr": 0.03, + "negative_goodness": 1.237937996641795, + "pair_accuracy": 0.9205555555555556, + "positive_goodness": 3.7138754042731392, + "runtime_seconds": 5.758692026138306 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7695717412524753, + "lr": 0.03, + "negative_goodness": 1.2457568622165256, + "pair_accuracy": 0.9239555555979411, + "positive_goodness": 3.7133022310468884, + "runtime_seconds": 5.753956317901611 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7692192128711277, + "lr": 0.03, + "negative_goodness": 1.2392812497297923, + "pair_accuracy": 0.9248666666560703, + "positive_goodness": 3.7232880932701957, + "runtime_seconds": 5.753502130508423 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7871191929923164, + "lr": 0.03, + "negative_goodness": 1.2514005773332384, + "pair_accuracy": 0.9153777777989706, + "positive_goodness": 3.6802447766621906, + "runtime_seconds": 5.7554357051849365 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7619459289550782, + "lr": 0.03, + "negative_goodness": 1.2249471339331732, + "pair_accuracy": 0.9239111111005147, + "positive_goodness": 3.768452564112345, + "runtime_seconds": 5.759975433349609 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7725860329522027, + "lr": 0.03, + "negative_goodness": 1.2347367746988933, + "pair_accuracy": 0.9166888889047835, + "positive_goodness": 3.7349421644422742, + "runtime_seconds": 5.754283905029297 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7419282669226328, + "lr": 0.03, + "negative_goodness": 1.2009725462383694, + "pair_accuracy": 0.9275333333333333, + "positive_goodness": 3.825543650054932, + "runtime_seconds": 5.753037691116333 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7607971534729004, + "lr": 0.03, + "negative_goodness": 1.2308227834489611, + "pair_accuracy": 0.9228444444391463, + "positive_goodness": 3.778584742779202, + "runtime_seconds": 5.756396532058716 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7326693570666843, + "lr": 0.03, + "negative_goodness": 1.1803621599409315, + "pair_accuracy": 0.9339555556085375, + "positive_goodness": 3.8612797172122533, + "runtime_seconds": 5.759248971939087 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7370321968396505, + "lr": 0.03, + "negative_goodness": 1.1999787329355875, + "pair_accuracy": 0.9340222222116258, + "positive_goodness": 3.864283059734768, + "runtime_seconds": 5.762724876403809 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7415303980297513, + "lr": 0.03, + "negative_goodness": 1.192693390613132, + "pair_accuracy": 0.9286888888570998, + "positive_goodness": 3.877204465145535, + "runtime_seconds": 5.756363153457642 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7595845201068454, + "lr": 0.03, + "negative_goodness": 1.2122657893922595, + "pair_accuracy": 0.9199777777565851, + "positive_goodness": 3.789339090177748, + "runtime_seconds": 5.756715536117554 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.7304526949776543, + "lr": 0.03, + "negative_goodness": 1.1795321429146661, + "pair_accuracy": 0.9338444444762336, + "positive_goodness": 3.8969675871955025, + "runtime_seconds": 5.716740369796753 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7490390420489841, + "lr": 0.03, + "negative_goodness": 1.2016959102630616, + "pair_accuracy": 0.929222222243415, + "positive_goodness": 3.837284291966756, + "runtime_seconds": 5.748619794845581 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7495264328638712, + "lr": 0.03, + "negative_goodness": 1.2034323207643296, + "pair_accuracy": 0.9255333333863153, + "positive_goodness": 3.83754851175944, + "runtime_seconds": 5.749948501586914 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.7507361323436101, + "lr": 0.03, + "negative_goodness": 1.2014748363521364, + "pair_accuracy": 0.9248, + "positive_goodness": 3.8406530301411945, + "runtime_seconds": 5.745687007904053 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.7651537483639187, + "lr": 0.03, + "negative_goodness": 1.213175236829122, + "pair_accuracy": 0.9163555555767483, + "positive_goodness": 3.815087097422282, + "runtime_seconds": 5.747126817703247 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7700360563225216, + "lr": 0.03, + "negative_goodness": 1.2193270669619243, + "pair_accuracy": 0.9196444444126552, + "positive_goodness": 3.7779588674757214, + "runtime_seconds": 5.748195648193359 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.7537023237546285, + "lr": 0.03, + "negative_goodness": 1.2080131790293587, + "pair_accuracy": 0.9239777777671814, + "positive_goodness": 3.837610082075331, + "runtime_seconds": 5.748215675354004 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7406288306183285, + "lr": 0.03, + "negative_goodness": 1.1885924572202895, + "pair_accuracy": 0.9239777777777778, + "positive_goodness": 3.891186388397217, + "runtime_seconds": 5.749351739883423 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.7275124936951531, + "lr": 0.03, + "negative_goodness": 1.174076169986195, + "pair_accuracy": 0.9312222222222222, + "positive_goodness": 3.9378487415737577, + "runtime_seconds": 5.741469860076904 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.7831378217644162, + "lr": 0.03, + "negative_goodness": 1.2253883186181387, + "pair_accuracy": 0.9108666666666667, + "positive_goodness": 3.742452353922526, + "runtime_seconds": 5.750660419464111 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.7348270265155369, + "lr": 0.03, + "negative_goodness": 1.1911831648932563, + "pair_accuracy": 0.930488888920678, + "positive_goodness": 3.9099100686391193, + "runtime_seconds": 5.749508857727051 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.7163754540761312, + "lr": 0.03, + "negative_goodness": 1.1526712789959377, + "pair_accuracy": 0.9356222222752041, + "positive_goodness": 4.017707122675578, + "runtime_seconds": 5.746120452880859 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.7742395129733616, + "lr": 0.03, + "negative_goodness": 1.222750423251258, + "pair_accuracy": 0.9097333333333333, + "positive_goodness": 3.757572292836507, + "runtime_seconds": 5.752737522125244 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.7167493282424079, + "lr": 0.03, + "negative_goodness": 1.1565696288850573, + "pair_accuracy": 0.9287555555343628, + "positive_goodness": 3.9827481270260283, + "runtime_seconds": 5.7479307651519775 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.7160431832843357, + "lr": 0.03, + "negative_goodness": 1.1405263184229533, + "pair_accuracy": 0.9252888888676961, + "positive_goodness": 4.021398482513428, + "runtime_seconds": 5.75000786781311 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.7246207553863525, + "lr": 0.03, + "negative_goodness": 1.148696636644999, + "pair_accuracy": 0.9256444444497426, + "positive_goodness": 4.009518666754828, + "runtime_seconds": 5.7468483448028564 + } + ], + "layer": 11 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9946849116113451, + "lr": 0.03, + "negative_goodness": 1.6064418148676554, + "pair_accuracy": 0.8706, + "positive_goodness": 3.140908454555935, + "runtime_seconds": 5.7846174240112305 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8225258633189731, + "lr": 0.03, + "negative_goodness": 1.3672663881407843, + "pair_accuracy": 0.9241333333333334, + "positive_goodness": 3.510426073964437, + "runtime_seconds": 5.763875246047974 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7418166275660197, + "lr": 0.03, + "negative_goodness": 1.260801914554172, + "pair_accuracy": 0.9457333333280351, + "positive_goodness": 3.791060314729479, + "runtime_seconds": 5.7638161182403564 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6873290313508775, + "lr": 0.03, + "negative_goodness": 1.1810527423646715, + "pair_accuracy": 0.9577999999894036, + "positive_goodness": 3.9831047383202445, + "runtime_seconds": 5.765203952789307 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7095690973493788, + "lr": 0.03, + "negative_goodness": 1.180944119728936, + "pair_accuracy": 0.9465111111111111, + "positive_goodness": 3.9367141470167373, + "runtime_seconds": 5.764171838760376 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7294610643121932, + "lr": 0.03, + "negative_goodness": 1.206970269203186, + "pair_accuracy": 0.9383777777777778, + "positive_goodness": 3.8433197354634605, + "runtime_seconds": 5.771553039550781 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6917828528192308, + "lr": 0.03, + "negative_goodness": 1.1540422094133165, + "pair_accuracy": 0.9470888888888889, + "positive_goodness": 3.9723238320244683, + "runtime_seconds": 5.763399362564087 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6961672931830089, + "lr": 0.03, + "negative_goodness": 1.1659416334152222, + "pair_accuracy": 0.9418888888888889, + "positive_goodness": 3.98491137898763, + "runtime_seconds": 5.764940977096558 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6857694083107843, + "lr": 0.03, + "negative_goodness": 1.136616614490085, + "pair_accuracy": 0.946244444433848, + "positive_goodness": 4.022404312578837, + "runtime_seconds": 5.763532638549805 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6687521117157407, + "lr": 0.03, + "negative_goodness": 1.1237163663652208, + "pair_accuracy": 0.9512222222222222, + "positive_goodness": 4.0792609921773275, + "runtime_seconds": 5.76363468170166 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6791578168127271, + "lr": 0.03, + "negative_goodness": 1.1190327644772, + "pair_accuracy": 0.9450888888782925, + "positive_goodness": 4.07531490711636, + "runtime_seconds": 5.767618417739868 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.703188109615114, + "lr": 0.03, + "negative_goodness": 1.1823407694180808, + "pair_accuracy": 0.9500666666666666, + "positive_goodness": 3.924278050994873, + "runtime_seconds": 5.764822244644165 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6511278497801887, + "lr": 0.03, + "negative_goodness": 1.1041512873331705, + "pair_accuracy": 0.9548666666454739, + "positive_goodness": 4.150717164654202, + "runtime_seconds": 5.765350580215454 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7176990300708347, + "lr": 0.03, + "negative_goodness": 1.1681682117674086, + "pair_accuracy": 0.9410222222540113, + "positive_goodness": 3.8786502331203883, + "runtime_seconds": 5.7711052894592285 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6387504645983378, + "lr": 0.03, + "negative_goodness": 1.0761802174462212, + "pair_accuracy": 0.9573555555555555, + "positive_goodness": 4.175900648583307, + "runtime_seconds": 5.766415596008301 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6278525077078078, + "lr": 0.03, + "negative_goodness": 1.0728917270342508, + "pair_accuracy": 0.9591333333333333, + "positive_goodness": 4.243845497809516, + "runtime_seconds": 5.764694690704346 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6759601549784342, + "lr": 0.03, + "negative_goodness": 1.127734753926595, + "pair_accuracy": 0.9463333333280352, + "positive_goodness": 4.0825395521799726, + "runtime_seconds": 5.76822304725647 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6984392398728265, + "lr": 0.03, + "negative_goodness": 1.1582882980876499, + "pair_accuracy": 0.9367555555449592, + "positive_goodness": 3.9520928218417697, + "runtime_seconds": 5.765625715255737 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6735327794816759, + "lr": 0.03, + "negative_goodness": 1.146242513232761, + "pair_accuracy": 0.947755555566152, + "positive_goodness": 4.042730136956109, + "runtime_seconds": 5.737646102905273 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6541542056030697, + "lr": 0.03, + "negative_goodness": 1.1060080804506938, + "pair_accuracy": 0.9545555555555556, + "positive_goodness": 4.111525553131104, + "runtime_seconds": 5.726773023605347 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6887589375389946, + "lr": 0.03, + "negative_goodness": 1.1397494913313124, + "pair_accuracy": 0.9416666666719649, + "positive_goodness": 4.001963252724542, + "runtime_seconds": 5.726088285446167 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6548081399705675, + "lr": 0.03, + "negative_goodness": 1.104850848367479, + "pair_accuracy": 0.9507999999894036, + "positive_goodness": 4.134027066040039, + "runtime_seconds": 5.728089809417725 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.619109881761339, + "lr": 0.03, + "negative_goodness": 1.0473229433801439, + "pair_accuracy": 0.9552666666454739, + "positive_goodness": 4.291938828892178, + "runtime_seconds": 5.72571063041687 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6439668169869317, + "lr": 0.03, + "negative_goodness": 1.0753313442336188, + "pair_accuracy": 0.9485555555555556, + "positive_goodness": 4.220159403313531, + "runtime_seconds": 5.766700983047485 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6316386740260654, + "lr": 0.03, + "negative_goodness": 1.0636431889428033, + "pair_accuracy": 0.954333333322737, + "positive_goodness": 4.2512381883409285, + "runtime_seconds": 5.77483057975769 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6365419530656603, + "lr": 0.03, + "negative_goodness": 1.0678073558383518, + "pair_accuracy": 0.9517777777883741, + "positive_goodness": 4.292520994991726, + "runtime_seconds": 5.771050214767456 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6528009257634481, + "lr": 0.03, + "negative_goodness": 1.0804360843658447, + "pair_accuracy": 0.9488000000158946, + "positive_goodness": 4.161495060306125, + "runtime_seconds": 5.77618408203125 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6637167902840508, + "lr": 0.03, + "negative_goodness": 1.109504618496365, + "pair_accuracy": 0.9416888888782925, + "positive_goodness": 4.07300619345771, + "runtime_seconds": 5.769900798797607 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6408851031992171, + "lr": 0.03, + "negative_goodness": 1.0659740923457675, + "pair_accuracy": 0.9514222222222222, + "positive_goodness": 4.206271760135227, + "runtime_seconds": 5.7810585498809814 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6386490150875516, + "lr": 0.03, + "negative_goodness": 1.0502550659073724, + "pair_accuracy": 0.9463555555449592, + "positive_goodness": 4.266929696655273, + "runtime_seconds": 5.7724769115448 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6373488094329834, + "lr": 0.03, + "negative_goodness": 1.075750215763516, + "pair_accuracy": 0.9538888888676961, + "positive_goodness": 4.250357792748345, + "runtime_seconds": 5.769796133041382 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6321178010834588, + "lr": 0.03, + "negative_goodness": 1.0783726112895542, + "pair_accuracy": 0.956, + "positive_goodness": 4.224235955641005, + "runtime_seconds": 5.768980264663696 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6117709699842665, + "lr": 0.03, + "negative_goodness": 1.02743972875807, + "pair_accuracy": 0.9555111111534966, + "positive_goodness": 4.3621201741112605, + "runtime_seconds": 5.768429279327393 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6244633092456394, + "lr": 0.03, + "negative_goodness": 1.0380465028127035, + "pair_accuracy": 0.9540888889312744, + "positive_goodness": 4.321101817237007, + "runtime_seconds": 5.772653818130493 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6545856227344937, + "lr": 0.03, + "negative_goodness": 1.0844174739837646, + "pair_accuracy": 0.9498222221904331, + "positive_goodness": 4.169097192043728, + "runtime_seconds": 5.767456293106079 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6039136610560947, + "lr": 0.03, + "negative_goodness": 1.0266005378723144, + "pair_accuracy": 0.9575333333333333, + "positive_goodness": 4.35531908692254, + "runtime_seconds": 5.7737648487091064 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6312355670928955, + "lr": 0.03, + "negative_goodness": 1.0482331006579928, + "pair_accuracy": 0.9475777777910233, + "positive_goodness": 4.273657191382514, + "runtime_seconds": 5.7705864906311035 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6051662317276001, + "lr": 0.03, + "negative_goodness": 1.0314987736596002, + "pair_accuracy": 0.9558000000529819, + "positive_goodness": 4.314997428152297, + "runtime_seconds": 5.772111177444458 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6133075955920749, + "lr": 0.03, + "negative_goodness": 1.033108330514696, + "pair_accuracy": 0.9508222222328186, + "positive_goodness": 4.3614936996459965, + "runtime_seconds": 5.772385358810425 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6551618481106228, + "lr": 0.03, + "negative_goodness": 1.0623282018449571, + "pair_accuracy": 0.9374444444550408, + "positive_goodness": 4.211365131674873, + "runtime_seconds": 5.760085821151733 + } + ], + "layer": 12 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8540332996050517, + "lr": 0.03, + "negative_goodness": 1.3434200750456915, + "pair_accuracy": 0.9036, + "positive_goodness": 3.2741741295284696, + "runtime_seconds": 5.749600172042847 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7653872764481439, + "lr": 0.03, + "negative_goodness": 1.3293893282148572, + "pair_accuracy": 0.9236666666666666, + "positive_goodness": 3.5647324121263293, + "runtime_seconds": 5.7505269050598145 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7531612387975057, + "lr": 0.03, + "negative_goodness": 1.295059016778734, + "pair_accuracy": 0.9164888888888889, + "positive_goodness": 3.6501341388278536, + "runtime_seconds": 5.75272798538208 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.734629892815484, + "lr": 0.03, + "negative_goodness": 1.2657844700495402, + "pair_accuracy": 0.9237555555555556, + "positive_goodness": 3.737789638264974, + "runtime_seconds": 5.751026391983032 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.708084321890937, + "lr": 0.03, + "negative_goodness": 1.22912751443651, + "pair_accuracy": 0.9326444444868299, + "positive_goodness": 3.837101155810886, + "runtime_seconds": 5.75607967376709 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.691583088737064, + "lr": 0.03, + "negative_goodness": 1.2262635633680556, + "pair_accuracy": 0.9376000000423855, + "positive_goodness": 3.911175416395399, + "runtime_seconds": 5.750149488449097 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6574644733005099, + "lr": 0.03, + "negative_goodness": 1.1783541355556912, + "pair_accuracy": 0.9451777777777778, + "positive_goodness": 4.028090018463135, + "runtime_seconds": 5.747611999511719 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6704959388309055, + "lr": 0.03, + "negative_goodness": 1.1793388313717312, + "pair_accuracy": 0.932155555597941, + "positive_goodness": 4.02892949417962, + "runtime_seconds": 5.744710683822632 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6713891165733338, + "lr": 0.03, + "negative_goodness": 1.1892348524623446, + "pair_accuracy": 0.9379555555449591, + "positive_goodness": 4.015346378156874, + "runtime_seconds": 5.749179363250732 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6922993989679549, + "lr": 0.03, + "negative_goodness": 1.1772714713838366, + "pair_accuracy": 0.9275333333333333, + "positive_goodness": 3.981441475507948, + "runtime_seconds": 5.758931636810303 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6631121809853447, + "lr": 0.03, + "negative_goodness": 1.1617356998019748, + "pair_accuracy": 0.939111111137602, + "positive_goodness": 4.093457502259149, + "runtime_seconds": 5.74536657333374 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6985048366016812, + "lr": 0.03, + "negative_goodness": 1.1786981843948365, + "pair_accuracy": 0.9222444444444444, + "positive_goodness": 3.9663596165974937, + "runtime_seconds": 5.7527947425842285 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6718570493062337, + "lr": 0.03, + "negative_goodness": 1.1572782375335693, + "pair_accuracy": 0.928488888891538, + "positive_goodness": 4.064365269957649, + "runtime_seconds": 5.753241777420044 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6507991463078393, + "lr": 0.03, + "negative_goodness": 1.150112991089291, + "pair_accuracy": 0.9440888888888889, + "positive_goodness": 4.161397470940484, + "runtime_seconds": 5.756378650665283 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6294966059949663, + "lr": 0.03, + "negative_goodness": 1.1285861448076036, + "pair_accuracy": 0.9526, + "positive_goodness": 4.219054242112901, + "runtime_seconds": 5.7610578536987305 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6198889340506659, + "lr": 0.03, + "negative_goodness": 1.1063506406148276, + "pair_accuracy": 0.948911111079322, + "positive_goodness": 4.318144995498657, + "runtime_seconds": 5.748613595962524 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6390460455576579, + "lr": 0.03, + "negative_goodness": 1.1237831602944268, + "pair_accuracy": 0.9466222222010294, + "positive_goodness": 4.247413605584039, + "runtime_seconds": 5.7496161460876465 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.645763427734375, + "lr": 0.03, + "negative_goodness": 1.1285826001909045, + "pair_accuracy": 0.9411555555555555, + "positive_goodness": 4.184084960301717, + "runtime_seconds": 5.751158714294434 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7102169564670987, + "lr": 0.03, + "negative_goodness": 1.1820358677334255, + "pair_accuracy": 0.9178444444550409, + "positive_goodness": 4.002140952089098, + "runtime_seconds": 5.744044065475464 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6498757873323229, + "lr": 0.03, + "negative_goodness": 1.1377857475916544, + "pair_accuracy": 0.9382444444444444, + "positive_goodness": 4.1498637727949355, + "runtime_seconds": 5.749158143997192 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6691546712027656, + "lr": 0.03, + "negative_goodness": 1.134167752371894, + "pair_accuracy": 0.930866666677263, + "positive_goodness": 4.153627344110277, + "runtime_seconds": 5.75012469291687 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6361577193366157, + "lr": 0.03, + "negative_goodness": 1.093573320749071, + "pair_accuracy": 0.9385333333333333, + "positive_goodness": 4.2881567003885905, + "runtime_seconds": 5.754344463348389 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6616242147975497, + "lr": 0.03, + "negative_goodness": 1.1272449854956732, + "pair_accuracy": 0.9337777777671814, + "positive_goodness": 4.175097025426229, + "runtime_seconds": 5.759347200393677 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6392924103207058, + "lr": 0.03, + "negative_goodness": 1.109362647098965, + "pair_accuracy": 0.9408666666560703, + "positive_goodness": 4.272357833862305, + "runtime_seconds": 5.757445573806763 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6581394183953603, + "lr": 0.03, + "negative_goodness": 1.1446731450928582, + "pair_accuracy": 0.9406666666666667, + "positive_goodness": 4.176545705032349, + "runtime_seconds": 5.749590158462524 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6383516111585829, + "lr": 0.03, + "negative_goodness": 1.0690929736243353, + "pair_accuracy": 0.9342444444232517, + "positive_goodness": 4.2912073562834, + "runtime_seconds": 5.749115943908691 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6483928392834134, + "lr": 0.03, + "negative_goodness": 1.1201879672580295, + "pair_accuracy": 0.9411333333015441, + "positive_goodness": 4.202414436001248, + "runtime_seconds": 5.756299018859863 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6586950920740763, + "lr": 0.03, + "negative_goodness": 1.1177345055474175, + "pair_accuracy": 0.9321555555449592, + "positive_goodness": 4.176719429736667, + "runtime_seconds": 5.735976219177246 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6692334099557664, + "lr": 0.03, + "negative_goodness": 1.131355964342753, + "pair_accuracy": 0.9290666666560703, + "positive_goodness": 4.142820126258003, + "runtime_seconds": 5.715628623962402 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.639711204179128, + "lr": 0.03, + "negative_goodness": 1.1028184026930068, + "pair_accuracy": 0.9421111111005147, + "positive_goodness": 4.2426062290615505, + "runtime_seconds": 5.688433408737183 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.642971858851115, + "lr": 0.03, + "negative_goodness": 1.0892483710077074, + "pair_accuracy": 0.9349333333333333, + "positive_goodness": 4.287008912913005, + "runtime_seconds": 5.694193601608276 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6314679561932882, + "lr": 0.03, + "negative_goodness": 1.0883652584287855, + "pair_accuracy": 0.9381555555555555, + "positive_goodness": 4.346467011430528, + "runtime_seconds": 5.689831972122192 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6223114574856228, + "lr": 0.03, + "negative_goodness": 1.082394452603658, + "pair_accuracy": 0.9423555555873447, + "positive_goodness": 4.325633566241794, + "runtime_seconds": 5.737508773803711 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6447513795216878, + "lr": 0.03, + "negative_goodness": 1.1047448252148098, + "pair_accuracy": 0.9359777777671814, + "positive_goodness": 4.25716221754286, + "runtime_seconds": 5.7426629066467285 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6329421992407904, + "lr": 0.03, + "negative_goodness": 1.0504063210805257, + "pair_accuracy": 0.9341999999682109, + "positive_goodness": 4.388163126118978, + "runtime_seconds": 5.73896598815918 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6461264427926805, + "lr": 0.03, + "negative_goodness": 1.0979877633412678, + "pair_accuracy": 0.9360000000211928, + "positive_goodness": 4.271173250198364, + "runtime_seconds": 5.718415975570679 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6703304821226332, + "lr": 0.03, + "negative_goodness": 1.1255891436841752, + "pair_accuracy": 0.9272, + "positive_goodness": 4.1366023515489365, + "runtime_seconds": 5.691599607467651 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6863593573040433, + "lr": 0.03, + "negative_goodness": 1.1471323379304674, + "pair_accuracy": 0.9231333333863152, + "positive_goodness": 4.119454989581638, + "runtime_seconds": 5.688931703567505 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6551330188963148, + "lr": 0.03, + "negative_goodness": 1.1229114174313015, + "pair_accuracy": 0.937133333322737, + "positive_goodness": 4.181014437272814, + "runtime_seconds": 5.690289497375488 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6331193715837267, + "lr": 0.03, + "negative_goodness": 1.1003547662946913, + "pair_accuracy": 0.9455333333333333, + "positive_goodness": 4.291617814805773, + "runtime_seconds": 5.691157102584839 + } + ], + "layer": 13 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.043642360051473, + "lr": 0.03, + "negative_goodness": 1.6823684190114339, + "pair_accuracy": 0.8501333333757188, + "positive_goodness": 3.229523801718818, + "runtime_seconds": 5.7214226722717285 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7994504263771904, + "lr": 0.03, + "negative_goodness": 1.3083217616611056, + "pair_accuracy": 0.9259999999788072, + "positive_goodness": 3.6307460513220895, + "runtime_seconds": 5.72455096244812 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7301259875615438, + "lr": 0.03, + "negative_goodness": 1.20170991897583, + "pair_accuracy": 0.9474444444444444, + "positive_goodness": 3.746752492099338, + "runtime_seconds": 5.719738245010376 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7419664233101739, + "lr": 0.03, + "negative_goodness": 1.1707582299232482, + "pair_accuracy": 0.9333333333439298, + "positive_goodness": 3.733354941262139, + "runtime_seconds": 5.723090410232544 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7391217812962002, + "lr": 0.03, + "negative_goodness": 1.1656699635293748, + "pair_accuracy": 0.9295333333333333, + "positive_goodness": 3.7003906007554797, + "runtime_seconds": 5.757087707519531 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7526978332625495, + "lr": 0.03, + "negative_goodness": 1.1922727178149752, + "pair_accuracy": 0.9367777777459886, + "positive_goodness": 3.6147367449018692, + "runtime_seconds": 5.7542641162872314 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7225359031889174, + "lr": 0.03, + "negative_goodness": 1.1504110637876723, + "pair_accuracy": 0.9377333333492279, + "positive_goodness": 3.743911797014872, + "runtime_seconds": 5.724207401275635 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.693683757019043, + "lr": 0.03, + "negative_goodness": 1.111594743792216, + "pair_accuracy": 0.9392000000105963, + "positive_goodness": 3.79741755065918, + "runtime_seconds": 5.742961645126343 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7051963830841912, + "lr": 0.03, + "negative_goodness": 1.143509324476454, + "pair_accuracy": 0.9417555555555556, + "positive_goodness": 3.777595842573378, + "runtime_seconds": 5.754719972610474 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6755006093872918, + "lr": 0.03, + "negative_goodness": 1.131213269466824, + "pair_accuracy": 0.9466666666666667, + "positive_goodness": 3.887035101021661, + "runtime_seconds": 5.753638505935669 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.703067866537306, + "lr": 0.03, + "negative_goodness": 1.1458185427135892, + "pair_accuracy": 0.9326888889100816, + "positive_goodness": 3.815826393975152, + "runtime_seconds": 5.755520820617676 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.670444846174452, + "lr": 0.03, + "negative_goodness": 1.1198077953762477, + "pair_accuracy": 0.9475333333333333, + "positive_goodness": 3.9486516887664793, + "runtime_seconds": 5.753145933151245 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6220295028474596, + "lr": 0.03, + "negative_goodness": 1.0430439650853476, + "pair_accuracy": 0.9500444444338481, + "positive_goodness": 4.164717975997925, + "runtime_seconds": 5.756787538528442 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6416215340216954, + "lr": 0.03, + "negative_goodness": 1.0653484756257798, + "pair_accuracy": 0.9426, + "positive_goodness": 4.083619950697157, + "runtime_seconds": 5.756726980209351 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6336193116611905, + "lr": 0.03, + "negative_goodness": 1.0341917546696133, + "pair_accuracy": 0.9474888888888889, + "positive_goodness": 4.108205336083306, + "runtime_seconds": 5.755108833312988 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6432069259219699, + "lr": 0.03, + "negative_goodness": 1.0753858481566112, + "pair_accuracy": 0.9499777777777778, + "positive_goodness": 4.0623186760372585, + "runtime_seconds": 5.756911039352417 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6444311626328363, + "lr": 0.03, + "negative_goodness": 1.0791118018892076, + "pair_accuracy": 0.9465555555449592, + "positive_goodness": 4.054011386108399, + "runtime_seconds": 5.736189842224121 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6400380184809367, + "lr": 0.03, + "negative_goodness": 1.0541164202372233, + "pair_accuracy": 0.9453777777777778, + "positive_goodness": 4.11759268307156, + "runtime_seconds": 5.7395524978637695 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6458212190839979, + "lr": 0.03, + "negative_goodness": 1.0626306294971042, + "pair_accuracy": 0.9437555555343627, + "positive_goodness": 4.091231670888265, + "runtime_seconds": 5.7568135261535645 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6180040195147196, + "lr": 0.03, + "negative_goodness": 1.0193807631280687, + "pair_accuracy": 0.9465333333121406, + "positive_goodness": 4.224441503228082, + "runtime_seconds": 5.754671812057495 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6398611927774217, + "lr": 0.03, + "negative_goodness": 1.0499626024987962, + "pair_accuracy": 0.9415555555449592, + "positive_goodness": 4.1161576704237195, + "runtime_seconds": 5.753520250320435 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6024826329814064, + "lr": 0.03, + "negative_goodness": 0.9863851332346598, + "pair_accuracy": 0.9490222222222222, + "positive_goodness": 4.300285968271892, + "runtime_seconds": 5.733676195144653 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6264154883066814, + "lr": 0.03, + "negative_goodness": 1.0206735513793097, + "pair_accuracy": 0.9396, + "positive_goodness": 4.2153838796827525, + "runtime_seconds": 5.720741510391235 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6320249613020155, + "lr": 0.03, + "negative_goodness": 1.0302738459693062, + "pair_accuracy": 0.9468444444974263, + "positive_goodness": 4.148143635813395, + "runtime_seconds": 5.723891496658325 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5902998833444384, + "lr": 0.03, + "negative_goodness": 0.9871009024090237, + "pair_accuracy": 0.952022222243415, + "positive_goodness": 4.290982182693481, + "runtime_seconds": 5.721240997314453 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5921229245927598, + "lr": 0.03, + "negative_goodness": 0.9616461636649237, + "pair_accuracy": 0.9558000000529819, + "positive_goodness": 4.291720061365764, + "runtime_seconds": 5.75085711479187 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5852482746442159, + "lr": 0.03, + "negative_goodness": 0.9669520259009468, + "pair_accuracy": 0.9572222222752041, + "positive_goodness": 4.325546126386854, + "runtime_seconds": 5.755417585372925 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5937615621195899, + "lr": 0.03, + "negative_goodness": 0.9702258494006263, + "pair_accuracy": 0.9528666666666666, + "positive_goodness": 4.2996888850741914, + "runtime_seconds": 5.720393896102905 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5959901544994778, + "lr": 0.03, + "negative_goodness": 0.9773693108664618, + "pair_accuracy": 0.9498666666984558, + "positive_goodness": 4.343172363959418, + "runtime_seconds": 5.838986158370972 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.641119163608551, + "lr": 0.03, + "negative_goodness": 0.9928368497318691, + "pair_accuracy": 0.9371555555555555, + "positive_goodness": 4.164507951524523, + "runtime_seconds": 5.724598169326782 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6001592504501343, + "lr": 0.03, + "negative_goodness": 0.9864484714402093, + "pair_accuracy": 0.9518, + "positive_goodness": 4.303451423814562, + "runtime_seconds": 5.735449314117432 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5691245153533088, + "lr": 0.03, + "negative_goodness": 0.9349674735016293, + "pair_accuracy": 0.9581333333333333, + "positive_goodness": 4.514090444776747, + "runtime_seconds": 5.72387170791626 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5926044074058533, + "lr": 0.03, + "negative_goodness": 0.9569485036426121, + "pair_accuracy": 0.9509999999894037, + "positive_goodness": 4.35850186636183, + "runtime_seconds": 5.741779804229736 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5859315923266941, + "lr": 0.03, + "negative_goodness": 0.9463732207854589, + "pair_accuracy": 0.9542888888888889, + "positive_goodness": 4.405624916627672, + "runtime_seconds": 5.755510330200195 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6243330943743388, + "lr": 0.03, + "negative_goodness": 0.9951585437986585, + "pair_accuracy": 0.9445333333227369, + "positive_goodness": 4.254890898174709, + "runtime_seconds": 5.758537292480469 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6087314392222298, + "lr": 0.03, + "negative_goodness": 0.9599108479420344, + "pair_accuracy": 0.9505555555555556, + "positive_goodness": 4.3173218822055395, + "runtime_seconds": 5.756553888320923 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5584267165395949, + "lr": 0.03, + "negative_goodness": 0.9183604367733001, + "pair_accuracy": 0.9568888888888889, + "positive_goodness": 4.532250287543403, + "runtime_seconds": 5.750408411026001 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6104720228513082, + "lr": 0.03, + "negative_goodness": 0.9605097610897488, + "pair_accuracy": 0.9504444444444444, + "positive_goodness": 4.31890218234592, + "runtime_seconds": 5.7401134967803955 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5734154635217454, + "lr": 0.03, + "negative_goodness": 0.9109228537665474, + "pair_accuracy": 0.9553555555555555, + "positive_goodness": 4.4389154182434085, + "runtime_seconds": 5.7187347412109375 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5856910566223992, + "lr": 0.03, + "negative_goodness": 0.9311357114685906, + "pair_accuracy": 0.9491777777777778, + "positive_goodness": 4.419648718092176, + "runtime_seconds": 5.728126287460327 + } + ], + "layer": 14 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8077292849858602, + "lr": 0.03, + "negative_goodness": 1.209199212604099, + "pair_accuracy": 0.9061999999682109, + "positive_goodness": 3.5330154333750405, + "runtime_seconds": 5.740723609924316 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7228240838898553, + "lr": 0.03, + "negative_goodness": 1.146035060013665, + "pair_accuracy": 0.9296222222222222, + "positive_goodness": 3.8675915012783473, + "runtime_seconds": 5.735582590103149 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7128756819407145, + "lr": 0.03, + "negative_goodness": 1.133212046262953, + "pair_accuracy": 0.9295555555608538, + "positive_goodness": 3.9272841211107044, + "runtime_seconds": 5.721423387527466 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6758963685671489, + "lr": 0.03, + "negative_goodness": 1.0977544721391466, + "pair_accuracy": 0.9378888888782925, + "positive_goodness": 3.9872726520114474, + "runtime_seconds": 5.686263084411621 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7077178281995985, + "lr": 0.03, + "negative_goodness": 1.1293531079822117, + "pair_accuracy": 0.927577777756585, + "positive_goodness": 3.969195002831353, + "runtime_seconds": 5.683469772338867 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6773271365536584, + "lr": 0.03, + "negative_goodness": 1.106495298576355, + "pair_accuracy": 0.9370222222222222, + "positive_goodness": 3.9921829366895887, + "runtime_seconds": 5.6845386028289795 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6687681733873155, + "lr": 0.03, + "negative_goodness": 1.0703529817369248, + "pair_accuracy": 0.931244444433848, + "positive_goodness": 4.093548706902398, + "runtime_seconds": 5.6848931312561035 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6623319101015727, + "lr": 0.03, + "negative_goodness": 1.0737663223266602, + "pair_accuracy": 0.9367111111111112, + "positive_goodness": 4.115773443433974, + "runtime_seconds": 5.684698104858398 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6755407812860277, + "lr": 0.03, + "negative_goodness": 1.0680918160676955, + "pair_accuracy": 0.9258888888888889, + "positive_goodness": 4.089353062184652, + "runtime_seconds": 5.6866044998168945 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6476636092503866, + "lr": 0.03, + "negative_goodness": 1.039709871588813, + "pair_accuracy": 0.939933333322737, + "positive_goodness": 4.1971469653235545, + "runtime_seconds": 5.685578107833862 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6419257877985637, + "lr": 0.03, + "negative_goodness": 1.0368057378345066, + "pair_accuracy": 0.9368222222646078, + "positive_goodness": 4.203251220279269, + "runtime_seconds": 5.684859275817871 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6086089744991726, + "lr": 0.03, + "negative_goodness": 0.9985459487491184, + "pair_accuracy": 0.9499777777797646, + "positive_goodness": 4.356890682983399, + "runtime_seconds": 5.6858227252960205 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6330047592057122, + "lr": 0.03, + "negative_goodness": 1.0138513754950629, + "pair_accuracy": 0.9384222222010294, + "positive_goodness": 4.242979154544407, + "runtime_seconds": 5.717362880706787 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5980334328757392, + "lr": 0.03, + "negative_goodness": 0.9774680274115668, + "pair_accuracy": 0.9461111111323038, + "positive_goodness": 4.458681714757284, + "runtime_seconds": 5.7007856369018555 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5740556812975142, + "lr": 0.03, + "negative_goodness": 0.9389847209082709, + "pair_accuracy": 0.9514, + "positive_goodness": 4.522922228834364, + "runtime_seconds": 5.685072898864746 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6079774571683672, + "lr": 0.03, + "negative_goodness": 0.9779205644395617, + "pair_accuracy": 0.9467777777777778, + "positive_goodness": 4.396587236022949, + "runtime_seconds": 5.7222983837127686 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5672759466171264, + "lr": 0.03, + "negative_goodness": 0.9072347744411893, + "pair_accuracy": 0.9553555555661519, + "positive_goodness": 4.624685880512661, + "runtime_seconds": 5.742198705673218 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.559445163400968, + "lr": 0.03, + "negative_goodness": 0.9082724655310312, + "pair_accuracy": 0.9559333333333333, + "positive_goodness": 4.639303219604492, + "runtime_seconds": 5.738272190093994 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6238925936169094, + "lr": 0.03, + "negative_goodness": 0.9985655376434326, + "pair_accuracy": 0.9442666666348776, + "positive_goodness": 4.373583974117703, + "runtime_seconds": 5.706565618515015 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6108534517923991, + "lr": 0.03, + "negative_goodness": 0.9665532493273418, + "pair_accuracy": 0.9414666666666667, + "positive_goodness": 4.444241522683038, + "runtime_seconds": 5.686985015869141 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6023656463305156, + "lr": 0.03, + "negative_goodness": 0.9389311989254422, + "pair_accuracy": 0.9376444444338481, + "positive_goodness": 4.483940943272908, + "runtime_seconds": 5.706195116043091 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5631650396717919, + "lr": 0.03, + "negative_goodness": 0.9183290605121188, + "pair_accuracy": 0.9565333333333333, + "positive_goodness": 4.639872148980035, + "runtime_seconds": 5.737342357635498 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6051714739057753, + "lr": 0.03, + "negative_goodness": 0.9504967876646254, + "pair_accuracy": 0.9439555556085375, + "positive_goodness": 4.45208524500529, + "runtime_seconds": 5.7364397048950195 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5848545355584887, + "lr": 0.03, + "negative_goodness": 0.9283776485350397, + "pair_accuracy": 0.9494444444444444, + "positive_goodness": 4.599643828413222, + "runtime_seconds": 5.726597785949707 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.598423991590076, + "lr": 0.03, + "negative_goodness": 0.9240547195381589, + "pair_accuracy": 0.9371555555555555, + "positive_goodness": 4.550133902147081, + "runtime_seconds": 5.737509727478027 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5989017757309808, + "lr": 0.03, + "negative_goodness": 0.9319571215894488, + "pair_accuracy": 0.9397555555449592, + "positive_goodness": 4.56941507174174, + "runtime_seconds": 5.7367494106292725 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6028853678173489, + "lr": 0.03, + "negative_goodness": 0.9269093066533407, + "pair_accuracy": 0.9402000000529819, + "positive_goodness": 4.532553838857015, + "runtime_seconds": 5.737213134765625 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.56592711464564, + "lr": 0.03, + "negative_goodness": 0.9052523978233338, + "pair_accuracy": 0.9507333333439297, + "positive_goodness": 4.710734250047472, + "runtime_seconds": 5.740714073181152 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5717803855472141, + "lr": 0.03, + "negative_goodness": 0.8847948793093363, + "pair_accuracy": 0.9469111111111111, + "positive_goodness": 4.71238114344279, + "runtime_seconds": 5.731891393661499 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5985497381316291, + "lr": 0.03, + "negative_goodness": 0.9327549852689108, + "pair_accuracy": 0.9429555555555555, + "positive_goodness": 4.564009812121921, + "runtime_seconds": 5.7237842082977295 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5576520248730977, + "lr": 0.03, + "negative_goodness": 0.8783806577788459, + "pair_accuracy": 0.9530888888888889, + "positive_goodness": 4.782075267834133, + "runtime_seconds": 5.737922668457031 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.573078599219852, + "lr": 0.03, + "negative_goodness": 0.9196779096338484, + "pair_accuracy": 0.9509777777777778, + "positive_goodness": 4.643351791127523, + "runtime_seconds": 5.719398021697998 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5811459675682915, + "lr": 0.03, + "negative_goodness": 0.9059310257593791, + "pair_accuracy": 0.9430444444447755, + "positive_goodness": 4.680737541474237, + "runtime_seconds": 5.6859142780303955 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5622108904573653, + "lr": 0.03, + "negative_goodness": 0.8781166532357534, + "pair_accuracy": 0.9538, + "positive_goodness": 4.712309113396539, + "runtime_seconds": 5.686975002288818 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5366445859008365, + "lr": 0.03, + "negative_goodness": 0.8430267969025506, + "pair_accuracy": 0.9529555555449591, + "positive_goodness": 4.906475472429063, + "runtime_seconds": 5.694453716278076 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5703455469873216, + "lr": 0.03, + "negative_goodness": 0.8877037631988526, + "pair_accuracy": 0.951333333322737, + "positive_goodness": 4.722531295055814, + "runtime_seconds": 5.735188722610474 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5644487661838532, + "lr": 0.03, + "negative_goodness": 0.8758599687788221, + "pair_accuracy": 0.9481333333333334, + "positive_goodness": 4.754723003217909, + "runtime_seconds": 5.734181880950928 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5876818772315979, + "lr": 0.03, + "negative_goodness": 0.8893282003190782, + "pair_accuracy": 0.9459555555343628, + "positive_goodness": 4.652373493703206, + "runtime_seconds": 5.7367942333221436 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5866580487145318, + "lr": 0.03, + "negative_goodness": 0.9005969947073195, + "pair_accuracy": 0.9383555555449592, + "positive_goodness": 4.644764848539564, + "runtime_seconds": 5.736835956573486 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5601717548370362, + "lr": 0.03, + "negative_goodness": 0.8695829068024953, + "pair_accuracy": 0.9463777777777778, + "positive_goodness": 4.79460900217692, + "runtime_seconds": 5.707314491271973 + } + ], + "layer": 15 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0813628235287136, + "lr": 0.03, + "negative_goodness": 1.6765438162273831, + "pair_accuracy": 0.8168444444444445, + "positive_goodness": 3.1979609195921155, + "runtime_seconds": 5.743719816207886 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8086916566848755, + "lr": 0.03, + "negative_goodness": 1.3351606928083632, + "pair_accuracy": 0.9083111111323039, + "positive_goodness": 3.699197622977363, + "runtime_seconds": 5.761765956878662 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7209807239956326, + "lr": 0.03, + "negative_goodness": 1.227506582959493, + "pair_accuracy": 0.9294888888676961, + "positive_goodness": 3.9776111350801258, + "runtime_seconds": 5.754207611083984 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6983618433210584, + "lr": 0.03, + "negative_goodness": 1.151666973368327, + "pair_accuracy": 0.9323111111323039, + "positive_goodness": 3.927937078603109, + "runtime_seconds": 5.755901336669922 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6652849786334568, + "lr": 0.03, + "negative_goodness": 1.099031655714247, + "pair_accuracy": 0.9333999999682109, + "positive_goodness": 4.130532081052992, + "runtime_seconds": 5.750417232513428 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6832835120413039, + "lr": 0.03, + "negative_goodness": 1.1277515489366319, + "pair_accuracy": 0.9311777778201633, + "positive_goodness": 4.028683422470093, + "runtime_seconds": 5.748272895812988 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6728859994093577, + "lr": 0.03, + "negative_goodness": 1.085537832567427, + "pair_accuracy": 0.9303111111111111, + "positive_goodness": 4.076718399556478, + "runtime_seconds": 5.744665145874023 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.5820480612172021, + "lr": 0.03, + "negative_goodness": 1.005483854760064, + "pair_accuracy": 0.9578, + "positive_goodness": 4.389581043836806, + "runtime_seconds": 5.762751340866089 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6399625000582801, + "lr": 0.03, + "negative_goodness": 1.0685314306471083, + "pair_accuracy": 0.9386888888888889, + "positive_goodness": 4.149139383443196, + "runtime_seconds": 5.765979766845703 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5982635676966773, + "lr": 0.03, + "negative_goodness": 1.0190116641044618, + "pair_accuracy": 0.9538222222222222, + "positive_goodness": 4.321236879052056, + "runtime_seconds": 5.777781248092651 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5641626050154368, + "lr": 0.03, + "negative_goodness": 0.9608929763423072, + "pair_accuracy": 0.9572, + "positive_goodness": 4.496902349260118, + "runtime_seconds": 5.763747215270996 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6123360478083293, + "lr": 0.03, + "negative_goodness": 0.9956898840374416, + "pair_accuracy": 0.9399999999894036, + "positive_goodness": 4.312644374126858, + "runtime_seconds": 5.766972303390503 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5851441229926215, + "lr": 0.03, + "negative_goodness": 0.9723166086408827, + "pair_accuracy": 0.9529555555582047, + "positive_goodness": 4.390237932078044, + "runtime_seconds": 5.767284154891968 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5873849428812663, + "lr": 0.03, + "negative_goodness": 0.9747640508439805, + "pair_accuracy": 0.9477777777797646, + "positive_goodness": 4.4128102148691815, + "runtime_seconds": 5.765705585479736 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6118840769873725, + "lr": 0.03, + "negative_goodness": 1.0197348164187536, + "pair_accuracy": 0.9501111111005147, + "positive_goodness": 4.288538654496935, + "runtime_seconds": 5.770308256149292 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6096735409736633, + "lr": 0.03, + "negative_goodness": 0.9845185489018758, + "pair_accuracy": 0.9465555555979411, + "positive_goodness": 4.339837828360664, + "runtime_seconds": 5.772658824920654 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5765107239511278, + "lr": 0.03, + "negative_goodness": 0.9463970541742113, + "pair_accuracy": 0.9518888888676962, + "positive_goodness": 4.4198666905721025, + "runtime_seconds": 5.7649359703063965 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5992112108336555, + "lr": 0.03, + "negative_goodness": 0.9818605893346999, + "pair_accuracy": 0.9423111111111111, + "positive_goodness": 4.3567831985473635, + "runtime_seconds": 5.76821231842041 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5827254225518969, + "lr": 0.03, + "negative_goodness": 0.9673968142403496, + "pair_accuracy": 0.9495555555449592, + "positive_goodness": 4.391504744466146, + "runtime_seconds": 5.7669360637664795 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5707711565282609, + "lr": 0.03, + "negative_goodness": 0.936334270977974, + "pair_accuracy": 0.9496444444444444, + "positive_goodness": 4.457128370920817, + "runtime_seconds": 5.764110565185547 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6012887466960483, + "lr": 0.03, + "negative_goodness": 0.9751232186423407, + "pair_accuracy": 0.9429555555555555, + "positive_goodness": 4.312078301239014, + "runtime_seconds": 5.7724974155426025 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5422316340976291, + "lr": 0.03, + "negative_goodness": 0.9275819087982178, + "pair_accuracy": 0.9591333333121406, + "positive_goodness": 4.578923684353299, + "runtime_seconds": 5.764500379562378 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5728404538260566, + "lr": 0.03, + "negative_goodness": 0.9397935285462273, + "pair_accuracy": 0.9548000000423855, + "positive_goodness": 4.4830408335367835, + "runtime_seconds": 5.763909101486206 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5455308247778151, + "lr": 0.03, + "negative_goodness": 0.9132291709899902, + "pair_accuracy": 0.9579333333333333, + "positive_goodness": 4.556146794806587, + "runtime_seconds": 5.763471364974976 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5753064369413587, + "lr": 0.03, + "negative_goodness": 0.9364274180306329, + "pair_accuracy": 0.9490444444338481, + "positive_goodness": 4.471171285925971, + "runtime_seconds": 5.75764536857605 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5735203237374623, + "lr": 0.03, + "negative_goodness": 0.9579669907199012, + "pair_accuracy": 0.9531333333333334, + "positive_goodness": 4.424232546403673, + "runtime_seconds": 5.723729133605957 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.565815491549174, + "lr": 0.03, + "negative_goodness": 0.9278921001752217, + "pair_accuracy": 0.9505333333333333, + "positive_goodness": 4.454271600087484, + "runtime_seconds": 5.727167129516602 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5849146755218506, + "lr": 0.03, + "negative_goodness": 0.948643079164293, + "pair_accuracy": 0.9425111111058129, + "positive_goodness": 4.418724549060397, + "runtime_seconds": 5.725392818450928 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5691687502543131, + "lr": 0.03, + "negative_goodness": 0.8885811158286201, + "pair_accuracy": 0.9488444444762336, + "positive_goodness": 4.485611656400892, + "runtime_seconds": 5.726901292800903 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5359419727272458, + "lr": 0.03, + "negative_goodness": 0.8915692127333747, + "pair_accuracy": 0.9627555555555556, + "positive_goodness": 4.562875947231716, + "runtime_seconds": 5.726470470428467 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5254947767310673, + "lr": 0.03, + "negative_goodness": 0.8622430050849914, + "pair_accuracy": 0.9568666666666666, + "positive_goodness": 4.751280461883545, + "runtime_seconds": 5.729458570480347 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5675594629711574, + "lr": 0.03, + "negative_goodness": 0.92706068549686, + "pair_accuracy": 0.949622222243415, + "positive_goodness": 4.473617198308309, + "runtime_seconds": 5.728997468948364 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.552798648929596, + "lr": 0.03, + "negative_goodness": 0.8800009325875177, + "pair_accuracy": 0.9482888888782925, + "positive_goodness": 4.588654892985026, + "runtime_seconds": 5.727462291717529 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5736042762756348, + "lr": 0.03, + "negative_goodness": 0.9037736954794989, + "pair_accuracy": 0.9421555555767483, + "positive_goodness": 4.543971029557122, + "runtime_seconds": 5.728877305984497 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5465184293958876, + "lr": 0.03, + "negative_goodness": 0.9004793021731906, + "pair_accuracy": 0.9550888888895511, + "positive_goodness": 4.588282968298594, + "runtime_seconds": 5.753023386001587 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5807900384691026, + "lr": 0.03, + "negative_goodness": 0.9028742028130425, + "pair_accuracy": 0.9453555555661519, + "positive_goodness": 4.44759680050744, + "runtime_seconds": 5.760590076446533 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5165767370435926, + "lr": 0.03, + "negative_goodness": 0.8714184447712369, + "pair_accuracy": 0.9663333333333334, + "positive_goodness": 4.670254816775852, + "runtime_seconds": 5.728963851928711 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5346224175400204, + "lr": 0.03, + "negative_goodness": 0.8656261479934056, + "pair_accuracy": 0.9531333333333334, + "positive_goodness": 4.684896028306749, + "runtime_seconds": 5.729176998138428 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5123895366933611, + "lr": 0.03, + "negative_goodness": 0.8475099859979418, + "pair_accuracy": 0.9615111111111111, + "positive_goodness": 4.771570641411675, + "runtime_seconds": 5.750804662704468 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5050272702111138, + "lr": 0.03, + "negative_goodness": 0.8420383647653792, + "pair_accuracy": 0.9654444444444444, + "positive_goodness": 4.795401853179932, + "runtime_seconds": 5.749722719192505 + } + ], + "layer": 16 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.759397067700492, + "lr": 0.03, + "negative_goodness": 1.1536874280187819, + "pair_accuracy": 0.9144888888888889, + "positive_goodness": 3.7950426486545137, + "runtime_seconds": 5.744027614593506 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.6649845866203308, + "lr": 0.03, + "negative_goodness": 1.1204468712382847, + "pair_accuracy": 0.9317333333015442, + "positive_goodness": 4.162667276467217, + "runtime_seconds": 5.759847402572632 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.6269203478071425, + "lr": 0.03, + "negative_goodness": 1.1035344711515638, + "pair_accuracy": 0.9468, + "positive_goodness": 4.272452491845025, + "runtime_seconds": 5.757518529891968 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6278893554369609, + "lr": 0.03, + "negative_goodness": 1.0771295362790425, + "pair_accuracy": 0.9357777777777778, + "positive_goodness": 4.360184202914768, + "runtime_seconds": 5.7599427700042725 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6119264091014862, + "lr": 0.03, + "negative_goodness": 1.0377744220627678, + "pair_accuracy": 0.9379333333333333, + "positive_goodness": 4.443644617377387, + "runtime_seconds": 5.755059003829956 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.5483569103505876, + "lr": 0.03, + "negative_goodness": 0.9715085257954068, + "pair_accuracy": 0.9556888888888889, + "positive_goodness": 4.731107317691379, + "runtime_seconds": 5.752534627914429 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.586669972843594, + "lr": 0.03, + "negative_goodness": 1.0072616982248095, + "pair_accuracy": 0.9374888888888889, + "positive_goodness": 4.606926610310873, + "runtime_seconds": 5.758478164672852 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.5493673117107816, + "lr": 0.03, + "negative_goodness": 0.9678855889426338, + "pair_accuracy": 0.9567333333121406, + "positive_goodness": 4.738526354132758, + "runtime_seconds": 5.749340057373047 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.5551696324772305, + "lr": 0.03, + "negative_goodness": 0.9687659836663141, + "pair_accuracy": 0.9470666666348775, + "positive_goodness": 4.7471492424011235, + "runtime_seconds": 5.74617600440979 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5498966087341308, + "lr": 0.03, + "negative_goodness": 0.9439172001308865, + "pair_accuracy": 0.9503777777777778, + "positive_goodness": 4.796833125485314, + "runtime_seconds": 5.7520387172698975 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5628774675793118, + "lr": 0.03, + "negative_goodness": 0.9417937134636774, + "pair_accuracy": 0.9469111111534967, + "positive_goodness": 4.777043729951647, + "runtime_seconds": 5.754746913909912 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5499605759514703, + "lr": 0.03, + "negative_goodness": 0.9517924171023898, + "pair_accuracy": 0.9495777777724795, + "positive_goodness": 4.813456586498685, + "runtime_seconds": 5.727785110473633 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.597411921538247, + "lr": 0.03, + "negative_goodness": 0.962941753408644, + "pair_accuracy": 0.9314222222222223, + "positive_goodness": 4.672578823428684, + "runtime_seconds": 5.6959068775177 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5598294277191163, + "lr": 0.03, + "negative_goodness": 0.9564417145199245, + "pair_accuracy": 0.948844444460339, + "positive_goodness": 4.756462690734863, + "runtime_seconds": 5.701268196105957 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5660446677313911, + "lr": 0.03, + "negative_goodness": 0.9488096901363796, + "pair_accuracy": 0.9433333333353201, + "positive_goodness": 4.764745210096571, + "runtime_seconds": 5.702724933624268 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5829398211797079, + "lr": 0.03, + "negative_goodness": 0.9629330648316278, + "pair_accuracy": 0.9402444444444444, + "positive_goodness": 4.684011428578694, + "runtime_seconds": 5.700514793395996 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5489764802561866, + "lr": 0.03, + "negative_goodness": 0.9250198801517486, + "pair_accuracy": 0.9481777777777778, + "positive_goodness": 4.852488317447238, + "runtime_seconds": 5.699199914932251 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5654425000296699, + "lr": 0.03, + "negative_goodness": 0.9304125044292874, + "pair_accuracy": 0.9416888888676961, + "positive_goodness": 4.854834271367391, + "runtime_seconds": 5.697219610214233 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5815076601134406, + "lr": 0.03, + "negative_goodness": 0.9703806310335795, + "pair_accuracy": 0.9403777777777778, + "positive_goodness": 4.713371356116401, + "runtime_seconds": 5.698409557342529 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5341214591926998, + "lr": 0.03, + "negative_goodness": 0.8892277530723148, + "pair_accuracy": 0.9484666666560703, + "positive_goodness": 5.048682787153456, + "runtime_seconds": 5.704947471618652 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5651182108402252, + "lr": 0.03, + "negative_goodness": 0.9217767543792724, + "pair_accuracy": 0.9394444444338481, + "positive_goodness": 4.796742873128255, + "runtime_seconds": 5.698298454284668 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5372714517911276, + "lr": 0.03, + "negative_goodness": 0.896261022928026, + "pair_accuracy": 0.9496666666984558, + "positive_goodness": 4.989602671771579, + "runtime_seconds": 5.701730966567993 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5548915547688802, + "lr": 0.03, + "negative_goodness": 0.9115036477618748, + "pair_accuracy": 0.9447111111005148, + "positive_goodness": 4.852162124803331, + "runtime_seconds": 5.69572639465332 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5528256639639536, + "lr": 0.03, + "negative_goodness": 0.9171851162645552, + "pair_accuracy": 0.9470888888888889, + "positive_goodness": 4.862711146121555, + "runtime_seconds": 5.728313684463501 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5342635036892361, + "lr": 0.03, + "negative_goodness": 0.8626411575847202, + "pair_accuracy": 0.9488666666666666, + "positive_goodness": 5.056239941575792, + "runtime_seconds": 5.754504203796387 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5534604049894545, + "lr": 0.03, + "negative_goodness": 0.8683919976764255, + "pair_accuracy": 0.9361333333651225, + "positive_goodness": 4.9679725218455, + "runtime_seconds": 5.754232168197632 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5546055186960432, + "lr": 0.03, + "negative_goodness": 0.9052549703915914, + "pair_accuracy": 0.9439555555555555, + "positive_goodness": 4.962640204111735, + "runtime_seconds": 5.749769926071167 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5476874506897397, + "lr": 0.03, + "negative_goodness": 0.8928324872016907, + "pair_accuracy": 0.9478888888888889, + "positive_goodness": 4.974921618313259, + "runtime_seconds": 5.71799373626709 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5480882154464721, + "lr": 0.03, + "negative_goodness": 0.8977161531024509, + "pair_accuracy": 0.9421111111429002, + "positive_goodness": 4.967288090303209, + "runtime_seconds": 5.69822883605957 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.4955062199804518, + "lr": 0.03, + "negative_goodness": 0.8484430962350633, + "pair_accuracy": 0.9650444444126554, + "positive_goodness": 5.221068635050456, + "runtime_seconds": 5.703501224517822 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5291912065294054, + "lr": 0.03, + "negative_goodness": 0.853339294136895, + "pair_accuracy": 0.9479333333439297, + "positive_goodness": 5.1649919900682235, + "runtime_seconds": 5.69917893409729 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5240054973337386, + "lr": 0.03, + "negative_goodness": 0.8295799037085639, + "pair_accuracy": 0.9487777777777778, + "positive_goodness": 5.120215635681152, + "runtime_seconds": 5.693878173828125 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5674657982296414, + "lr": 0.03, + "negative_goodness": 0.8763314692815145, + "pair_accuracy": 0.9329777778307596, + "positive_goodness": 4.988220212512546, + "runtime_seconds": 5.7157697677612305 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5198648954285515, + "lr": 0.03, + "negative_goodness": 0.8462434901979234, + "pair_accuracy": 0.9549777777459887, + "positive_goodness": 5.094926131778293, + "runtime_seconds": 5.699906587600708 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5554430743959214, + "lr": 0.03, + "negative_goodness": 0.8792623123592801, + "pair_accuracy": 0.9373999999894036, + "positive_goodness": 4.955067873382569, + "runtime_seconds": 5.695971965789795 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5148544895172119, + "lr": 0.03, + "negative_goodness": 0.838081872897678, + "pair_accuracy": 0.9515333333333333, + "positive_goodness": 5.165329999499851, + "runtime_seconds": 5.701892375946045 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.47652245382732816, + "lr": 0.03, + "negative_goodness": 0.8037202909257677, + "pair_accuracy": 0.9668222222222222, + "positive_goodness": 5.355902731577555, + "runtime_seconds": 5.744146823883057 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5314608515209622, + "lr": 0.03, + "negative_goodness": 0.8662011790805393, + "pair_accuracy": 0.9478, + "positive_goodness": 5.13519300910102, + "runtime_seconds": 5.740092992782593 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.542677180035909, + "lr": 0.03, + "negative_goodness": 0.8578786113315159, + "pair_accuracy": 0.9424000000158945, + "positive_goodness": 5.104390487925212, + "runtime_seconds": 5.6904296875 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5238069118075901, + "lr": 0.03, + "negative_goodness": 0.8350118919796414, + "pair_accuracy": 0.9493111111117734, + "positive_goodness": 5.243276686710781, + "runtime_seconds": 5.688002109527588 + } + ], + "layer": 17 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9577057856241862, + "lr": 0.03, + "negative_goodness": 1.4619166843626235, + "pair_accuracy": 0.850377777756585, + "positive_goodness": 3.5938261953141954, + "runtime_seconds": 5.725976228713989 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7676047029071384, + "lr": 0.03, + "negative_goodness": 1.1529164318932428, + "pair_accuracy": 0.9129555555343628, + "positive_goodness": 3.840557531314426, + "runtime_seconds": 5.724540948867798 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.760160752805074, + "lr": 0.03, + "negative_goodness": 1.1440867092980278, + "pair_accuracy": 0.9139333333333334, + "positive_goodness": 3.7870904688941107, + "runtime_seconds": 5.725184917449951 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6946383789963192, + "lr": 0.03, + "negative_goodness": 1.0596212670432197, + "pair_accuracy": 0.9297555555555556, + "positive_goodness": 4.012601882595486, + "runtime_seconds": 5.77173924446106 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6998100776778328, + "lr": 0.03, + "negative_goodness": 1.0615760700331793, + "pair_accuracy": 0.9315555555555556, + "positive_goodness": 3.950152573055691, + "runtime_seconds": 5.7675440311431885 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6187991501808167, + "lr": 0.03, + "negative_goodness": 0.9679103818363614, + "pair_accuracy": 0.9475333333333333, + "positive_goodness": 4.244551625315348, + "runtime_seconds": 5.764724493026733 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6265819016244676, + "lr": 0.03, + "negative_goodness": 0.9553032650205824, + "pair_accuracy": 0.9406222222752041, + "positive_goodness": 4.236839876217312, + "runtime_seconds": 5.766928195953369 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6157694844086965, + "lr": 0.03, + "negative_goodness": 0.9631993121041192, + "pair_accuracy": 0.9494888888888889, + "positive_goodness": 4.210624705335829, + "runtime_seconds": 5.7653257846832275 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6150836736096276, + "lr": 0.03, + "negative_goodness": 0.9448088834550645, + "pair_accuracy": 0.9454888888888889, + "positive_goodness": 4.309662520853679, + "runtime_seconds": 5.739302635192871 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6250465776231554, + "lr": 0.03, + "negative_goodness": 0.9754710626178318, + "pair_accuracy": 0.9442444444126553, + "positive_goodness": 4.190601248380873, + "runtime_seconds": 5.729222059249878 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.616706213511361, + "lr": 0.03, + "negative_goodness": 0.9448618777487013, + "pair_accuracy": 0.943, + "positive_goodness": 4.268950197601319, + "runtime_seconds": 5.762585401535034 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6182784839895037, + "lr": 0.03, + "negative_goodness": 0.9673233775509729, + "pair_accuracy": 0.9471555555555555, + "positive_goodness": 4.264080620956421, + "runtime_seconds": 5.768152952194214 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6252033799701266, + "lr": 0.03, + "negative_goodness": 0.9682959156036377, + "pair_accuracy": 0.9394222222116259, + "positive_goodness": 4.200671216540867, + "runtime_seconds": 5.7657692432403564 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6171026913881302, + "lr": 0.03, + "negative_goodness": 0.944217780160904, + "pair_accuracy": 0.9423999999894036, + "positive_goodness": 4.240507948303223, + "runtime_seconds": 5.760894298553467 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5920481569157706, + "lr": 0.03, + "negative_goodness": 0.9278996078411738, + "pair_accuracy": 0.9472666666666667, + "positive_goodness": 4.3110301497565375, + "runtime_seconds": 5.766883850097656 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5825462632311715, + "lr": 0.03, + "negative_goodness": 0.8934798438138432, + "pair_accuracy": 0.9501111111111111, + "positive_goodness": 4.473941497717964, + "runtime_seconds": 5.761570930480957 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6214004647043017, + "lr": 0.03, + "negative_goodness": 0.9687958250257704, + "pair_accuracy": 0.9418222222222222, + "positive_goodness": 4.228186459096273, + "runtime_seconds": 5.762316703796387 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6264773794915941, + "lr": 0.03, + "negative_goodness": 0.9528502398014068, + "pair_accuracy": 0.9402666666666667, + "positive_goodness": 4.239637517632379, + "runtime_seconds": 5.7641990184783936 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5797003578768836, + "lr": 0.03, + "negative_goodness": 0.8870079243342082, + "pair_accuracy": 0.9453111111111111, + "positive_goodness": 4.496932288275825, + "runtime_seconds": 5.750097274780273 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5576246177355448, + "lr": 0.03, + "negative_goodness": 0.8574539640638563, + "pair_accuracy": 0.9534, + "positive_goodness": 4.5296410458882646, + "runtime_seconds": 5.727483034133911 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5876478322664896, + "lr": 0.03, + "negative_goodness": 0.9071962789959378, + "pair_accuracy": 0.9443777777671813, + "positive_goodness": 4.391317828538683, + "runtime_seconds": 5.7311060428619385 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6013725753466288, + "lr": 0.03, + "negative_goodness": 0.9259518835915459, + "pair_accuracy": 0.9420444444444445, + "positive_goodness": 4.3152334604899085, + "runtime_seconds": 5.759221792221069 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5911739395989312, + "lr": 0.03, + "negative_goodness": 0.8947589077419705, + "pair_accuracy": 0.9401111111111111, + "positive_goodness": 4.437092357296414, + "runtime_seconds": 5.764971971511841 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5793496297491921, + "lr": 0.03, + "negative_goodness": 0.889075196062194, + "pair_accuracy": 0.948, + "positive_goodness": 4.466226176028782, + "runtime_seconds": 5.761889457702637 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5606113241540062, + "lr": 0.03, + "negative_goodness": 0.876453778802024, + "pair_accuracy": 0.9548666666666666, + "positive_goodness": 4.492634482150607, + "runtime_seconds": 5.726857662200928 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5657969213750628, + "lr": 0.03, + "negative_goodness": 0.8765245838695103, + "pair_accuracy": 0.9517777777671814, + "positive_goodness": 4.567536749013265, + "runtime_seconds": 5.752737283706665 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5701282953156366, + "lr": 0.03, + "negative_goodness": 0.853758389515347, + "pair_accuracy": 0.9424222222222223, + "positive_goodness": 4.557413884056939, + "runtime_seconds": 5.7640221118927 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5731945234351687, + "lr": 0.03, + "negative_goodness": 0.8700037080287933, + "pair_accuracy": 0.9417333333333333, + "positive_goodness": 4.49355393040975, + "runtime_seconds": 5.75255274772644 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5326536716037327, + "lr": 0.03, + "negative_goodness": 0.8530720695839988, + "pair_accuracy": 0.9645333333333334, + "positive_goodness": 4.638807901594374, + "runtime_seconds": 5.732581377029419 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5405603237893847, + "lr": 0.03, + "negative_goodness": 0.8305870422575209, + "pair_accuracy": 0.95604444448683, + "positive_goodness": 4.62909386706882, + "runtime_seconds": 5.773881435394287 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5690969479031033, + "lr": 0.03, + "negative_goodness": 0.8520229380713569, + "pair_accuracy": 0.9428666666666666, + "positive_goodness": 4.559475392935011, + "runtime_seconds": 5.771756410598755 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.605961528661516, + "lr": 0.03, + "negative_goodness": 0.8993984793345133, + "pair_accuracy": 0.9373777777777778, + "positive_goodness": 4.335358750067817, + "runtime_seconds": 5.77569580078125 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5701895580185784, + "lr": 0.03, + "negative_goodness": 0.8677478248702155, + "pair_accuracy": 0.9469111111111111, + "positive_goodness": 4.527780914730496, + "runtime_seconds": 5.768612861633301 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5654286276181539, + "lr": 0.03, + "negative_goodness": 0.8391345565054151, + "pair_accuracy": 0.9466888888676961, + "positive_goodness": 4.5921472433302135, + "runtime_seconds": 5.775753021240234 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5498311881860097, + "lr": 0.03, + "negative_goodness": 0.8323346718947092, + "pair_accuracy": 0.9496222222222223, + "positive_goodness": 4.64699726647271, + "runtime_seconds": 5.743304252624512 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5570753570344713, + "lr": 0.03, + "negative_goodness": 0.8535927840550741, + "pair_accuracy": 0.9517111111429003, + "positive_goodness": 4.614758460828993, + "runtime_seconds": 5.727975368499756 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5406476256158617, + "lr": 0.03, + "negative_goodness": 0.8104295343822903, + "pair_accuracy": 0.9509111111005147, + "positive_goodness": 4.694618868891398, + "runtime_seconds": 5.732553243637085 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5749263187408448, + "lr": 0.03, + "negative_goodness": 0.8490812020513746, + "pair_accuracy": 0.9409111111534967, + "positive_goodness": 4.533195079421997, + "runtime_seconds": 5.731459140777588 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5563022260983785, + "lr": 0.03, + "negative_goodness": 0.8223408967547946, + "pair_accuracy": 0.9474666666560703, + "positive_goodness": 4.695275541093614, + "runtime_seconds": 5.729041814804077 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5279654003355239, + "lr": 0.03, + "negative_goodness": 0.8059647903760274, + "pair_accuracy": 0.9544666666878594, + "positive_goodness": 4.842536381997003, + "runtime_seconds": 5.7299346923828125 + } + ], + "layer": 18 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9263122883690728, + "lr": 0.03, + "negative_goodness": 1.3150742263158162, + "pair_accuracy": 0.8644222221904331, + "positive_goodness": 3.2163579455693565, + "runtime_seconds": 5.700520992279053 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8144938829421997, + "lr": 0.03, + "negative_goodness": 1.2262532011667888, + "pair_accuracy": 0.8998888888782925, + "positive_goodness": 3.5648745544009737, + "runtime_seconds": 5.715801477432251 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7779525783750746, + "lr": 0.03, + "negative_goodness": 1.1706740433163112, + "pair_accuracy": 0.9040444444126553, + "positive_goodness": 3.6433769714779323, + "runtime_seconds": 5.704484462738037 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7956858524958292, + "lr": 0.03, + "negative_goodness": 1.203554916381836, + "pair_accuracy": 0.8999555556085375, + "positive_goodness": 3.6070161174350313, + "runtime_seconds": 5.742747068405151 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8035456697252061, + "lr": 0.03, + "negative_goodness": 1.207407422097524, + "pair_accuracy": 0.8949333333121405, + "positive_goodness": 3.57438130001492, + "runtime_seconds": 5.746243238449097 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7906303107367622, + "lr": 0.03, + "negative_goodness": 1.1886490882025824, + "pair_accuracy": 0.8971111111164093, + "positive_goodness": 3.611208499018351, + "runtime_seconds": 5.749932765960693 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7992349120352004, + "lr": 0.03, + "negative_goodness": 1.2099979325612387, + "pair_accuracy": 0.8982666666454739, + "positive_goodness": 3.603535381825765, + "runtime_seconds": 5.747341871261597 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7509078476164076, + "lr": 0.03, + "negative_goodness": 1.1430456893497043, + "pair_accuracy": 0.9111111111111111, + "positive_goodness": 3.7245486774020726, + "runtime_seconds": 5.751100540161133 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7682752063857184, + "lr": 0.03, + "negative_goodness": 1.168561129781935, + "pair_accuracy": 0.9093333333015442, + "positive_goodness": 3.685987099753486, + "runtime_seconds": 5.744611978530884 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7641701204088, + "lr": 0.03, + "negative_goodness": 1.1371417885038588, + "pair_accuracy": 0.8991333333333333, + "positive_goodness": 3.715414608256022, + "runtime_seconds": 5.744386911392212 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7519543470965492, + "lr": 0.03, + "negative_goodness": 1.1299019293573167, + "pair_accuracy": 0.9104888888888889, + "positive_goodness": 3.770201301574707, + "runtime_seconds": 5.742348670959473 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7365872945361667, + "lr": 0.03, + "negative_goodness": 1.1029301561779445, + "pair_accuracy": 0.9091555555767483, + "positive_goodness": 3.818439960522122, + "runtime_seconds": 5.748476982116699 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7522697380065918, + "lr": 0.03, + "negative_goodness": 1.1199044268290201, + "pair_accuracy": 0.9040888888888889, + "positive_goodness": 3.7853669555028278, + "runtime_seconds": 5.742859840393066 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7376326025433011, + "lr": 0.03, + "negative_goodness": 1.1045116954379612, + "pair_accuracy": 0.9102444444338481, + "positive_goodness": 3.835060037104289, + "runtime_seconds": 5.753012180328369 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7516712121751573, + "lr": 0.03, + "negative_goodness": 1.1052641821755302, + "pair_accuracy": 0.9039777777989706, + "positive_goodness": 3.780563330629137, + "runtime_seconds": 5.745338678359985 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7064394462585449, + "lr": 0.03, + "negative_goodness": 1.062743245739407, + "pair_accuracy": 0.9157111111117734, + "positive_goodness": 3.9345241401778326, + "runtime_seconds": 5.7456419467926025 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.705780578607983, + "lr": 0.03, + "negative_goodness": 1.0511516462537978, + "pair_accuracy": 0.9148666666666667, + "positive_goodness": 3.927437756008572, + "runtime_seconds": 5.7497735023498535 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7077267515659332, + "lr": 0.03, + "negative_goodness": 1.059475143845876, + "pair_accuracy": 0.9153777777777777, + "positive_goodness": 3.929382805803087, + "runtime_seconds": 5.732954263687134 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6953863581021626, + "lr": 0.03, + "negative_goodness": 1.0467672288894654, + "pair_accuracy": 0.9247555555237664, + "positive_goodness": 3.9707774541642933, + "runtime_seconds": 5.703260660171509 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7029471530278524, + "lr": 0.03, + "negative_goodness": 1.050344552230835, + "pair_accuracy": 0.9192666667090522, + "positive_goodness": 3.9585252838134766, + "runtime_seconds": 5.700374603271484 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7431428763283624, + "lr": 0.03, + "negative_goodness": 1.0897467143906487, + "pair_accuracy": 0.9063333333227369, + "positive_goodness": 3.8230730514526368, + "runtime_seconds": 5.701416015625 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7529144206576878, + "lr": 0.03, + "negative_goodness": 1.1038630726708307, + "pair_accuracy": 0.9018222222222222, + "positive_goodness": 3.780073408847385, + "runtime_seconds": 5.703928470611572 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7223552401012845, + "lr": 0.03, + "negative_goodness": 1.0739998178693984, + "pair_accuracy": 0.9121333333757189, + "positive_goodness": 3.8990577644771998, + "runtime_seconds": 5.703487873077393 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7000790270010631, + "lr": 0.03, + "negative_goodness": 1.0425852892663743, + "pair_accuracy": 0.9170222222222222, + "positive_goodness": 3.978319629245334, + "runtime_seconds": 5.701123237609863 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6776921324412029, + "lr": 0.03, + "negative_goodness": 1.0013853470908272, + "pair_accuracy": 0.9210444444232517, + "positive_goodness": 4.088877335484822, + "runtime_seconds": 5.702106714248657 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7048524278852675, + "lr": 0.03, + "negative_goodness": 1.0390299512015448, + "pair_accuracy": 0.9173333333598243, + "positive_goodness": 3.9789804444419015, + "runtime_seconds": 5.713441371917725 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7477828614976672, + "lr": 0.03, + "negative_goodness": 1.0793396636539034, + "pair_accuracy": 0.9012888888676961, + "positive_goodness": 3.821340954759386, + "runtime_seconds": 5.738520383834839 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.7432786608060201, + "lr": 0.03, + "negative_goodness": 1.089103878153695, + "pair_accuracy": 0.9074666666666666, + "positive_goodness": 3.82734944390191, + "runtime_seconds": 5.7341978549957275 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.7136080302291447, + "lr": 0.03, + "negative_goodness": 1.0517639005237156, + "pair_accuracy": 0.9140666666666667, + "positive_goodness": 3.9776181325276694, + "runtime_seconds": 5.700990676879883 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7355094401041666, + "lr": 0.03, + "negative_goodness": 1.0879976016256545, + "pair_accuracy": 0.9077777777777778, + "positive_goodness": 3.873377038913303, + "runtime_seconds": 5.70186972618103 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.7212884676827325, + "lr": 0.03, + "negative_goodness": 1.0509944243537055, + "pair_accuracy": 0.910577777756585, + "positive_goodness": 3.95149168709649, + "runtime_seconds": 5.697181224822998 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6970755260043674, + "lr": 0.03, + "negative_goodness": 1.0241277264065212, + "pair_accuracy": 0.9153111111429002, + "positive_goodness": 4.004135726314121, + "runtime_seconds": 5.700494289398193 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6941078568140666, + "lr": 0.03, + "negative_goodness": 1.0202261762830946, + "pair_accuracy": 0.9190222222646077, + "positive_goodness": 4.052033515824212, + "runtime_seconds": 5.70062255859375 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6733550859027438, + "lr": 0.03, + "negative_goodness": 0.9832621575249566, + "pair_accuracy": 0.9201777777777778, + "positive_goodness": 4.125832840177748, + "runtime_seconds": 5.70168399810791 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6820949530495538, + "lr": 0.03, + "negative_goodness": 0.9914625395668878, + "pair_accuracy": 0.9166888888782925, + "positive_goodness": 4.127924662441678, + "runtime_seconds": 5.700541734695435 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6959327753490872, + "lr": 0.03, + "negative_goodness": 1.004011077033149, + "pair_accuracy": 0.9135555555979411, + "positive_goodness": 4.054662399588691, + "runtime_seconds": 5.700461387634277 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.703707708867391, + "lr": 0.03, + "negative_goodness": 1.0231371240191989, + "pair_accuracy": 0.910311111164093, + "positive_goodness": 4.011140643819173, + "runtime_seconds": 5.700968027114868 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.7324818994734023, + "lr": 0.03, + "negative_goodness": 1.0543110136244032, + "pair_accuracy": 0.9032888888676961, + "positive_goodness": 3.874596724955241, + "runtime_seconds": 5.704450845718384 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6877946521441142, + "lr": 0.03, + "negative_goodness": 1.0013150321642559, + "pair_accuracy": 0.913933333322737, + "positive_goodness": 4.0827318420410155, + "runtime_seconds": 5.730523586273193 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6711762631522284, + "lr": 0.03, + "negative_goodness": 0.9616798298729791, + "pair_accuracy": 0.9168666666984558, + "positive_goodness": 4.176285707855224, + "runtime_seconds": 5.748501300811768 + } + ], + "layer": 19 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.746504917467965, + "lr": 0.03, + "negative_goodness": 1.192281888877021, + "pair_accuracy": 0.9039777777777778, + "positive_goodness": 3.9958398749457467, + "runtime_seconds": 5.774827480316162 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.646177405526903, + "lr": 0.03, + "negative_goodness": 0.9895106896400452, + "pair_accuracy": 0.9129555555555555, + "positive_goodness": 4.407964524332682, + "runtime_seconds": 5.757359504699707 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.6568763248231676, + "lr": 0.03, + "negative_goodness": 0.9700846069547865, + "pair_accuracy": 0.9104444444444444, + "positive_goodness": 4.427821077728272, + "runtime_seconds": 5.773834466934204 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.5926674130121867, + "lr": 0.03, + "negative_goodness": 0.8937411913977729, + "pair_accuracy": 0.9234222222328186, + "positive_goodness": 4.739267477713691, + "runtime_seconds": 5.775673866271973 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.5593587899949816, + "lr": 0.03, + "negative_goodness": 0.8305445675320096, + "pair_accuracy": 0.9309555555714502, + "positive_goodness": 4.936140268241035, + "runtime_seconds": 5.7756028175354 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6206921752399869, + "lr": 0.03, + "negative_goodness": 0.9061093681971232, + "pair_accuracy": 0.9174222222116258, + "positive_goodness": 4.59780991651747, + "runtime_seconds": 5.7785162925720215 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6550905184825262, + "lr": 0.03, + "negative_goodness": 0.9513076179345449, + "pair_accuracy": 0.9055777777777778, + "positive_goodness": 4.522358288065592, + "runtime_seconds": 5.761461973190308 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.600490606546402, + "lr": 0.03, + "negative_goodness": 0.8504249448776245, + "pair_accuracy": 0.9177777777777778, + "positive_goodness": 4.605003950246175, + "runtime_seconds": 5.7276811599731445 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.598178320227729, + "lr": 0.03, + "negative_goodness": 0.8327650664859347, + "pair_accuracy": 0.9155999999894037, + "positive_goodness": 4.765759118609958, + "runtime_seconds": 5.744890928268433 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5660859701050652, + "lr": 0.03, + "negative_goodness": 0.8217740489853753, + "pair_accuracy": 0.9264888888888889, + "positive_goodness": 4.904238153500027, + "runtime_seconds": 5.766097784042358 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5968672676934136, + "lr": 0.03, + "negative_goodness": 0.8403316151936849, + "pair_accuracy": 0.9148888888941871, + "positive_goodness": 4.758172670661079, + "runtime_seconds": 5.766343593597412 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5452184511184692, + "lr": 0.03, + "negative_goodness": 0.8035976674185858, + "pair_accuracy": 0.9300222222646077, + "positive_goodness": 4.988195556343927, + "runtime_seconds": 5.750545978546143 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5511276573816936, + "lr": 0.03, + "negative_goodness": 0.8055549056159126, + "pair_accuracy": 0.9351777777724796, + "positive_goodness": 5.0777185972213745, + "runtime_seconds": 5.731651067733765 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6227538257651859, + "lr": 0.03, + "negative_goodness": 0.8734592464658949, + "pair_accuracy": 0.9055777777777778, + "positive_goodness": 4.608577184295655, + "runtime_seconds": 5.7704432010650635 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5595903827243381, + "lr": 0.03, + "negative_goodness": 0.8160647285355462, + "pair_accuracy": 0.9306888888888889, + "positive_goodness": 4.904559693230523, + "runtime_seconds": 5.771132230758667 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5529223177512487, + "lr": 0.03, + "negative_goodness": 0.7982207537280188, + "pair_accuracy": 0.9275777777777777, + "positive_goodness": 5.005473762936062, + "runtime_seconds": 5.77310848236084 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5539961952633328, + "lr": 0.03, + "negative_goodness": 0.8039123396555583, + "pair_accuracy": 0.9328222222434149, + "positive_goodness": 5.063785285801358, + "runtime_seconds": 5.772217273712158 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5701248872333102, + "lr": 0.03, + "negative_goodness": 0.8137362314648099, + "pair_accuracy": 0.9215111111323039, + "positive_goodness": 4.93994214108785, + "runtime_seconds": 5.766020774841309 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5540773571517732, + "lr": 0.03, + "negative_goodness": 0.8050570064968533, + "pair_accuracy": 0.9279111111111111, + "positive_goodness": 4.952839305877686, + "runtime_seconds": 5.771812200546265 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5525768010033502, + "lr": 0.03, + "negative_goodness": 0.7871935642666287, + "pair_accuracy": 0.9246222222381167, + "positive_goodness": 4.986049685711331, + "runtime_seconds": 5.770998001098633 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5760082555611928, + "lr": 0.03, + "negative_goodness": 0.8226467085520427, + "pair_accuracy": 0.9247111111111112, + "positive_goodness": 4.953537122090657, + "runtime_seconds": 5.77115797996521 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.549348661661148, + "lr": 0.03, + "negative_goodness": 0.7881570786476135, + "pair_accuracy": 0.9242666666666667, + "positive_goodness": 5.023995748901367, + "runtime_seconds": 5.768463611602783 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5505292378266653, + "lr": 0.03, + "negative_goodness": 0.777209182829327, + "pair_accuracy": 0.9190222222222222, + "positive_goodness": 4.987499838002523, + "runtime_seconds": 5.765850782394409 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5734816552983389, + "lr": 0.03, + "negative_goodness": 0.804159465985828, + "pair_accuracy": 0.9219111111111111, + "positive_goodness": 5.018432483418782, + "runtime_seconds": 5.767985105514526 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5605978979375628, + "lr": 0.03, + "negative_goodness": 0.8136387375248803, + "pair_accuracy": 0.9279777777777778, + "positive_goodness": 4.99617195332845, + "runtime_seconds": 5.769669532775879 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6041693143420749, + "lr": 0.03, + "negative_goodness": 0.8280545444700453, + "pair_accuracy": 0.9228000000317892, + "positive_goodness": 4.735340146340264, + "runtime_seconds": 5.7687695026397705 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5882692878405253, + "lr": 0.03, + "negative_goodness": 0.8327075459215376, + "pair_accuracy": 0.9202666666666667, + "positive_goodness": 4.725605791007148, + "runtime_seconds": 5.76594090461731 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6047397191047669, + "lr": 0.03, + "negative_goodness": 0.8388884000539779, + "pair_accuracy": 0.914, + "positive_goodness": 4.749292483859592, + "runtime_seconds": 5.767653465270996 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5888007972505358, + "lr": 0.03, + "negative_goodness": 0.8504074037763808, + "pair_accuracy": 0.9139555555237664, + "positive_goodness": 4.825941466437445, + "runtime_seconds": 5.76636266708374 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5017724362214406, + "lr": 0.03, + "negative_goodness": 0.7300211783938938, + "pair_accuracy": 0.9462444444444444, + "positive_goodness": 5.367614581129286, + "runtime_seconds": 5.764364004135132 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5638237345324623, + "lr": 0.03, + "negative_goodness": 0.8071941965315077, + "pair_accuracy": 0.9198, + "positive_goodness": 5.015726544443766, + "runtime_seconds": 5.764991044998169 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6134024450090196, + "lr": 0.03, + "negative_goodness": 0.8318457206090292, + "pair_accuracy": 0.9151333333121405, + "positive_goodness": 4.6745345777299665, + "runtime_seconds": 5.765571117401123 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5911979225688511, + "lr": 0.03, + "negative_goodness": 0.8311339224391513, + "pair_accuracy": 0.9107777777777778, + "positive_goodness": 4.837755716493395, + "runtime_seconds": 5.763781547546387 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5734446211338043, + "lr": 0.03, + "negative_goodness": 0.8079844941986931, + "pair_accuracy": 0.9194222222222223, + "positive_goodness": 5.008301584540473, + "runtime_seconds": 5.76393461227417 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5882066487550736, + "lr": 0.03, + "negative_goodness": 0.8229006448056962, + "pair_accuracy": 0.9154222222222222, + "positive_goodness": 4.79484088244968, + "runtime_seconds": 5.762035369873047 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5087898074785868, + "lr": 0.03, + "negative_goodness": 0.7638894797854954, + "pair_accuracy": 0.9420444444126553, + "positive_goodness": 5.252660886891683, + "runtime_seconds": 5.7634053230285645 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5656262221230401, + "lr": 0.03, + "negative_goodness": 0.7853118505689833, + "pair_accuracy": 0.9261333333439297, + "positive_goodness": 5.08302936905755, + "runtime_seconds": 5.769062519073486 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5379645358085632, + "lr": 0.03, + "negative_goodness": 0.7569934474150339, + "pair_accuracy": 0.9272444444444444, + "positive_goodness": 4.961096735805936, + "runtime_seconds": 5.759426593780518 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5772296222448349, + "lr": 0.03, + "negative_goodness": 0.8174748180601332, + "pair_accuracy": 0.9187777777777778, + "positive_goodness": 4.834553493075901, + "runtime_seconds": 5.76745343208313 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5691998064782884, + "lr": 0.03, + "negative_goodness": 0.8075627181424035, + "pair_accuracy": 0.9249777777777778, + "positive_goodness": 4.9651389528910315, + "runtime_seconds": 5.767188310623169 + } + ], + "layer": 20 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.7428148556868235, + "lr": 0.03, + "negative_goodness": 1.009572798013687, + "pair_accuracy": 0.8984, + "positive_goodness": 4.102617980363634, + "runtime_seconds": 5.7527196407318115 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.6908857504155901, + "lr": 0.03, + "negative_goodness": 1.0154289303514692, + "pair_accuracy": 0.9023777777777777, + "positive_goodness": 4.158219267951117, + "runtime_seconds": 5.737620830535889 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.6017744602282842, + "lr": 0.03, + "negative_goodness": 0.8892395595259137, + "pair_accuracy": 0.9221555555555555, + "positive_goodness": 4.50834197362264, + "runtime_seconds": 5.69761323928833 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6504314399825202, + "lr": 0.03, + "negative_goodness": 0.9282438901371426, + "pair_accuracy": 0.909933333322737, + "positive_goodness": 4.3891806492275665, + "runtime_seconds": 5.69562292098999 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.647202275466919, + "lr": 0.03, + "negative_goodness": 0.9071986725701227, + "pair_accuracy": 0.9031777777459886, + "positive_goodness": 4.376440244759453, + "runtime_seconds": 5.699294567108154 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6588519474824269, + "lr": 0.03, + "negative_goodness": 0.9321461787382762, + "pair_accuracy": 0.8982222222222223, + "positive_goodness": 4.425626082780626, + "runtime_seconds": 5.746535539627075 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.5501342499044206, + "lr": 0.03, + "negative_goodness": 0.8164014209217495, + "pair_accuracy": 0.9315111111111111, + "positive_goodness": 4.787025345696343, + "runtime_seconds": 5.732232093811035 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6618823233498468, + "lr": 0.03, + "negative_goodness": 0.9286178450690376, + "pair_accuracy": 0.8976888889312744, + "positive_goodness": 4.237853141784668, + "runtime_seconds": 5.70439887046814 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.5623268986384073, + "lr": 0.03, + "negative_goodness": 0.8138733287175497, + "pair_accuracy": 0.9268888889047835, + "positive_goodness": 4.716569514656067, + "runtime_seconds": 5.697106599807739 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5823449679480659, + "lr": 0.03, + "negative_goodness": 0.8388727773454454, + "pair_accuracy": 0.9215999999682108, + "positive_goodness": 4.681861766857571, + "runtime_seconds": 5.703355312347412 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5873574942906697, + "lr": 0.03, + "negative_goodness": 0.855556444952223, + "pair_accuracy": 0.9170888888729943, + "positive_goodness": 4.635685315216912, + "runtime_seconds": 5.711586952209473 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5673231011311213, + "lr": 0.03, + "negative_goodness": 0.8229846584161122, + "pair_accuracy": 0.9206888888888889, + "positive_goodness": 4.781351800367568, + "runtime_seconds": 5.712363958358765 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5668479283279843, + "lr": 0.03, + "negative_goodness": 0.8155678075578477, + "pair_accuracy": 0.9264888888888889, + "positive_goodness": 4.7963551353454585, + "runtime_seconds": 5.711527585983276 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6276424763997396, + "lr": 0.03, + "negative_goodness": 0.886898921140035, + "pair_accuracy": 0.9092666666666667, + "positive_goodness": 4.46294089296129, + "runtime_seconds": 5.7123799324035645 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.553037655766805, + "lr": 0.03, + "negative_goodness": 0.805893057568868, + "pair_accuracy": 0.9280222222222222, + "positive_goodness": 4.834404620446099, + "runtime_seconds": 5.7656543254852295 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5718262065887452, + "lr": 0.03, + "negative_goodness": 0.8320633211983575, + "pair_accuracy": 0.9215777777797646, + "positive_goodness": 4.805002007039388, + "runtime_seconds": 5.711460590362549 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5639220238155789, + "lr": 0.03, + "negative_goodness": 0.8121826200856103, + "pair_accuracy": 0.9259777777777778, + "positive_goodness": 4.70165281592475, + "runtime_seconds": 5.701445817947388 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5747581143485175, + "lr": 0.03, + "negative_goodness": 0.8526498676935832, + "pair_accuracy": 0.924711111164093, + "positive_goodness": 4.7317989138285315, + "runtime_seconds": 5.701628923416138 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.548208193869061, + "lr": 0.03, + "negative_goodness": 0.8123380743450589, + "pair_accuracy": 0.9297555555555556, + "positive_goodness": 4.886739801449246, + "runtime_seconds": 5.71523642539978 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5780948365953233, + "lr": 0.03, + "negative_goodness": 0.824911177444458, + "pair_accuracy": 0.9234888889100816, + "positive_goodness": 4.795563508372837, + "runtime_seconds": 5.761465787887573 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5119505446645949, + "lr": 0.03, + "negative_goodness": 0.724637806002299, + "pair_accuracy": 0.9310666666560703, + "positive_goodness": 5.140401090918647, + "runtime_seconds": 5.7056567668914795 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5648887824800279, + "lr": 0.03, + "negative_goodness": 0.8111314103020562, + "pair_accuracy": 0.918400000052982, + "positive_goodness": 4.822147284317016, + "runtime_seconds": 5.70621657371521 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5848804532263014, + "lr": 0.03, + "negative_goodness": 0.8322466042412652, + "pair_accuracy": 0.918311111079322, + "positive_goodness": 4.7340495305379235, + "runtime_seconds": 5.699137449264526 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5501536971410116, + "lr": 0.03, + "negative_goodness": 0.8113299584706625, + "pair_accuracy": 0.9325777778201633, + "positive_goodness": 4.818579499181111, + "runtime_seconds": 5.703050136566162 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5794388729519314, + "lr": 0.03, + "negative_goodness": 0.8310262003156874, + "pair_accuracy": 0.9184444444656372, + "positive_goodness": 4.768697421095107, + "runtime_seconds": 5.700189113616943 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5103833306206598, + "lr": 0.03, + "negative_goodness": 0.7649363394843207, + "pair_accuracy": 0.9416666666825613, + "positive_goodness": 5.0120662427902225, + "runtime_seconds": 5.71618914604187 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5267006521092521, + "lr": 0.03, + "negative_goodness": 0.7389593008968566, + "pair_accuracy": 0.9328666666666666, + "positive_goodness": 5.082619032287598, + "runtime_seconds": 5.7171690464019775 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5790581131352318, + "lr": 0.03, + "negative_goodness": 0.8057224964062373, + "pair_accuracy": 0.9178222222222222, + "positive_goodness": 4.772455982293023, + "runtime_seconds": 5.71695613861084 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5317489037381278, + "lr": 0.03, + "negative_goodness": 0.7651599706437853, + "pair_accuracy": 0.9370888888888889, + "positive_goodness": 5.056510584343804, + "runtime_seconds": 5.809706687927246 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5552971263037788, + "lr": 0.03, + "negative_goodness": 0.7850613664415148, + "pair_accuracy": 0.9250888889153799, + "positive_goodness": 4.931492949167887, + "runtime_seconds": 5.695232391357422 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5287609558502833, + "lr": 0.03, + "negative_goodness": 0.7696848704576492, + "pair_accuracy": 0.9360444444444445, + "positive_goodness": 4.987255048709446, + "runtime_seconds": 5.700540542602539 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5446992869165208, + "lr": 0.03, + "negative_goodness": 0.7900620350731744, + "pair_accuracy": 0.9252666666666667, + "positive_goodness": 4.956562307061089, + "runtime_seconds": 5.701624631881714 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5390034709903929, + "lr": 0.03, + "negative_goodness": 0.7622774787320031, + "pair_accuracy": 0.9246666666666666, + "positive_goodness": 4.99262759721544, + "runtime_seconds": 5.6979615688323975 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.558179290983412, + "lr": 0.03, + "negative_goodness": 0.7974225768831041, + "pair_accuracy": 0.9300000000211928, + "positive_goodness": 4.92750273276435, + "runtime_seconds": 5.700984716415405 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5580386633184221, + "lr": 0.03, + "negative_goodness": 0.8075221111721462, + "pair_accuracy": 0.9243111111005148, + "positive_goodness": 4.848821037716336, + "runtime_seconds": 5.743110179901123 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5571057445949978, + "lr": 0.03, + "negative_goodness": 0.8010746526930067, + "pair_accuracy": 0.9212666666454739, + "positive_goodness": 4.842429424879286, + "runtime_seconds": 5.700196027755737 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5527871266471015, + "lr": 0.03, + "negative_goodness": 0.7777534755812751, + "pair_accuracy": 0.9251777777989706, + "positive_goodness": 4.976997320344713, + "runtime_seconds": 5.699245929718018 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5758044769260618, + "lr": 0.03, + "negative_goodness": 0.8222665736569299, + "pair_accuracy": 0.9216666666666666, + "positive_goodness": 4.764862040117052, + "runtime_seconds": 5.699931383132935 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.59571982044114, + "lr": 0.03, + "negative_goodness": 0.8492936598141988, + "pair_accuracy": 0.9165333333227369, + "positive_goodness": 4.634458578830295, + "runtime_seconds": 5.6991448402404785 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.536081352127923, + "lr": 0.03, + "negative_goodness": 0.7800501566780939, + "pair_accuracy": 0.9333333333545261, + "positive_goodness": 5.021082417000665, + "runtime_seconds": 5.71549129486084 + } + ], + "layer": 21 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9158022232479519, + "lr": 0.03, + "negative_goodness": 1.2748495734532674, + "pair_accuracy": 0.8560222222275204, + "positive_goodness": 3.422542005856832, + "runtime_seconds": 5.739056587219238 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7845718409114414, + "lr": 0.03, + "negative_goodness": 1.1204705835872226, + "pair_accuracy": 0.8862666666666666, + "positive_goodness": 3.689883165825738, + "runtime_seconds": 5.726155042648315 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7610200921164618, + "lr": 0.03, + "negative_goodness": 1.0855861891640557, + "pair_accuracy": 0.8958666666613685, + "positive_goodness": 3.846726592402988, + "runtime_seconds": 5.72219443321228 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7368108070320554, + "lr": 0.03, + "negative_goodness": 1.0391365828461117, + "pair_accuracy": 0.8998222222222222, + "positive_goodness": 3.9068157988654244, + "runtime_seconds": 5.722283840179443 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7389519119474623, + "lr": 0.03, + "negative_goodness": 1.0338325393464831, + "pair_accuracy": 0.8933555555767483, + "positive_goodness": 3.9571500148349337, + "runtime_seconds": 5.786516189575195 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6956501661300659, + "lr": 0.03, + "negative_goodness": 0.988950027402242, + "pair_accuracy": 0.9054222222222222, + "positive_goodness": 4.066734867604573, + "runtime_seconds": 5.727678298950195 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7121242268721263, + "lr": 0.03, + "negative_goodness": 1.0112374235365127, + "pair_accuracy": 0.9056, + "positive_goodness": 4.002680098554823, + "runtime_seconds": 5.726747989654541 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6338653253502315, + "lr": 0.03, + "negative_goodness": 0.9314497576236724, + "pair_accuracy": 0.9262444444444444, + "positive_goodness": 4.34117100253635, + "runtime_seconds": 5.7441582679748535 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7241813119305505, + "lr": 0.03, + "negative_goodness": 1.0266012931929693, + "pair_accuracy": 0.8995555555555556, + "positive_goodness": 3.9528741822136775, + "runtime_seconds": 5.781005382537842 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6606558270348443, + "lr": 0.03, + "negative_goodness": 0.9708483539581299, + "pair_accuracy": 0.9220222222235468, + "positive_goodness": 4.175279452938503, + "runtime_seconds": 5.772340536117554 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6694381414837307, + "lr": 0.03, + "negative_goodness": 0.9708366502126058, + "pair_accuracy": 0.9137555555449591, + "positive_goodness": 4.171933107333713, + "runtime_seconds": 5.772831916809082 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6639566351678636, + "lr": 0.03, + "negative_goodness": 0.9466765246815152, + "pair_accuracy": 0.9153999999682109, + "positive_goodness": 4.252376545376248, + "runtime_seconds": 5.776269912719727 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6824605131997002, + "lr": 0.03, + "negative_goodness": 0.9938160099453396, + "pair_accuracy": 0.9142222222381168, + "positive_goodness": 4.146780004628499, + "runtime_seconds": 5.769298791885376 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6236625256114536, + "lr": 0.03, + "negative_goodness": 0.9062442581282721, + "pair_accuracy": 0.9302222222646077, + "positive_goodness": 4.418360205671522, + "runtime_seconds": 5.773621082305908 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6602642025629679, + "lr": 0.03, + "negative_goodness": 0.9487414491547479, + "pair_accuracy": 0.9176666666666666, + "positive_goodness": 4.305140459272597, + "runtime_seconds": 5.771631717681885 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6737460733625624, + "lr": 0.03, + "negative_goodness": 0.9616782094955444, + "pair_accuracy": 0.9143111111429002, + "positive_goodness": 4.222655561998155, + "runtime_seconds": 5.7715535163879395 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6045331915113661, + "lr": 0.03, + "negative_goodness": 0.8842248244073656, + "pair_accuracy": 0.9313777777777777, + "positive_goodness": 4.538564077928331, + "runtime_seconds": 5.7782301902771 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6269677652570936, + "lr": 0.03, + "negative_goodness": 0.9195878952980041, + "pair_accuracy": 0.9270222222222222, + "positive_goodness": 4.390675894673666, + "runtime_seconds": 5.776055097579956 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6404883182313708, + "lr": 0.03, + "negative_goodness": 0.9094976113213433, + "pair_accuracy": 0.9165999999894037, + "positive_goodness": 4.413618849923876, + "runtime_seconds": 5.771401405334473 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6531680600166321, + "lr": 0.03, + "negative_goodness": 0.921480109861162, + "pair_accuracy": 0.9150000000529819, + "positive_goodness": 4.335416521453857, + "runtime_seconds": 5.77219820022583 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6708007087071737, + "lr": 0.03, + "negative_goodness": 0.9711178276909722, + "pair_accuracy": 0.9138000000211928, + "positive_goodness": 4.181431842252943, + "runtime_seconds": 5.770110607147217 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6300383978949653, + "lr": 0.03, + "negative_goodness": 0.9158248655743069, + "pair_accuracy": 0.9244444444550408, + "positive_goodness": 4.463575002119276, + "runtime_seconds": 5.774263381958008 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6507966896480984, + "lr": 0.03, + "negative_goodness": 0.9129124313990276, + "pair_accuracy": 0.9142666666772631, + "positive_goodness": 4.303360291459826, + "runtime_seconds": 5.775014638900757 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.646243994612164, + "lr": 0.03, + "negative_goodness": 0.9374561444494459, + "pair_accuracy": 0.9183333333333333, + "positive_goodness": 4.2953485142178005, + "runtime_seconds": 5.776580572128296 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6751826821962993, + "lr": 0.03, + "negative_goodness": 0.9374526592254638, + "pair_accuracy": 0.9046, + "positive_goodness": 4.231614378611247, + "runtime_seconds": 5.7749247550964355 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6404543456183539, + "lr": 0.03, + "negative_goodness": 0.9101236077626547, + "pair_accuracy": 0.9187999999894036, + "positive_goodness": 4.343031591457791, + "runtime_seconds": 5.766470909118652 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6371370430019166, + "lr": 0.03, + "negative_goodness": 0.9137476211388906, + "pair_accuracy": 0.9202444444444444, + "positive_goodness": 4.454691663275825, + "runtime_seconds": 5.769125699996948 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6727831812964545, + "lr": 0.03, + "negative_goodness": 0.9835692531161838, + "pair_accuracy": 0.9112, + "positive_goodness": 4.212615512932671, + "runtime_seconds": 5.769660949707031 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6381023444122739, + "lr": 0.03, + "negative_goodness": 0.9219874803172218, + "pair_accuracy": 0.9248222222222222, + "positive_goodness": 4.328705134328207, + "runtime_seconds": 5.770522117614746 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6355500014411078, + "lr": 0.03, + "negative_goodness": 0.9136206320020888, + "pair_accuracy": 0.9163333333174387, + "positive_goodness": 4.379345751550463, + "runtime_seconds": 5.772338390350342 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6542871587435405, + "lr": 0.03, + "negative_goodness": 0.9244012186474271, + "pair_accuracy": 0.9164888889100816, + "positive_goodness": 4.359427176751031, + "runtime_seconds": 5.773272514343262 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.664187037732866, + "lr": 0.03, + "negative_goodness": 0.9077838017039829, + "pair_accuracy": 0.9018888888888889, + "positive_goodness": 4.31265775197347, + "runtime_seconds": 5.770962953567505 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6576355831782024, + "lr": 0.03, + "negative_goodness": 0.9510465106540256, + "pair_accuracy": 0.9162888889100816, + "positive_goodness": 4.307362069447835, + "runtime_seconds": 5.77982759475708 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6067596122953627, + "lr": 0.03, + "negative_goodness": 0.8841717366324531, + "pair_accuracy": 0.928311111164093, + "positive_goodness": 4.483722672865126, + "runtime_seconds": 5.773644208908081 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6280429982503255, + "lr": 0.03, + "negative_goodness": 0.9026978871663411, + "pair_accuracy": 0.9225777777777778, + "positive_goodness": 4.423261332978143, + "runtime_seconds": 5.772738456726074 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6102971201578776, + "lr": 0.03, + "negative_goodness": 0.8861939358181423, + "pair_accuracy": 0.9298, + "positive_goodness": 4.540893933063083, + "runtime_seconds": 5.774813652038574 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6057277827474806, + "lr": 0.03, + "negative_goodness": 0.8534703689575195, + "pair_accuracy": 0.9218444444444445, + "positive_goodness": 4.565166687435574, + "runtime_seconds": 5.765909671783447 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6036392381668091, + "lr": 0.03, + "negative_goodness": 0.8551051323784722, + "pair_accuracy": 0.9289111111323038, + "positive_goodness": 4.648189417775472, + "runtime_seconds": 5.765926122665405 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5939441831800673, + "lr": 0.03, + "negative_goodness": 0.8431892071617975, + "pair_accuracy": 0.9248888888941871, + "positive_goodness": 4.655638651360406, + "runtime_seconds": 5.772392272949219 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.666224861158265, + "lr": 0.03, + "negative_goodness": 0.9504491801526811, + "pair_accuracy": 0.9126444444444445, + "positive_goodness": 4.332281086730957, + "runtime_seconds": 5.770372152328491 + } + ], + "layer": 22 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.7597865676350064, + "lr": 0.03, + "negative_goodness": 1.1032827728589376, + "pair_accuracy": 0.8925777777777778, + "positive_goodness": 3.7613116094801162, + "runtime_seconds": 5.723416090011597 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.5378080727736155, + "lr": 0.03, + "negative_goodness": 0.8410175452868144, + "pair_accuracy": 0.9443999999894036, + "positive_goodness": 4.782545659467909, + "runtime_seconds": 5.733130693435669 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.5298088203854031, + "lr": 0.03, + "negative_goodness": 0.7922448548210992, + "pair_accuracy": 0.938066666677263, + "positive_goodness": 4.880243152109782, + "runtime_seconds": 5.752476453781128 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.5581458579328326, + "lr": 0.03, + "negative_goodness": 0.8023349838309818, + "pair_accuracy": 0.9277777777777778, + "positive_goodness": 4.780967174699572, + "runtime_seconds": 5.7525458335876465 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.5566692732281155, + "lr": 0.03, + "negative_goodness": 0.8046165510177612, + "pair_accuracy": 0.930688888920678, + "positive_goodness": 4.873879135174222, + "runtime_seconds": 5.753463983535767 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.5398722642474705, + "lr": 0.03, + "negative_goodness": 0.7858881309085423, + "pair_accuracy": 0.9389999999894036, + "positive_goodness": 4.941533960236443, + "runtime_seconds": 5.752246618270874 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.5635275330066681, + "lr": 0.03, + "negative_goodness": 0.806165946843889, + "pair_accuracy": 0.926, + "positive_goodness": 4.812397022162544, + "runtime_seconds": 5.752748250961304 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.5005361453692119, + "lr": 0.03, + "negative_goodness": 0.7184887111451891, + "pair_accuracy": 0.9444444444444444, + "positive_goodness": 5.094257862345378, + "runtime_seconds": 5.749958276748657 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.47864682670699227, + "lr": 0.03, + "negative_goodness": 0.6869535485267639, + "pair_accuracy": 0.9493777777777778, + "positive_goodness": 5.274619487508138, + "runtime_seconds": 5.753445148468018 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5387196906937494, + "lr": 0.03, + "negative_goodness": 0.7727665319866605, + "pair_accuracy": 0.9365333333333333, + "positive_goodness": 4.940619318517049, + "runtime_seconds": 5.761058568954468 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5085024002657996, + "lr": 0.03, + "negative_goodness": 0.7271389644993677, + "pair_accuracy": 0.9392444444444444, + "positive_goodness": 5.02051760474311, + "runtime_seconds": 5.756248712539673 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5147156748983596, + "lr": 0.03, + "negative_goodness": 0.70807825595008, + "pair_accuracy": 0.9348222222116258, + "positive_goodness": 5.024937693617079, + "runtime_seconds": 5.752288103103638 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5023562056382497, + "lr": 0.03, + "negative_goodness": 0.6949789440260993, + "pair_accuracy": 0.9369555555449591, + "positive_goodness": 5.182579813554552, + "runtime_seconds": 5.755751132965088 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5229635483211941, + "lr": 0.03, + "negative_goodness": 0.7276155243661668, + "pair_accuracy": 0.9310888888888889, + "positive_goodness": 5.072771271006267, + "runtime_seconds": 5.756684303283691 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.4945284484439426, + "lr": 0.03, + "negative_goodness": 0.6754602149009704, + "pair_accuracy": 0.9334888888570998, + "positive_goodness": 5.2694120451185436, + "runtime_seconds": 5.754324436187744 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5400338170395957, + "lr": 0.03, + "negative_goodness": 0.7586257376909256, + "pair_accuracy": 0.9270222222222222, + "positive_goodness": 4.925584698571099, + "runtime_seconds": 5.754494905471802 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.4549777421871821, + "lr": 0.03, + "negative_goodness": 0.6432049429019292, + "pair_accuracy": 0.9510444444444445, + "positive_goodness": 5.380245847320556, + "runtime_seconds": 5.750392913818359 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.4567881935093138, + "lr": 0.03, + "negative_goodness": 0.6335200913959079, + "pair_accuracy": 0.9436444444444444, + "positive_goodness": 5.468264590369331, + "runtime_seconds": 5.753018617630005 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5508346231248644, + "lr": 0.03, + "negative_goodness": 0.7470167620870802, + "pair_accuracy": 0.925844444445769, + "positive_goodness": 4.919353705957201, + "runtime_seconds": 5.752810478210449 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.49250969404909345, + "lr": 0.03, + "negative_goodness": 0.6969889356825086, + "pair_accuracy": 0.9417555555555556, + "positive_goodness": 5.18569611714681, + "runtime_seconds": 5.754855155944824 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5000248632510503, + "lr": 0.03, + "negative_goodness": 0.6846590441332923, + "pair_accuracy": 0.9376, + "positive_goodness": 5.224633295525445, + "runtime_seconds": 5.759576082229614 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.4794923448350694, + "lr": 0.03, + "negative_goodness": 0.6798613057666355, + "pair_accuracy": 0.9444666666666667, + "positive_goodness": 5.222694150967068, + "runtime_seconds": 5.755669355392456 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.4911986413107978, + "lr": 0.03, + "negative_goodness": 0.6752415509753757, + "pair_accuracy": 0.9399111111111111, + "positive_goodness": 5.252689332411024, + "runtime_seconds": 5.7521138191223145 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.4554533041636149, + "lr": 0.03, + "negative_goodness": 0.6402307418399387, + "pair_accuracy": 0.9497555555555556, + "positive_goodness": 5.387772097735935, + "runtime_seconds": 5.755874395370483 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.44476763022210863, + "lr": 0.03, + "negative_goodness": 0.6148889507955975, + "pair_accuracy": 0.9512, + "positive_goodness": 5.502230835639106, + "runtime_seconds": 5.752506256103516 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.44758611760669287, + "lr": 0.03, + "negative_goodness": 0.6132875776555803, + "pair_accuracy": 0.9486222222222223, + "positive_goodness": 5.467687959882948, + "runtime_seconds": 5.753904581069946 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.4575349214977688, + "lr": 0.03, + "negative_goodness": 0.6168965344587962, + "pair_accuracy": 0.9474222222540114, + "positive_goodness": 5.490343040508694, + "runtime_seconds": 5.751749753952026 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5303549104690551, + "lr": 0.03, + "negative_goodness": 0.7121720992724101, + "pair_accuracy": 0.9321777778201633, + "positive_goodness": 5.0366641922420925, + "runtime_seconds": 5.754856824874878 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.43890498575899334, + "lr": 0.03, + "negative_goodness": 0.6249061688489385, + "pair_accuracy": 0.9594222222222222, + "positive_goodness": 5.472611295233833, + "runtime_seconds": 5.753361701965332 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.43858586450417836, + "lr": 0.03, + "negative_goodness": 0.5943158519877327, + "pair_accuracy": 0.9515333333333333, + "positive_goodness": 5.5877202363755964, + "runtime_seconds": 5.751518249511719 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.43950404228634304, + "lr": 0.03, + "negative_goodness": 0.5939206649462382, + "pair_accuracy": 0.9499777778201633, + "positive_goodness": 5.535954093382093, + "runtime_seconds": 5.753729343414307 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5283757723066542, + "lr": 0.03, + "negative_goodness": 0.7100444530354606, + "pair_accuracy": 0.9306222222222222, + "positive_goodness": 5.058100440979004, + "runtime_seconds": 5.758105039596558 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.43558644897143045, + "lr": 0.03, + "negative_goodness": 0.599708374218146, + "pair_accuracy": 0.9562, + "positive_goodness": 5.540505610571968, + "runtime_seconds": 5.755064010620117 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.47109462331136065, + "lr": 0.03, + "negative_goodness": 0.612176609939999, + "pair_accuracy": 0.9434888889100816, + "positive_goodness": 5.434012820180257, + "runtime_seconds": 5.7572760581970215 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.4773023063023885, + "lr": 0.03, + "negative_goodness": 0.6236825687461429, + "pair_accuracy": 0.9533333333651225, + "positive_goodness": 5.3205304876539445, + "runtime_seconds": 5.755882024765015 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.4702939012395011, + "lr": 0.03, + "negative_goodness": 0.6169281500915687, + "pair_accuracy": 0.9428444444444445, + "positive_goodness": 5.302540976715088, + "runtime_seconds": 5.754346132278442 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.45481620633337233, + "lr": 0.03, + "negative_goodness": 0.6111852466212379, + "pair_accuracy": 0.9459555555555555, + "positive_goodness": 5.448009305742052, + "runtime_seconds": 5.759153604507446 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.44342759282853866, + "lr": 0.03, + "negative_goodness": 0.6258061089727613, + "pair_accuracy": 0.9551333333333333, + "positive_goodness": 5.441068546549479, + "runtime_seconds": 5.7518298625946045 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.42877328499688044, + "lr": 0.03, + "negative_goodness": 0.557102456996176, + "pair_accuracy": 0.9529777777777778, + "positive_goodness": 5.652599124230279, + "runtime_seconds": 5.7256083488464355 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.47371829874250626, + "lr": 0.03, + "negative_goodness": 0.620463962369495, + "pair_accuracy": 0.9402888888676961, + "positive_goodness": 5.438083211474948, + "runtime_seconds": 5.75152063369751 + } + ], + "layer": 23 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8236368554751078, + "lr": 0.03, + "negative_goodness": 1.2394282296392654, + "pair_accuracy": 0.8860222222222223, + "positive_goodness": 3.891671777767605, + "runtime_seconds": 5.768295526504517 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7312335719691383, + "lr": 0.03, + "negative_goodness": 1.1024558968332079, + "pair_accuracy": 0.8998222222222222, + "positive_goodness": 4.047236139000787, + "runtime_seconds": 5.775110244750977 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7119262755182054, + "lr": 0.03, + "negative_goodness": 1.0597153746922812, + "pair_accuracy": 0.8987999999894036, + "positive_goodness": 4.138730842166477, + "runtime_seconds": 5.776895523071289 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6540865354220072, + "lr": 0.03, + "negative_goodness": 0.9948378452089097, + "pair_accuracy": 0.9178666666666667, + "positive_goodness": 4.3115215834723575, + "runtime_seconds": 5.77028489112854 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6435201360755497, + "lr": 0.03, + "negative_goodness": 0.9917075356536441, + "pair_accuracy": 0.9234666666666667, + "positive_goodness": 4.324371408759223, + "runtime_seconds": 5.779892206192017 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.5941727326975929, + "lr": 0.03, + "negative_goodness": 0.92133263247278, + "pair_accuracy": 0.9356222222222222, + "positive_goodness": 4.515372006522284, + "runtime_seconds": 5.77513575553894 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.576439141315884, + "lr": 0.03, + "negative_goodness": 0.8724753833770752, + "pair_accuracy": 0.9361111111111111, + "positive_goodness": 4.600004647869534, + "runtime_seconds": 5.775555849075317 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.5684406332333882, + "lr": 0.03, + "negative_goodness": 0.8490902305179172, + "pair_accuracy": 0.9329777778307596, + "positive_goodness": 4.638883229827881, + "runtime_seconds": 5.777878522872925 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.5428425333870782, + "lr": 0.03, + "negative_goodness": 0.818441285472446, + "pair_accuracy": 0.9413333333333334, + "positive_goodness": 4.806975565973918, + "runtime_seconds": 5.773029804229736 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5809653398301866, + "lr": 0.03, + "negative_goodness": 0.8635092038048638, + "pair_accuracy": 0.930866666677263, + "positive_goodness": 4.601922847747803, + "runtime_seconds": 5.775029182434082 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5667288667149014, + "lr": 0.03, + "negative_goodness": 0.8404975007163153, + "pair_accuracy": 0.9322888889206781, + "positive_goodness": 4.671448984061347, + "runtime_seconds": 5.776838302612305 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5850747918234931, + "lr": 0.03, + "negative_goodness": 0.8619788706037733, + "pair_accuracy": 0.9272888889312744, + "positive_goodness": 4.48115041847229, + "runtime_seconds": 5.777906656265259 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5430946508407593, + "lr": 0.03, + "negative_goodness": 0.8100481690300836, + "pair_accuracy": 0.9421555555555555, + "positive_goodness": 4.7249588662041555, + "runtime_seconds": 5.778505086898804 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5564938066588507, + "lr": 0.03, + "negative_goodness": 0.8302659429868062, + "pair_accuracy": 0.9371777777671814, + "positive_goodness": 4.70767255660163, + "runtime_seconds": 5.772287845611572 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5581318636152479, + "lr": 0.03, + "negative_goodness": 0.8105740712536705, + "pair_accuracy": 0.9330444444444445, + "positive_goodness": 4.723146457163493, + "runtime_seconds": 5.776129484176636 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.49099857121043733, + "lr": 0.03, + "negative_goodness": 0.730945943429735, + "pair_accuracy": 0.9542666666666667, + "positive_goodness": 5.044725779808892, + "runtime_seconds": 5.724509000778198 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5727126641061571, + "lr": 0.03, + "negative_goodness": 0.8098070030106439, + "pair_accuracy": 0.9320444444444445, + "positive_goodness": 4.67849929055108, + "runtime_seconds": 5.7135984897613525 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5818944966634114, + "lr": 0.03, + "negative_goodness": 0.8625370081159803, + "pair_accuracy": 0.9360666666719648, + "positive_goodness": 4.6052071649339466, + "runtime_seconds": 5.712658405303955 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5505909269332886, + "lr": 0.03, + "negative_goodness": 0.8124216866599189, + "pair_accuracy": 0.9441333333757188, + "positive_goodness": 4.695871950531006, + "runtime_seconds": 5.720596790313721 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5240447458161248, + "lr": 0.03, + "negative_goodness": 0.7711205706702339, + "pair_accuracy": 0.9422444444974264, + "positive_goodness": 4.924391558159722, + "runtime_seconds": 5.7157392501831055 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5178044118669298, + "lr": 0.03, + "negative_goodness": 0.7339083123842876, + "pair_accuracy": 0.9432888888902135, + "positive_goodness": 4.954157479667663, + "runtime_seconds": 5.718006372451782 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5451760921637218, + "lr": 0.03, + "negative_goodness": 0.752618955538008, + "pair_accuracy": 0.9319111111111111, + "positive_goodness": 4.866562509875827, + "runtime_seconds": 5.720115423202515 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5730807255956861, + "lr": 0.03, + "negative_goodness": 0.8286298975626628, + "pair_accuracy": 0.9284222222010294, + "positive_goodness": 4.636204884847005, + "runtime_seconds": 5.765565395355225 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.4890427827676137, + "lr": 0.03, + "negative_goodness": 0.7203051743295458, + "pair_accuracy": 0.9542666666666667, + "positive_goodness": 5.033970663282607, + "runtime_seconds": 5.7737696170806885 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5419051138983833, + "lr": 0.03, + "negative_goodness": 0.7932650146272447, + "pair_accuracy": 0.9349555555555555, + "positive_goodness": 4.760821913772159, + "runtime_seconds": 5.751305341720581 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5617631230672201, + "lr": 0.03, + "negative_goodness": 0.7720039734098646, + "pair_accuracy": 0.922977777756585, + "positive_goodness": 4.761349189885458, + "runtime_seconds": 5.7551915645599365 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.4782078629917569, + "lr": 0.03, + "negative_goodness": 0.6986148377140363, + "pair_accuracy": 0.9542888888888889, + "positive_goodness": 5.17087609363132, + "runtime_seconds": 5.776471138000488 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.49179869425031875, + "lr": 0.03, + "negative_goodness": 0.7068904551612006, + "pair_accuracy": 0.948755555566152, + "positive_goodness": 5.082620434655084, + "runtime_seconds": 5.772497653961182 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5140071848975287, + "lr": 0.03, + "negative_goodness": 0.7315652355088128, + "pair_accuracy": 0.9460000000079473, + "positive_goodness": 4.9646019449657866, + "runtime_seconds": 5.775348424911499 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5114593678368462, + "lr": 0.03, + "negative_goodness": 0.7201736993895637, + "pair_accuracy": 0.9410222221904331, + "positive_goodness": 5.019007443873088, + "runtime_seconds": 5.773955821990967 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5434818915473091, + "lr": 0.03, + "negative_goodness": 0.7923279324001736, + "pair_accuracy": 0.9350222222222222, + "positive_goodness": 4.83517402581109, + "runtime_seconds": 5.770503759384155 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5316653374698427, + "lr": 0.03, + "negative_goodness": 0.7542348668045468, + "pair_accuracy": 0.9415777777777777, + "positive_goodness": 4.974805640241835, + "runtime_seconds": 5.779271125793457 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.49914046280119156, + "lr": 0.03, + "negative_goodness": 0.6958955371485817, + "pair_accuracy": 0.9470444444444445, + "positive_goodness": 5.193199293263754, + "runtime_seconds": 5.7765772342681885 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.47235146526760524, + "lr": 0.03, + "negative_goodness": 0.6541238987869686, + "pair_accuracy": 0.9505111111111111, + "positive_goodness": 5.297444203948975, + "runtime_seconds": 5.777455806732178 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5087363817109002, + "lr": 0.03, + "negative_goodness": 0.7079899822447036, + "pair_accuracy": 0.9399333333333333, + "positive_goodness": 5.110126774427626, + "runtime_seconds": 5.778188705444336 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5123800580713485, + "lr": 0.03, + "negative_goodness": 0.731172288968828, + "pair_accuracy": 0.9425555555555556, + "positive_goodness": 4.991833383517795, + "runtime_seconds": 5.785386800765991 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5537448647181193, + "lr": 0.03, + "negative_goodness": 0.756736624537574, + "pair_accuracy": 0.9325555555449592, + "positive_goodness": 4.894533478207058, + "runtime_seconds": 5.77578067779541 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5019587519380782, + "lr": 0.03, + "negative_goodness": 0.7085958245595296, + "pair_accuracy": 0.9477777777777778, + "positive_goodness": 5.111588397894965, + "runtime_seconds": 5.7799506187438965 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5520952100223965, + "lr": 0.03, + "negative_goodness": 0.7796248361799452, + "pair_accuracy": 0.9296444444656372, + "positive_goodness": 4.847157195154826, + "runtime_seconds": 5.77272891998291 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5562735932985942, + "lr": 0.03, + "negative_goodness": 0.7908286596616109, + "pair_accuracy": 0.9371111111111111, + "positive_goodness": 4.832965255228679, + "runtime_seconds": 5.774224281311035 + } + ], + "layer": 24 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.7759124896526337, + "lr": 0.03, + "negative_goodness": 1.1005417030599383, + "pair_accuracy": 0.8830666666666667, + "positive_goodness": 3.782617977057563, + "runtime_seconds": 5.778293132781982 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.6930125774966346, + "lr": 0.03, + "negative_goodness": 1.0255520627392662, + "pair_accuracy": 0.8987111111111111, + "positive_goodness": 4.0999359624226885, + "runtime_seconds": 5.8612096309661865 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.6418475437376234, + "lr": 0.03, + "negative_goodness": 0.9681947346581353, + "pair_accuracy": 0.9147111111164093, + "positive_goodness": 4.343923709551493, + "runtime_seconds": 5.956259489059448 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6267523051288393, + "lr": 0.03, + "negative_goodness": 0.9178170670535829, + "pair_accuracy": 0.9103111111111111, + "positive_goodness": 4.539555975680881, + "runtime_seconds": 5.759544134140015 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6266075184292264, + "lr": 0.03, + "negative_goodness": 0.9118377769682142, + "pair_accuracy": 0.9166888888941871, + "positive_goodness": 4.462033198229472, + "runtime_seconds": 5.8082075119018555 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6052466165542603, + "lr": 0.03, + "negative_goodness": 0.8931362322065566, + "pair_accuracy": 0.9232666666772631, + "positive_goodness": 4.499123220570882, + "runtime_seconds": 5.758336305618286 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.5909486871401469, + "lr": 0.03, + "negative_goodness": 0.896508368704054, + "pair_accuracy": 0.9285555555449592, + "positive_goodness": 4.546556822035048, + "runtime_seconds": 5.868626832962036 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.556941071150038, + "lr": 0.03, + "negative_goodness": 0.8163719156265259, + "pair_accuracy": 0.9345333333439297, + "positive_goodness": 4.7897312363518605, + "runtime_seconds": 5.842290639877319 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.5722500378131866, + "lr": 0.03, + "negative_goodness": 0.7971140820172098, + "pair_accuracy": 0.9256222222222222, + "positive_goodness": 4.699186923641629, + "runtime_seconds": 5.765608549118042 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.614142225546307, + "lr": 0.03, + "negative_goodness": 0.8862986117270257, + "pair_accuracy": 0.9114444444444444, + "positive_goodness": 4.480087195078532, + "runtime_seconds": 5.772972345352173 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.5901966046015421, + "lr": 0.03, + "negative_goodness": 0.8452876313951281, + "pair_accuracy": 0.9255333333333333, + "positive_goodness": 4.692075071207682, + "runtime_seconds": 5.7524354457855225 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5893073957761129, + "lr": 0.03, + "negative_goodness": 0.8268574403497908, + "pair_accuracy": 0.9221555555767483, + "positive_goodness": 4.664543419350518, + "runtime_seconds": 5.760410785675049 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5531603454298444, + "lr": 0.03, + "negative_goodness": 0.787946060736974, + "pair_accuracy": 0.9330222222222222, + "positive_goodness": 4.825251840040419, + "runtime_seconds": 5.750154256820679 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5560770448923111, + "lr": 0.03, + "negative_goodness": 0.8139467695858743, + "pair_accuracy": 0.9370666666666667, + "positive_goodness": 4.760530293697781, + "runtime_seconds": 5.750669240951538 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5636890802568859, + "lr": 0.03, + "negative_goodness": 0.8124043542716238, + "pair_accuracy": 0.9322444444444444, + "positive_goodness": 4.704586854553223, + "runtime_seconds": 5.753316879272461 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5903086110962762, + "lr": 0.03, + "negative_goodness": 0.8468953497621748, + "pair_accuracy": 0.9223555555449592, + "positive_goodness": 4.548601989279853, + "runtime_seconds": 5.749436855316162 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5629288630379571, + "lr": 0.03, + "negative_goodness": 0.7921164613458845, + "pair_accuracy": 0.924822222243415, + "positive_goodness": 4.748258216328091, + "runtime_seconds": 5.752376317977905 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5861538963397344, + "lr": 0.03, + "negative_goodness": 0.8348544835911856, + "pair_accuracy": 0.9253777777777777, + "positive_goodness": 4.553587959713406, + "runtime_seconds": 5.751805305480957 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5220763342036141, + "lr": 0.03, + "negative_goodness": 0.7375006920907232, + "pair_accuracy": 0.9355111111111111, + "positive_goodness": 4.930164765591091, + "runtime_seconds": 5.747009038925171 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5801256591796875, + "lr": 0.03, + "negative_goodness": 0.7988071659194098, + "pair_accuracy": 0.9224666666772631, + "positive_goodness": 4.649323149066501, + "runtime_seconds": 5.754092216491699 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5724094608041975, + "lr": 0.03, + "negative_goodness": 0.817235540129741, + "pair_accuracy": 0.9273333333333333, + "positive_goodness": 4.703226401435004, + "runtime_seconds": 5.748584270477295 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5066130536397299, + "lr": 0.03, + "negative_goodness": 0.7043162304666307, + "pair_accuracy": 0.9402888889100817, + "positive_goodness": 5.053125237443712, + "runtime_seconds": 5.748387098312378 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5256743421289656, + "lr": 0.03, + "negative_goodness": 0.7240827173338996, + "pair_accuracy": 0.9327555555449591, + "positive_goodness": 4.944138068644206, + "runtime_seconds": 5.75201678276062 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5179924608442519, + "lr": 0.03, + "negative_goodness": 0.7405056825637817, + "pair_accuracy": 0.9409555555568802, + "positive_goodness": 5.019666278478835, + "runtime_seconds": 5.751461505889893 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.49154040834638807, + "lr": 0.03, + "negative_goodness": 0.6858152725060781, + "pair_accuracy": 0.9440444444232516, + "positive_goodness": 5.19856129667494, + "runtime_seconds": 5.752009868621826 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.47378333235846626, + "lr": 0.03, + "negative_goodness": 0.6641707968155544, + "pair_accuracy": 0.9505111111111111, + "positive_goodness": 5.266504046122233, + "runtime_seconds": 5.739060401916504 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5225185769372516, + "lr": 0.03, + "negative_goodness": 0.7078885957479477, + "pair_accuracy": 0.9342444444444444, + "positive_goodness": 5.01822876180013, + "runtime_seconds": 5.699687719345093 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5668482038391961, + "lr": 0.03, + "negative_goodness": 0.7948266682200962, + "pair_accuracy": 0.9274, + "positive_goodness": 4.7111146354675295, + "runtime_seconds": 5.701916217803955 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.4794558560901218, + "lr": 0.03, + "negative_goodness": 0.6595524687078264, + "pair_accuracy": 0.9460000000158946, + "positive_goodness": 5.294395875295003, + "runtime_seconds": 5.752354860305786 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5305708394156562, + "lr": 0.03, + "negative_goodness": 0.7212659503195021, + "pair_accuracy": 0.9360222222222222, + "positive_goodness": 4.988705612013075, + "runtime_seconds": 5.749047756195068 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5126206144756741, + "lr": 0.03, + "negative_goodness": 0.6893335135009554, + "pair_accuracy": 0.9355555555767483, + "positive_goodness": 5.1314979195912676, + "runtime_seconds": 5.753125905990601 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5124257130304972, + "lr": 0.03, + "negative_goodness": 0.7081922082795037, + "pair_accuracy": 0.9385999999788073, + "positive_goodness": 5.021304100884332, + "runtime_seconds": 5.750704288482666 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5173896270540026, + "lr": 0.03, + "negative_goodness": 0.7179763283729553, + "pair_accuracy": 0.9358666666560703, + "positive_goodness": 5.094697735171848, + "runtime_seconds": 5.756037950515747 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5297377308739556, + "lr": 0.03, + "negative_goodness": 0.7070946801622708, + "pair_accuracy": 0.9313555555555556, + "positive_goodness": 5.103248092312283, + "runtime_seconds": 5.754626989364624 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5112835858239068, + "lr": 0.03, + "negative_goodness": 0.7053363431241777, + "pair_accuracy": 0.9391111111640931, + "positive_goodness": 5.111908788553873, + "runtime_seconds": 5.758772850036621 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5803942813449435, + "lr": 0.03, + "negative_goodness": 0.7857773713429769, + "pair_accuracy": 0.9227333333545261, + "positive_goodness": 4.746806896167331, + "runtime_seconds": 5.752537250518799 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5024096036487156, + "lr": 0.03, + "negative_goodness": 0.7004105557441711, + "pair_accuracy": 0.9412888889418708, + "positive_goodness": 5.201705381181505, + "runtime_seconds": 5.758629560470581 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5590816033151415, + "lr": 0.03, + "negative_goodness": 0.7719264529863993, + "pair_accuracy": 0.931400000052982, + "positive_goodness": 4.8967368620554605, + "runtime_seconds": 5.756474733352661 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5352963430616591, + "lr": 0.03, + "negative_goodness": 0.7242437351173825, + "pair_accuracy": 0.9311555555555555, + "positive_goodness": 5.093951758702596, + "runtime_seconds": 5.749353647232056 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5310772362497118, + "lr": 0.03, + "negative_goodness": 0.7346345748477512, + "pair_accuracy": 0.939044444433848, + "positive_goodness": 5.0013569332970516, + "runtime_seconds": 5.7513651847839355 + } + ], + "layer": 25 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8691449199888441, + "lr": 0.03, + "negative_goodness": 1.2756023330900403, + "pair_accuracy": 0.8686888888994853, + "positive_goodness": 3.9053985872480603, + "runtime_seconds": 5.768747091293335 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7758376006656222, + "lr": 0.03, + "negative_goodness": 1.1540445440928142, + "pair_accuracy": 0.8844, + "positive_goodness": 3.994426626841227, + "runtime_seconds": 5.767136573791504 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7164606677903069, + "lr": 0.03, + "negative_goodness": 1.0821673134909735, + "pair_accuracy": 0.8980000000105964, + "positive_goodness": 4.1766401920318605, + "runtime_seconds": 5.76836895942688 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6538357098314497, + "lr": 0.03, + "negative_goodness": 0.9493291177219815, + "pair_accuracy": 0.9024888888782925, + "positive_goodness": 4.484349713304308, + "runtime_seconds": 5.775793790817261 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6312372475306193, + "lr": 0.03, + "negative_goodness": 0.9609345058017307, + "pair_accuracy": 0.9194888888782925, + "positive_goodness": 4.376230959955851, + "runtime_seconds": 5.776683807373047 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6498057890468174, + "lr": 0.03, + "negative_goodness": 0.9703240114212036, + "pair_accuracy": 0.9090666667196485, + "positive_goodness": 4.363021802605523, + "runtime_seconds": 5.774123191833496 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6344572637346055, + "lr": 0.03, + "negative_goodness": 0.9434865324444242, + "pair_accuracy": 0.9139111111111111, + "positive_goodness": 4.460641666603088, + "runtime_seconds": 5.766280174255371 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.5852469966808955, + "lr": 0.03, + "negative_goodness": 0.8488293397823969, + "pair_accuracy": 0.9225111111111111, + "positive_goodness": 4.661031088680692, + "runtime_seconds": 5.7741522789001465 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.5938080990685357, + "lr": 0.03, + "negative_goodness": 0.8557149709701538, + "pair_accuracy": 0.919777777756585, + "positive_goodness": 4.6006266020033095, + "runtime_seconds": 5.768979787826538 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6599421416229672, + "lr": 0.03, + "negative_goodness": 0.9553777626037597, + "pair_accuracy": 0.9074444444444445, + "positive_goodness": 4.24195304192437, + "runtime_seconds": 5.772446632385254 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6138882786962722, + "lr": 0.03, + "negative_goodness": 0.8959124916500515, + "pair_accuracy": 0.9178666666560703, + "positive_goodness": 4.478489092254638, + "runtime_seconds": 5.770900249481201 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6137114641825359, + "lr": 0.03, + "negative_goodness": 0.8920736785676744, + "pair_accuracy": 0.920088888894187, + "positive_goodness": 4.3998442653656005, + "runtime_seconds": 5.770014524459839 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6269636445628273, + "lr": 0.03, + "negative_goodness": 0.9029143361462487, + "pair_accuracy": 0.9173111111111111, + "positive_goodness": 4.3661135003831655, + "runtime_seconds": 5.775274753570557 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6198454562717014, + "lr": 0.03, + "negative_goodness": 0.889662273534139, + "pair_accuracy": 0.9166666666984558, + "positive_goodness": 4.4683930170694985, + "runtime_seconds": 5.771220684051514 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6087410496711732, + "lr": 0.03, + "negative_goodness": 0.8822437804010179, + "pair_accuracy": 0.9191777777671813, + "positive_goodness": 4.467032770156861, + "runtime_seconds": 5.7681355476379395 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6197687394301097, + "lr": 0.03, + "negative_goodness": 0.8556080218791962, + "pair_accuracy": 0.9070222222222222, + "positive_goodness": 4.517776809946696, + "runtime_seconds": 5.772353649139404 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5607915094799466, + "lr": 0.03, + "negative_goodness": 0.7988726832389832, + "pair_accuracy": 0.9251999999947018, + "positive_goodness": 4.737890431722005, + "runtime_seconds": 5.767745733261108 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.522432948536343, + "lr": 0.03, + "negative_goodness": 0.7470611872037252, + "pair_accuracy": 0.941222222190433, + "positive_goodness": 4.835689142396715, + "runtime_seconds": 5.772491693496704 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5631403236839506, + "lr": 0.03, + "negative_goodness": 0.8243457999971178, + "pair_accuracy": 0.9363777777777778, + "positive_goodness": 4.629994623396132, + "runtime_seconds": 5.772245407104492 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5791719722165002, + "lr": 0.03, + "negative_goodness": 0.8255993426216973, + "pair_accuracy": 0.9270888888888889, + "positive_goodness": 4.666686462741428, + "runtime_seconds": 5.766655921936035 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5463564126544529, + "lr": 0.03, + "negative_goodness": 0.7712688382254707, + "pair_accuracy": 0.9309999999682108, + "positive_goodness": 4.882754359859891, + "runtime_seconds": 5.767024755477905 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5945090368588766, + "lr": 0.03, + "negative_goodness": 0.8292215109295316, + "pair_accuracy": 0.9159555555555555, + "positive_goodness": 4.672410916307237, + "runtime_seconds": 5.769819498062134 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.59597626630995, + "lr": 0.03, + "negative_goodness": 0.8142696750640869, + "pair_accuracy": 0.914533333337307, + "positive_goodness": 4.629728762732611, + "runtime_seconds": 5.766554117202759 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6200673864682515, + "lr": 0.03, + "negative_goodness": 0.8569040145026313, + "pair_accuracy": 0.9122888888782925, + "positive_goodness": 4.516442606565687, + "runtime_seconds": 5.768168926239014 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5824712764104207, + "lr": 0.03, + "negative_goodness": 0.8023208283318414, + "pair_accuracy": 0.9270888888782925, + "positive_goodness": 4.652998281775581, + "runtime_seconds": 5.766993999481201 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5819081634521485, + "lr": 0.03, + "negative_goodness": 0.8234504105832842, + "pair_accuracy": 0.9234444444444444, + "positive_goodness": 4.6172042268541125, + "runtime_seconds": 5.769309759140015 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5943271397511164, + "lr": 0.03, + "negative_goodness": 0.8276812359465493, + "pair_accuracy": 0.9181777777777778, + "positive_goodness": 4.5660211331685385, + "runtime_seconds": 5.769578695297241 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5458791923734877, + "lr": 0.03, + "negative_goodness": 0.7966393275154962, + "pair_accuracy": 0.9367333333651224, + "positive_goodness": 4.7359977319505475, + "runtime_seconds": 5.7630510330200195 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5853455590354072, + "lr": 0.03, + "negative_goodness": 0.8122709909545051, + "pair_accuracy": 0.9188888888782925, + "positive_goodness": 4.576144072638618, + "runtime_seconds": 5.768775224685669 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5806422383997175, + "lr": 0.03, + "negative_goodness": 0.8164612790849474, + "pair_accuracy": 0.9214666666666667, + "positive_goodness": 4.640560384284125, + "runtime_seconds": 5.770317792892456 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5575711984422472, + "lr": 0.03, + "negative_goodness": 0.7637210708936055, + "pair_accuracy": 0.9227111111323039, + "positive_goodness": 4.804096196195815, + "runtime_seconds": 5.76486611366272 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5730281701935662, + "lr": 0.03, + "negative_goodness": 0.7695296943982443, + "pair_accuracy": 0.9177555555555555, + "positive_goodness": 4.736537072414822, + "runtime_seconds": 5.764469146728516 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5357508070733812, + "lr": 0.03, + "negative_goodness": 0.753422713359197, + "pair_accuracy": 0.9373777777777778, + "positive_goodness": 4.893970614878336, + "runtime_seconds": 5.772696495056152 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5140173654556275, + "lr": 0.03, + "negative_goodness": 0.7080116315205892, + "pair_accuracy": 0.9354222222222223, + "positive_goodness": 5.121292691802979, + "runtime_seconds": 5.796298265457153 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.568180827442805, + "lr": 0.03, + "negative_goodness": 0.811039190003607, + "pair_accuracy": 0.9308, + "positive_goodness": 4.660574131308661, + "runtime_seconds": 5.773202896118164 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5678786044120788, + "lr": 0.03, + "negative_goodness": 0.7782195916175842, + "pair_accuracy": 0.9236888889418708, + "positive_goodness": 4.7139827150556775, + "runtime_seconds": 5.766860723495483 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5737603149414062, + "lr": 0.03, + "negative_goodness": 0.7597439681477017, + "pair_accuracy": 0.9248, + "positive_goodness": 4.724123015848796, + "runtime_seconds": 5.762269020080566 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5348647080315484, + "lr": 0.03, + "negative_goodness": 0.7362920667860243, + "pair_accuracy": 0.9331555556085375, + "positive_goodness": 4.936836877356635, + "runtime_seconds": 5.758923053741455 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5286555137316385, + "lr": 0.03, + "negative_goodness": 0.7166437354710367, + "pair_accuracy": 0.9275333333333333, + "positive_goodness": 5.0739412954542376, + "runtime_seconds": 5.76154637336731 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5942570487817128, + "lr": 0.03, + "negative_goodness": 0.8131109452035692, + "pair_accuracy": 0.9192888888888889, + "positive_goodness": 4.592604352654351, + "runtime_seconds": 5.764978885650635 + } + ], + "layer": 26 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8084363707118564, + "lr": 0.03, + "negative_goodness": 1.158089070028729, + "pair_accuracy": 0.8747777777777778, + "positive_goodness": 3.6323737016465927, + "runtime_seconds": 5.741919040679932 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7812008749749926, + "lr": 0.03, + "negative_goodness": 1.1368345398690964, + "pair_accuracy": 0.8716666666666667, + "positive_goodness": 3.75166996705797, + "runtime_seconds": 5.743672609329224 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7586057490242852, + "lr": 0.03, + "negative_goodness": 1.1275699659347533, + "pair_accuracy": 0.8818666667196485, + "positive_goodness": 3.8800418702867296, + "runtime_seconds": 5.743683099746704 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.762867280377282, + "lr": 0.03, + "negative_goodness": 1.1046923269483777, + "pair_accuracy": 0.8772444444444445, + "positive_goodness": 3.8508194448682995, + "runtime_seconds": 5.739527463912964 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7285458332697551, + "lr": 0.03, + "negative_goodness": 1.1014811985015869, + "pair_accuracy": 0.8931111111534966, + "positive_goodness": 3.871954575051202, + "runtime_seconds": 5.741718292236328 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6350302108870612, + "lr": 0.03, + "negative_goodness": 0.9185178550932143, + "pair_accuracy": 0.9082, + "positive_goodness": 4.286509584638807, + "runtime_seconds": 5.737320423126221 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6474988093058268, + "lr": 0.03, + "negative_goodness": 0.9531331206003825, + "pair_accuracy": 0.9076, + "positive_goodness": 4.272821172247992, + "runtime_seconds": 5.73823618888855 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6424919864442613, + "lr": 0.03, + "negative_goodness": 0.938532027053833, + "pair_accuracy": 0.9066888889418708, + "positive_goodness": 4.292170397482978, + "runtime_seconds": 5.734927177429199 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6774927584542169, + "lr": 0.03, + "negative_goodness": 0.9869739438798693, + "pair_accuracy": 0.8984666666984558, + "positive_goodness": 4.216594154781766, + "runtime_seconds": 5.939295053482056 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.604319073761834, + "lr": 0.03, + "negative_goodness": 0.8784277821858724, + "pair_accuracy": 0.9174444444444444, + "positive_goodness": 4.512238707987468, + "runtime_seconds": 5.858438491821289 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6406278072357178, + "lr": 0.03, + "negative_goodness": 0.9156859415160286, + "pair_accuracy": 0.907866666667329, + "positive_goodness": 4.368158476140764, + "runtime_seconds": 5.7537267208099365 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6249927485995822, + "lr": 0.03, + "negative_goodness": 0.8989114084455702, + "pair_accuracy": 0.912822222243415, + "positive_goodness": 4.374211318567064, + "runtime_seconds": 5.736196041107178 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6283692262861463, + "lr": 0.03, + "negative_goodness": 0.8801180156495836, + "pair_accuracy": 0.907777777756585, + "positive_goodness": 4.466059360165066, + "runtime_seconds": 5.738134384155273 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6479576886177063, + "lr": 0.03, + "negative_goodness": 0.9078564077907139, + "pair_accuracy": 0.8974666666666666, + "positive_goodness": 4.31240752597385, + "runtime_seconds": 5.740265130996704 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6614159726672703, + "lr": 0.03, + "negative_goodness": 0.9189221441056993, + "pair_accuracy": 0.9035111111217075, + "positive_goodness": 4.337638670984904, + "runtime_seconds": 5.739917039871216 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5695430955171585, + "lr": 0.03, + "negative_goodness": 0.8376247229019801, + "pair_accuracy": 0.9362444444444444, + "positive_goodness": 4.52653044052124, + "runtime_seconds": 5.741122245788574 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6225008974737591, + "lr": 0.03, + "negative_goodness": 0.8878242182440228, + "pair_accuracy": 0.9137111111111111, + "positive_goodness": 4.470306767527262, + "runtime_seconds": 5.743023157119751 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5872150650077396, + "lr": 0.03, + "negative_goodness": 0.8164392730077108, + "pair_accuracy": 0.9191333333333334, + "positive_goodness": 4.5605950826856825, + "runtime_seconds": 5.740209102630615 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6179196581628588, + "lr": 0.03, + "negative_goodness": 0.856479850323995, + "pair_accuracy": 0.911311111164093, + "positive_goodness": 4.45063169025845, + "runtime_seconds": 5.7347190380096436 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5950813184367286, + "lr": 0.03, + "negative_goodness": 0.8454543294694689, + "pair_accuracy": 0.9223111111111111, + "positive_goodness": 4.525326748318142, + "runtime_seconds": 5.743018865585327 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5963388442145453, + "lr": 0.03, + "negative_goodness": 0.8480051206270853, + "pair_accuracy": 0.9179333333306843, + "positive_goodness": 4.518809627946218, + "runtime_seconds": 5.738779306411743 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5884139969083998, + "lr": 0.03, + "negative_goodness": 0.8114695014847649, + "pair_accuracy": 0.9153777777671814, + "positive_goodness": 4.634938375430637, + "runtime_seconds": 5.736543416976929 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5956029255443149, + "lr": 0.03, + "negative_goodness": 0.8498534440040588, + "pair_accuracy": 0.9212444444444444, + "positive_goodness": 4.543190793016222, + "runtime_seconds": 5.738102912902832 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6034532033920288, + "lr": 0.03, + "negative_goodness": 0.8340802776866489, + "pair_accuracy": 0.9133555555449592, + "positive_goodness": 4.559225820922852, + "runtime_seconds": 5.736464023590088 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6232241544935438, + "lr": 0.03, + "negative_goodness": 0.8802603209071689, + "pair_accuracy": 0.9126444444550408, + "positive_goodness": 4.379986967171563, + "runtime_seconds": 5.73792028427124 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5929411944468816, + "lr": 0.03, + "negative_goodness": 0.8272830882072448, + "pair_accuracy": 0.9171555555555555, + "positive_goodness": 4.629040351782905, + "runtime_seconds": 5.7388317584991455 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5601581744935777, + "lr": 0.03, + "negative_goodness": 0.7927962332831489, + "pair_accuracy": 0.930400000052982, + "positive_goodness": 4.782227538469103, + "runtime_seconds": 5.954761028289795 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5947156613085005, + "lr": 0.03, + "negative_goodness": 0.7967705048984951, + "pair_accuracy": 0.9128444444444445, + "positive_goodness": 4.617085298665365, + "runtime_seconds": 5.7491819858551025 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5620013471497429, + "lr": 0.03, + "negative_goodness": 0.762116616015964, + "pair_accuracy": 0.9207555555555555, + "positive_goodness": 4.8450752510070805, + "runtime_seconds": 5.742101192474365 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5722757049878439, + "lr": 0.03, + "negative_goodness": 0.8086085693041484, + "pair_accuracy": 0.9259555555873447, + "positive_goodness": 4.726776081212361, + "runtime_seconds": 5.738970041275024 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5244141721857919, + "lr": 0.03, + "negative_goodness": 0.7207129636089007, + "pair_accuracy": 0.9323111111111111, + "positive_goodness": 4.952565743933784, + "runtime_seconds": 5.76630711555481 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5970580485449897, + "lr": 0.03, + "negative_goodness": 0.8011925642119514, + "pair_accuracy": 0.910755555566152, + "positive_goodness": 4.666813331137763, + "runtime_seconds": 5.736765384674072 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6391966626697116, + "lr": 0.03, + "negative_goodness": 0.8763367379930285, + "pair_accuracy": 0.9048000000529819, + "positive_goodness": 4.395986998960707, + "runtime_seconds": 5.751223564147949 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.58930176548428, + "lr": 0.03, + "negative_goodness": 0.8275755999830035, + "pair_accuracy": 0.9205777777777778, + "positive_goodness": 4.546788506825765, + "runtime_seconds": 5.740467071533203 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5690433528476291, + "lr": 0.03, + "negative_goodness": 0.7271618698014153, + "pair_accuracy": 0.9130222222275204, + "positive_goodness": 4.866904794184367, + "runtime_seconds": 5.736985206604004 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6157973398685456, + "lr": 0.03, + "negative_goodness": 0.826984855026669, + "pair_accuracy": 0.9068888888888889, + "positive_goodness": 4.642722849443223, + "runtime_seconds": 5.741124153137207 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5890981058385637, + "lr": 0.03, + "negative_goodness": 0.8016071583588918, + "pair_accuracy": 0.9131555555555556, + "positive_goodness": 4.67902118250529, + "runtime_seconds": 5.6878650188446045 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6060756579319636, + "lr": 0.03, + "negative_goodness": 0.8098992373572456, + "pair_accuracy": 0.9122, + "positive_goodness": 4.546348466406928, + "runtime_seconds": 5.684890031814575 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5611272255632612, + "lr": 0.03, + "negative_goodness": 0.7661255999618106, + "pair_accuracy": 0.9277555555343628, + "positive_goodness": 4.751028854327732, + "runtime_seconds": 5.693612813949585 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5680051735083262, + "lr": 0.03, + "negative_goodness": 0.7760852172533671, + "pair_accuracy": 0.9241333333333334, + "positive_goodness": 4.758114740668403, + "runtime_seconds": 5.703518390655518 + } + ], + "layer": 27 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8815447125699785, + "lr": 0.03, + "negative_goodness": 1.2807901606241863, + "pair_accuracy": 0.8719333333333333, + "positive_goodness": 3.7282654830084905, + "runtime_seconds": 5.72236442565918 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7608266480763753, + "lr": 0.03, + "negative_goodness": 1.1562179145177205, + "pair_accuracy": 0.8961555555714501, + "positive_goodness": 3.9495015362845525, + "runtime_seconds": 5.722447633743286 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7032649840619829, + "lr": 0.03, + "negative_goodness": 1.0452431730482314, + "pair_accuracy": 0.8971777777777777, + "positive_goodness": 4.204037189144558, + "runtime_seconds": 5.738828897476196 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7062967882050408, + "lr": 0.03, + "negative_goodness": 1.0505365156597561, + "pair_accuracy": 0.8992222222222223, + "positive_goodness": 4.121507434251574, + "runtime_seconds": 5.705208778381348 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6708004700554742, + "lr": 0.03, + "negative_goodness": 0.9830680129792955, + "pair_accuracy": 0.9057999999894036, + "positive_goodness": 4.251535476599799, + "runtime_seconds": 5.707974433898926 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6418411757787069, + "lr": 0.03, + "negative_goodness": 0.9823238424725003, + "pair_accuracy": 0.9196444444444445, + "positive_goodness": 4.33984765179952, + "runtime_seconds": 5.703347444534302 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6696057551701864, + "lr": 0.03, + "negative_goodness": 0.9536093906402587, + "pair_accuracy": 0.9006888888888889, + "positive_goodness": 4.228073384518093, + "runtime_seconds": 5.705111742019653 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.674973356850942, + "lr": 0.03, + "negative_goodness": 0.9641228446324667, + "pair_accuracy": 0.8968444444444444, + "positive_goodness": 4.194057128312853, + "runtime_seconds": 5.706455945968628 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6434441016885969, + "lr": 0.03, + "negative_goodness": 0.9228303441047668, + "pair_accuracy": 0.9055111111111112, + "positive_goodness": 4.334781834750705, + "runtime_seconds": 5.70423698425293 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6125797572877671, + "lr": 0.03, + "negative_goodness": 0.9015872220569187, + "pair_accuracy": 0.918555555566152, + "positive_goodness": 4.565852890290154, + "runtime_seconds": 5.706712484359741 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6186572722540962, + "lr": 0.03, + "negative_goodness": 0.9331881972842746, + "pair_accuracy": 0.9254000000039736, + "positive_goodness": 4.359093622228834, + "runtime_seconds": 5.7481019496917725 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6222645626809862, + "lr": 0.03, + "negative_goodness": 0.8649206262164646, + "pair_accuracy": 0.9093333333333333, + "positive_goodness": 4.458748987240262, + "runtime_seconds": 5.759721517562866 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6206678662829929, + "lr": 0.03, + "negative_goodness": 0.875505788591173, + "pair_accuracy": 0.90404444448683, + "positive_goodness": 4.468356991789076, + "runtime_seconds": 5.759978532791138 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6003919833342234, + "lr": 0.03, + "negative_goodness": 0.8836962394714355, + "pair_accuracy": 0.9215777777777778, + "positive_goodness": 4.533370751359728, + "runtime_seconds": 5.757293701171875 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6052229743745592, + "lr": 0.03, + "negative_goodness": 0.8656185893005794, + "pair_accuracy": 0.9237555555555556, + "positive_goodness": 4.520285724809435, + "runtime_seconds": 5.7599732875823975 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6088279895040724, + "lr": 0.03, + "negative_goodness": 0.8530358399285211, + "pair_accuracy": 0.9135777777671814, + "positive_goodness": 4.48956459189521, + "runtime_seconds": 5.757725715637207 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5744700448142158, + "lr": 0.03, + "negative_goodness": 0.816550196287367, + "pair_accuracy": 0.9216444444444445, + "positive_goodness": 4.648743999226888, + "runtime_seconds": 5.7558534145355225 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6233847510311339, + "lr": 0.03, + "negative_goodness": 0.8555568226099014, + "pair_accuracy": 0.9080888888888888, + "positive_goodness": 4.47492395401001, + "runtime_seconds": 5.757293939590454 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6515179951349894, + "lr": 0.03, + "negative_goodness": 0.9347524477852716, + "pair_accuracy": 0.9131555555767483, + "positive_goodness": 4.231623632770114, + "runtime_seconds": 5.752676486968994 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5841715709262424, + "lr": 0.03, + "negative_goodness": 0.8575441345055898, + "pair_accuracy": 0.9292666666666667, + "positive_goodness": 4.571445810360379, + "runtime_seconds": 5.755805969238281 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5880509582095677, + "lr": 0.03, + "negative_goodness": 0.8357745949639215, + "pair_accuracy": 0.9223333333333333, + "positive_goodness": 4.582530876837836, + "runtime_seconds": 5.776163101196289 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5901695616828071, + "lr": 0.03, + "negative_goodness": 0.8191283634079827, + "pair_accuracy": 0.9224222222328186, + "positive_goodness": 4.621841793251037, + "runtime_seconds": 5.759082078933716 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6146634568108452, + "lr": 0.03, + "negative_goodness": 0.8475224419063992, + "pair_accuracy": 0.9159333333121406, + "positive_goodness": 4.387584690433078, + "runtime_seconds": 5.785030841827393 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5291019570668538, + "lr": 0.03, + "negative_goodness": 0.758983005036248, + "pair_accuracy": 0.9355111111111111, + "positive_goodness": 4.946957145690918, + "runtime_seconds": 5.7629313468933105 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5746219203525119, + "lr": 0.03, + "negative_goodness": 0.8195406008826361, + "pair_accuracy": 0.9260444444444444, + "positive_goodness": 4.689784757826064, + "runtime_seconds": 5.760458469390869 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5249632368246714, + "lr": 0.03, + "negative_goodness": 0.763694635216395, + "pair_accuracy": 0.9405555555555556, + "positive_goodness": 4.874673622979058, + "runtime_seconds": 5.7578489780426025 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5882698161655002, + "lr": 0.03, + "negative_goodness": 0.8141186389711168, + "pair_accuracy": 0.9159555555582046, + "positive_goodness": 4.568861750857035, + "runtime_seconds": 5.76453161239624 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6041157677968343, + "lr": 0.03, + "negative_goodness": 0.8102451239585876, + "pair_accuracy": 0.9093333333333333, + "positive_goodness": 4.651557633166843, + "runtime_seconds": 5.761082410812378 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5721777939372592, + "lr": 0.03, + "negative_goodness": 0.8019275480482313, + "pair_accuracy": 0.9234888888676961, + "positive_goodness": 4.722862887827556, + "runtime_seconds": 5.764195919036865 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.566209872004721, + "lr": 0.03, + "negative_goodness": 0.821934363497628, + "pair_accuracy": 0.9301111111111111, + "positive_goodness": 4.693826862250434, + "runtime_seconds": 5.762457370758057 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5390209814071655, + "lr": 0.03, + "negative_goodness": 0.761849317211575, + "pair_accuracy": 0.9339333333651225, + "positive_goodness": 4.848205346849229, + "runtime_seconds": 5.762497186660767 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5524235251002841, + "lr": 0.03, + "negative_goodness": 0.788948677306705, + "pair_accuracy": 0.9339555555449591, + "positive_goodness": 4.809712139426337, + "runtime_seconds": 5.765229940414429 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6114796046680875, + "lr": 0.03, + "negative_goodness": 0.8480206505669488, + "pair_accuracy": 0.9153777777459886, + "positive_goodness": 4.539419817988078, + "runtime_seconds": 5.762238502502441 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5916456183195115, + "lr": 0.03, + "negative_goodness": 0.8106230479982164, + "pair_accuracy": 0.9168888888888889, + "positive_goodness": 4.681875444030761, + "runtime_seconds": 5.7574622631073 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.535443944062127, + "lr": 0.03, + "negative_goodness": 0.7640542104932997, + "pair_accuracy": 0.9366444444550408, + "positive_goodness": 4.827931089146932, + "runtime_seconds": 5.756945848464966 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5511569157388475, + "lr": 0.03, + "negative_goodness": 0.7688426042026943, + "pair_accuracy": 0.9305333333121406, + "positive_goodness": 4.756971312204997, + "runtime_seconds": 5.728097677230835 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5296367491404216, + "lr": 0.03, + "negative_goodness": 0.7444688649071588, + "pair_accuracy": 0.9355555555595292, + "positive_goodness": 5.063383630636003, + "runtime_seconds": 5.703036785125732 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5510098451296488, + "lr": 0.03, + "negative_goodness": 0.7512083364592658, + "pair_accuracy": 0.9299111111111111, + "positive_goodness": 4.825047082773844, + "runtime_seconds": 5.745945692062378 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6020484879599677, + "lr": 0.03, + "negative_goodness": 0.8148801503711277, + "pair_accuracy": 0.9109555555555555, + "positive_goodness": 4.786112586720784, + "runtime_seconds": 5.7677001953125 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5608703695615133, + "lr": 0.03, + "negative_goodness": 0.7804486520449321, + "pair_accuracy": 0.9262888888888889, + "positive_goodness": 4.77632561700609, + "runtime_seconds": 5.775120973587036 + } + ], + "layer": 28 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8240096335623, + "lr": 0.03, + "negative_goodness": 1.1244661579873827, + "pair_accuracy": 0.8692222222222222, + "positive_goodness": 3.5858012979295517, + "runtime_seconds": 5.7532289028167725 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7313504621505738, + "lr": 0.03, + "negative_goodness": 1.0636509520848592, + "pair_accuracy": 0.8860222222116259, + "positive_goodness": 3.982748972023858, + "runtime_seconds": 5.706995010375977 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7161583251317342, + "lr": 0.03, + "negative_goodness": 1.0243039344363742, + "pair_accuracy": 0.8799111111534966, + "positive_goodness": 4.028418170928955, + "runtime_seconds": 5.686343669891357 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7478894294420878, + "lr": 0.03, + "negative_goodness": 1.0447714681201512, + "pair_accuracy": 0.8639555555555556, + "positive_goodness": 3.9302858566708037, + "runtime_seconds": 5.682456016540527 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7138433705647786, + "lr": 0.03, + "negative_goodness": 1.0388259237077502, + "pair_accuracy": 0.8918222222261959, + "positive_goodness": 4.049128533501095, + "runtime_seconds": 5.681676387786865 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.706796869775984, + "lr": 0.03, + "negative_goodness": 1.0163372108883328, + "pair_accuracy": 0.8886222222646077, + "positive_goodness": 4.183341834682889, + "runtime_seconds": 5.727728605270386 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6413479445987278, + "lr": 0.03, + "negative_goodness": 0.9390928334765964, + "pair_accuracy": 0.9109333333863152, + "positive_goodness": 4.370119054328071, + "runtime_seconds": 5.748652458190918 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6368805636882782, + "lr": 0.03, + "negative_goodness": 0.930874518181218, + "pair_accuracy": 0.9161555555555555, + "positive_goodness": 4.398961093224419, + "runtime_seconds": 5.755783557891846 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6775689103894763, + "lr": 0.03, + "negative_goodness": 0.9376054365793864, + "pair_accuracy": 0.8926666666666667, + "positive_goodness": 4.304339673529731, + "runtime_seconds": 5.832923173904419 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6031974662144979, + "lr": 0.03, + "negative_goodness": 0.8643321345541213, + "pair_accuracy": 0.9127777777777778, + "positive_goodness": 4.705496149359809, + "runtime_seconds": 5.756362199783325 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6911547184838189, + "lr": 0.03, + "negative_goodness": 0.9562649085998535, + "pair_accuracy": 0.8971111111111111, + "positive_goodness": 4.210168125640021, + "runtime_seconds": 5.727161407470703 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5969146053817537, + "lr": 0.03, + "negative_goodness": 0.8442959854417377, + "pair_accuracy": 0.9165333333333333, + "positive_goodness": 4.58834968202379, + "runtime_seconds": 5.7101380825042725 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6190997353023953, + "lr": 0.03, + "negative_goodness": 0.8450684779802958, + "pair_accuracy": 0.9093999999894036, + "positive_goodness": 4.574237760331895, + "runtime_seconds": 5.753710508346558 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6314372679180569, + "lr": 0.03, + "negative_goodness": 0.8374366881582472, + "pair_accuracy": 0.8972, + "positive_goodness": 4.545659521145291, + "runtime_seconds": 5.758972644805908 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6360067891438802, + "lr": 0.03, + "negative_goodness": 0.8859276418262058, + "pair_accuracy": 0.9041555555767483, + "positive_goodness": 4.491482976765103, + "runtime_seconds": 5.7513532638549805 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6776475423177083, + "lr": 0.03, + "negative_goodness": 0.9267222124523586, + "pair_accuracy": 0.8897777777724796, + "positive_goodness": 4.2520882218466864, + "runtime_seconds": 5.753249168395996 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6165931335131327, + "lr": 0.03, + "negative_goodness": 0.8667723801188999, + "pair_accuracy": 0.9094444444444445, + "positive_goodness": 4.535691222296821, + "runtime_seconds": 5.756136655807495 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6153933380762736, + "lr": 0.03, + "negative_goodness": 0.8743690500365363, + "pair_accuracy": 0.9094888889100816, + "positive_goodness": 4.490080358462864, + "runtime_seconds": 5.7542338371276855 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6369866584142049, + "lr": 0.03, + "negative_goodness": 0.8486668114344279, + "pair_accuracy": 0.8976888889312744, + "positive_goodness": 4.572417034509447, + "runtime_seconds": 5.753452301025391 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6430036381615533, + "lr": 0.03, + "negative_goodness": 0.8972125413073434, + "pair_accuracy": 0.9006222222222222, + "positive_goodness": 4.323217069244385, + "runtime_seconds": 5.758290767669678 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5860103961520725, + "lr": 0.03, + "negative_goodness": 0.8205743580129411, + "pair_accuracy": 0.9209111111111111, + "positive_goodness": 4.558798409186469, + "runtime_seconds": 5.754262924194336 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6272836796336704, + "lr": 0.03, + "negative_goodness": 0.861072411886851, + "pair_accuracy": 0.9102666666348775, + "positive_goodness": 4.415864261288113, + "runtime_seconds": 5.787163496017456 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6465262879689535, + "lr": 0.03, + "negative_goodness": 0.8482853006256951, + "pair_accuracy": 0.8931555556085374, + "positive_goodness": 4.520222092776828, + "runtime_seconds": 5.75012731552124 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.590275842248069, + "lr": 0.03, + "negative_goodness": 0.7983322786357667, + "pair_accuracy": 0.9088444444444445, + "positive_goodness": 4.787737795766195, + "runtime_seconds": 5.754158020019531 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6343805305904813, + "lr": 0.03, + "negative_goodness": 0.8649373079299927, + "pair_accuracy": 0.9008888888729943, + "positive_goodness": 4.474457183943854, + "runtime_seconds": 5.7537620067596436 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6302807067871093, + "lr": 0.03, + "negative_goodness": 0.8540389270358616, + "pair_accuracy": 0.9034666666666666, + "positive_goodness": 4.62692865202162, + "runtime_seconds": 5.750718355178833 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5879753572516971, + "lr": 0.03, + "negative_goodness": 0.7971228520393372, + "pair_accuracy": 0.9159333333333334, + "positive_goodness": 4.783544617970785, + "runtime_seconds": 5.753721237182617 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5692773025724623, + "lr": 0.03, + "negative_goodness": 0.7918763537194994, + "pair_accuracy": 0.9173555555979411, + "positive_goodness": 4.732267215304905, + "runtime_seconds": 5.751240253448486 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6064081617461311, + "lr": 0.03, + "negative_goodness": 0.8061328511344062, + "pair_accuracy": 0.9075555555555556, + "positive_goodness": 4.634950497521294, + "runtime_seconds": 5.750380516052246 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5583288639916314, + "lr": 0.03, + "negative_goodness": 0.7773953559663561, + "pair_accuracy": 0.9214666667196486, + "positive_goodness": 4.924488716379801, + "runtime_seconds": 5.751892566680908 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6080972855038113, + "lr": 0.03, + "negative_goodness": 0.8155856942282783, + "pair_accuracy": 0.9081333333439298, + "positive_goodness": 4.581304925240411, + "runtime_seconds": 5.754560470581055 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6077495155864292, + "lr": 0.03, + "negative_goodness": 0.8082417677455478, + "pair_accuracy": 0.9118666666348775, + "positive_goodness": 4.601687281163533, + "runtime_seconds": 5.744723081588745 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5684889970964856, + "lr": 0.03, + "negative_goodness": 0.7699438978645536, + "pair_accuracy": 0.9181555555555555, + "positive_goodness": 4.750456239318848, + "runtime_seconds": 5.714291334152222 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5747301972707113, + "lr": 0.03, + "negative_goodness": 0.7719677573521933, + "pair_accuracy": 0.9170888888570997, + "positive_goodness": 4.833439463848538, + "runtime_seconds": 5.733037710189819 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5874912351820204, + "lr": 0.03, + "negative_goodness": 0.7862294746716817, + "pair_accuracy": 0.9121555555555556, + "positive_goodness": 4.674688023715549, + "runtime_seconds": 5.7502992153167725 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6041825548383924, + "lr": 0.03, + "negative_goodness": 0.8026528259277343, + "pair_accuracy": 0.9093777778307597, + "positive_goodness": 4.685105075475905, + "runtime_seconds": 5.747925758361816 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6279419397989909, + "lr": 0.03, + "negative_goodness": 0.8349605463398827, + "pair_accuracy": 0.9044666666666666, + "positive_goodness": 4.545225892554389, + "runtime_seconds": 5.749473810195923 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5831956124623616, + "lr": 0.03, + "negative_goodness": 0.7669739714198642, + "pair_accuracy": 0.9150888889206781, + "positive_goodness": 4.861039180246989, + "runtime_seconds": 5.751079797744751 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5829895033518473, + "lr": 0.03, + "negative_goodness": 0.8049851893107096, + "pair_accuracy": 0.9172888888570997, + "positive_goodness": 4.727388479359945, + "runtime_seconds": 5.754021167755127 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5872321266863081, + "lr": 0.03, + "negative_goodness": 0.7726147991816202, + "pair_accuracy": 0.9063333333333333, + "positive_goodness": 4.820214025794135, + "runtime_seconds": 5.756152391433716 + } + ], + "layer": 29 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8839181497573853, + "lr": 0.03, + "negative_goodness": 1.2886588450537788, + "pair_accuracy": 0.8773111110793219, + "positive_goodness": 3.677121157031589, + "runtime_seconds": 5.7707343101501465 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7486571011861165, + "lr": 0.03, + "negative_goodness": 1.0817694777594673, + "pair_accuracy": 0.8871333333333333, + "positive_goodness": 4.064350888760885, + "runtime_seconds": 5.774807453155518 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7335625515805351, + "lr": 0.03, + "negative_goodness": 1.0509015742937724, + "pair_accuracy": 0.8864666666666666, + "positive_goodness": 4.0217693567064075, + "runtime_seconds": 5.769707918167114 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6666985371642643, + "lr": 0.03, + "negative_goodness": 0.9719207290649414, + "pair_accuracy": 0.906844444433848, + "positive_goodness": 4.246028872511122, + "runtime_seconds": 5.769813299179077 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6583646146774292, + "lr": 0.03, + "negative_goodness": 0.9771668761359321, + "pair_accuracy": 0.9105555555767483, + "positive_goodness": 4.373249750285678, + "runtime_seconds": 5.772747755050659 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6426836492220561, + "lr": 0.03, + "negative_goodness": 0.8774957171228197, + "pair_accuracy": 0.9054666666666666, + "positive_goodness": 4.461831588660346, + "runtime_seconds": 5.770208835601807 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6281152543650733, + "lr": 0.03, + "negative_goodness": 0.8870124119758606, + "pair_accuracy": 0.9060888888888889, + "positive_goodness": 4.534120838419597, + "runtime_seconds": 5.770803928375244 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6704565930154589, + "lr": 0.03, + "negative_goodness": 0.9532290557861328, + "pair_accuracy": 0.8988222222222222, + "positive_goodness": 4.267465506490072, + "runtime_seconds": 5.77040433883667 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6261120307736927, + "lr": 0.03, + "negative_goodness": 0.9097580998102824, + "pair_accuracy": 0.9084, + "positive_goodness": 4.516037893083361, + "runtime_seconds": 5.766118288040161 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6408558435281118, + "lr": 0.03, + "negative_goodness": 0.8964108575503031, + "pair_accuracy": 0.9006444444444445, + "positive_goodness": 4.428207152133518, + "runtime_seconds": 5.773272514343262 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6562798737843831, + "lr": 0.03, + "negative_goodness": 0.9255687695609198, + "pair_accuracy": 0.9051555555661519, + "positive_goodness": 4.370465622923109, + "runtime_seconds": 5.772752046585083 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5986478869014316, + "lr": 0.03, + "negative_goodness": 0.8601013723691304, + "pair_accuracy": 0.9177999999894036, + "positive_goodness": 4.54815641547309, + "runtime_seconds": 5.772681951522827 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6585124092313979, + "lr": 0.03, + "negative_goodness": 0.9192739942126804, + "pair_accuracy": 0.8967111110899183, + "positive_goodness": 4.376741173977322, + "runtime_seconds": 5.774482011795044 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.634193971713384, + "lr": 0.03, + "negative_goodness": 0.8951374372906155, + "pair_accuracy": 0.9073555555555556, + "positive_goodness": 4.368209583791097, + "runtime_seconds": 5.770380973815918 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5771423038482666, + "lr": 0.03, + "negative_goodness": 0.7997183230294121, + "pair_accuracy": 0.9199333333121406, + "positive_goodness": 4.796635635736253, + "runtime_seconds": 5.772207260131836 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6100533631748624, + "lr": 0.03, + "negative_goodness": 0.8696117891947428, + "pair_accuracy": 0.9151777777671813, + "positive_goodness": 4.519509595150418, + "runtime_seconds": 5.769837141036987 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6006618497583601, + "lr": 0.03, + "negative_goodness": 0.86713392223782, + "pair_accuracy": 0.9195333333333333, + "positive_goodness": 4.498237360551622, + "runtime_seconds": 5.768229961395264 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6079024824089474, + "lr": 0.03, + "negative_goodness": 0.8358721492714352, + "pair_accuracy": 0.9073555555555556, + "positive_goodness": 4.615041595289442, + "runtime_seconds": 5.77214503288269 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6055260875489976, + "lr": 0.03, + "negative_goodness": 0.8677692352082994, + "pair_accuracy": 0.926933333322737, + "positive_goodness": 4.503363872146607, + "runtime_seconds": 5.766549110412598 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.604856105656094, + "lr": 0.03, + "negative_goodness": 0.8507391162024603, + "pair_accuracy": 0.9121777777936724, + "positive_goodness": 4.549069806755914, + "runtime_seconds": 5.769184589385986 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5495101394679811, + "lr": 0.03, + "negative_goodness": 0.7838471796777513, + "pair_accuracy": 0.9293333333333333, + "positive_goodness": 4.844970642683241, + "runtime_seconds": 5.772970676422119 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5894508729961183, + "lr": 0.03, + "negative_goodness": 0.8121974825011359, + "pair_accuracy": 0.9155111111111112, + "positive_goodness": 4.656916767120362, + "runtime_seconds": 5.7681214809417725 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6254463623682658, + "lr": 0.03, + "negative_goodness": 0.8591351911968655, + "pair_accuracy": 0.9024888888782925, + "positive_goodness": 4.567636774571737, + "runtime_seconds": 5.769543886184692 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5938733110825221, + "lr": 0.03, + "negative_goodness": 0.7944394828531477, + "pair_accuracy": 0.907, + "positive_goodness": 4.655256319851346, + "runtime_seconds": 5.765979051589966 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6099664942529467, + "lr": 0.03, + "negative_goodness": 0.8564848399162293, + "pair_accuracy": 0.9178888888888889, + "positive_goodness": 4.482119104088677, + "runtime_seconds": 5.769772529602051 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5499596232838101, + "lr": 0.03, + "negative_goodness": 0.7736451981438531, + "pair_accuracy": 0.9264666666666667, + "positive_goodness": 4.792301180818346, + "runtime_seconds": 5.774900197982788 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5559413334767024, + "lr": 0.03, + "negative_goodness": 0.760579228258133, + "pair_accuracy": 0.9231777777777778, + "positive_goodness": 4.919445445929633, + "runtime_seconds": 5.769733905792236 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.592491348224216, + "lr": 0.03, + "negative_goodness": 0.8223145419862535, + "pair_accuracy": 0.913822222243415, + "positive_goodness": 4.5972154591030545, + "runtime_seconds": 5.7720417976379395 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.57064817782508, + "lr": 0.03, + "negative_goodness": 0.7766496880637275, + "pair_accuracy": 0.9169555556085375, + "positive_goodness": 4.791650563303629, + "runtime_seconds": 5.7751829624176025 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5511611660480499, + "lr": 0.03, + "negative_goodness": 0.7757307043764327, + "pair_accuracy": 0.9266, + "positive_goodness": 4.786701979912652, + "runtime_seconds": 5.7719457149505615 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6135997237523397, + "lr": 0.03, + "negative_goodness": 0.8315623671849569, + "pair_accuracy": 0.9085555555555556, + "positive_goodness": 4.574039168633355, + "runtime_seconds": 5.769151210784912 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5918812077416314, + "lr": 0.03, + "negative_goodness": 0.8262545558929444, + "pair_accuracy": 0.9195333333863153, + "positive_goodness": 4.610591309907702, + "runtime_seconds": 5.776021718978882 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.577402552541097, + "lr": 0.03, + "negative_goodness": 0.7831381676355997, + "pair_accuracy": 0.9184888889047834, + "positive_goodness": 4.7875724257787065, + "runtime_seconds": 5.773364067077637 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5523063559320238, + "lr": 0.03, + "negative_goodness": 0.7443826806704203, + "pair_accuracy": 0.9211111111323039, + "positive_goodness": 5.049980414369371, + "runtime_seconds": 5.7766029834747314 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6004825270970663, + "lr": 0.03, + "negative_goodness": 0.8049077553219265, + "pair_accuracy": 0.9092222222116259, + "positive_goodness": 4.6209863855361935, + "runtime_seconds": 5.772159099578857 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6083383402082655, + "lr": 0.03, + "negative_goodness": 0.8246537428538004, + "pair_accuracy": 0.9149777777671814, + "positive_goodness": 4.584843021731906, + "runtime_seconds": 5.7733190059661865 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5861595803578694, + "lr": 0.03, + "negative_goodness": 0.8059644411828782, + "pair_accuracy": 0.9229777778201633, + "positive_goodness": 4.664937475628323, + "runtime_seconds": 5.771757125854492 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5753527121517393, + "lr": 0.03, + "negative_goodness": 0.7658315282106399, + "pair_accuracy": 0.9142888888888889, + "positive_goodness": 4.772544534556071, + "runtime_seconds": 5.76759934425354 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5616926870902379, + "lr": 0.03, + "negative_goodness": 0.788593441886372, + "pair_accuracy": 0.9239333333333334, + "positive_goodness": 4.827094375016954, + "runtime_seconds": 5.76919150352478 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.603986334461636, + "lr": 0.03, + "negative_goodness": 0.81092396291097, + "pair_accuracy": 0.9048000000105964, + "positive_goodness": 4.769885769822863, + "runtime_seconds": 5.767747163772583 + } + ], + "layer": 30 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0806074440108404, + "lr": 0.03, + "negative_goodness": 1.1196772415267096, + "pair_accuracy": 0.8391333333121406, + "positive_goodness": 3.008626028262244, + "runtime_seconds": 5.517334222793579 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.991377209493849, + "lr": 0.03, + "negative_goodness": 1.1442320047060648, + "pair_accuracy": 0.8633111111217074, + "positive_goodness": 3.2885721189286974, + "runtime_seconds": 5.50715970993042 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.9106916055679322, + "lr": 0.03, + "negative_goodness": 0.9665998303042518, + "pair_accuracy": 0.8867777778095669, + "positive_goodness": 3.7205429278479683, + "runtime_seconds": 5.511929273605347 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0035151354471843, + "lr": 0.03, + "negative_goodness": 1.096169355858697, + "pair_accuracy": 0.8596444444444444, + "positive_goodness": 3.314085551177131, + "runtime_seconds": 5.511645555496216 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.9407366711669498, + "lr": 0.03, + "negative_goodness": 1.1069256670527987, + "pair_accuracy": 0.8885777777777778, + "positive_goodness": 3.5933003251817492, + "runtime_seconds": 5.5112974643707275 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.951607786374622, + "lr": 0.03, + "negative_goodness": 1.1173641214688619, + "pair_accuracy": 0.8808666666666667, + "positive_goodness": 3.598687186686198, + "runtime_seconds": 5.51159405708313 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.9350339941660564, + "lr": 0.03, + "negative_goodness": 1.086389090262519, + "pair_accuracy": 0.8921777777671814, + "positive_goodness": 3.5422951685587565, + "runtime_seconds": 5.5087549686431885 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.929724402809143, + "lr": 0.03, + "negative_goodness": 1.088092151843177, + "pair_accuracy": 0.8840444444444444, + "positive_goodness": 3.603536415863037, + "runtime_seconds": 5.509066104888916 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.9378244368129306, + "lr": 0.03, + "negative_goodness": 1.1211033556090462, + "pair_accuracy": 0.8909555555555555, + "positive_goodness": 3.5732413273705377, + "runtime_seconds": 5.507841348648071 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8893966476440429, + "lr": 0.03, + "negative_goodness": 1.0551479596032036, + "pair_accuracy": 0.8966888888570997, + "positive_goodness": 3.6580216370476615, + "runtime_seconds": 5.508056402206421 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.8680827240837945, + "lr": 0.03, + "negative_goodness": 1.0383475747002495, + "pair_accuracy": 0.9140222222752041, + "positive_goodness": 3.8057690994262696, + "runtime_seconds": 5.507979869842529 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.9229250279108683, + "lr": 0.03, + "negative_goodness": 1.0954879327138265, + "pair_accuracy": 0.8910222222010294, + "positive_goodness": 3.554184342871772, + "runtime_seconds": 5.503948211669922 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.919588608593411, + "lr": 0.03, + "negative_goodness": 1.1103112631903755, + "pair_accuracy": 0.8989777777883742, + "positive_goodness": 3.6381345487806533, + "runtime_seconds": 5.505334377288818 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.9128944606781005, + "lr": 0.03, + "negative_goodness": 1.0675211286756727, + "pair_accuracy": 0.904577777785725, + "positive_goodness": 3.6259599681854247, + "runtime_seconds": 5.5084638595581055 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.8191538105858697, + "lr": 0.03, + "negative_goodness": 0.9897986520131429, + "pair_accuracy": 0.9194444444285499, + "positive_goodness": 3.9234300924725, + "runtime_seconds": 5.508040189743042 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.8930603544208738, + "lr": 0.03, + "negative_goodness": 1.1132122848775652, + "pair_accuracy": 0.9017111111111111, + "positive_goodness": 3.717992191145155, + "runtime_seconds": 5.507583856582642 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.9267324110454983, + "lr": 0.03, + "negative_goodness": 1.10939549853007, + "pair_accuracy": 0.894533333322737, + "positive_goodness": 3.5849625977834068, + "runtime_seconds": 5.506751298904419 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.9217814283794827, + "lr": 0.03, + "negative_goodness": 1.0943154345830282, + "pair_accuracy": 0.8930666666666667, + "positive_goodness": 3.5185422654045952, + "runtime_seconds": 5.50817608833313 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.88256396867964, + "lr": 0.03, + "negative_goodness": 1.0500105741182963, + "pair_accuracy": 0.9086666667090522, + "positive_goodness": 3.6818674172295465, + "runtime_seconds": 5.511053800582886 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.8855476172765097, + "lr": 0.03, + "negative_goodness": 1.034508775308397, + "pair_accuracy": 0.8881777777989706, + "positive_goodness": 3.835855992995368, + "runtime_seconds": 5.510568857192993 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.9427857373131646, + "lr": 0.03, + "negative_goodness": 1.1594156524658203, + "pair_accuracy": 0.8918666667090522, + "positive_goodness": 3.4744130424499513, + "runtime_seconds": 5.508163690567017 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.9139693277994791, + "lr": 0.03, + "negative_goodness": 1.108639943069882, + "pair_accuracy": 0.8942666666666667, + "positive_goodness": 3.5837798483954537, + "runtime_seconds": 5.508002042770386 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.9084175142500136, + "lr": 0.03, + "negative_goodness": 1.11495112212499, + "pair_accuracy": 0.8920444444656372, + "positive_goodness": 3.5240115464634365, + "runtime_seconds": 5.509254693984985 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.9045965575536092, + "lr": 0.03, + "negative_goodness": 1.0725294980261062, + "pair_accuracy": 0.889911111079322, + "positive_goodness": 3.669958193545871, + "runtime_seconds": 5.5122435092926025 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.9035706564744314, + "lr": 0.03, + "negative_goodness": 1.0877270147111682, + "pair_accuracy": 0.8976, + "positive_goodness": 3.665256595272488, + "runtime_seconds": 5.507101058959961 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9392848833243053, + "lr": 0.03, + "negative_goodness": 1.1295184087965224, + "pair_accuracy": 0.8915777777777778, + "positive_goodness": 3.507970821931627, + "runtime_seconds": 5.464153528213501 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.9539917216618856, + "lr": 0.03, + "negative_goodness": 1.1507166154225668, + "pair_accuracy": 0.8832222222222222, + "positive_goodness": 3.364953783019384, + "runtime_seconds": 5.4838175773620605 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9400826879925198, + "lr": 0.03, + "negative_goodness": 1.125582417233785, + "pair_accuracy": 0.893666666677263, + "positive_goodness": 3.550314247279697, + "runtime_seconds": 5.497881650924683 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.8745653021229638, + "lr": 0.03, + "negative_goodness": 1.1050881310039096, + "pair_accuracy": 0.9109111111111111, + "positive_goodness": 3.668231116739909, + "runtime_seconds": 5.496273040771484 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.8346634846925736, + "lr": 0.03, + "negative_goodness": 1.0130252832094828, + "pair_accuracy": 0.9097555555555555, + "positive_goodness": 3.9823453586154516, + "runtime_seconds": 5.4924821853637695 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.8746226275549994, + "lr": 0.03, + "negative_goodness": 1.0400563873079087, + "pair_accuracy": 0.9047333333651225, + "positive_goodness": 3.856266967264811, + "runtime_seconds": 5.498166561126709 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.9212238523271349, + "lr": 0.03, + "negative_goodness": 1.1284536643981933, + "pair_accuracy": 0.9016222222222222, + "positive_goodness": 3.6469815513186985, + "runtime_seconds": 5.494876384735107 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9080663626988729, + "lr": 0.03, + "negative_goodness": 1.0822988918516372, + "pair_accuracy": 0.8881555555767483, + "positive_goodness": 3.623752948824565, + "runtime_seconds": 5.50894570350647 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.8942194516499837, + "lr": 0.03, + "negative_goodness": 1.0872526880370246, + "pair_accuracy": 0.893311111105813, + "positive_goodness": 3.714408939827813, + "runtime_seconds": 5.509981870651245 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9485488543722365, + "lr": 0.03, + "negative_goodness": 1.1977286613994174, + "pair_accuracy": 0.8882888888888889, + "positive_goodness": 3.464136240386963, + "runtime_seconds": 5.511054992675781 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9284373806211683, + "lr": 0.03, + "negative_goodness": 1.1387328002399868, + "pair_accuracy": 0.8982444444232517, + "positive_goodness": 3.5337524319118923, + "runtime_seconds": 5.5095374584198 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9270698877546523, + "lr": 0.03, + "negative_goodness": 1.11444564581977, + "pair_accuracy": 0.9011333333863153, + "positive_goodness": 3.6845912597232395, + "runtime_seconds": 5.5098090171813965 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.9024464615080092, + "lr": 0.03, + "negative_goodness": 1.1000949136628044, + "pair_accuracy": 0.9041555555714501, + "positive_goodness": 3.62350832356347, + "runtime_seconds": 5.487447738647461 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.8898744351704916, + "lr": 0.03, + "negative_goodness": 1.047657404083676, + "pair_accuracy": 0.9057111111111111, + "positive_goodness": 3.610856678305732, + "runtime_seconds": 5.475712060928345 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.8929182352277968, + "lr": 0.03, + "negative_goodness": 1.0770899312761095, + "pair_accuracy": 0.9100222222261959, + "positive_goodness": 3.5730012862523397, + "runtime_seconds": 5.473771095275879 + } + ], + "layer": 31 + } + ], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 7469.690595626831, + "work": { + "candidate_label_evaluation_presentations": 50000, + "completed_global_epochs": 1280, + "feedforward_example_passes": 115200000, + "forward_macs_per_example": 68862592, + "forward_parameter_count": 464154, + "local_target_backward_example_evaluations": 115200000, + "local_vjp_example_evaluations": 0, + "logical_task_loss_queries": 0, + "num_trainable_layers": 32, + "ordinary_training_examples": 57600000, + "ordinary_validation_examples": 5000, + "relaxation_example_passes": 0, + "training_example_presentations": 115200000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json.manifest.json new file mode 100644 index 0000000..d070bc2 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "ff", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "40", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.03", + "--output_lr", + "0.03", + "--lr_schedule", + "constant", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--ff_threshold", + "2.0", + "--ff_score_from_layer", + "1" + ], + "completed_unix_time": 1786935215.9924834, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 7484.067112207413, + "epochs": 40, + "experiment_name": "resnet-r2-ff-d32-lr0p03", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "constant", + "method": "ff", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d32-lr0p03.json", + "output_exists": true, + "output_sha256": "d48f2e4651502170feb4afc4665bebb9625f077335db9f59c129748c4fdbf731", + "rate": 0.03, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json b/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json new file mode 100644 index 0000000..d5e9716 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json @@ -0,0 +1,22845 @@ +{ + "architecture": { + "base_width": 16, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 853018, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 40, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "constant", + "max_steps": 0, + "method": "ff", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json", + "output_lr": 0.03, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.1122, + "evaluation_seconds": 4.29287576675415, + "examples": 5000, + "finite": true, + "loss": null + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 13807850496, + "peak_memory_reserved_bytes": 14566817792, + "torch_version": "2.3.1+cu118" + }, + "layers": [ + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.4005121646457248, + "lr": 0.03, + "negative_goodness": 1.9601310249752468, + "pair_accuracy": 0.47391111113760204, + "positive_goodness": 1.960143265914917, + "runtime_seconds": 11.888412475585938 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3870076300091214, + "lr": 0.03, + "negative_goodness": 1.9987232397291395, + "pair_accuracy": 0.47784444443384805, + "positive_goodness": 1.998744415834215, + "runtime_seconds": 11.472265481948853 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3869588095559013, + "lr": 0.03, + "negative_goodness": 1.9987902501000299, + "pair_accuracy": 0.48128888891537985, + "positive_goodness": 1.9988118084377713, + "runtime_seconds": 11.478168487548828 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.3869192699644302, + "lr": 0.03, + "negative_goodness": 1.9989530040317112, + "pair_accuracy": 0.47242222225401137, + "positive_goodness": 1.9989705482906766, + "runtime_seconds": 11.497294902801514 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3867757711410522, + "lr": 0.03, + "negative_goodness": 1.9991573665195042, + "pair_accuracy": 0.47540000001589455, + "positive_goodness": 1.9991776463402642, + "runtime_seconds": 11.505736112594604 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3866989361868964, + "lr": 0.03, + "negative_goodness": 1.9993364716423883, + "pair_accuracy": 0.4778000000211928, + "positive_goodness": 1.9993548786587185, + "runtime_seconds": 11.502387523651123 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3865385047488743, + "lr": 0.03, + "negative_goodness": 1.9995932444254558, + "pair_accuracy": 0.4744666666825612, + "positive_goodness": 1.9996111411624484, + "runtime_seconds": 11.520817279815674 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3864329044977823, + "lr": 0.03, + "negative_goodness": 1.9997635805977716, + "pair_accuracy": 0.4780666666560703, + "positive_goodness": 1.9997798171361287, + "runtime_seconds": 11.526031494140625 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3863745672225951, + "lr": 0.03, + "negative_goodness": 1.999873269314236, + "pair_accuracy": 0.48231111111111113, + "positive_goodness": 1.999887440278795, + "runtime_seconds": 11.544586420059204 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.38633431869083, + "lr": 0.03, + "negative_goodness": 1.9999270545959473, + "pair_accuracy": 0.4776444444126553, + "positive_goodness": 1.9999369700537788, + "runtime_seconds": 11.547145128250122 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.386317720519172, + "lr": 0.03, + "negative_goodness": 1.9999535422854953, + "pair_accuracy": 0.4809555555820465, + "positive_goodness": 1.9999625154707168, + "runtime_seconds": 11.537258625030518 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.386313130569458, + "lr": 0.03, + "negative_goodness": 1.9999653950585259, + "pair_accuracy": 0.47904444444444444, + "positive_goodness": 1.9999740001042683, + "runtime_seconds": 11.545739650726318 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.386307244237264, + "lr": 0.03, + "negative_goodness": 1.9999784156799316, + "pair_accuracy": 0.48011111108991833, + "positive_goodness": 1.9999848660998873, + "runtime_seconds": 11.529833555221558 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3863020586861505, + "lr": 0.03, + "negative_goodness": 1.999984213871426, + "pair_accuracy": 0.4806888888888889, + "positive_goodness": 1.999991203795539, + "runtime_seconds": 11.53919506072998 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.3862991127649944, + "lr": 0.03, + "negative_goodness": 1.9999898811128405, + "pair_accuracy": 0.4789777778095669, + "positive_goodness": 1.999996231036716, + "runtime_seconds": 11.530725955963135 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3862972239176432, + "lr": 0.03, + "negative_goodness": 1.999991698943244, + "pair_accuracy": 0.4777777777671814, + "positive_goodness": 1.9999973920398288, + "runtime_seconds": 11.51472806930542 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3862957444296944, + "lr": 0.03, + "negative_goodness": 1.999992286491394, + "pair_accuracy": 0.47711111113230387, + "positive_goodness": 1.999999624739753, + "runtime_seconds": 11.52843189239502 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3862954870223998, + "lr": 0.03, + "negative_goodness": 1.9999921376546224, + "pair_accuracy": 0.47268888890478344, + "positive_goodness": 2.0000018501281738, + "runtime_seconds": 11.527636051177979 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.3862940211402046, + "lr": 0.03, + "negative_goodness": 1.9999834799236722, + "pair_accuracy": 0.4655999999894036, + "positive_goodness": 2.000004312239753, + "runtime_seconds": 11.527149677276611 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3862847721099854, + "lr": 0.03, + "negative_goodness": 1.9999535123401218, + "pair_accuracy": 0.47613333332803515, + "positive_goodness": 2.0000252869500055, + "runtime_seconds": 11.554929494857788 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3862584415859647, + "lr": 0.03, + "negative_goodness": 1.9998969139522977, + "pair_accuracy": 0.48017777774598863, + "positive_goodness": 2.0000922549353706, + "runtime_seconds": 11.5234215259552 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.386219581158956, + "lr": 0.03, + "negative_goodness": 1.9998349396599664, + "pair_accuracy": 0.48055555557145013, + "positive_goodness": 2.0001358089023165, + "runtime_seconds": 11.532071352005005 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3861581343544853, + "lr": 0.03, + "negative_goodness": 1.9997466002782185, + "pair_accuracy": 0.47713333332273694, + "positive_goodness": 2.0002426638709174, + "runtime_seconds": 11.537826299667358 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3861012099160088, + "lr": 0.03, + "negative_goodness": 1.9995884713490804, + "pair_accuracy": 0.4808888888994853, + "positive_goodness": 2.000364315838284, + "runtime_seconds": 11.541010856628418 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3860186934577094, + "lr": 0.03, + "negative_goodness": 1.99942697974311, + "pair_accuracy": 0.4763777777618832, + "positive_goodness": 2.0004105749342176, + "runtime_seconds": 11.547834396362305 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3858750409868028, + "lr": 0.03, + "negative_goodness": 1.999180720117357, + "pair_accuracy": 0.4766888888835907, + "positive_goodness": 2.000694658385383, + "runtime_seconds": 11.552971124649048 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3857830055448743, + "lr": 0.03, + "negative_goodness": 1.9990499725129869, + "pair_accuracy": 0.4768666666454739, + "positive_goodness": 2.0009320565541584, + "runtime_seconds": 11.545578241348267 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.385942573038737, + "lr": 0.03, + "negative_goodness": 1.9993100320180257, + "pair_accuracy": 0.4802888888782925, + "positive_goodness": 2.000648820537991, + "runtime_seconds": 11.52512502670288 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3855780379189386, + "lr": 0.03, + "negative_goodness": 1.9986798217985364, + "pair_accuracy": 0.4787111111217075, + "positive_goodness": 2.0011903016408286, + "runtime_seconds": 11.547099590301514 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3854929528342352, + "lr": 0.03, + "negative_goodness": 1.9987340570025973, + "pair_accuracy": 0.48042222224341496, + "positive_goodness": 2.0014954757266574, + "runtime_seconds": 11.53250789642334 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.38529529554579, + "lr": 0.03, + "negative_goodness": 1.9982106337441339, + "pair_accuracy": 0.4801555555767483, + "positive_goodness": 2.0017501200781926, + "runtime_seconds": 11.543595552444458 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.38551191851298, + "lr": 0.03, + "negative_goodness": 1.9983050478829278, + "pair_accuracy": 0.48117777780426874, + "positive_goodness": 2.0014268269644844, + "runtime_seconds": 11.542173862457275 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3853183434804281, + "lr": 0.03, + "negative_goodness": 1.9983176961686877, + "pair_accuracy": 0.4829777778095669, + "positive_goodness": 2.001822560564677, + "runtime_seconds": 11.527581453323364 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3851959697723388, + "lr": 0.03, + "negative_goodness": 1.9979889679167007, + "pair_accuracy": 0.48308888893127444, + "positive_goodness": 2.001833200454712, + "runtime_seconds": 11.530505895614624 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.3853187643263074, + "lr": 0.03, + "negative_goodness": 1.9982384596294827, + "pair_accuracy": 0.48664444447623356, + "positive_goodness": 2.0019903350194297, + "runtime_seconds": 11.537710905075073 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.3852804603364732, + "lr": 0.03, + "negative_goodness": 1.9981059277428521, + "pair_accuracy": 0.4867555555767483, + "positive_goodness": 2.001842661497328, + "runtime_seconds": 11.532411575317383 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3851466642379762, + "lr": 0.03, + "negative_goodness": 1.997938426484002, + "pair_accuracy": 0.4864000000529819, + "positive_goodness": 2.0020966667599147, + "runtime_seconds": 11.535484790802002 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.385352158885532, + "lr": 0.03, + "negative_goodness": 1.9982086719089085, + "pair_accuracy": 0.48704444444444445, + "positive_goodness": 2.0017527420043946, + "runtime_seconds": 11.529053688049316 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.385140723016527, + "lr": 0.03, + "negative_goodness": 1.9978433038499621, + "pair_accuracy": 0.48646666666136845, + "positive_goodness": 2.0021842341529, + "runtime_seconds": 11.540180921554565 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3851511835098267, + "lr": 0.03, + "negative_goodness": 1.9978831863403321, + "pair_accuracy": 0.4845333333492279, + "positive_goodness": 2.0020109354654947, + "runtime_seconds": 11.532382726669312 + } + ], + "layer": 0 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.393239182387458, + "lr": 0.03, + "negative_goodness": 1.9616807654910617, + "pair_accuracy": 0.5257555555449591, + "positive_goodness": 1.974690376747979, + "runtime_seconds": 11.670281648635864 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3788410871505736, + "lr": 0.03, + "negative_goodness": 1.989470316865709, + "pair_accuracy": 0.5428888888782925, + "positive_goodness": 2.0147465257432726, + "runtime_seconds": 11.653087139129639 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3786003628836738, + "lr": 0.03, + "negative_goodness": 1.9887916625552706, + "pair_accuracy": 0.5456666666666666, + "positive_goodness": 2.0154076199001736, + "runtime_seconds": 11.651921510696411 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.377793459171719, + "lr": 0.03, + "negative_goodness": 1.9879681592941285, + "pair_accuracy": 0.5493333333598243, + "positive_goodness": 2.0172291335635717, + "runtime_seconds": 11.659631252288818 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3765557661056518, + "lr": 0.03, + "negative_goodness": 1.9865790831883747, + "pair_accuracy": 0.55524444448683, + "positive_goodness": 2.0195820301055907, + "runtime_seconds": 11.654523372650146 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3761253327475653, + "lr": 0.03, + "negative_goodness": 1.9854320661332872, + "pair_accuracy": 0.5534222222116258, + "positive_goodness": 2.0206570768568253, + "runtime_seconds": 11.646361589431763 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3752493978712295, + "lr": 0.03, + "negative_goodness": 1.9841625955793594, + "pair_accuracy": 0.5591111111164093, + "positive_goodness": 2.0231010535769993, + "runtime_seconds": 11.651077270507812 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3758854344050089, + "lr": 0.03, + "negative_goodness": 1.9852601196712918, + "pair_accuracy": 0.5582666666772631, + "positive_goodness": 2.0212522205352785, + "runtime_seconds": 11.661660432815552 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3745526916503907, + "lr": 0.03, + "negative_goodness": 1.9828639489915636, + "pair_accuracy": 0.5611111111005147, + "positive_goodness": 2.0239674207051594, + "runtime_seconds": 11.64214825630188 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3749197245491875, + "lr": 0.03, + "negative_goodness": 1.9827560615963407, + "pair_accuracy": 0.5600444444656372, + "positive_goodness": 2.0232069963667128, + "runtime_seconds": 11.64255690574646 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.374197341325548, + "lr": 0.03, + "negative_goodness": 1.9827787133322823, + "pair_accuracy": 0.5670888889312744, + "positive_goodness": 2.025035506990221, + "runtime_seconds": 11.64520263671875 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.373631821738349, + "lr": 0.03, + "negative_goodness": 1.9816444267484876, + "pair_accuracy": 0.5680444444974263, + "positive_goodness": 2.0259911195966933, + "runtime_seconds": 11.63614559173584 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.373728506533305, + "lr": 0.03, + "negative_goodness": 1.981421699057685, + "pair_accuracy": 0.5689333333545261, + "positive_goodness": 2.025994673156738, + "runtime_seconds": 11.642465353012085 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3729024719662137, + "lr": 0.03, + "negative_goodness": 1.980512024921841, + "pair_accuracy": 0.574777777809567, + "positive_goodness": 2.0276968816545273, + "runtime_seconds": 11.642671823501587 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.3738867102305095, + "lr": 0.03, + "negative_goodness": 1.980746754413181, + "pair_accuracy": 0.5673111111005147, + "positive_goodness": 2.0251719367133245, + "runtime_seconds": 11.649094104766846 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3735298875172932, + "lr": 0.03, + "negative_goodness": 1.9805400116814507, + "pair_accuracy": 0.5680222222328186, + "positive_goodness": 2.0267051574283177, + "runtime_seconds": 11.649485349655151 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3731885256873237, + "lr": 0.03, + "negative_goodness": 1.980150555144416, + "pair_accuracy": 0.5710444444656372, + "positive_goodness": 2.0270660296122234, + "runtime_seconds": 11.631670236587524 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.3723228305180868, + "lr": 0.03, + "negative_goodness": 1.9786794797261555, + "pair_accuracy": 0.5710666667090522, + "positive_goodness": 2.0289484508514404, + "runtime_seconds": 11.639246702194214 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.3720224593480428, + "lr": 0.03, + "negative_goodness": 1.978490402730306, + "pair_accuracy": 0.5724666666772631, + "positive_goodness": 2.029432863108317, + "runtime_seconds": 11.624900817871094 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3724742984347873, + "lr": 0.03, + "negative_goodness": 1.9781931638293797, + "pair_accuracy": 0.566088888920678, + "positive_goodness": 2.027853514099121, + "runtime_seconds": 11.62791657447815 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3718375903023614, + "lr": 0.03, + "negative_goodness": 1.9782027191797893, + "pair_accuracy": 0.5666222222328186, + "positive_goodness": 2.0302766784879895, + "runtime_seconds": 11.628828763961792 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.370922574488322, + "lr": 0.03, + "negative_goodness": 1.976829414600796, + "pair_accuracy": 0.5660888889153799, + "positive_goodness": 2.031816070429484, + "runtime_seconds": 11.631505966186523 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3707920121510824, + "lr": 0.03, + "negative_goodness": 1.9763365146213108, + "pair_accuracy": 0.5638444444338481, + "positive_goodness": 2.03183480313619, + "runtime_seconds": 11.62484097480774 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.370685884475708, + "lr": 0.03, + "negative_goodness": 1.9758300384521483, + "pair_accuracy": 0.5675777777459886, + "positive_goodness": 2.0316514954249065, + "runtime_seconds": 11.635108947753906 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3703188287734986, + "lr": 0.03, + "negative_goodness": 1.97615784096188, + "pair_accuracy": 0.5665333333651225, + "positive_goodness": 2.033067085054186, + "runtime_seconds": 11.628687143325806 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3708816438039144, + "lr": 0.03, + "negative_goodness": 1.9758337425443862, + "pair_accuracy": 0.5721111111217075, + "positive_goodness": 2.0320555086347794, + "runtime_seconds": 11.630337238311768 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3702506418652005, + "lr": 0.03, + "negative_goodness": 1.9754018872790866, + "pair_accuracy": 0.5706666666348775, + "positive_goodness": 2.0323763962215846, + "runtime_seconds": 11.633752346038818 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3704272895177205, + "lr": 0.03, + "negative_goodness": 1.9755978047476874, + "pair_accuracy": 0.5698888889100816, + "positive_goodness": 2.03263863474528, + "runtime_seconds": 11.64438009262085 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3695679747899374, + "lr": 0.03, + "negative_goodness": 1.9746400522443983, + "pair_accuracy": 0.5712, + "positive_goodness": 2.0343042987823488, + "runtime_seconds": 11.636109828948975 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3694282942030165, + "lr": 0.03, + "negative_goodness": 1.9735730887095133, + "pair_accuracy": 0.5731999999682108, + "positive_goodness": 2.0342647095150417, + "runtime_seconds": 11.643722772598267 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3695027187983195, + "lr": 0.03, + "negative_goodness": 1.9734778780195448, + "pair_accuracy": 0.5751111111429003, + "positive_goodness": 2.03434226375156, + "runtime_seconds": 11.642555713653564 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.368854802746243, + "lr": 0.03, + "negative_goodness": 1.973143243492974, + "pair_accuracy": 0.5815555555343628, + "positive_goodness": 2.0353850836435954, + "runtime_seconds": 11.635533571243286 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3700896099090576, + "lr": 0.03, + "negative_goodness": 1.974072251701355, + "pair_accuracy": 0.5722222222752041, + "positive_goodness": 2.0335351993984645, + "runtime_seconds": 11.643234014511108 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3683025502522785, + "lr": 0.03, + "negative_goodness": 1.9725307209014893, + "pair_accuracy": 0.5773333333015442, + "positive_goodness": 2.0372370908525257, + "runtime_seconds": 11.642822265625 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.3687897600809733, + "lr": 0.03, + "negative_goodness": 1.972279850599501, + "pair_accuracy": 0.5769555555237664, + "positive_goodness": 2.0358499195098876, + "runtime_seconds": 11.634697198867798 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.367533031548394, + "lr": 0.03, + "negative_goodness": 1.970652756076389, + "pair_accuracy": 0.5771333333757188, + "positive_goodness": 2.0385554161495634, + "runtime_seconds": 11.618282794952393 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3692380514992608, + "lr": 0.03, + "negative_goodness": 1.9719833724127875, + "pair_accuracy": 0.5803777778307597, + "positive_goodness": 2.034656539577908, + "runtime_seconds": 11.61963677406311 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.3680272641499838, + "lr": 0.03, + "negative_goodness": 1.971846527481079, + "pair_accuracy": 0.5758666666666666, + "positive_goodness": 2.0372249233669706, + "runtime_seconds": 11.616859436035156 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.367805988269382, + "lr": 0.03, + "negative_goodness": 1.9715043141682942, + "pair_accuracy": 0.5773333333333334, + "positive_goodness": 2.038396928299798, + "runtime_seconds": 11.619065046310425 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3674915272394816, + "lr": 0.03, + "negative_goodness": 1.9700829806645712, + "pair_accuracy": 0.5748444444338481, + "positive_goodness": 2.03831669637892, + "runtime_seconds": 11.623569965362549 + } + ], + "layer": 1 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.3779288852055867, + "lr": 0.03, + "negative_goodness": 1.9887657958136664, + "pair_accuracy": 0.6058444444391462, + "positive_goodness": 2.019618115785387, + "runtime_seconds": 11.730591297149658 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3687814564387004, + "lr": 0.03, + "negative_goodness": 1.9779838721593221, + "pair_accuracy": 0.6417333333757188, + "positive_goodness": 2.0357369383070205, + "runtime_seconds": 11.721787452697754 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3644064165327283, + "lr": 0.03, + "negative_goodness": 1.9714145876778497, + "pair_accuracy": 0.6431555555449592, + "positive_goodness": 2.0440731387456257, + "runtime_seconds": 11.723405361175537 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.362065449333191, + "lr": 0.03, + "negative_goodness": 1.9683866706848145, + "pair_accuracy": 0.6455111111005147, + "positive_goodness": 2.0485361354404024, + "runtime_seconds": 11.722576141357422 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3612158682293363, + "lr": 0.03, + "negative_goodness": 1.966961821768019, + "pair_accuracy": 0.6397111111111111, + "positive_goodness": 2.0508648213068645, + "runtime_seconds": 11.722963571548462 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3568571471956041, + "lr": 0.03, + "negative_goodness": 1.9619198601616754, + "pair_accuracy": 0.6581333333651225, + "positive_goodness": 2.058422923448351, + "runtime_seconds": 11.722264528274536 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3563150974061755, + "lr": 0.03, + "negative_goodness": 1.9595249438179865, + "pair_accuracy": 0.6472444444868299, + "positive_goodness": 2.06053877398173, + "runtime_seconds": 11.710577249526978 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3549759705649482, + "lr": 0.03, + "negative_goodness": 1.9585947766198053, + "pair_accuracy": 0.6510444444974264, + "positive_goodness": 2.0625272004445394, + "runtime_seconds": 11.708112239837646 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3543412515640258, + "lr": 0.03, + "negative_goodness": 1.9554468627929686, + "pair_accuracy": 0.6513111111429003, + "positive_goodness": 2.063941818194919, + "runtime_seconds": 11.70679521560669 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3528763632244534, + "lr": 0.03, + "negative_goodness": 1.9547743957095676, + "pair_accuracy": 0.6574666666772631, + "positive_goodness": 2.0672829138014053, + "runtime_seconds": 11.705687999725342 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.3560978654649523, + "lr": 0.03, + "negative_goodness": 1.955414313337538, + "pair_accuracy": 0.6444444444497426, + "positive_goodness": 2.0601611454433866, + "runtime_seconds": 11.711697340011597 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3524977211210463, + "lr": 0.03, + "negative_goodness": 1.9540758435143364, + "pair_accuracy": 0.6544888889100816, + "positive_goodness": 2.0679304470909967, + "runtime_seconds": 11.713745594024658 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.3518627265930176, + "lr": 0.03, + "negative_goodness": 1.9525753909640842, + "pair_accuracy": 0.6546666666348775, + "positive_goodness": 2.0690668973710804, + "runtime_seconds": 11.70510745048523 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3506922488530477, + "lr": 0.03, + "negative_goodness": 1.9494787661446464, + "pair_accuracy": 0.6551999999947018, + "positive_goodness": 2.0711264187494915, + "runtime_seconds": 11.703119039535522 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.3493246259265477, + "lr": 0.03, + "negative_goodness": 1.9490591440836589, + "pair_accuracy": 0.6634000000105964, + "positive_goodness": 2.0734378031836616, + "runtime_seconds": 11.700402975082397 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3488970048480564, + "lr": 0.03, + "negative_goodness": 1.9485145690282186, + "pair_accuracy": 0.6548000000317892, + "positive_goodness": 2.075845837656657, + "runtime_seconds": 11.71081805229187 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.3495160894817777, + "lr": 0.03, + "negative_goodness": 1.9472817274517482, + "pair_accuracy": 0.654288888920678, + "positive_goodness": 2.0735929873572454, + "runtime_seconds": 11.695086479187012 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.350903788333469, + "lr": 0.03, + "negative_goodness": 1.949051123131646, + "pair_accuracy": 0.6490888888676961, + "positive_goodness": 2.0715569050682916, + "runtime_seconds": 11.716677904129028 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.3471432321972316, + "lr": 0.03, + "negative_goodness": 1.945978702460395, + "pair_accuracy": 0.6623555555979411, + "positive_goodness": 2.077838640107049, + "runtime_seconds": 11.702491760253906 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3479180399364896, + "lr": 0.03, + "negative_goodness": 1.9449413262261284, + "pair_accuracy": 0.6597555555873447, + "positive_goodness": 2.076523918363783, + "runtime_seconds": 11.701437711715698 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3481328354941473, + "lr": 0.03, + "negative_goodness": 1.9450641219245064, + "pair_accuracy": 0.6556444444762336, + "positive_goodness": 2.077333525127835, + "runtime_seconds": 11.710411071777344 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.3472794145584106, + "lr": 0.03, + "negative_goodness": 1.9443302415211996, + "pair_accuracy": 0.6572444444391462, + "positive_goodness": 2.0780227456410727, + "runtime_seconds": 11.709099769592285 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3464267033259074, + "lr": 0.03, + "negative_goodness": 1.9418150040944417, + "pair_accuracy": 0.6620222222010295, + "positive_goodness": 2.0812279297934637, + "runtime_seconds": 11.71731686592102 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.343720426834954, + "lr": 0.03, + "negative_goodness": 1.9397735955132378, + "pair_accuracy": 0.6637999999894036, + "positive_goodness": 2.0842563710530597, + "runtime_seconds": 11.7159583568573 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.34674928372701, + "lr": 0.03, + "negative_goodness": 1.9431276210784911, + "pair_accuracy": 0.6601555555608537, + "positive_goodness": 2.0791513670603434, + "runtime_seconds": 11.722659349441528 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3443423208236693, + "lr": 0.03, + "negative_goodness": 1.9409642622205947, + "pair_accuracy": 0.6657777777989705, + "positive_goodness": 2.0844644080691865, + "runtime_seconds": 11.722656488418579 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3468330902735393, + "lr": 0.03, + "negative_goodness": 1.941868440161811, + "pair_accuracy": 0.652422222216924, + "positive_goodness": 2.0789949456532795, + "runtime_seconds": 11.718188047409058 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3456064592573378, + "lr": 0.03, + "negative_goodness": 1.9406791231367324, + "pair_accuracy": 0.6624444444656372, + "positive_goodness": 2.0815360538482666, + "runtime_seconds": 11.720701694488525 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.345142516877916, + "lr": 0.03, + "negative_goodness": 1.9398507436964247, + "pair_accuracy": 0.6605555555979411, + "positive_goodness": 2.082879420598348, + "runtime_seconds": 11.716899633407593 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3442338406456842, + "lr": 0.03, + "negative_goodness": 1.9395842061572606, + "pair_accuracy": 0.6600444444550408, + "positive_goodness": 2.084769582663642, + "runtime_seconds": 11.721388816833496 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3441558550516763, + "lr": 0.03, + "negative_goodness": 1.9395059670766195, + "pair_accuracy": 0.6571111111429002, + "positive_goodness": 2.0850807479010687, + "runtime_seconds": 11.718357563018799 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3446083188798692, + "lr": 0.03, + "negative_goodness": 1.9379974900139703, + "pair_accuracy": 0.6592000000211927, + "positive_goodness": 2.0834518875545927, + "runtime_seconds": 11.723237037658691 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3431040344662137, + "lr": 0.03, + "negative_goodness": 1.9392359955045912, + "pair_accuracy": 0.6648666666666667, + "positive_goodness": 2.0866791542053225, + "runtime_seconds": 11.720881462097168 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.3437935701158312, + "lr": 0.03, + "negative_goodness": 1.9389615159140692, + "pair_accuracy": 0.6619333333863152, + "positive_goodness": 2.0859078858269586, + "runtime_seconds": 11.72426986694336 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.343361407746209, + "lr": 0.03, + "negative_goodness": 1.9375534372541638, + "pair_accuracy": 0.66304444448683, + "positive_goodness": 2.085440637927585, + "runtime_seconds": 11.723474979400635 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.3424837616814507, + "lr": 0.03, + "negative_goodness": 1.935817343436347, + "pair_accuracy": 0.6673777777989706, + "positive_goodness": 2.0889858891381157, + "runtime_seconds": 11.72346019744873 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3436959777832032, + "lr": 0.03, + "negative_goodness": 1.937473092481825, + "pair_accuracy": 0.667333333354526, + "positive_goodness": 2.086082305293613, + "runtime_seconds": 11.729148864746094 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.341579214117262, + "lr": 0.03, + "negative_goodness": 1.9371237266116672, + "pair_accuracy": 0.6674000000105964, + "positive_goodness": 2.0890570577409533, + "runtime_seconds": 11.723307847976685 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.3419573312759399, + "lr": 0.03, + "negative_goodness": 1.9358948572794596, + "pair_accuracy": 0.6640222222328186, + "positive_goodness": 2.0892184382544623, + "runtime_seconds": 11.72895884513855 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.3413844200346206, + "lr": 0.03, + "negative_goodness": 1.9352415680143569, + "pair_accuracy": 0.6656222222540114, + "positive_goodness": 2.09022466085222, + "runtime_seconds": 11.722878217697144 + } + ], + "layer": 2 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.367459845161438, + "lr": 0.03, + "negative_goodness": 1.9330516387727525, + "pair_accuracy": 0.641066666645474, + "positive_goodness": 2.0261028359307183, + "runtime_seconds": 11.667250394821167 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.3472698350694445, + "lr": 0.03, + "negative_goodness": 1.9507722727033827, + "pair_accuracy": 0.6807777777459887, + "positive_goodness": 2.0758965493943955, + "runtime_seconds": 11.661813259124756 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.3402033028072782, + "lr": 0.03, + "negative_goodness": 1.9453956406911215, + "pair_accuracy": 0.7050666666984559, + "positive_goodness": 2.0905503288269043, + "runtime_seconds": 11.645031690597534 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.338367722553677, + "lr": 0.03, + "negative_goodness": 1.9432355700174968, + "pair_accuracy": 0.7085777778201633, + "positive_goodness": 2.0938557069566515, + "runtime_seconds": 11.642852544784546 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.3344461128022935, + "lr": 0.03, + "negative_goodness": 1.9399895871056452, + "pair_accuracy": 0.7206666666772631, + "positive_goodness": 2.1013122850630017, + "runtime_seconds": 11.651227951049805 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.3315617384380765, + "lr": 0.03, + "negative_goodness": 1.9371667368994818, + "pair_accuracy": 0.7284222222540113, + "positive_goodness": 2.1069872475941978, + "runtime_seconds": 11.643929719924927 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.3285587941699557, + "lr": 0.03, + "negative_goodness": 1.9342771880679661, + "pair_accuracy": 0.736, + "positive_goodness": 2.113783571243286, + "runtime_seconds": 11.64828085899353 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.3265849003050063, + "lr": 0.03, + "negative_goodness": 1.9326733703613281, + "pair_accuracy": 0.7379777778201633, + "positive_goodness": 2.117318302324083, + "runtime_seconds": 11.645737886428833 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.3253709215799967, + "lr": 0.03, + "negative_goodness": 1.931463662677341, + "pair_accuracy": 0.7440888889312745, + "positive_goodness": 2.119651028611925, + "runtime_seconds": 11.646366119384766 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.3253197524388631, + "lr": 0.03, + "negative_goodness": 1.931640121799045, + "pair_accuracy": 0.745755555597941, + "positive_goodness": 2.1195050912221274, + "runtime_seconds": 11.636082172393799 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.322776247490777, + "lr": 0.03, + "negative_goodness": 1.9290033570395575, + "pair_accuracy": 0.7493111111429003, + "positive_goodness": 2.1253040307362876, + "runtime_seconds": 11.645835161209106 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.3204720533794827, + "lr": 0.03, + "negative_goodness": 1.9231762857437134, + "pair_accuracy": 0.7449111111217075, + "positive_goodness": 2.130285669369168, + "runtime_seconds": 11.63994026184082 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.3175998697280884, + "lr": 0.03, + "negative_goodness": 1.923855586666531, + "pair_accuracy": 0.7539777778042688, + "positive_goodness": 2.1357438816070555, + "runtime_seconds": 11.644152879714966 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.3211536750157675, + "lr": 0.03, + "negative_goodness": 1.9226992101245457, + "pair_accuracy": 0.746911111164093, + "positive_goodness": 2.1281302406522964, + "runtime_seconds": 11.623492956161499 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.319751652675205, + "lr": 0.03, + "negative_goodness": 1.9243443406422933, + "pair_accuracy": 0.748422222243415, + "positive_goodness": 2.132717906697591, + "runtime_seconds": 11.628913402557373 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.3150872550328572, + "lr": 0.03, + "negative_goodness": 1.9186897555881077, + "pair_accuracy": 0.7639555555767483, + "positive_goodness": 2.140177485699124, + "runtime_seconds": 11.627901554107666 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.316307385190328, + "lr": 0.03, + "negative_goodness": 1.9201749829186334, + "pair_accuracy": 0.7561111111534966, + "positive_goodness": 2.1396531886206733, + "runtime_seconds": 11.629817724227905 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.313331455569797, + "lr": 0.03, + "negative_goodness": 1.9130325802909003, + "pair_accuracy": 0.7633555555555556, + "positive_goodness": 2.1434914997524683, + "runtime_seconds": 11.623202323913574 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.3123902754041883, + "lr": 0.03, + "negative_goodness": 1.915258291223314, + "pair_accuracy": 0.7598888888888888, + "positive_goodness": 2.14806330625746, + "runtime_seconds": 11.619752407073975 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.3130006041844686, + "lr": 0.03, + "negative_goodness": 1.9159666412777372, + "pair_accuracy": 0.7623111111111112, + "positive_goodness": 2.1457001810709637, + "runtime_seconds": 11.637765169143677 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.3154597177929348, + "lr": 0.03, + "negative_goodness": 1.916594235950046, + "pair_accuracy": 0.7538888889206781, + "positive_goodness": 2.1410362669203016, + "runtime_seconds": 11.641167640686035 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.312836881934272, + "lr": 0.03, + "negative_goodness": 1.914357154973348, + "pair_accuracy": 0.7585777777883742, + "positive_goodness": 2.1469469069374933, + "runtime_seconds": 11.63805603981018 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.3125800873862372, + "lr": 0.03, + "negative_goodness": 1.913197738901774, + "pair_accuracy": 0.758822222243415, + "positive_goodness": 2.146700999153985, + "runtime_seconds": 11.621946096420288 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.3095291776869031, + "lr": 0.03, + "negative_goodness": 1.9085307179556952, + "pair_accuracy": 0.7667555555449592, + "positive_goodness": 2.1535462416966755, + "runtime_seconds": 11.623175144195557 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.3048832671695285, + "lr": 0.03, + "negative_goodness": 1.9060974386427139, + "pair_accuracy": 0.7748666666348776, + "positive_goodness": 2.1630530720180934, + "runtime_seconds": 11.618128776550293 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.3096126085917155, + "lr": 0.03, + "negative_goodness": 1.9098250877804226, + "pair_accuracy": 0.7658444444232517, + "positive_goodness": 2.152207693311903, + "runtime_seconds": 11.616456270217896 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.3087055231730143, + "lr": 0.03, + "negative_goodness": 1.908130583275689, + "pair_accuracy": 0.7731333333439298, + "positive_goodness": 2.1546057190789116, + "runtime_seconds": 11.623927354812622 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.3108820977105036, + "lr": 0.03, + "negative_goodness": 1.908169064034356, + "pair_accuracy": 0.7585111111111111, + "positive_goodness": 2.1519969491322835, + "runtime_seconds": 11.628703355789185 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.3094999146355524, + "lr": 0.03, + "negative_goodness": 1.908427742852105, + "pair_accuracy": 0.7626, + "positive_goodness": 2.154376592042711, + "runtime_seconds": 11.618270874023438 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.3071542667812772, + "lr": 0.03, + "negative_goodness": 1.9072853684531317, + "pair_accuracy": 0.7695111111429003, + "positive_goodness": 2.156296457672119, + "runtime_seconds": 11.612388849258423 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.3037435710695056, + "lr": 0.03, + "negative_goodness": 1.9053281052907307, + "pair_accuracy": 0.77824444448683, + "positive_goodness": 2.164063990995619, + "runtime_seconds": 11.618407487869263 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.3051587497499253, + "lr": 0.03, + "negative_goodness": 1.903402171919081, + "pair_accuracy": 0.7731333333863153, + "positive_goodness": 2.163484990141127, + "runtime_seconds": 11.620911121368408 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.3064426875008477, + "lr": 0.03, + "negative_goodness": 1.9060485151926676, + "pair_accuracy": 0.7691999999894036, + "positive_goodness": 2.159691228315565, + "runtime_seconds": 11.610280752182007 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.308355294058058, + "lr": 0.03, + "negative_goodness": 1.9051784443537394, + "pair_accuracy": 0.7619111111217075, + "positive_goodness": 2.1564514309777154, + "runtime_seconds": 11.614599466323853 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.3042696283128528, + "lr": 0.03, + "negative_goodness": 1.9028806007385255, + "pair_accuracy": 0.7675333333863152, + "positive_goodness": 2.1636217622968887, + "runtime_seconds": 11.604770421981812 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.300296650483873, + "lr": 0.03, + "negative_goodness": 1.8978835581673517, + "pair_accuracy": 0.7782666666560702, + "positive_goodness": 2.1727986998240154, + "runtime_seconds": 11.609273195266724 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.3086429328494602, + "lr": 0.03, + "negative_goodness": 1.9046085112253825, + "pair_accuracy": 0.7616888888994853, + "positive_goodness": 2.1568697394053142, + "runtime_seconds": 11.605615139007568 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.3027260330412123, + "lr": 0.03, + "negative_goodness": 1.9004781151453654, + "pair_accuracy": 0.7750222222646077, + "positive_goodness": 2.1668782386779784, + "runtime_seconds": 11.606534481048584 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.302135106171502, + "lr": 0.03, + "negative_goodness": 1.9000673748228285, + "pair_accuracy": 0.7743111110793219, + "positive_goodness": 2.170170675532023, + "runtime_seconds": 11.611836671829224 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.2992284558614096, + "lr": 0.03, + "negative_goodness": 1.8986741865581935, + "pair_accuracy": 0.7775777777883741, + "positive_goodness": 2.1736937514199153, + "runtime_seconds": 11.604755640029907 + } + ], + "layer": 3 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2974100570042928, + "lr": 0.03, + "negative_goodness": 1.9269117997487386, + "pair_accuracy": 0.8021111111429002, + "positive_goodness": 2.2028569863637286, + "runtime_seconds": 11.71943974494934 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.266902316877577, + "lr": 0.03, + "negative_goodness": 1.8903473185857138, + "pair_accuracy": 0.8629555555979411, + "positive_goodness": 2.2574358168707955, + "runtime_seconds": 11.705174684524536 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.2602848402235243, + "lr": 0.03, + "negative_goodness": 1.8844839577568901, + "pair_accuracy": 0.8586444444656373, + "positive_goodness": 2.27039889348348, + "runtime_seconds": 11.700181007385254 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.2541868543836805, + "lr": 0.03, + "negative_goodness": 1.8768214871724447, + "pair_accuracy": 0.8630444444656372, + "positive_goodness": 2.279573274697198, + "runtime_seconds": 11.697310209274292 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.258516755104065, + "lr": 0.03, + "negative_goodness": 1.8788381919648913, + "pair_accuracy": 0.8528222222010294, + "positive_goodness": 2.2712957264794245, + "runtime_seconds": 11.685123682022095 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.2459681421279907, + "lr": 0.03, + "negative_goodness": 1.8724061972300212, + "pair_accuracy": 0.8662888889418708, + "positive_goodness": 2.295175547366672, + "runtime_seconds": 11.678721189498901 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.2476011573791503, + "lr": 0.03, + "negative_goodness": 1.8705129869249133, + "pair_accuracy": 0.8556222222222222, + "positive_goodness": 2.2910348448435465, + "runtime_seconds": 11.68252682685852 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.2391502485911052, + "lr": 0.03, + "negative_goodness": 1.8617994691001045, + "pair_accuracy": 0.8643777778307596, + "positive_goodness": 2.3087576757219104, + "runtime_seconds": 11.697506427764893 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.2430077379862468, + "lr": 0.03, + "negative_goodness": 1.8647374381171333, + "pair_accuracy": 0.8567111111111111, + "positive_goodness": 2.2991002980126276, + "runtime_seconds": 11.691073179244995 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.2425199806213378, + "lr": 0.03, + "negative_goodness": 1.8643790057076348, + "pair_accuracy": 0.8580888888676961, + "positive_goodness": 2.3020848465389676, + "runtime_seconds": 11.688095331192017 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.2373072817060682, + "lr": 0.03, + "negative_goodness": 1.8596033768759834, + "pair_accuracy": 0.8605111111429002, + "positive_goodness": 2.312140870581733, + "runtime_seconds": 11.709283590316772 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.2283480828179254, + "lr": 0.03, + "negative_goodness": 1.8550180793338351, + "pair_accuracy": 0.8710888889312745, + "positive_goodness": 2.3313412198384604, + "runtime_seconds": 11.686227560043335 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.2346090318467882, + "lr": 0.03, + "negative_goodness": 1.8553644347720677, + "pair_accuracy": 0.8640888888888889, + "positive_goodness": 2.3175848207685683, + "runtime_seconds": 11.6819748878479 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.2267238519456651, + "lr": 0.03, + "negative_goodness": 1.8520526095496284, + "pair_accuracy": 0.8676222222116259, + "positive_goodness": 2.332529445054796, + "runtime_seconds": 11.692780017852783 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.231348532019721, + "lr": 0.03, + "negative_goodness": 1.8511238772710163, + "pair_accuracy": 0.859711111079322, + "positive_goodness": 2.3248994780222576, + "runtime_seconds": 11.623886823654175 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.224970084063212, + "lr": 0.03, + "negative_goodness": 1.8484964473512437, + "pair_accuracy": 0.8694222222540113, + "positive_goodness": 2.337208944193522, + "runtime_seconds": 11.623820066452026 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.2337086119969687, + "lr": 0.03, + "negative_goodness": 1.8517006676355998, + "pair_accuracy": 0.8582222222540113, + "positive_goodness": 2.320470022455851, + "runtime_seconds": 11.622841596603394 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.2233535760243734, + "lr": 0.03, + "negative_goodness": 1.8441259371439616, + "pair_accuracy": 0.8621111111005147, + "positive_goodness": 2.341385015911526, + "runtime_seconds": 11.623075723648071 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.2238338752746583, + "lr": 0.03, + "negative_goodness": 1.8431409721798366, + "pair_accuracy": 0.863977777809567, + "positive_goodness": 2.341117366748386, + "runtime_seconds": 11.620038032531738 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.2320029454549153, + "lr": 0.03, + "negative_goodness": 1.8481553483115303, + "pair_accuracy": 0.8548444444550408, + "positive_goodness": 2.323660628721449, + "runtime_seconds": 11.623883724212646 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.2252296841727364, + "lr": 0.03, + "negative_goodness": 1.8471482981363931, + "pair_accuracy": 0.8634666666878594, + "positive_goodness": 2.339702256562975, + "runtime_seconds": 11.620970010757446 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.216896374575297, + "lr": 0.03, + "negative_goodness": 1.8399128238677978, + "pair_accuracy": 0.8714444444126553, + "positive_goodness": 2.352192191865709, + "runtime_seconds": 11.624725818634033 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.2211043091668023, + "lr": 0.03, + "negative_goodness": 1.837358495839437, + "pair_accuracy": 0.8575111111005147, + "positive_goodness": 2.347506913757324, + "runtime_seconds": 11.621871709823608 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.2195650589201186, + "lr": 0.03, + "negative_goodness": 1.8400406060112848, + "pair_accuracy": 0.861977777809567, + "positive_goodness": 2.3512246698167587, + "runtime_seconds": 11.618400573730469 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.2172330122417874, + "lr": 0.03, + "negative_goodness": 1.8378624129825167, + "pair_accuracy": 0.8661111110793219, + "positive_goodness": 2.356418167453342, + "runtime_seconds": 11.623245239257812 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.2202847961213854, + "lr": 0.03, + "negative_goodness": 1.8398439022911919, + "pair_accuracy": 0.8641111111534966, + "positive_goodness": 2.350682380549113, + "runtime_seconds": 11.619524002075195 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.2184630412419637, + "lr": 0.03, + "negative_goodness": 1.8356134390936958, + "pair_accuracy": 0.8619111110899184, + "positive_goodness": 2.3530470195346407, + "runtime_seconds": 11.62349557876587 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.2198888775931465, + "lr": 0.03, + "negative_goodness": 1.8395154721577962, + "pair_accuracy": 0.8632888889312744, + "positive_goodness": 2.3478364072587756, + "runtime_seconds": 11.625456094741821 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.2210440651787653, + "lr": 0.03, + "negative_goodness": 1.8363555541144476, + "pair_accuracy": 0.8597111111534966, + "positive_goodness": 2.3461366112179225, + "runtime_seconds": 11.632477521896362 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.2153547239091662, + "lr": 0.03, + "negative_goodness": 1.835810061009725, + "pair_accuracy": 0.870333333322737, + "positive_goodness": 2.359571813074748, + "runtime_seconds": 11.631120920181274 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.210071067958408, + "lr": 0.03, + "negative_goodness": 1.8302204285727608, + "pair_accuracy": 0.8749111111429002, + "positive_goodness": 2.371207626809014, + "runtime_seconds": 11.626039981842041 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.2212485893249512, + "lr": 0.03, + "negative_goodness": 1.8375002615186904, + "pair_accuracy": 0.8562000000211928, + "positive_goodness": 2.3504805737813315, + "runtime_seconds": 11.622323513031006 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.2096435585869683, + "lr": 0.03, + "negative_goodness": 1.8281557895236544, + "pair_accuracy": 0.8686888888994853, + "positive_goodness": 2.370964261288113, + "runtime_seconds": 11.618919610977173 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.2151046886232164, + "lr": 0.03, + "negative_goodness": 1.8310402765274048, + "pair_accuracy": 0.8636666667196485, + "positive_goodness": 2.3598729967329235, + "runtime_seconds": 11.615501165390015 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.2134665329827203, + "lr": 0.03, + "negative_goodness": 1.8323688009474013, + "pair_accuracy": 0.8668000000105964, + "positive_goodness": 2.3668355489518906, + "runtime_seconds": 11.621449947357178 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.2125865678151448, + "lr": 0.03, + "negative_goodness": 1.8306454999923707, + "pair_accuracy": 0.8675111111217075, + "positive_goodness": 2.367877593188816, + "runtime_seconds": 11.615777730941772 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.215733489312066, + "lr": 0.03, + "negative_goodness": 1.8316720685111152, + "pair_accuracy": 0.8602222222752042, + "positive_goodness": 2.3616803107791475, + "runtime_seconds": 11.618388652801514 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.2051207848654852, + "lr": 0.03, + "negative_goodness": 1.8211462195078532, + "pair_accuracy": 0.8687999999894036, + "positive_goodness": 2.383752703942193, + "runtime_seconds": 11.614291667938232 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.2088908529069688, + "lr": 0.03, + "negative_goodness": 1.8280805202907986, + "pair_accuracy": 0.8697333333121405, + "positive_goodness": 2.3775679295433894, + "runtime_seconds": 11.61533522605896 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.2186238932715523, + "lr": 0.03, + "negative_goodness": 1.8293737785551283, + "pair_accuracy": 0.8589111111534966, + "positive_goodness": 2.3541522352430557, + "runtime_seconds": 11.629469633102417 + } + ], + "layer": 4 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2640943185806275, + "lr": 0.03, + "negative_goodness": 1.837949653010898, + "pair_accuracy": 0.817422222190433, + "positive_goodness": 2.2383010758294, + "runtime_seconds": 11.543519496917725 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.2257584898206924, + "lr": 0.03, + "negative_goodness": 1.8485151755226983, + "pair_accuracy": 0.8469777777671814, + "positive_goodness": 2.3276434290568035, + "runtime_seconds": 11.528894186019897 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.2186515785853067, + "lr": 0.03, + "negative_goodness": 1.8423954274495442, + "pair_accuracy": 0.8486000000423856, + "positive_goodness": 2.3426477941301136, + "runtime_seconds": 11.536607503890991 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.2148487347284953, + "lr": 0.03, + "negative_goodness": 1.839344624710083, + "pair_accuracy": 0.8546888888782925, + "positive_goodness": 2.3535570917341446, + "runtime_seconds": 11.529999256134033 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.2094115921444364, + "lr": 0.03, + "negative_goodness": 1.8373536109076607, + "pair_accuracy": 0.8564666666666667, + "positive_goodness": 2.3640496288723414, + "runtime_seconds": 11.537490367889404 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.2084835014979045, + "lr": 0.03, + "negative_goodness": 1.833891646364, + "pair_accuracy": 0.856422222243415, + "positive_goodness": 2.364744267442491, + "runtime_seconds": 11.527695655822754 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.1991506695217558, + "lr": 0.03, + "negative_goodness": 1.8261526878992715, + "pair_accuracy": 0.8599999999682109, + "positive_goodness": 2.3870917737325033, + "runtime_seconds": 11.528368949890137 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.1879227691226535, + "lr": 0.03, + "negative_goodness": 1.8220267570919462, + "pair_accuracy": 0.8696888888888888, + "positive_goodness": 2.414912501186795, + "runtime_seconds": 11.524341106414795 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.196601596238878, + "lr": 0.03, + "negative_goodness": 1.8190533699247573, + "pair_accuracy": 0.8565777777565851, + "positive_goodness": 2.3921807975345186, + "runtime_seconds": 11.521955251693726 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.192830986234877, + "lr": 0.03, + "negative_goodness": 1.8233083541234334, + "pair_accuracy": 0.8683111111005147, + "positive_goodness": 2.400154271274143, + "runtime_seconds": 11.524983882904053 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.195799821556939, + "lr": 0.03, + "negative_goodness": 1.828004918776618, + "pair_accuracy": 0.8581777777777778, + "positive_goodness": 2.395688100687663, + "runtime_seconds": 11.531301259994507 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.1882898572497897, + "lr": 0.03, + "negative_goodness": 1.83034230533176, + "pair_accuracy": 0.8659333333333333, + "positive_goodness": 2.4098353417714438, + "runtime_seconds": 11.529027462005615 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.1830635355419583, + "lr": 0.03, + "negative_goodness": 1.8295861261791653, + "pair_accuracy": 0.8722444444232517, + "positive_goodness": 2.4185531385209824, + "runtime_seconds": 11.53736162185669 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.1830593860626222, + "lr": 0.03, + "negative_goodness": 1.8295215478685167, + "pair_accuracy": 0.8675333333015441, + "positive_goodness": 2.4190804520077176, + "runtime_seconds": 11.542793989181519 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.1783839426252578, + "lr": 0.03, + "negative_goodness": 1.8269362415737576, + "pair_accuracy": 0.8726666666348776, + "positive_goodness": 2.4326939543830024, + "runtime_seconds": 11.541524648666382 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.1670567652172512, + "lr": 0.03, + "negative_goodness": 1.8228577316708034, + "pair_accuracy": 0.8855777778307596, + "positive_goodness": 2.451684148322211, + "runtime_seconds": 11.542285442352295 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.1646284853193496, + "lr": 0.03, + "negative_goodness": 1.8171203642527263, + "pair_accuracy": 0.8821777777777777, + "positive_goodness": 2.459286468908522, + "runtime_seconds": 11.53899359703064 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.1734953879568313, + "lr": 0.03, + "negative_goodness": 1.8188947138044569, + "pair_accuracy": 0.8769999999788073, + "positive_goodness": 2.4421356020185683, + "runtime_seconds": 11.531331539154053 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.1733903111139934, + "lr": 0.03, + "negative_goodness": 1.8217276794009738, + "pair_accuracy": 0.8745333333863152, + "positive_goodness": 2.443972422027588, + "runtime_seconds": 11.543113946914673 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.1579077375517952, + "lr": 0.03, + "negative_goodness": 1.8098951486587525, + "pair_accuracy": 0.8828666666666667, + "positive_goodness": 2.474400340652466, + "runtime_seconds": 11.534554243087769 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.1619299196243287, + "lr": 0.03, + "negative_goodness": 1.8154900616539849, + "pair_accuracy": 0.8755555555237664, + "positive_goodness": 2.4672131683349607, + "runtime_seconds": 11.540829181671143 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.1672344216452704, + "lr": 0.03, + "negative_goodness": 1.8176004131529067, + "pair_accuracy": 0.8785777778201633, + "positive_goodness": 2.455469316482544, + "runtime_seconds": 11.553859949111938 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.1715546137068007, + "lr": 0.03, + "negative_goodness": 1.8171795966042412, + "pair_accuracy": 0.8706222222646077, + "positive_goodness": 2.4485380153656005, + "runtime_seconds": 11.540721416473389 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.176495985306634, + "lr": 0.03, + "negative_goodness": 1.8202688810984293, + "pair_accuracy": 0.8626888888676961, + "positive_goodness": 2.4382167087131075, + "runtime_seconds": 11.554187536239624 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.1624179238637289, + "lr": 0.03, + "negative_goodness": 1.8158309200498792, + "pair_accuracy": 0.8846888888570997, + "positive_goodness": 2.468046030553182, + "runtime_seconds": 11.549007892608643 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.1736109365887113, + "lr": 0.03, + "negative_goodness": 1.8189527132458156, + "pair_accuracy": 0.873777777756585, + "positive_goodness": 2.4418824298858643, + "runtime_seconds": 11.554255485534668 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.164602993657854, + "lr": 0.03, + "negative_goodness": 1.8128246152453953, + "pair_accuracy": 0.8739777777671814, + "positive_goodness": 2.4650518174489338, + "runtime_seconds": 11.552649736404419 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.1558470350689358, + "lr": 0.03, + "negative_goodness": 1.81203887424469, + "pair_accuracy": 0.8863111111005147, + "positive_goodness": 2.4806192494710286, + "runtime_seconds": 11.541622877120972 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.1705708314259846, + "lr": 0.03, + "negative_goodness": 1.8168842864990233, + "pair_accuracy": 0.8754222222328186, + "positive_goodness": 2.453692120827569, + "runtime_seconds": 11.541551113128662 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.167636282602946, + "lr": 0.03, + "negative_goodness": 1.8136602623621623, + "pair_accuracy": 0.8702444444126552, + "positive_goodness": 2.457043643230862, + "runtime_seconds": 11.533794403076172 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.1491878471586439, + "lr": 0.03, + "negative_goodness": 1.8019826803631254, + "pair_accuracy": 0.8855333333015442, + "positive_goodness": 2.495717789883084, + "runtime_seconds": 11.55349349975586 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.1570207036972047, + "lr": 0.03, + "negative_goodness": 1.8057381370968288, + "pair_accuracy": 0.8806444444232516, + "positive_goodness": 2.4827624225616454, + "runtime_seconds": 11.5486421585083 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.1579727396011352, + "lr": 0.03, + "negative_goodness": 1.8047701614379883, + "pair_accuracy": 0.8808666666772631, + "positive_goodness": 2.4783652513715957, + "runtime_seconds": 11.551050424575806 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.1590536931355795, + "lr": 0.03, + "negative_goodness": 1.8106206292894151, + "pair_accuracy": 0.8805555555237664, + "positive_goodness": 2.4739725797865124, + "runtime_seconds": 11.549458026885986 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.1479945594787597, + "lr": 0.03, + "negative_goodness": 1.798618041059706, + "pair_accuracy": 0.8868888888676961, + "positive_goodness": 2.503169287829929, + "runtime_seconds": 11.551609754562378 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.145593791601393, + "lr": 0.03, + "negative_goodness": 1.793120276260376, + "pair_accuracy": 0.8898222222540114, + "positive_goodness": 2.506599712329441, + "runtime_seconds": 11.540189027786255 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.1569806369569566, + "lr": 0.03, + "negative_goodness": 1.8061028752009074, + "pair_accuracy": 0.883733333322737, + "positive_goodness": 2.4832660158369277, + "runtime_seconds": 11.541802167892456 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.1494661441379124, + "lr": 0.03, + "negative_goodness": 1.7960279945585462, + "pair_accuracy": 0.8794000000529819, + "positive_goodness": 2.499471570417616, + "runtime_seconds": 11.542744159698486 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.1600474132325913, + "lr": 0.03, + "negative_goodness": 1.8039143653021918, + "pair_accuracy": 0.8668888888676961, + "positive_goodness": 2.4775163215213354, + "runtime_seconds": 11.535056829452515 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.1499775709788005, + "lr": 0.03, + "negative_goodness": 1.797292807091607, + "pair_accuracy": 0.8822888889312744, + "positive_goodness": 2.500367884190877, + "runtime_seconds": 11.545231819152832 + } + ], + "layer": 5 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2831310903549193, + "lr": 0.03, + "negative_goodness": 1.9032643579271105, + "pair_accuracy": 0.7616444444444445, + "positive_goodness": 2.256997769970364, + "runtime_seconds": 11.64335560798645 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.2216866991043092, + "lr": 0.03, + "negative_goodness": 1.8548423623402914, + "pair_accuracy": 0.8301777777724796, + "positive_goodness": 2.3521213873969185, + "runtime_seconds": 11.635652542114258 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.2185263077841866, + "lr": 0.03, + "negative_goodness": 1.84158028594123, + "pair_accuracy": 0.8302666667090521, + "positive_goodness": 2.3502636603461373, + "runtime_seconds": 11.63414216041565 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.2001175803926256, + "lr": 0.03, + "negative_goodness": 1.8359826647228665, + "pair_accuracy": 0.8460888889100816, + "positive_goodness": 2.390721460511949, + "runtime_seconds": 11.637096881866455 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.1992333313624064, + "lr": 0.03, + "negative_goodness": 1.8327802218119302, + "pair_accuracy": 0.8430000000423855, + "positive_goodness": 2.3909603073543972, + "runtime_seconds": 11.629005670547485 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.1939834526485866, + "lr": 0.03, + "negative_goodness": 1.8216703240712484, + "pair_accuracy": 0.8457111110899184, + "positive_goodness": 2.3978169802771676, + "runtime_seconds": 11.636808156967163 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.2042369902928671, + "lr": 0.03, + "negative_goodness": 1.8351795131259494, + "pair_accuracy": 0.8393777777565851, + "positive_goodness": 2.37671755794949, + "runtime_seconds": 11.644983530044556 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.1937279999838935, + "lr": 0.03, + "negative_goodness": 1.8255864045461019, + "pair_accuracy": 0.8449777777459886, + "positive_goodness": 2.395956175189548, + "runtime_seconds": 11.647197484970093 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.1822545893563166, + "lr": 0.03, + "negative_goodness": 1.8136470387988621, + "pair_accuracy": 0.8522666666878594, + "positive_goodness": 2.4208717231750487, + "runtime_seconds": 11.639488458633423 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.1753739503860474, + "lr": 0.03, + "negative_goodness": 1.8126982564502292, + "pair_accuracy": 0.8579333333333333, + "positive_goodness": 2.436575340270996, + "runtime_seconds": 11.637054920196533 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.1734663257598876, + "lr": 0.03, + "negative_goodness": 1.8063731282975939, + "pair_accuracy": 0.8558888888782925, + "positive_goodness": 2.4396613458421497, + "runtime_seconds": 11.62238597869873 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.1676012566248575, + "lr": 0.03, + "negative_goodness": 1.7997524366590711, + "pair_accuracy": 0.8590222222752041, + "positive_goodness": 2.454410974036323, + "runtime_seconds": 11.634993553161621 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.1617865560955472, + "lr": 0.03, + "negative_goodness": 1.7935044751061333, + "pair_accuracy": 0.8613555555343628, + "positive_goodness": 2.4669948513454862, + "runtime_seconds": 11.620763301849365 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.1595107731501262, + "lr": 0.03, + "negative_goodness": 1.7922387006547715, + "pair_accuracy": 0.8613333333492279, + "positive_goodness": 2.4709676604800754, + "runtime_seconds": 11.62801194190979 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.1818918980492485, + "lr": 0.03, + "negative_goodness": 1.8059306574927436, + "pair_accuracy": 0.8468666667196486, + "positive_goodness": 2.4239385548061794, + "runtime_seconds": 11.63556694984436 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.158981059243944, + "lr": 0.03, + "negative_goodness": 1.7916974442164104, + "pair_accuracy": 0.8702222221904331, + "positive_goodness": 2.469469474750095, + "runtime_seconds": 11.630403757095337 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.1553620676252576, + "lr": 0.03, + "negative_goodness": 1.7864330818388197, + "pair_accuracy": 0.8642666667090522, + "positive_goodness": 2.481314300282796, + "runtime_seconds": 11.634033918380737 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.1619070245954726, + "lr": 0.03, + "negative_goodness": 1.7875835597568088, + "pair_accuracy": 0.8564444444497427, + "positive_goodness": 2.4678259824328954, + "runtime_seconds": 11.632471084594727 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.1670588163375855, + "lr": 0.03, + "negative_goodness": 1.7894921004825168, + "pair_accuracy": 0.852711111164093, + "positive_goodness": 2.4588375985039606, + "runtime_seconds": 11.631313562393188 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.1522335155698988, + "lr": 0.03, + "negative_goodness": 1.7806840529971653, + "pair_accuracy": 0.8617777777989706, + "positive_goodness": 2.4897556865268284, + "runtime_seconds": 11.63663935661316 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.153405696996053, + "lr": 0.03, + "negative_goodness": 1.77760267607371, + "pair_accuracy": 0.8664444444550409, + "positive_goodness": 2.483673891576131, + "runtime_seconds": 11.635491132736206 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.1688775838851928, + "lr": 0.03, + "negative_goodness": 1.7865968455844454, + "pair_accuracy": 0.8437555555449592, + "positive_goodness": 2.4552407937791614, + "runtime_seconds": 11.639838933944702 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.1512686656739977, + "lr": 0.03, + "negative_goodness": 1.7732483816358777, + "pair_accuracy": 0.862177777809567, + "positive_goodness": 2.4921748797098795, + "runtime_seconds": 11.650378227233887 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.140595461294386, + "lr": 0.03, + "negative_goodness": 1.7676722120285033, + "pair_accuracy": 0.8703777778201633, + "positive_goodness": 2.5145021452585854, + "runtime_seconds": 11.648691892623901 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.1509253253830805, + "lr": 0.03, + "negative_goodness": 1.7751500752978855, + "pair_accuracy": 0.8640888888570998, + "positive_goodness": 2.4923622044033475, + "runtime_seconds": 11.647879600524902 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.1450816348605686, + "lr": 0.03, + "negative_goodness": 1.7687745539135402, + "pair_accuracy": 0.8655333333545261, + "positive_goodness": 2.505045714272393, + "runtime_seconds": 11.650751113891602 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.1506344610638088, + "lr": 0.03, + "negative_goodness": 1.7733274820963543, + "pair_accuracy": 0.8662222221904331, + "positive_goodness": 2.498063173336453, + "runtime_seconds": 11.654037475585938 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.1537141801834105, + "lr": 0.03, + "negative_goodness": 1.767330745485094, + "pair_accuracy": 0.85384444448683, + "positive_goodness": 2.4871170323265925, + "runtime_seconds": 11.644893169403076 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.1407649997923108, + "lr": 0.03, + "negative_goodness": 1.7592718709309896, + "pair_accuracy": 0.8563555555237664, + "positive_goodness": 2.5183498206032646, + "runtime_seconds": 11.654298782348633 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.150628978093465, + "lr": 0.03, + "negative_goodness": 1.7694000733481514, + "pair_accuracy": 0.865177777809567, + "positive_goodness": 2.4955952262454564, + "runtime_seconds": 11.647338151931763 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.155311982568105, + "lr": 0.03, + "negative_goodness": 1.7679326303482055, + "pair_accuracy": 0.8567111110899184, + "positive_goodness": 2.486476365322537, + "runtime_seconds": 11.653135299682617 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.1231622149149576, + "lr": 0.03, + "negative_goodness": 1.751640970802307, + "pair_accuracy": 0.880355555566152, + "positive_goodness": 2.55521278690762, + "runtime_seconds": 11.647217750549316 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.1323195665783352, + "lr": 0.03, + "negative_goodness": 1.7516998313056098, + "pair_accuracy": 0.8678888888782925, + "positive_goodness": 2.541685751724243, + "runtime_seconds": 11.650207996368408 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.1406212789217631, + "lr": 0.03, + "negative_goodness": 1.755116664038764, + "pair_accuracy": 0.8613777777671814, + "positive_goodness": 2.5187380920410156, + "runtime_seconds": 11.6449134349823 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.1541338058259751, + "lr": 0.03, + "negative_goodness": 1.7651160417132907, + "pair_accuracy": 0.8551333333545261, + "positive_goodness": 2.4896445960998537, + "runtime_seconds": 11.63179612159729 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.1490392410914103, + "lr": 0.03, + "negative_goodness": 1.7605886145485772, + "pair_accuracy": 0.85544444448683, + "positive_goodness": 2.5015436003367104, + "runtime_seconds": 11.640091896057129 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.1550054531945122, + "lr": 0.03, + "negative_goodness": 1.7668716177198622, + "pair_accuracy": 0.8542666667090522, + "positive_goodness": 2.4880130575815835, + "runtime_seconds": 11.63948392868042 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.143092616589864, + "lr": 0.03, + "negative_goodness": 1.7522311424043444, + "pair_accuracy": 0.8588666666984558, + "positive_goodness": 2.515844226328532, + "runtime_seconds": 11.651279926300049 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.1281243333922493, + "lr": 0.03, + "negative_goodness": 1.7438618380652533, + "pair_accuracy": 0.8678222222222223, + "positive_goodness": 2.5548484908209907, + "runtime_seconds": 11.647342443466187 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.1336454075283475, + "lr": 0.03, + "negative_goodness": 1.7440049746407402, + "pair_accuracy": 0.8655999999682109, + "positive_goodness": 2.534890666241116, + "runtime_seconds": 11.651652574539185 + } + ], + "layer": 6 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.193708923933241, + "lr": 0.03, + "negative_goodness": 1.8106541033214993, + "pair_accuracy": 0.8111333333015441, + "positive_goodness": 2.3650558641009862, + "runtime_seconds": 11.578234910964966 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.0638215680440266, + "lr": 0.03, + "negative_goodness": 1.7233453978644477, + "pair_accuracy": 0.848711111108462, + "positive_goodness": 2.693716177813212, + "runtime_seconds": 11.563432693481445 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.0232853580050998, + "lr": 0.03, + "negative_goodness": 1.6599879977332221, + "pair_accuracy": 0.853177777785725, + "positive_goodness": 2.8395551266988117, + "runtime_seconds": 11.570576429367065 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0067333543141683, + "lr": 0.03, + "negative_goodness": 1.6444733361138237, + "pair_accuracy": 0.8414666666666667, + "positive_goodness": 2.865559132003784, + "runtime_seconds": 11.565989255905151 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.0417599753167894, + "lr": 0.03, + "negative_goodness": 1.6617913422054715, + "pair_accuracy": 0.8300444444444445, + "positive_goodness": 2.793796153089735, + "runtime_seconds": 11.559894561767578 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.0057484053505792, + "lr": 0.03, + "negative_goodness": 1.636474609014723, + "pair_accuracy": 0.8531333333121406, + "positive_goodness": 2.8901943898095026, + "runtime_seconds": 11.572447538375854 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.9935006383895874, + "lr": 0.03, + "negative_goodness": 1.6029610666486953, + "pair_accuracy": 0.8399777777777778, + "positive_goodness": 2.917819483311971, + "runtime_seconds": 11.565543174743652 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0588986869388157, + "lr": 0.03, + "negative_goodness": 1.6799850389268662, + "pair_accuracy": 0.8277777777777777, + "positive_goodness": 2.7682005643208822, + "runtime_seconds": 11.566135883331299 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.9779198942608304, + "lr": 0.03, + "negative_goodness": 1.5964112702475655, + "pair_accuracy": 0.8558888888888889, + "positive_goodness": 2.9650085633595786, + "runtime_seconds": 11.5684494972229 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.0030526588863797, + "lr": 0.03, + "negative_goodness": 1.625244273715549, + "pair_accuracy": 0.8526000000001656, + "positive_goodness": 2.917096414756775, + "runtime_seconds": 11.571086168289185 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.9862556213590834, + "lr": 0.03, + "negative_goodness": 1.6023293881734213, + "pair_accuracy": 0.8589111111429003, + "positive_goodness": 2.953614741388957, + "runtime_seconds": 11.57690691947937 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.9585694722387526, + "lr": 0.03, + "negative_goodness": 1.5739991123623318, + "pair_accuracy": 0.8665999999894036, + "positive_goodness": 3.0354659536573623, + "runtime_seconds": 11.572102785110474 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.0127696116765341, + "lr": 0.03, + "negative_goodness": 1.6235073079427083, + "pair_accuracy": 0.8515555555449592, + "positive_goodness": 2.8888150266859265, + "runtime_seconds": 11.570863008499146 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.967100605392456, + "lr": 0.03, + "negative_goodness": 1.5838099830203587, + "pair_accuracy": 0.8675777777777778, + "positive_goodness": 3.0391008648766413, + "runtime_seconds": 11.573936700820923 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.9884672917683919, + "lr": 0.03, + "negative_goodness": 1.5855790413114759, + "pair_accuracy": 0.8501333333333333, + "positive_goodness": 2.9624402432335746, + "runtime_seconds": 11.571308135986328 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.9441820137023926, + "lr": 0.03, + "negative_goodness": 1.5726106774012247, + "pair_accuracy": 0.8805555555449591, + "positive_goodness": 3.0805527334425182, + "runtime_seconds": 11.566744565963745 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.9422643315633138, + "lr": 0.03, + "negative_goodness": 1.554574377017551, + "pair_accuracy": 0.8677777777777778, + "positive_goodness": 3.0937586096021863, + "runtime_seconds": 11.56650185585022 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.9714892594761319, + "lr": 0.03, + "negative_goodness": 1.5747348880873786, + "pair_accuracy": 0.8614, + "positive_goodness": 3.005356151029799, + "runtime_seconds": 11.567492485046387 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.9531869136598375, + "lr": 0.03, + "negative_goodness": 1.558459246084425, + "pair_accuracy": 0.8646222222222222, + "positive_goodness": 3.0524219104766845, + "runtime_seconds": 11.56470012664795 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.9599510569466485, + "lr": 0.03, + "negative_goodness": 1.5595230391608343, + "pair_accuracy": 0.8500666666560703, + "positive_goodness": 3.06916668226454, + "runtime_seconds": 11.56752061843872 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.9879240493562487, + "lr": 0.03, + "negative_goodness": 1.593888588778178, + "pair_accuracy": 0.8585777778307597, + "positive_goodness": 2.9737403921339247, + "runtime_seconds": 11.562905311584473 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.9790169833925035, + "lr": 0.03, + "negative_goodness": 1.5751507808473375, + "pair_accuracy": 0.8551777777459887, + "positive_goodness": 3.0164108688778346, + "runtime_seconds": 11.56494140625 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.989998945151435, + "lr": 0.03, + "negative_goodness": 1.6011156029595268, + "pair_accuracy": 0.8581333333386315, + "positive_goodness": 2.9544246815787423, + "runtime_seconds": 11.572411060333252 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.9608355660332574, + "lr": 0.03, + "negative_goodness": 1.571561061859131, + "pair_accuracy": 0.8742000000211928, + "positive_goodness": 3.076666064707438, + "runtime_seconds": 11.562913656234741 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.0016085388183593, + "lr": 0.03, + "negative_goodness": 1.5987449869367811, + "pair_accuracy": 0.8460444444709354, + "positive_goodness": 2.957156473244561, + "runtime_seconds": 11.571542024612427 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.0047700121773613, + "lr": 0.03, + "negative_goodness": 1.5937798593309191, + "pair_accuracy": 0.8335999999788072, + "positive_goodness": 2.926841056611803, + "runtime_seconds": 11.559342384338379 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.9821756483289931, + "lr": 0.03, + "negative_goodness": 1.588885865826077, + "pair_accuracy": 0.8530666666668323, + "positive_goodness": 3.015258138296339, + "runtime_seconds": 11.549248456954956 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9736262245496115, + "lr": 0.03, + "negative_goodness": 1.5798177282121446, + "pair_accuracy": 0.8628888888782925, + "positive_goodness": 2.9795200074513755, + "runtime_seconds": 11.552801132202148 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.9665005274454752, + "lr": 0.03, + "negative_goodness": 1.5645531705644395, + "pair_accuracy": 0.8609777778042688, + "positive_goodness": 3.0491655886544122, + "runtime_seconds": 11.558199167251587 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9634738037321302, + "lr": 0.03, + "negative_goodness": 1.5632364145914714, + "pair_accuracy": 0.8576666666666667, + "positive_goodness": 3.050362683614095, + "runtime_seconds": 11.567423105239868 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.9823360467274984, + "lr": 0.03, + "negative_goodness": 1.5871771996816, + "pair_accuracy": 0.8691111111112767, + "positive_goodness": 3.0120031104829574, + "runtime_seconds": 11.569977045059204 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.9679386521233453, + "lr": 0.03, + "negative_goodness": 1.567070337846544, + "pair_accuracy": 0.8556444444464313, + "positive_goodness": 3.0237515992906356, + "runtime_seconds": 11.564167022705078 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9779824444452921, + "lr": 0.03, + "negative_goodness": 1.5751856358210246, + "pair_accuracy": 0.8552000000211928, + "positive_goodness": 3.0002514718797473, + "runtime_seconds": 11.56067967414856 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.0016394863976372, + "lr": 0.03, + "negative_goodness": 1.5976119382646348, + "pair_accuracy": 0.8445555555661519, + "positive_goodness": 2.9667820676167804, + "runtime_seconds": 11.553045511245728 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9779250389628941, + "lr": 0.03, + "negative_goodness": 1.585612048043145, + "pair_accuracy": 0.8632222222222222, + "positive_goodness": 2.9753170187208386, + "runtime_seconds": 11.557290077209473 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9621538004345364, + "lr": 0.03, + "negative_goodness": 1.5585800388336182, + "pair_accuracy": 0.8648222222328186, + "positive_goodness": 3.068522462251451, + "runtime_seconds": 11.56400465965271 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9708220187717014, + "lr": 0.03, + "negative_goodness": 1.5743880742814806, + "pair_accuracy": 0.8726, + "positive_goodness": 3.0283388798183863, + "runtime_seconds": 11.559566736221313 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.9609143774244521, + "lr": 0.03, + "negative_goodness": 1.5623675180223253, + "pair_accuracy": 0.8712, + "positive_goodness": 3.0584021805233426, + "runtime_seconds": 11.565445184707642 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.9456150897767809, + "lr": 0.03, + "negative_goodness": 1.5447191306432089, + "pair_accuracy": 0.8699777777777777, + "positive_goodness": 3.083082226774428, + "runtime_seconds": 11.56197452545166 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.9991903157975939, + "lr": 0.03, + "negative_goodness": 1.5931006571875679, + "pair_accuracy": 0.8445777777989706, + "positive_goodness": 2.981040947341919, + "runtime_seconds": 11.557490587234497 + } + ], + "layer": 7 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1624050590727064, + "lr": 0.03, + "negative_goodness": 1.8071102701399062, + "pair_accuracy": 0.8214000000423856, + "positive_goodness": 2.655324779849582, + "runtime_seconds": 11.598166704177856 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.0629239166259765, + "lr": 0.03, + "negative_goodness": 1.6357996010674372, + "pair_accuracy": 0.8666, + "positive_goodness": 2.817196292241414, + "runtime_seconds": 11.58745265007019 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.0363019378662108, + "lr": 0.03, + "negative_goodness": 1.5878252516004774, + "pair_accuracy": 0.8704444444656372, + "positive_goodness": 2.8713181153615315, + "runtime_seconds": 11.586557865142822 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0552016740904915, + "lr": 0.03, + "negative_goodness": 1.5901202859560648, + "pair_accuracy": 0.862644444433848, + "positive_goodness": 2.8061738337198894, + "runtime_seconds": 11.575703382492065 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.0504920744366115, + "lr": 0.03, + "negative_goodness": 1.5821493306054009, + "pair_accuracy": 0.8666888889206781, + "positive_goodness": 2.816741194576687, + "runtime_seconds": 11.573366165161133 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.0386215751859877, + "lr": 0.03, + "negative_goodness": 1.5553311055925156, + "pair_accuracy": 0.8686222222116259, + "positive_goodness": 2.8537127347310385, + "runtime_seconds": 11.564469814300537 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.0253396486388313, + "lr": 0.03, + "negative_goodness": 1.5309403247197468, + "pair_accuracy": 0.8680222221904331, + "positive_goodness": 2.882759758928087, + "runtime_seconds": 11.557538509368896 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0303579194598729, + "lr": 0.03, + "negative_goodness": 1.5341225226084392, + "pair_accuracy": 0.8652222222434149, + "positive_goodness": 2.868284925672743, + "runtime_seconds": 11.574856519699097 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.0322211239496868, + "lr": 0.03, + "negative_goodness": 1.5345511588732401, + "pair_accuracy": 0.8675555555343628, + "positive_goodness": 2.856819558885362, + "runtime_seconds": 11.56475305557251 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.9952091615147061, + "lr": 0.03, + "negative_goodness": 1.5007011742273966, + "pair_accuracy": 0.8832444444868299, + "positive_goodness": 2.9690300074683296, + "runtime_seconds": 11.569388151168823 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.9917958272298177, + "lr": 0.03, + "negative_goodness": 1.4761433100382486, + "pair_accuracy": 0.8778444444444444, + "positive_goodness": 2.987101358795166, + "runtime_seconds": 11.571244716644287 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.039844977982839, + "lr": 0.03, + "negative_goodness": 1.5173656932195028, + "pair_accuracy": 0.8602888888782925, + "positive_goodness": 2.853926255544027, + "runtime_seconds": 11.573226690292358 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.0180194044854907, + "lr": 0.03, + "negative_goodness": 1.5093747937096489, + "pair_accuracy": 0.8664, + "positive_goodness": 2.8982726995256214, + "runtime_seconds": 11.569424390792847 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.011065776930915, + "lr": 0.03, + "negative_goodness": 1.4920133019765218, + "pair_accuracy": 0.8685777777459887, + "positive_goodness": 2.935240157784356, + "runtime_seconds": 11.56815481185913 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.0105102202945286, + "lr": 0.03, + "negative_goodness": 1.498405114724901, + "pair_accuracy": 0.8750666666348775, + "positive_goodness": 2.9148550056033664, + "runtime_seconds": 11.56531810760498 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.028764500448439, + "lr": 0.03, + "negative_goodness": 1.497916492038303, + "pair_accuracy": 0.8588222222010294, + "positive_goodness": 2.8848984497070314, + "runtime_seconds": 11.55925989151001 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.0003451054467096, + "lr": 0.03, + "negative_goodness": 1.4748300237549676, + "pair_accuracy": 0.8726888888782925, + "positive_goodness": 2.965740839513143, + "runtime_seconds": 11.55771279335022 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.0174301494174534, + "lr": 0.03, + "negative_goodness": 1.4962157827801175, + "pair_accuracy": 0.8704, + "positive_goodness": 2.9021022871229385, + "runtime_seconds": 11.561007499694824 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.0022147780736288, + "lr": 0.03, + "negative_goodness": 1.466903915585412, + "pair_accuracy": 0.8721333333227369, + "positive_goodness": 2.9632861269632977, + "runtime_seconds": 11.561740636825562 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.0017209449768067, + "lr": 0.03, + "negative_goodness": 1.4662789082845051, + "pair_accuracy": 0.8718222222752041, + "positive_goodness": 2.964099035771688, + "runtime_seconds": 11.563477993011475 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.0031634598202175, + "lr": 0.03, + "negative_goodness": 1.480427396244473, + "pair_accuracy": 0.8725111111429003, + "positive_goodness": 2.948700109948052, + "runtime_seconds": 11.548043489456177 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.9904079232957628, + "lr": 0.03, + "negative_goodness": 1.4524435780419245, + "pair_accuracy": 0.8762444444232517, + "positive_goodness": 3.0080146090613473, + "runtime_seconds": 11.549242734909058 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.9848939821031358, + "lr": 0.03, + "negative_goodness": 1.4544583805084228, + "pair_accuracy": 0.8799555555237664, + "positive_goodness": 3.008424222395155, + "runtime_seconds": 11.543025255203247 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.0136781842761569, + "lr": 0.03, + "negative_goodness": 1.4835673369513618, + "pair_accuracy": 0.8705111111534967, + "positive_goodness": 2.930361731677585, + "runtime_seconds": 11.541553258895874 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.0013109457651774, + "lr": 0.03, + "negative_goodness": 1.4673859877056545, + "pair_accuracy": 0.8748666667090522, + "positive_goodness": 2.957102272966173, + "runtime_seconds": 11.538407564163208 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9949440238740709, + "lr": 0.03, + "negative_goodness": 1.460557360511356, + "pair_accuracy": 0.8784888888676962, + "positive_goodness": 2.9625886800554064, + "runtime_seconds": 11.55015230178833 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.0037332155015735, + "lr": 0.03, + "negative_goodness": 1.4624052653630575, + "pair_accuracy": 0.8649555555767483, + "positive_goodness": 2.965547286817763, + "runtime_seconds": 11.549099683761597 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9747168774922689, + "lr": 0.03, + "negative_goodness": 1.447528918202718, + "pair_accuracy": 0.8853777777777778, + "positive_goodness": 3.03338418252733, + "runtime_seconds": 11.543556213378906 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.9968160237842136, + "lr": 0.03, + "negative_goodness": 1.4604224563598633, + "pair_accuracy": 0.8732222222646078, + "positive_goodness": 2.9775408834669324, + "runtime_seconds": 11.54002046585083 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9909024478700426, + "lr": 0.03, + "negative_goodness": 1.4434840935389202, + "pair_accuracy": 0.8768444444126553, + "positive_goodness": 2.996407509909736, + "runtime_seconds": 11.538000583648682 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.995808433787028, + "lr": 0.03, + "negative_goodness": 1.4555070829603407, + "pair_accuracy": 0.8726222222328186, + "positive_goodness": 2.974818309995863, + "runtime_seconds": 11.53914213180542 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.9698203016704983, + "lr": 0.03, + "negative_goodness": 1.4334495524724324, + "pair_accuracy": 0.8817111111534967, + "positive_goodness": 3.0549879611968995, + "runtime_seconds": 11.537705659866333 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9846197265836928, + "lr": 0.03, + "negative_goodness": 1.4376219977908664, + "pair_accuracy": 0.8766222222010295, + "positive_goodness": 3.027612776268853, + "runtime_seconds": 11.54435420036316 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.0217437257342867, + "lr": 0.03, + "negative_goodness": 1.4807908700095282, + "pair_accuracy": 0.8633555555343628, + "positive_goodness": 2.898221781200833, + "runtime_seconds": 11.543360471725464 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.970083560233646, + "lr": 0.03, + "negative_goodness": 1.435541660255856, + "pair_accuracy": 0.8848222222010295, + "positive_goodness": 3.051855604807536, + "runtime_seconds": 11.548787355422974 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9876901002671984, + "lr": 0.03, + "negative_goodness": 1.4438267154163784, + "pair_accuracy": 0.8764666666666666, + "positive_goodness": 3.0055871031019423, + "runtime_seconds": 11.54691481590271 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9870826393445333, + "lr": 0.03, + "negative_goodness": 1.444716289138794, + "pair_accuracy": 0.8794222222116258, + "positive_goodness": 3.0086579487270777, + "runtime_seconds": 11.538259506225586 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.9842266766442193, + "lr": 0.03, + "negative_goodness": 1.439598520681593, + "pair_accuracy": 0.8746888888676961, + "positive_goodness": 3.0224814715067545, + "runtime_seconds": 11.546049356460571 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.980545025338067, + "lr": 0.03, + "negative_goodness": 1.4359055313534206, + "pair_accuracy": 0.8818000000529819, + "positive_goodness": 3.0254759242163765, + "runtime_seconds": 11.547362565994263 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.9949075287077161, + "lr": 0.03, + "negative_goodness": 1.4519662930806478, + "pair_accuracy": 0.8751777777777778, + "positive_goodness": 2.9924401444753013, + "runtime_seconds": 11.546504735946655 + } + ], + "layer": 8 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2810424160427518, + "lr": 0.03, + "negative_goodness": 1.802178020837572, + "pair_accuracy": 0.7330666666666666, + "positive_goodness": 2.2089679333580863, + "runtime_seconds": 11.46892237663269 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.2350619206746418, + "lr": 0.03, + "negative_goodness": 1.800489911609226, + "pair_accuracy": 0.7762888888994852, + "positive_goodness": 2.3197671134524875, + "runtime_seconds": 11.449081182479858 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.1918770500183105, + "lr": 0.03, + "negative_goodness": 1.7551822923872207, + "pair_accuracy": 0.805733333322737, + "positive_goodness": 2.412999385664198, + "runtime_seconds": 11.452248334884644 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.195995113796658, + "lr": 0.03, + "negative_goodness": 1.7533206011030409, + "pair_accuracy": 0.8047777777618832, + "positive_goodness": 2.4064073320388792, + "runtime_seconds": 11.448771715164185 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.1977013872994318, + "lr": 0.03, + "negative_goodness": 1.7538935129801432, + "pair_accuracy": 0.802711111164093, + "positive_goodness": 2.4096489748213026, + "runtime_seconds": 11.446816682815552 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.1811967791451348, + "lr": 0.03, + "negative_goodness": 1.7293699423048232, + "pair_accuracy": 0.8111333333545261, + "positive_goodness": 2.4435169648064505, + "runtime_seconds": 11.452430963516235 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.1790878586239284, + "lr": 0.03, + "negative_goodness": 1.7312912804073757, + "pair_accuracy": 0.812133333322737, + "positive_goodness": 2.4459274131351045, + "runtime_seconds": 11.44127631187439 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.176241363504198, + "lr": 0.03, + "negative_goodness": 1.72111992670695, + "pair_accuracy": 0.8104222222328186, + "positive_goodness": 2.4576332804785834, + "runtime_seconds": 11.445200443267822 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.171912247742547, + "lr": 0.03, + "negative_goodness": 1.719698671743605, + "pair_accuracy": 0.8106000000317891, + "positive_goodness": 2.4678126288943822, + "runtime_seconds": 11.446125984191895 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.1697931928422716, + "lr": 0.03, + "negative_goodness": 1.7140349965837267, + "pair_accuracy": 0.8142444444126553, + "positive_goodness": 2.4738440177493626, + "runtime_seconds": 11.442464351654053 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.1639880314932929, + "lr": 0.03, + "negative_goodness": 1.707286149851481, + "pair_accuracy": 0.8147333333863153, + "positive_goodness": 2.491516338517931, + "runtime_seconds": 11.443772554397583 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.1682570800145466, + "lr": 0.03, + "negative_goodness": 1.7109500538508098, + "pair_accuracy": 0.8109333333015442, + "positive_goodness": 2.479049278216892, + "runtime_seconds": 11.438893795013428 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.158277620930142, + "lr": 0.03, + "negative_goodness": 1.7035367208268908, + "pair_accuracy": 0.8190888888676962, + "positive_goodness": 2.5051479439205595, + "runtime_seconds": 11.43858528137207 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.153733840115865, + "lr": 0.03, + "negative_goodness": 1.6892236789279513, + "pair_accuracy": 0.8217999999682108, + "positive_goodness": 2.5151157393137615, + "runtime_seconds": 11.440085411071777 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.1513904504352146, + "lr": 0.03, + "negative_goodness": 1.6906933756934273, + "pair_accuracy": 0.8217555555979411, + "positive_goodness": 2.5319919010162355, + "runtime_seconds": 11.448254346847534 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.1535588036855062, + "lr": 0.03, + "negative_goodness": 1.6881935676998563, + "pair_accuracy": 0.8189777777671814, + "positive_goodness": 2.516611200544569, + "runtime_seconds": 11.442718982696533 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.1491351672490437, + "lr": 0.03, + "negative_goodness": 1.685635793685913, + "pair_accuracy": 0.8221555555767484, + "positive_goodness": 2.530864345889621, + "runtime_seconds": 11.439991474151611 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.1529820975197687, + "lr": 0.03, + "negative_goodness": 1.6918790591133965, + "pair_accuracy": 0.8192888888782925, + "positive_goodness": 2.5204783033159046, + "runtime_seconds": 11.440810203552246 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.1468147493362426, + "lr": 0.03, + "negative_goodness": 1.680855159166124, + "pair_accuracy": 0.8232888889312744, + "positive_goodness": 2.5330645046657985, + "runtime_seconds": 11.444422245025635 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.1376396603902181, + "lr": 0.03, + "negative_goodness": 1.6703671563042535, + "pair_accuracy": 0.826377777756585, + "positive_goodness": 2.5668101516299777, + "runtime_seconds": 11.447741031646729 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.1440726331922744, + "lr": 0.03, + "negative_goodness": 1.6744962484571668, + "pair_accuracy": 0.8226222222116258, + "positive_goodness": 2.5453936746385364, + "runtime_seconds": 11.452205657958984 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.129628098350101, + "lr": 0.03, + "negative_goodness": 1.6670555954403348, + "pair_accuracy": 0.8305333333651225, + "positive_goodness": 2.583802520328098, + "runtime_seconds": 11.458752870559692 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.118846660815345, + "lr": 0.03, + "negative_goodness": 1.6437995262145997, + "pair_accuracy": 0.8386888889312745, + "positive_goodness": 2.6126211353302002, + "runtime_seconds": 11.451538324356079 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.136883087963528, + "lr": 0.03, + "negative_goodness": 1.6610396134694416, + "pair_accuracy": 0.8276666666560702, + "positive_goodness": 2.5688784614562987, + "runtime_seconds": 11.45877194404602 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.1423543485005696, + "lr": 0.03, + "negative_goodness": 1.6723464031219482, + "pair_accuracy": 0.8225555555449592, + "positive_goodness": 2.555529827457004, + "runtime_seconds": 11.457077741622925 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.1361837560865615, + "lr": 0.03, + "negative_goodness": 1.663235276667277, + "pair_accuracy": 0.8265333333121405, + "positive_goodness": 2.5756415714263916, + "runtime_seconds": 11.453002452850342 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.129178019258711, + "lr": 0.03, + "negative_goodness": 1.6576433699925741, + "pair_accuracy": 0.8287111110793219, + "positive_goodness": 2.5982341188642715, + "runtime_seconds": 11.454869985580444 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.1362602837244669, + "lr": 0.03, + "negative_goodness": 1.6666065088272095, + "pair_accuracy": 0.8230888888676962, + "positive_goodness": 2.567364109463162, + "runtime_seconds": 11.454508543014526 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.1289022867414686, + "lr": 0.03, + "negative_goodness": 1.6530009918424817, + "pair_accuracy": 0.8249111111111112, + "positive_goodness": 2.595009131749471, + "runtime_seconds": 11.450302839279175 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.1459694388283623, + "lr": 0.03, + "negative_goodness": 1.6723779571957058, + "pair_accuracy": 0.8193777777989706, + "positive_goodness": 2.554299385113186, + "runtime_seconds": 11.457003593444824 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.1413544043858845, + "lr": 0.03, + "negative_goodness": 1.660772195900811, + "pair_accuracy": 0.8193999999894036, + "positive_goodness": 2.5553354953765868, + "runtime_seconds": 11.452372074127197 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.1231734962675306, + "lr": 0.03, + "negative_goodness": 1.6484852498796252, + "pair_accuracy": 0.8328222222646078, + "positive_goodness": 2.603844694222344, + "runtime_seconds": 11.451504468917847 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.120835010125902, + "lr": 0.03, + "negative_goodness": 1.6438771489461264, + "pair_accuracy": 0.8301555555343628, + "positive_goodness": 2.622349441401164, + "runtime_seconds": 11.44959306716919 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.120575396876865, + "lr": 0.03, + "negative_goodness": 1.6458346774631076, + "pair_accuracy": 0.8324666666348776, + "positive_goodness": 2.609006208674113, + "runtime_seconds": 11.453397750854492 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.131725577905443, + "lr": 0.03, + "negative_goodness": 1.6563217774497139, + "pair_accuracy": 0.8281999999894036, + "positive_goodness": 2.5826770040724014, + "runtime_seconds": 11.45354676246643 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.1212047446250915, + "lr": 0.03, + "negative_goodness": 1.6368803334765964, + "pair_accuracy": 0.8329333333757188, + "positive_goodness": 2.6214402926974825, + "runtime_seconds": 11.455111503601074 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.1331447929594252, + "lr": 0.03, + "negative_goodness": 1.6545736312442356, + "pair_accuracy": 0.826955555566152, + "positive_goodness": 2.586926565170288, + "runtime_seconds": 11.44940996170044 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.1293285662121242, + "lr": 0.03, + "negative_goodness": 1.6496625779893663, + "pair_accuracy": 0.827600000052982, + "positive_goodness": 2.5900656791263157, + "runtime_seconds": 11.453441619873047 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.1123874522950914, + "lr": 0.03, + "negative_goodness": 1.6318904026667278, + "pair_accuracy": 0.8390444444126554, + "positive_goodness": 2.645440456559923, + "runtime_seconds": 11.447323083877563 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.1248836096233792, + "lr": 0.03, + "negative_goodness": 1.639298873647054, + "pair_accuracy": 0.8311777777989705, + "positive_goodness": 2.6058933215671116, + "runtime_seconds": 11.43995976448059 + } + ], + "layer": 9 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0197470476044548, + "lr": 0.03, + "negative_goodness": 1.6526911909739177, + "pair_accuracy": 0.847644444433848, + "positive_goodness": 2.951335395050049, + "runtime_seconds": 11.542145013809204 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7907591996086968, + "lr": 0.03, + "negative_goodness": 1.3622017312261794, + "pair_accuracy": 0.9112444444444444, + "positive_goodness": 3.6709078497568766, + "runtime_seconds": 11.546712160110474 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.6900565280490452, + "lr": 0.03, + "negative_goodness": 1.186734651798672, + "pair_accuracy": 0.9222888889206781, + "positive_goodness": 4.092241453086006, + "runtime_seconds": 11.539007186889648 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6624583934148153, + "lr": 0.03, + "negative_goodness": 1.1085671440972222, + "pair_accuracy": 0.9295333333333333, + "positive_goodness": 4.444590751690335, + "runtime_seconds": 11.545347452163696 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7045376744588216, + "lr": 0.03, + "negative_goodness": 1.123039866399765, + "pair_accuracy": 0.8970888888888889, + "positive_goodness": 4.350357008828057, + "runtime_seconds": 11.54257869720459 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6816732353369395, + "lr": 0.03, + "negative_goodness": 1.0913058404975466, + "pair_accuracy": 0.9028, + "positive_goodness": 4.50996727027893, + "runtime_seconds": 11.544751644134521 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.662585324690077, + "lr": 0.03, + "negative_goodness": 1.0427874615987143, + "pair_accuracy": 0.915111111164093, + "positive_goodness": 4.7954829383850095, + "runtime_seconds": 11.540824890136719 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7140578371895684, + "lr": 0.03, + "negative_goodness": 1.090511099073622, + "pair_accuracy": 0.8993333333015442, + "positive_goodness": 4.901268071683248, + "runtime_seconds": 11.548352479934692 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6068567491849264, + "lr": 0.03, + "negative_goodness": 0.9661160415437486, + "pair_accuracy": 0.9287333333333333, + "positive_goodness": 5.15535404985216, + "runtime_seconds": 11.540897607803345 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6580246758672926, + "lr": 0.03, + "negative_goodness": 0.9854307993782891, + "pair_accuracy": 0.9078666666666667, + "positive_goodness": 4.996160660468208, + "runtime_seconds": 11.549391031265259 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6577051205237706, + "lr": 0.03, + "negative_goodness": 1.0049718107276493, + "pair_accuracy": 0.9204222222222223, + "positive_goodness": 5.127144548543295, + "runtime_seconds": 11.556092500686646 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6702429601987203, + "lr": 0.03, + "negative_goodness": 1.0247871730274625, + "pair_accuracy": 0.9098222222222222, + "positive_goodness": 5.014241579606798, + "runtime_seconds": 11.550369262695312 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6464180569701724, + "lr": 0.03, + "negative_goodness": 0.9823185089429219, + "pair_accuracy": 0.9215777777777778, + "positive_goodness": 5.0761317892286515, + "runtime_seconds": 11.545844554901123 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6934963181177775, + "lr": 0.03, + "negative_goodness": 0.9848393710268868, + "pair_accuracy": 0.8901555555555556, + "positive_goodness": 4.932096115790473, + "runtime_seconds": 11.541527271270752 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6492117781347698, + "lr": 0.03, + "negative_goodness": 0.9524534037325117, + "pair_accuracy": 0.9084, + "positive_goodness": 5.232229871453179, + "runtime_seconds": 11.540297508239746 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6205342916276719, + "lr": 0.03, + "negative_goodness": 0.941739904499054, + "pair_accuracy": 0.9239555555873447, + "positive_goodness": 5.723918665631612, + "runtime_seconds": 11.544211149215698 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6852208897908528, + "lr": 0.03, + "negative_goodness": 1.0010174025853475, + "pair_accuracy": 0.8980222222116259, + "positive_goodness": 5.182245096164279, + "runtime_seconds": 11.544406414031982 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6388488392088149, + "lr": 0.03, + "negative_goodness": 0.9354392488585578, + "pair_accuracy": 0.9185333333757189, + "positive_goodness": 5.438502935557895, + "runtime_seconds": 11.54707670211792 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6191577568901909, + "lr": 0.03, + "negative_goodness": 0.8940483794318305, + "pair_accuracy": 0.9144444444656372, + "positive_goodness": 5.333071939510769, + "runtime_seconds": 11.548838138580322 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6374720425181919, + "lr": 0.03, + "negative_goodness": 0.9527080087237888, + "pair_accuracy": 0.921, + "positive_goodness": 5.2867589215596515, + "runtime_seconds": 11.546690464019775 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6304635513623555, + "lr": 0.03, + "negative_goodness": 0.9147475657780966, + "pair_accuracy": 0.9188666666666667, + "positive_goodness": 5.348030254957411, + "runtime_seconds": 11.540030479431152 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6009789626200994, + "lr": 0.03, + "negative_goodness": 0.8905735390265783, + "pair_accuracy": 0.9219111111111111, + "positive_goodness": 5.76654562081231, + "runtime_seconds": 11.548795700073242 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.637808436732822, + "lr": 0.03, + "negative_goodness": 0.9180165583716499, + "pair_accuracy": 0.9139111111323038, + "positive_goodness": 5.5876707636727225, + "runtime_seconds": 11.552531719207764 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6646475955645244, + "lr": 0.03, + "negative_goodness": 0.9482181676016913, + "pair_accuracy": 0.8996444444444445, + "positive_goodness": 5.3838495855543345, + "runtime_seconds": 11.54880142211914 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6471510716014438, + "lr": 0.03, + "negative_goodness": 0.9467136119630601, + "pair_accuracy": 0.9124444444444444, + "positive_goodness": 5.511436414591471, + "runtime_seconds": 11.544897317886353 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6659424202813042, + "lr": 0.03, + "negative_goodness": 0.9556736569298638, + "pair_accuracy": 0.9001999999894036, + "positive_goodness": 5.8521287917243106, + "runtime_seconds": 11.544965982437134 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6273209865146213, + "lr": 0.03, + "negative_goodness": 0.917988824780782, + "pair_accuracy": 0.9129333333333334, + "positive_goodness": 5.399079176266988, + "runtime_seconds": 11.546667337417603 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6245543173472087, + "lr": 0.03, + "negative_goodness": 0.9247534974416097, + "pair_accuracy": 0.9281333333333334, + "positive_goodness": 5.799899732462565, + "runtime_seconds": 11.543205976486206 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6154186265945435, + "lr": 0.03, + "negative_goodness": 0.8889251074155172, + "pair_accuracy": 0.9105111111111112, + "positive_goodness": 5.677125831858317, + "runtime_seconds": 11.5482759475708 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6431973449336158, + "lr": 0.03, + "negative_goodness": 0.9009304652531942, + "pair_accuracy": 0.9076444444444445, + "positive_goodness": 5.518554148186578, + "runtime_seconds": 11.550745248794556 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6259825389226278, + "lr": 0.03, + "negative_goodness": 0.9247956595314873, + "pair_accuracy": 0.9272666666666667, + "positive_goodness": 5.851179406907823, + "runtime_seconds": 11.542673587799072 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6367674411667718, + "lr": 0.03, + "negative_goodness": 0.9143038021511501, + "pair_accuracy": 0.9151555555343628, + "positive_goodness": 5.499129385036892, + "runtime_seconds": 11.554342269897461 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.654088089413113, + "lr": 0.03, + "negative_goodness": 0.9111127433776856, + "pair_accuracy": 0.9102444444447756, + "positive_goodness": 5.704829903051588, + "runtime_seconds": 11.558852910995483 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6400204045613607, + "lr": 0.03, + "negative_goodness": 0.8942479709201389, + "pair_accuracy": 0.9070444444444444, + "positive_goodness": 5.250745907423231, + "runtime_seconds": 11.560565948486328 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6269667870415582, + "lr": 0.03, + "negative_goodness": 0.8882603498458862, + "pair_accuracy": 0.9103555555767483, + "positive_goodness": 5.647836947504679, + "runtime_seconds": 11.550877094268799 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6594994717703925, + "lr": 0.03, + "negative_goodness": 0.917987692811754, + "pair_accuracy": 0.9032222222222223, + "positive_goodness": 5.656359895155165, + "runtime_seconds": 11.551506519317627 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6464732962926228, + "lr": 0.03, + "negative_goodness": 0.9117061728795369, + "pair_accuracy": 0.9037555555237664, + "positive_goodness": 5.554154005813599, + "runtime_seconds": 11.549391746520996 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6122911452770233, + "lr": 0.03, + "negative_goodness": 0.8720993827237024, + "pair_accuracy": 0.9242444444444444, + "positive_goodness": 5.73562278857761, + "runtime_seconds": 11.54864501953125 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6550111597379049, + "lr": 0.03, + "negative_goodness": 0.9351782784779866, + "pair_accuracy": 0.9081555555555556, + "positive_goodness": 5.339910694461398, + "runtime_seconds": 11.551912784576416 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6805609974013435, + "lr": 0.03, + "negative_goodness": 0.9338013343175252, + "pair_accuracy": 0.8934888888782925, + "positive_goodness": 5.71300556725396, + "runtime_seconds": 11.550026655197144 + } + ], + "layer": 10 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2111847195095486, + "lr": 0.03, + "negative_goodness": 1.7805507250044081, + "pair_accuracy": 0.8187333333492279, + "positive_goodness": 2.4150893653445773, + "runtime_seconds": 11.493102073669434 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.1470107851664226, + "lr": 0.03, + "negative_goodness": 1.7760346396976048, + "pair_accuracy": 0.8523333333863152, + "positive_goodness": 2.570447665447659, + "runtime_seconds": 11.481266736984253 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.1411677047305637, + "lr": 0.03, + "negative_goodness": 1.7542013695822822, + "pair_accuracy": 0.8479111110793219, + "positive_goodness": 2.604704242706299, + "runtime_seconds": 11.477240324020386 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.1318981454637316, + "lr": 0.03, + "negative_goodness": 1.7480286489698622, + "pair_accuracy": 0.8531777777936723, + "positive_goodness": 2.623071604368422, + "runtime_seconds": 11.47891116142273 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.1347760075887043, + "lr": 0.03, + "negative_goodness": 1.747025266096327, + "pair_accuracy": 0.8509333333333333, + "positive_goodness": 2.610262600241767, + "runtime_seconds": 11.474208116531372 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.139761929427253, + "lr": 0.03, + "negative_goodness": 1.764079036670261, + "pair_accuracy": 0.8462666666719648, + "positive_goodness": 2.5945838677724202, + "runtime_seconds": 11.484982252120972 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.1245643644650778, + "lr": 0.03, + "negative_goodness": 1.7378508670594957, + "pair_accuracy": 0.8535999999682109, + "positive_goodness": 2.6366177987416584, + "runtime_seconds": 11.479062795639038 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.1505384425057306, + "lr": 0.03, + "negative_goodness": 1.7634596385320027, + "pair_accuracy": 0.8369111110899183, + "positive_goodness": 2.5739212081485325, + "runtime_seconds": 11.472409963607788 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.1314235587014092, + "lr": 0.03, + "negative_goodness": 1.7503205571068658, + "pair_accuracy": 0.8502444444550408, + "positive_goodness": 2.627887469143338, + "runtime_seconds": 11.476337194442749 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.1353526366551716, + "lr": 0.03, + "negative_goodness": 1.7411861402299669, + "pair_accuracy": 0.8436888889100816, + "positive_goodness": 2.6166478427039253, + "runtime_seconds": 11.474050283432007 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.146920807202657, + "lr": 0.03, + "negative_goodness": 1.759958772828844, + "pair_accuracy": 0.8427333333121406, + "positive_goodness": 2.58356425904168, + "runtime_seconds": 11.468706846237183 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.1298980909135607, + "lr": 0.03, + "negative_goodness": 1.7453139710532295, + "pair_accuracy": 0.8465333333651225, + "positive_goodness": 2.633159346347385, + "runtime_seconds": 11.466829061508179 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.1251611360973781, + "lr": 0.03, + "negative_goodness": 1.7346969876183405, + "pair_accuracy": 0.8562888889312744, + "positive_goodness": 2.6565152744293212, + "runtime_seconds": 11.475943088531494 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.1286665219836765, + "lr": 0.03, + "negative_goodness": 1.743789956325955, + "pair_accuracy": 0.8478888888676961, + "positive_goodness": 2.622329004796346, + "runtime_seconds": 11.474309921264648 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.112904544152154, + "lr": 0.03, + "negative_goodness": 1.7235551546096801, + "pair_accuracy": 0.8581333333757188, + "positive_goodness": 2.699226770231459, + "runtime_seconds": 11.478260278701782 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.1275536583794488, + "lr": 0.03, + "negative_goodness": 1.734653325653076, + "pair_accuracy": 0.8523777777777778, + "positive_goodness": 2.6365874323103164, + "runtime_seconds": 11.472405910491943 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.1227410986582438, + "lr": 0.03, + "negative_goodness": 1.7343719259474013, + "pair_accuracy": 0.8482888888782925, + "positive_goodness": 2.6537717417822946, + "runtime_seconds": 11.478098630905151 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.1464650655640496, + "lr": 0.03, + "negative_goodness": 1.7630697198020087, + "pair_accuracy": 0.8402888888676961, + "positive_goodness": 2.5824177754296196, + "runtime_seconds": 11.47934079170227 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.111057382382287, + "lr": 0.03, + "negative_goodness": 1.720278442319234, + "pair_accuracy": 0.8605555555343628, + "positive_goodness": 2.6738229498121475, + "runtime_seconds": 11.472656726837158 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.1271803352567884, + "lr": 0.03, + "negative_goodness": 1.7342051865047878, + "pair_accuracy": 0.8418888889100816, + "positive_goodness": 2.6485416014353436, + "runtime_seconds": 11.473180532455444 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.1258867945988973, + "lr": 0.03, + "negative_goodness": 1.7378132394154866, + "pair_accuracy": 0.8485555555873447, + "positive_goodness": 2.636404772567749, + "runtime_seconds": 11.46953821182251 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.0941530153486463, + "lr": 0.03, + "negative_goodness": 1.7002550345950656, + "pair_accuracy": 0.861911111079322, + "positive_goodness": 2.7547994374169242, + "runtime_seconds": 11.475491285324097 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.1096769365946453, + "lr": 0.03, + "negative_goodness": 1.7158749684863621, + "pair_accuracy": 0.8588666666984558, + "positive_goodness": 2.7105133900536433, + "runtime_seconds": 11.476313591003418 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.1156801295598349, + "lr": 0.03, + "negative_goodness": 1.7280204755995008, + "pair_accuracy": 0.8566000000158945, + "positive_goodness": 2.667748974778917, + "runtime_seconds": 11.472839593887329 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.1373390385309856, + "lr": 0.03, + "negative_goodness": 1.7505225822872585, + "pair_accuracy": 0.8373555555979411, + "positive_goodness": 2.6083269564734564, + "runtime_seconds": 11.467719793319702 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.113638251262241, + "lr": 0.03, + "negative_goodness": 1.7335415981928508, + "pair_accuracy": 0.8614888888994853, + "positive_goodness": 2.672837751812405, + "runtime_seconds": 11.457533836364746 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.14028365427653, + "lr": 0.03, + "negative_goodness": 1.751970525656806, + "pair_accuracy": 0.8428888889418708, + "positive_goodness": 2.6018163433498804, + "runtime_seconds": 11.458572149276733 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.1191765787124635, + "lr": 0.03, + "negative_goodness": 1.7386669406043158, + "pair_accuracy": 0.8546222222646077, + "positive_goodness": 2.656929662238227, + "runtime_seconds": 11.464133977890015 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.1087499629762438, + "lr": 0.03, + "negative_goodness": 1.7135384314643012, + "pair_accuracy": 0.8551777777671814, + "positive_goodness": 2.698761953269111, + "runtime_seconds": 11.468998193740845 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.1387642513275147, + "lr": 0.03, + "negative_goodness": 1.741636189757453, + "pair_accuracy": 0.8383999999894036, + "positive_goodness": 2.6132325296613903, + "runtime_seconds": 11.474290370941162 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.1418560830858018, + "lr": 0.03, + "negative_goodness": 1.7599976495530871, + "pair_accuracy": 0.8410222222222222, + "positive_goodness": 2.59442695757548, + "runtime_seconds": 11.480457305908203 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.1352289052115547, + "lr": 0.03, + "negative_goodness": 1.7457155104955038, + "pair_accuracy": 0.8372444444974263, + "positive_goodness": 2.6203082054985893, + "runtime_seconds": 11.473753213882446 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.12519342534807, + "lr": 0.03, + "negative_goodness": 1.7426836949242486, + "pair_accuracy": 0.8506888889418708, + "positive_goodness": 2.6408929210238985, + "runtime_seconds": 11.469693183898926 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.131998105737898, + "lr": 0.03, + "negative_goodness": 1.7331860170152453, + "pair_accuracy": 0.8381777777671814, + "positive_goodness": 2.6304886586083307, + "runtime_seconds": 11.468692541122437 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.1242666631274754, + "lr": 0.03, + "negative_goodness": 1.7418215518315634, + "pair_accuracy": 0.8563555556085375, + "positive_goodness": 2.6452883426242404, + "runtime_seconds": 11.470277309417725 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.1082207261615329, + "lr": 0.03, + "negative_goodness": 1.7185913586298625, + "pair_accuracy": 0.8551555555873447, + "positive_goodness": 2.692173753950331, + "runtime_seconds": 11.46796703338623 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.122600012758043, + "lr": 0.03, + "negative_goodness": 1.728596531316969, + "pair_accuracy": 0.8499555555767483, + "positive_goodness": 2.6711099163479277, + "runtime_seconds": 11.465506553649902 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.130996280447642, + "lr": 0.03, + "negative_goodness": 1.7388997368282741, + "pair_accuracy": 0.8429111111111112, + "positive_goodness": 2.6309451173994276, + "runtime_seconds": 11.46432614326477 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.113902264383104, + "lr": 0.03, + "negative_goodness": 1.7193142015457152, + "pair_accuracy": 0.85244444448683, + "positive_goodness": 2.67905240872701, + "runtime_seconds": 11.46338176727295 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.1208294565200805, + "lr": 0.03, + "negative_goodness": 1.7318235557556152, + "pair_accuracy": 0.8502666667090522, + "positive_goodness": 2.6726195462968616, + "runtime_seconds": 11.4618501663208 + } + ], + "layer": 11 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.5193677131546868, + "lr": 0.03, + "negative_goodness": 1.9320369598388671, + "pair_accuracy": 0.6643333333651225, + "positive_goodness": 2.272864971669515, + "runtime_seconds": 11.545727729797363 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.262697262679206, + "lr": 0.03, + "negative_goodness": 1.8270866785049438, + "pair_accuracy": 0.793822222243415, + "positive_goodness": 2.4918087616814506, + "runtime_seconds": 11.532617092132568 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.2503030692842272, + "lr": 0.03, + "negative_goodness": 1.824347703022427, + "pair_accuracy": 0.8150444444762336, + "positive_goodness": 2.557272047000461, + "runtime_seconds": 11.534370183944702 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.198259717686971, + "lr": 0.03, + "negative_goodness": 1.766563485675388, + "pair_accuracy": 0.8423333333651225, + "positive_goodness": 2.624625918960571, + "runtime_seconds": 11.535468578338623 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.201768271721734, + "lr": 0.03, + "negative_goodness": 1.7708571797053019, + "pair_accuracy": 0.8450666666825613, + "positive_goodness": 2.6667771935568916, + "runtime_seconds": 11.53644871711731 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.1574914231406317, + "lr": 0.03, + "negative_goodness": 1.7371545761956109, + "pair_accuracy": 0.8379999999894037, + "positive_goodness": 2.6323825813717314, + "runtime_seconds": 11.527211666107178 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.0629911703109742, + "lr": 0.03, + "negative_goodness": 1.6635253148820666, + "pair_accuracy": 0.8816444444550409, + "positive_goodness": 2.7562135391659206, + "runtime_seconds": 11.532263278961182 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0960080075158014, + "lr": 0.03, + "negative_goodness": 1.6929194522857667, + "pair_accuracy": 0.8638888888888889, + "positive_goodness": 2.7558015445285373, + "runtime_seconds": 11.534942865371704 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.051277198600769, + "lr": 0.03, + "negative_goodness": 1.6284768035464816, + "pair_accuracy": 0.8736222222116259, + "positive_goodness": 2.8061870825237696, + "runtime_seconds": 11.534278392791748 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.0479209172566732, + "lr": 0.03, + "negative_goodness": 1.6194205465952556, + "pair_accuracy": 0.8683111111429003, + "positive_goodness": 2.8155720493740506, + "runtime_seconds": 11.534322023391724 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.0175448275672065, + "lr": 0.03, + "negative_goodness": 1.6099820223914252, + "pair_accuracy": 0.8781555555449592, + "positive_goodness": 2.8649584050920276, + "runtime_seconds": 11.529380083084106 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.0546869684113396, + "lr": 0.03, + "negative_goodness": 1.6383192995283338, + "pair_accuracy": 0.8728222222222223, + "positive_goodness": 2.7737451714409724, + "runtime_seconds": 11.532482862472534 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.0507381366305881, + "lr": 0.03, + "negative_goodness": 1.6313363731384278, + "pair_accuracy": 0.8712000000006622, + "positive_goodness": 2.8201868965572783, + "runtime_seconds": 11.520715236663818 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.2124960111194187, + "lr": 0.03, + "negative_goodness": 1.7440821873558892, + "pair_accuracy": 0.7999777777459887, + "positive_goodness": 2.528493297068278, + "runtime_seconds": 11.52906584739685 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.0764184854507446, + "lr": 0.03, + "negative_goodness": 1.6542337481604683, + "pair_accuracy": 0.875266666677263, + "positive_goodness": 2.728055406528049, + "runtime_seconds": 11.53321123123169 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.0372030908584595, + "lr": 0.03, + "negative_goodness": 1.6146424490398832, + "pair_accuracy": 0.8815777778095669, + "positive_goodness": 2.834608659193251, + "runtime_seconds": 11.528754234313965 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.0495546552446153, + "lr": 0.03, + "negative_goodness": 1.6183708804236518, + "pair_accuracy": 0.8681777778201633, + "positive_goodness": 2.818839939710829, + "runtime_seconds": 11.534511804580688 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.0224570914268494, + "lr": 0.03, + "negative_goodness": 1.604338450855679, + "pair_accuracy": 0.8677111111111111, + "positive_goodness": 2.8641202906714547, + "runtime_seconds": 11.532573699951172 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.0609682135052152, + "lr": 0.03, + "negative_goodness": 1.5934994101630318, + "pair_accuracy": 0.8545111111534966, + "positive_goodness": 2.8040942819807264, + "runtime_seconds": 11.541002750396729 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.9765424047682021, + "lr": 0.03, + "negative_goodness": 1.5594088678148057, + "pair_accuracy": 0.8885111111217074, + "positive_goodness": 2.9834167423248292, + "runtime_seconds": 11.543180465698242 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.0344724028905232, + "lr": 0.03, + "negative_goodness": 1.6050069938659668, + "pair_accuracy": 0.8675555555608537, + "positive_goodness": 2.833638962788052, + "runtime_seconds": 11.535996675491333 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.9731153658760918, + "lr": 0.03, + "negative_goodness": 1.546537456003825, + "pair_accuracy": 0.879355555566152, + "positive_goodness": 2.9921038664923776, + "runtime_seconds": 11.53177547454834 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.0334494757970174, + "lr": 0.03, + "negative_goodness": 1.564444471783108, + "pair_accuracy": 0.8624666666560703, + "positive_goodness": 2.889780890274048, + "runtime_seconds": 11.534960508346558 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.0005535671446057, + "lr": 0.03, + "negative_goodness": 1.5636444779925875, + "pair_accuracy": 0.8819333333121405, + "positive_goodness": 2.8960605008867053, + "runtime_seconds": 11.530638456344604 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.9914159838570489, + "lr": 0.03, + "negative_goodness": 1.537100967449612, + "pair_accuracy": 0.8833999999788072, + "positive_goodness": 2.95103758676317, + "runtime_seconds": 11.53454875946045 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9677256444719102, + "lr": 0.03, + "negative_goodness": 1.4994336482153998, + "pair_accuracy": 0.8764, + "positive_goodness": 3.036872087478638, + "runtime_seconds": 11.54128384590149 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.9541039331648085, + "lr": 0.03, + "negative_goodness": 1.4899488023122152, + "pair_accuracy": 0.8982666666348775, + "positive_goodness": 3.0348831945207384, + "runtime_seconds": 11.535680770874023 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.0193664511786567, + "lr": 0.03, + "negative_goodness": 1.5590861376444498, + "pair_accuracy": 0.8748666666666667, + "positive_goodness": 2.9056128875308564, + "runtime_seconds": 11.537053346633911 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.0129731191105313, + "lr": 0.03, + "negative_goodness": 1.5907342933442858, + "pair_accuracy": 0.8856444444444445, + "positive_goodness": 2.86142134636773, + "runtime_seconds": 11.543936014175415 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.9810234522925483, + "lr": 0.03, + "negative_goodness": 1.5207439072926838, + "pair_accuracy": 0.8833555555555556, + "positive_goodness": 2.9707189182281493, + "runtime_seconds": 11.54027771949768 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.9732660224278767, + "lr": 0.03, + "negative_goodness": 1.5146448860910204, + "pair_accuracy": 0.8822888888676961, + "positive_goodness": 2.982214727274577, + "runtime_seconds": 11.541056632995605 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.010890556102329, + "lr": 0.03, + "negative_goodness": 1.5369498457802666, + "pair_accuracy": 0.8769555555661519, + "positive_goodness": 2.891353311835395, + "runtime_seconds": 11.543065786361694 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9595183154635959, + "lr": 0.03, + "negative_goodness": 1.4837897325727674, + "pair_accuracy": 0.8780222222540114, + "positive_goodness": 3.021982015270657, + "runtime_seconds": 11.531642198562622 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.9663335095299614, + "lr": 0.03, + "negative_goodness": 1.494070506392585, + "pair_accuracy": 0.8884444444444445, + "positive_goodness": 3.0065826642354327, + "runtime_seconds": 11.532119512557983 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9965985849380493, + "lr": 0.03, + "negative_goodness": 1.5008866308848063, + "pair_accuracy": 0.8673555555237664, + "positive_goodness": 2.9677204121059844, + "runtime_seconds": 11.53207540512085 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9590903651131524, + "lr": 0.03, + "negative_goodness": 1.5074019980960423, + "pair_accuracy": 0.8905333333333333, + "positive_goodness": 2.9884808716668023, + "runtime_seconds": 11.534015655517578 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.9817744597964817, + "lr": 0.03, + "negative_goodness": 1.507597320599026, + "pair_accuracy": 0.8742666666613684, + "positive_goodness": 2.9812523196114435, + "runtime_seconds": 11.542096614837646 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.9429059699588351, + "lr": 0.03, + "negative_goodness": 1.4726391412099202, + "pair_accuracy": 0.8867111110899184, + "positive_goodness": 3.063178432973226, + "runtime_seconds": 11.539838314056396 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.947278528605567, + "lr": 0.03, + "negative_goodness": 1.4637559517436558, + "pair_accuracy": 0.8852222222010294, + "positive_goodness": 3.0461649700164797, + "runtime_seconds": 11.537808656692505 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.9708946546448601, + "lr": 0.03, + "negative_goodness": 1.4801661722183228, + "pair_accuracy": 0.8747333333015442, + "positive_goodness": 2.996933869849311, + "runtime_seconds": 11.52981424331665 + } + ], + "layer": 12 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.187055238840315, + "lr": 0.03, + "negative_goodness": 1.6645303972456191, + "pair_accuracy": 0.7825999999788072, + "positive_goodness": 2.499781099743313, + "runtime_seconds": 11.463204860687256 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.1523983831299676, + "lr": 0.03, + "negative_goodness": 1.690518280114068, + "pair_accuracy": 0.7923999999788073, + "positive_goodness": 2.6134464219411213, + "runtime_seconds": 11.472151279449463 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.1095500896559822, + "lr": 0.03, + "negative_goodness": 1.6503688806321886, + "pair_accuracy": 0.8131777777777778, + "positive_goodness": 2.7411831461588543, + "runtime_seconds": 11.4614896774292 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0960903608746, + "lr": 0.03, + "negative_goodness": 1.6540416673660279, + "pair_accuracy": 0.8274888889100817, + "positive_goodness": 2.764807811694675, + "runtime_seconds": 11.454071044921875 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.086757717259725, + "lr": 0.03, + "negative_goodness": 1.637558436287774, + "pair_accuracy": 0.8319777777671814, + "positive_goodness": 2.8068922745174834, + "runtime_seconds": 11.45148754119873 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.085557995584276, + "lr": 0.03, + "negative_goodness": 1.6419683685090807, + "pair_accuracy": 0.8273111110899184, + "positive_goodness": 2.804148233583238, + "runtime_seconds": 11.45022177696228 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.0840575093799167, + "lr": 0.03, + "negative_goodness": 1.6507474847581651, + "pair_accuracy": 0.8308, + "positive_goodness": 2.785093954510159, + "runtime_seconds": 11.462250471115112 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.081087223794725, + "lr": 0.03, + "negative_goodness": 1.624083229192098, + "pair_accuracy": 0.818066666677263, + "positive_goodness": 2.826162412050035, + "runtime_seconds": 11.450571298599243 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.0556726303524442, + "lr": 0.03, + "negative_goodness": 1.6192741134643556, + "pair_accuracy": 0.8330000000529819, + "positive_goodness": 2.8789256189134385, + "runtime_seconds": 11.444692373275757 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.0726550150553384, + "lr": 0.03, + "negative_goodness": 1.6443405077828301, + "pair_accuracy": 0.8364, + "positive_goodness": 2.818095349968804, + "runtime_seconds": 11.44552993774414 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.0427760321087307, + "lr": 0.03, + "negative_goodness": 1.5975865573035346, + "pair_accuracy": 0.8475333333386316, + "positive_goodness": 2.952888010872735, + "runtime_seconds": 11.446805953979492 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.052150032096439, + "lr": 0.03, + "negative_goodness": 1.5999639821794298, + "pair_accuracy": 0.8324, + "positive_goodness": 2.8899134225209555, + "runtime_seconds": 11.447792053222656 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.0554455427593654, + "lr": 0.03, + "negative_goodness": 1.610533689541287, + "pair_accuracy": 0.8328666666348775, + "positive_goodness": 2.8702683297899036, + "runtime_seconds": 11.45159649848938 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.0346028754764134, + "lr": 0.03, + "negative_goodness": 1.5802373492134942, + "pair_accuracy": 0.8349555555449591, + "positive_goodness": 2.9839156155904134, + "runtime_seconds": 11.445923089981079 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.0392125293519763, + "lr": 0.03, + "negative_goodness": 1.5930798063278198, + "pair_accuracy": 0.8459333333651224, + "positive_goodness": 2.9511025960710313, + "runtime_seconds": 11.447638988494873 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.0169939028104147, + "lr": 0.03, + "negative_goodness": 1.5689969771067303, + "pair_accuracy": 0.8481555555555556, + "positive_goodness": 3.0117116711086696, + "runtime_seconds": 11.442813873291016 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.010998498386807, + "lr": 0.03, + "negative_goodness": 1.5512925944222344, + "pair_accuracy": 0.8469333333757189, + "positive_goodness": 3.0556963124593097, + "runtime_seconds": 11.44613265991211 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.0347159330791897, + "lr": 0.03, + "negative_goodness": 1.589198858388265, + "pair_accuracy": 0.8449999999841055, + "positive_goodness": 2.9811453842163087, + "runtime_seconds": 11.44748330116272 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.0492824290487501, + "lr": 0.03, + "negative_goodness": 1.5961732657326593, + "pair_accuracy": 0.8365555555237664, + "positive_goodness": 2.9069758478800454, + "runtime_seconds": 11.445242643356323 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.0269272917747498, + "lr": 0.03, + "negative_goodness": 1.5823925561904908, + "pair_accuracy": 0.8436444444762335, + "positive_goodness": 2.958928258726332, + "runtime_seconds": 11.446783065795898 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.0133206390168932, + "lr": 0.03, + "negative_goodness": 1.558656491004096, + "pair_accuracy": 0.8404222222328186, + "positive_goodness": 3.0242538442399765, + "runtime_seconds": 11.445029497146606 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.0227636487113105, + "lr": 0.03, + "negative_goodness": 1.5659161832809447, + "pair_accuracy": 0.8399333333121406, + "positive_goodness": 2.9868255658043754, + "runtime_seconds": 11.44289493560791 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.0144517639584012, + "lr": 0.03, + "negative_goodness": 1.5586056084844802, + "pair_accuracy": 0.8520666666666666, + "positive_goodness": 3.0069104994455973, + "runtime_seconds": 11.444070816040039 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.0178563988155789, + "lr": 0.03, + "negative_goodness": 1.5712008424123127, + "pair_accuracy": 0.8473111111217075, + "positive_goodness": 2.9877158690134684, + "runtime_seconds": 11.444390535354614 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.009598030863868, + "lr": 0.03, + "negative_goodness": 1.5575954394234552, + "pair_accuracy": 0.8479111111111111, + "positive_goodness": 3.0114525914086236, + "runtime_seconds": 11.441192626953125 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.9922846456739638, + "lr": 0.03, + "negative_goodness": 1.5375812773810493, + "pair_accuracy": 0.8528000000052982, + "positive_goodness": 3.096474609798855, + "runtime_seconds": 11.449247121810913 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.0007158332400852, + "lr": 0.03, + "negative_goodness": 1.5485197892506917, + "pair_accuracy": 0.8458666666560702, + "positive_goodness": 3.054336033206516, + "runtime_seconds": 11.445556879043579 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.9866950854195489, + "lr": 0.03, + "negative_goodness": 1.5284585861206055, + "pair_accuracy": 0.8569333333280351, + "positive_goodness": 3.086672363726298, + "runtime_seconds": 11.446619272232056 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.0286698531044853, + "lr": 0.03, + "negative_goodness": 1.5724010879940458, + "pair_accuracy": 0.835177777756585, + "positive_goodness": 2.965653760655721, + "runtime_seconds": 11.450146913528442 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.0056273728794523, + "lr": 0.03, + "negative_goodness": 1.5462565071105958, + "pair_accuracy": 0.8435555555767483, + "positive_goodness": 3.0423256553226046, + "runtime_seconds": 11.44895887374878 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.0329135367923312, + "lr": 0.03, + "negative_goodness": 1.56774684937795, + "pair_accuracy": 0.8332888889418708, + "positive_goodness": 2.9468816632588704, + "runtime_seconds": 11.444929122924805 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.9821040502760146, + "lr": 0.03, + "negative_goodness": 1.5241473830329046, + "pair_accuracy": 0.8547555555449592, + "positive_goodness": 3.0982081828223333, + "runtime_seconds": 11.442040205001831 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.9710932619942559, + "lr": 0.03, + "negative_goodness": 1.5032350939008925, + "pair_accuracy": 0.8558444444232517, + "positive_goodness": 3.1679354300181073, + "runtime_seconds": 11.439621210098267 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.9669278212229411, + "lr": 0.03, + "negative_goodness": 1.502143272950914, + "pair_accuracy": 0.865866666645474, + "positive_goodness": 3.14518706495497, + "runtime_seconds": 11.43754243850708 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.9803796142790052, + "lr": 0.03, + "negative_goodness": 1.5132447395748563, + "pair_accuracy": 0.8569333333015442, + "positive_goodness": 3.148890071105957, + "runtime_seconds": 11.446901559829712 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.9944175962871975, + "lr": 0.03, + "negative_goodness": 1.5301775277455647, + "pair_accuracy": 0.8550444444232517, + "positive_goodness": 3.075522162967258, + "runtime_seconds": 11.452233076095581 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.0029264811939664, + "lr": 0.03, + "negative_goodness": 1.5208887918684217, + "pair_accuracy": 0.8440444444232517, + "positive_goodness": 3.112905707338121, + "runtime_seconds": 11.456154108047485 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.0034538180033366, + "lr": 0.03, + "negative_goodness": 1.5342007928212484, + "pair_accuracy": 0.8444222222275204, + "positive_goodness": 3.0684139039145575, + "runtime_seconds": 11.450307846069336 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.9921989512337579, + "lr": 0.03, + "negative_goodness": 1.5271187964333428, + "pair_accuracy": 0.8532666666666666, + "positive_goodness": 3.094704305436876, + "runtime_seconds": 11.452314853668213 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.0027156937917074, + "lr": 0.03, + "negative_goodness": 1.5299429417292276, + "pair_accuracy": 0.8407111111323039, + "positive_goodness": 3.1001334541744656, + "runtime_seconds": 11.453917741775513 + } + ], + "layer": 13 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.111645887374878, + "lr": 0.03, + "negative_goodness": 1.6989062956492107, + "pair_accuracy": 0.815955555566152, + "positive_goodness": 2.8191925273683336, + "runtime_seconds": 11.543596267700195 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.0011963349872166, + "lr": 0.03, + "negative_goodness": 1.5664288255055745, + "pair_accuracy": 0.8482444444656372, + "positive_goodness": 2.994190947002835, + "runtime_seconds": 11.531896114349365 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.9220980641629961, + "lr": 0.03, + "negative_goodness": 1.4505082081794738, + "pair_accuracy": 0.8635333333333334, + "positive_goodness": 3.1924099999321833, + "runtime_seconds": 11.527493000030518 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.897832439994812, + "lr": 0.03, + "negative_goodness": 1.410964274957445, + "pair_accuracy": 0.8691111111005148, + "positive_goodness": 3.2579684660169814, + "runtime_seconds": 11.526431560516357 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8859938145319621, + "lr": 0.03, + "negative_goodness": 1.3830814940770466, + "pair_accuracy": 0.8694444444656372, + "positive_goodness": 3.324379167344835, + "runtime_seconds": 11.527417182922363 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.8292007525761922, + "lr": 0.03, + "negative_goodness": 1.3079636148929596, + "pair_accuracy": 0.8879777777777778, + "positive_goodness": 3.468393700917562, + "runtime_seconds": 11.529961824417114 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.8048682911554972, + "lr": 0.03, + "negative_goodness": 1.29324195728302, + "pair_accuracy": 0.8976666666348775, + "positive_goodness": 3.5935117740207247, + "runtime_seconds": 11.52757716178894 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.8163558831744724, + "lr": 0.03, + "negative_goodness": 1.2915357350667318, + "pair_accuracy": 0.8918888888888888, + "positive_goodness": 3.4752817547268338, + "runtime_seconds": 11.535110712051392 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.8429549293412103, + "lr": 0.03, + "negative_goodness": 1.3027161559210882, + "pair_accuracy": 0.8931111111534966, + "positive_goodness": 3.43379563615587, + "runtime_seconds": 11.53310775756836 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8074124051200019, + "lr": 0.03, + "negative_goodness": 1.2818419242223105, + "pair_accuracy": 0.8972222222222223, + "positive_goodness": 3.5444242057376436, + "runtime_seconds": 11.530567646026611 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.8308450179206001, + "lr": 0.03, + "negative_goodness": 1.3154840785132513, + "pair_accuracy": 0.8957555555237664, + "positive_goodness": 3.4643086980607776, + "runtime_seconds": 11.532241106033325 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7833781020694309, + "lr": 0.03, + "negative_goodness": 1.2436315359327528, + "pair_accuracy": 0.8894000000423855, + "positive_goodness": 3.6214107580396866, + "runtime_seconds": 11.53185248374939 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7653297730339897, + "lr": 0.03, + "negative_goodness": 1.2300197507646349, + "pair_accuracy": 0.9009111111217075, + "positive_goodness": 3.6803789240095353, + "runtime_seconds": 11.525169610977173 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7482272415267096, + "lr": 0.03, + "negative_goodness": 1.1958177578661178, + "pair_accuracy": 0.9051333333333333, + "positive_goodness": 3.7746791979471843, + "runtime_seconds": 11.536474466323853 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7747451221572028, + "lr": 0.03, + "negative_goodness": 1.2465944431093003, + "pair_accuracy": 0.9072444444444444, + "positive_goodness": 3.6578682426028784, + "runtime_seconds": 11.530051946640015 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7383128370708889, + "lr": 0.03, + "negative_goodness": 1.206661610953013, + "pair_accuracy": 0.9217111111111111, + "positive_goodness": 3.7737226357777915, + "runtime_seconds": 11.526041984558105 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7033318257596758, + "lr": 0.03, + "negative_goodness": 1.1429789894633824, + "pair_accuracy": 0.9264888888888889, + "positive_goodness": 3.9141218212551543, + "runtime_seconds": 11.52195429801941 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6770591852347057, + "lr": 0.03, + "negative_goodness": 1.0913595321655274, + "pair_accuracy": 0.9238444444444445, + "positive_goodness": 4.031760748799642, + "runtime_seconds": 11.52651834487915 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7043597082773845, + "lr": 0.03, + "negative_goodness": 1.1239887936909994, + "pair_accuracy": 0.9187777778095669, + "positive_goodness": 3.961696693971422, + "runtime_seconds": 11.527292728424072 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7019581976148818, + "lr": 0.03, + "negative_goodness": 1.1249590590159098, + "pair_accuracy": 0.9174666666878595, + "positive_goodness": 3.9796235538482665, + "runtime_seconds": 11.529160261154175 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7117597900920444, + "lr": 0.03, + "negative_goodness": 1.1449455026626587, + "pair_accuracy": 0.9230888888888888, + "positive_goodness": 3.87799734471639, + "runtime_seconds": 11.525985717773438 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6948300093544854, + "lr": 0.03, + "negative_goodness": 1.0941705544577705, + "pair_accuracy": 0.9208000000529819, + "positive_goodness": 3.9913914607577854, + "runtime_seconds": 11.52907419204712 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.71239867155287, + "lr": 0.03, + "negative_goodness": 1.132381079673767, + "pair_accuracy": 0.9174444444762336, + "positive_goodness": 3.9150877708858913, + "runtime_seconds": 11.528342723846436 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.72788271917767, + "lr": 0.03, + "negative_goodness": 1.1406093189080557, + "pair_accuracy": 0.9096444444444445, + "positive_goodness": 3.890463505469428, + "runtime_seconds": 11.532347202301025 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6943158844841851, + "lr": 0.03, + "negative_goodness": 1.1045252775828043, + "pair_accuracy": 0.9209777777777778, + "positive_goodness": 4.044179633182949, + "runtime_seconds": 11.532017230987549 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7128617346975539, + "lr": 0.03, + "negative_goodness": 1.1105486018286812, + "pair_accuracy": 0.9103111110899184, + "positive_goodness": 3.938532731543647, + "runtime_seconds": 11.52856993675232 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7172038160218133, + "lr": 0.03, + "negative_goodness": 1.108133947383033, + "pair_accuracy": 0.9172222222116259, + "positive_goodness": 3.9613190709431967, + "runtime_seconds": 11.532332420349121 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.7270684468269348, + "lr": 0.03, + "negative_goodness": 1.11441917875078, + "pair_accuracy": 0.9062444444444444, + "positive_goodness": 3.9057641464233397, + "runtime_seconds": 11.536449193954468 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6927592328071595, + "lr": 0.03, + "negative_goodness": 1.0796056401252747, + "pair_accuracy": 0.9173111111111111, + "positive_goodness": 4.0139652704026965, + "runtime_seconds": 11.536830425262451 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.705089494864146, + "lr": 0.03, + "negative_goodness": 1.1069290477116902, + "pair_accuracy": 0.9127333333863152, + "positive_goodness": 3.9546617326524522, + "runtime_seconds": 11.54232120513916 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6804569045172797, + "lr": 0.03, + "negative_goodness": 1.0695598226335314, + "pair_accuracy": 0.9214444444497426, + "positive_goodness": 4.040312901645237, + "runtime_seconds": 11.539996147155762 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7062436036639743, + "lr": 0.03, + "negative_goodness": 1.1095148333655462, + "pair_accuracy": 0.92004444448683, + "positive_goodness": 3.912372166442871, + "runtime_seconds": 11.538929462432861 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.7260941573143005, + "lr": 0.03, + "negative_goodness": 1.144557462861803, + "pair_accuracy": 0.9176222222222222, + "positive_goodness": 3.8939497937096488, + "runtime_seconds": 11.539717197418213 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6370376238716973, + "lr": 0.03, + "negative_goodness": 1.0135473838594224, + "pair_accuracy": 0.9360222222222222, + "positive_goodness": 4.184457355033027, + "runtime_seconds": 11.53159785270691 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6624158107333713, + "lr": 0.03, + "negative_goodness": 1.0262183599472046, + "pair_accuracy": 0.922844444433848, + "positive_goodness": 4.141648948287964, + "runtime_seconds": 11.533844947814941 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6857576194551256, + "lr": 0.03, + "negative_goodness": 1.0480208604812622, + "pair_accuracy": 0.9131555555555556, + "positive_goodness": 4.004562062411838, + "runtime_seconds": 11.531554698944092 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6624427671432496, + "lr": 0.03, + "negative_goodness": 1.0368959489186604, + "pair_accuracy": 0.9227777778095669, + "positive_goodness": 4.168289960267809, + "runtime_seconds": 11.520925283432007 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6858035319010417, + "lr": 0.03, + "negative_goodness": 1.0797759267171223, + "pair_accuracy": 0.9192222222116259, + "positive_goodness": 4.0289274618784585, + "runtime_seconds": 11.5270414352417 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.7038402543809679, + "lr": 0.03, + "negative_goodness": 1.0788289434009128, + "pair_accuracy": 0.9147777777777778, + "positive_goodness": 3.9996849560631644, + "runtime_seconds": 11.533000230789185 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.7360875101036496, + "lr": 0.03, + "negative_goodness": 1.1174624177508885, + "pair_accuracy": 0.9041111111005147, + "positive_goodness": 3.889640283457438, + "runtime_seconds": 11.538204669952393 + } + ], + "layer": 14 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.960599424648285, + "lr": 0.03, + "negative_goodness": 1.3601893191655476, + "pair_accuracy": 0.8577333333651225, + "positive_goodness": 3.1071103331247967, + "runtime_seconds": 11.472905158996582 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.893838206439548, + "lr": 0.03, + "negative_goodness": 1.3041049480438232, + "pair_accuracy": 0.8824444444391463, + "positive_goodness": 3.325065190209283, + "runtime_seconds": 11.470533609390259 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8470166777292888, + "lr": 0.03, + "negative_goodness": 1.2622030092451308, + "pair_accuracy": 0.893022222243415, + "positive_goodness": 3.4148729423099096, + "runtime_seconds": 11.461101293563843 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8440505016326905, + "lr": 0.03, + "negative_goodness": 1.2520615111033122, + "pair_accuracy": 0.9014444444497426, + "positive_goodness": 3.4577885539160835, + "runtime_seconds": 11.456033945083618 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8499263882637024, + "lr": 0.03, + "negative_goodness": 1.260185123931037, + "pair_accuracy": 0.8941777778201633, + "positive_goodness": 3.4137213859558106, + "runtime_seconds": 11.462000370025635 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.8343270510779487, + "lr": 0.03, + "negative_goodness": 1.251692674308353, + "pair_accuracy": 0.9048444444126553, + "positive_goodness": 3.4664847854614256, + "runtime_seconds": 11.453728199005127 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7994651244693333, + "lr": 0.03, + "negative_goodness": 1.192407287809584, + "pair_accuracy": 0.9154444444444444, + "positive_goodness": 3.6229033004760742, + "runtime_seconds": 11.45452618598938 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.8373943489922417, + "lr": 0.03, + "negative_goodness": 1.2462273060692681, + "pair_accuracy": 0.9039555555555555, + "positive_goodness": 3.4637114920722114, + "runtime_seconds": 11.453918218612671 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.8108619611899058, + "lr": 0.03, + "negative_goodness": 1.2151891308572558, + "pair_accuracy": 0.9128222222116258, + "positive_goodness": 3.583334741634793, + "runtime_seconds": 11.453421831130981 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.8086799982600742, + "lr": 0.03, + "negative_goodness": 1.189872332657708, + "pair_accuracy": 0.9118222222222222, + "positive_goodness": 3.5657381619347466, + "runtime_seconds": 11.451894998550415 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7959217234505548, + "lr": 0.03, + "negative_goodness": 1.167801177957323, + "pair_accuracy": 0.9140888888888888, + "positive_goodness": 3.6095679395887585, + "runtime_seconds": 11.451181411743164 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.839419930627611, + "lr": 0.03, + "negative_goodness": 1.2239483438703749, + "pair_accuracy": 0.9031333333492279, + "positive_goodness": 3.4576898874494764, + "runtime_seconds": 11.446404933929443 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.8035131971465217, + "lr": 0.03, + "negative_goodness": 1.184665754614936, + "pair_accuracy": 0.9126222222646078, + "positive_goodness": 3.5427846087137858, + "runtime_seconds": 11.451668977737427 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7885926164627075, + "lr": 0.03, + "negative_goodness": 1.1589632986916436, + "pair_accuracy": 0.9125333333465788, + "positive_goodness": 3.622052189212375, + "runtime_seconds": 11.452984094619751 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7415854180759853, + "lr": 0.03, + "negative_goodness": 1.087257624520196, + "pair_accuracy": 0.9289777778307596, + "positive_goodness": 3.773742953491211, + "runtime_seconds": 11.44685959815979 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.795190030362871, + "lr": 0.03, + "negative_goodness": 1.167963710361057, + "pair_accuracy": 0.9103333333651225, + "positive_goodness": 3.61634488758511, + "runtime_seconds": 11.44825530052185 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7802685107972887, + "lr": 0.03, + "negative_goodness": 1.1353812955247031, + "pair_accuracy": 0.9179777777671814, + "positive_goodness": 3.6351017626868356, + "runtime_seconds": 11.449008703231812 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7477353561825223, + "lr": 0.03, + "negative_goodness": 1.068502224222819, + "pair_accuracy": 0.9243555555343628, + "positive_goodness": 3.768948446146647, + "runtime_seconds": 11.445704936981201 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7811429981019762, + "lr": 0.03, + "negative_goodness": 1.1284749468697441, + "pair_accuracy": 0.9151555556085375, + "positive_goodness": 3.6467212786568535, + "runtime_seconds": 11.443086862564087 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7454070212046305, + "lr": 0.03, + "negative_goodness": 1.0898615379333496, + "pair_accuracy": 0.925577777756585, + "positive_goodness": 3.7467219937642415, + "runtime_seconds": 11.43971848487854 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7840086373117235, + "lr": 0.03, + "negative_goodness": 1.1121802319420708, + "pair_accuracy": 0.9104444444126553, + "positive_goodness": 3.619318712404039, + "runtime_seconds": 11.443312406539917 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.804413391992781, + "lr": 0.03, + "negative_goodness": 1.1421119336658054, + "pair_accuracy": 0.9056444444974263, + "positive_goodness": 3.53049066212972, + "runtime_seconds": 11.44547414779663 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7980234428193834, + "lr": 0.03, + "negative_goodness": 1.1304966497633193, + "pair_accuracy": 0.9042444444126553, + "positive_goodness": 3.5897183646731907, + "runtime_seconds": 11.445657014846802 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7879173007647197, + "lr": 0.03, + "negative_goodness": 1.1245691056887308, + "pair_accuracy": 0.9140888888888888, + "positive_goodness": 3.6375263810899523, + "runtime_seconds": 11.45618486404419 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.7168285492367215, + "lr": 0.03, + "negative_goodness": 1.034459923860762, + "pair_accuracy": 0.9242222222222222, + "positive_goodness": 3.8313823131985134, + "runtime_seconds": 11.45240831375122 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.773394932132297, + "lr": 0.03, + "negative_goodness": 1.0790360607412126, + "pair_accuracy": 0.9116666666560703, + "positive_goodness": 3.738573385365804, + "runtime_seconds": 11.45563554763794 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7250216270446778, + "lr": 0.03, + "negative_goodness": 1.0744536408318415, + "pair_accuracy": 0.9332888888994852, + "positive_goodness": 3.814101518122355, + "runtime_seconds": 11.447478294372559 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.7365434669176737, + "lr": 0.03, + "negative_goodness": 1.0655198247061834, + "pair_accuracy": 0.9263777777671814, + "positive_goodness": 3.7857794466230605, + "runtime_seconds": 11.445266485214233 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.7271356018066406, + "lr": 0.03, + "negative_goodness": 1.04058972852495, + "pair_accuracy": 0.9259777777459887, + "positive_goodness": 3.8308889725579154, + "runtime_seconds": 11.448951482772827 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7597741550233629, + "lr": 0.03, + "negative_goodness": 1.0877258750279744, + "pair_accuracy": 0.9201333333333334, + "positive_goodness": 3.7495145966847736, + "runtime_seconds": 11.441141843795776 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.7833733652538724, + "lr": 0.03, + "negative_goodness": 1.1024183316018847, + "pair_accuracy": 0.9139333333439297, + "positive_goodness": 3.629896656587389, + "runtime_seconds": 11.440154314041138 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7618070243623521, + "lr": 0.03, + "negative_goodness": 1.123231649992201, + "pair_accuracy": 0.9265999999682109, + "positive_goodness": 3.6738633215586343, + "runtime_seconds": 11.437780141830444 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.7808926001230876, + "lr": 0.03, + "negative_goodness": 1.1159286908679538, + "pair_accuracy": 0.9134666666772631, + "positive_goodness": 3.667033295143975, + "runtime_seconds": 11.43369436264038 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.740237443584866, + "lr": 0.03, + "negative_goodness": 1.0216862137476603, + "pair_accuracy": 0.914844444433848, + "positive_goodness": 3.811989743423462, + "runtime_seconds": 11.433335781097412 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.7057032071643405, + "lr": 0.03, + "negative_goodness": 1.0403056070009868, + "pair_accuracy": 0.9363333333439298, + "positive_goodness": 3.873667058096992, + "runtime_seconds": 11.434491395950317 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.7296898908191257, + "lr": 0.03, + "negative_goodness": 1.0639901039123536, + "pair_accuracy": 0.9287777778095669, + "positive_goodness": 3.826322735044691, + "runtime_seconds": 11.445716142654419 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6879556561999851, + "lr": 0.03, + "negative_goodness": 1.0024969339582654, + "pair_accuracy": 0.9415333333227369, + "positive_goodness": 3.986411351860894, + "runtime_seconds": 11.457069158554077 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.7360889429569244, + "lr": 0.03, + "negative_goodness": 1.051177152188619, + "pair_accuracy": 0.9272888888888889, + "positive_goodness": 3.8191451993306478, + "runtime_seconds": 11.44845175743103 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.7195213839795854, + "lr": 0.03, + "negative_goodness": 1.0524153537591299, + "pair_accuracy": 0.9299111111111111, + "positive_goodness": 3.848191736009386, + "runtime_seconds": 11.443453788757324 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.7165089974986182, + "lr": 0.03, + "negative_goodness": 1.0419043377717336, + "pair_accuracy": 0.9297111111111112, + "positive_goodness": 3.8490470304701065, + "runtime_seconds": 11.440377712249756 + } + ], + "layer": 15 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.2142588294559056, + "lr": 0.03, + "negative_goodness": 1.7784797133339776, + "pair_accuracy": 0.759977777756585, + "positive_goodness": 2.738842090140449, + "runtime_seconds": 11.527745962142944 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.9354296149571737, + "lr": 0.03, + "negative_goodness": 1.483514446788364, + "pair_accuracy": 0.8699333333333333, + "positive_goodness": 3.2499016259087457, + "runtime_seconds": 11.513649225234985 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8704373075061375, + "lr": 0.03, + "negative_goodness": 1.378172203816308, + "pair_accuracy": 0.877, + "positive_goodness": 3.381556701066759, + "runtime_seconds": 11.508040428161621 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8229865861044989, + "lr": 0.03, + "negative_goodness": 1.3317320293426513, + "pair_accuracy": 0.8982444444656372, + "positive_goodness": 3.499170958328247, + "runtime_seconds": 11.511051416397095 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7556780715518527, + "lr": 0.03, + "negative_goodness": 1.2231418136384753, + "pair_accuracy": 0.9109777777777778, + "positive_goodness": 3.72832291937934, + "runtime_seconds": 11.50023603439331 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7866978704452514, + "lr": 0.03, + "negative_goodness": 1.2526698908487957, + "pair_accuracy": 0.9023333333863153, + "positive_goodness": 3.621622983656989, + "runtime_seconds": 11.50180172920227 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7663542337311638, + "lr": 0.03, + "negative_goodness": 1.2292960315174526, + "pair_accuracy": 0.8977555555555555, + "positive_goodness": 3.7069101987202964, + "runtime_seconds": 11.495861530303955 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7693121594164106, + "lr": 0.03, + "negative_goodness": 1.2191637406137255, + "pair_accuracy": 0.9021111111005148, + "positive_goodness": 3.6307792204115126, + "runtime_seconds": 11.497413873672485 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7334240840699937, + "lr": 0.03, + "negative_goodness": 1.1952915573332046, + "pair_accuracy": 0.9166222222222222, + "positive_goodness": 3.785205938551161, + "runtime_seconds": 11.489555835723877 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7409562177658081, + "lr": 0.03, + "negative_goodness": 1.1804754957516987, + "pair_accuracy": 0.9150888888676961, + "positive_goodness": 3.7516286460876467, + "runtime_seconds": 11.489777565002441 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7434640794330173, + "lr": 0.03, + "negative_goodness": 1.1920327616638609, + "pair_accuracy": 0.9134666666666666, + "positive_goodness": 3.7195192142910427, + "runtime_seconds": 11.493113040924072 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7461305735270183, + "lr": 0.03, + "negative_goodness": 1.1808386295318603, + "pair_accuracy": 0.9081111111111111, + "positive_goodness": 3.749010198889838, + "runtime_seconds": 11.498711109161377 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7303895598941379, + "lr": 0.03, + "negative_goodness": 1.1501327410380044, + "pair_accuracy": 0.9122000000003312, + "positive_goodness": 3.8424873844570584, + "runtime_seconds": 11.50261402130127 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7366996428171794, + "lr": 0.03, + "negative_goodness": 1.1709251528422038, + "pair_accuracy": 0.9139555555555555, + "positive_goodness": 3.772987537574768, + "runtime_seconds": 11.500129699707031 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7298393931918674, + "lr": 0.03, + "negative_goodness": 1.1728227408091227, + "pair_accuracy": 0.9147777777777778, + "positive_goodness": 3.7541645527733696, + "runtime_seconds": 11.490344285964966 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7243116877237956, + "lr": 0.03, + "negative_goodness": 1.1846681736628215, + "pair_accuracy": 0.9181555555661519, + "positive_goodness": 3.7792230934143065, + "runtime_seconds": 11.48521375656128 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7047620616700914, + "lr": 0.03, + "negative_goodness": 1.1382841284010146, + "pair_accuracy": 0.919044444433848, + "positive_goodness": 3.892727417924669, + "runtime_seconds": 11.484813213348389 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.698278495364719, + "lr": 0.03, + "negative_goodness": 1.1572799570295547, + "pair_accuracy": 0.9281555555767483, + "positive_goodness": 3.8422365626441106, + "runtime_seconds": 11.485666036605835 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6816179650412666, + "lr": 0.03, + "negative_goodness": 1.108322289106581, + "pair_accuracy": 0.9172222222010294, + "positive_goodness": 3.94772786271837, + "runtime_seconds": 11.492088794708252 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7241153572718303, + "lr": 0.03, + "negative_goodness": 1.1659487962722779, + "pair_accuracy": 0.9127777777671814, + "positive_goodness": 3.7765286272684735, + "runtime_seconds": 11.489299297332764 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.694065003352695, + "lr": 0.03, + "negative_goodness": 1.1338835131327312, + "pair_accuracy": 0.9194888888782925, + "positive_goodness": 3.889447196875678, + "runtime_seconds": 11.490039825439453 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6958048059675429, + "lr": 0.03, + "negative_goodness": 1.1172033810297648, + "pair_accuracy": 0.9248444444762336, + "positive_goodness": 3.9132805143992107, + "runtime_seconds": 11.499722003936768 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.658401736164093, + "lr": 0.03, + "negative_goodness": 1.0791171964857313, + "pair_accuracy": 0.9279111110899183, + "positive_goodness": 4.008408717303806, + "runtime_seconds": 11.488586664199829 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6692716567357381, + "lr": 0.03, + "negative_goodness": 1.1071652904086642, + "pair_accuracy": 0.9376444444974263, + "positive_goodness": 3.9199858462015786, + "runtime_seconds": 11.489335060119629 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6946061289045545, + "lr": 0.03, + "negative_goodness": 1.1132158311420017, + "pair_accuracy": 0.9227555555873447, + "positive_goodness": 3.9045609526316323, + "runtime_seconds": 11.486608505249023 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6776372696346706, + "lr": 0.03, + "negative_goodness": 1.0977235158496432, + "pair_accuracy": 0.9276444444974263, + "positive_goodness": 3.9345812582227917, + "runtime_seconds": 11.484783411026001 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6727552057796055, + "lr": 0.03, + "negative_goodness": 1.102965490023295, + "pair_accuracy": 0.9286888889418708, + "positive_goodness": 3.9431814825693765, + "runtime_seconds": 11.491783142089844 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6589089241080814, + "lr": 0.03, + "negative_goodness": 1.0784030436515808, + "pair_accuracy": 0.9324, + "positive_goodness": 4.020370868343777, + "runtime_seconds": 11.48865008354187 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6809759548611111, + "lr": 0.03, + "negative_goodness": 1.083020683034261, + "pair_accuracy": 0.9203555555979411, + "positive_goodness": 3.9640463020748564, + "runtime_seconds": 11.484764814376831 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6796425803555383, + "lr": 0.03, + "negative_goodness": 1.1007697193834516, + "pair_accuracy": 0.9258444444444445, + "positive_goodness": 3.926277626122369, + "runtime_seconds": 11.489139556884766 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6481052728441027, + "lr": 0.03, + "negative_goodness": 1.0620961283445358, + "pair_accuracy": 0.9392666666666667, + "positive_goodness": 4.009515909067789, + "runtime_seconds": 11.480816841125488 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7188689867867364, + "lr": 0.03, + "negative_goodness": 1.1390384341981676, + "pair_accuracy": 0.914977777783076, + "positive_goodness": 3.838976155026754, + "runtime_seconds": 11.489631652832031 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6848822917461396, + "lr": 0.03, + "negative_goodness": 1.0784242279688516, + "pair_accuracy": 0.9120666666666667, + "positive_goodness": 3.9276069274054635, + "runtime_seconds": 11.495017766952515 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.7053797570122613, + "lr": 0.03, + "negative_goodness": 1.1098642331865098, + "pair_accuracy": 0.9159333333280352, + "positive_goodness": 3.8565343021816676, + "runtime_seconds": 11.490397214889526 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.637313116423289, + "lr": 0.03, + "negative_goodness": 1.038259095064799, + "pair_accuracy": 0.9331555555449592, + "positive_goodness": 4.0959934928046335, + "runtime_seconds": 11.484040260314941 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6696493131478628, + "lr": 0.03, + "negative_goodness": 1.0935873567634158, + "pair_accuracy": 0.9292666666666667, + "positive_goodness": 3.9808066704644096, + "runtime_seconds": 11.480796337127686 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6352302144474453, + "lr": 0.03, + "negative_goodness": 1.037945770793491, + "pair_accuracy": 0.9305777777671814, + "positive_goodness": 4.060438124084473, + "runtime_seconds": 11.479879379272461 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6306862574683295, + "lr": 0.03, + "negative_goodness": 1.0322141410509744, + "pair_accuracy": 0.9397999999788073, + "positive_goodness": 4.117844483651055, + "runtime_seconds": 11.486491918563843 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6247682672447629, + "lr": 0.03, + "negative_goodness": 1.0095059126747978, + "pair_accuracy": 0.9324888888888889, + "positive_goodness": 4.203084432729085, + "runtime_seconds": 11.495095252990723 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6266565242343479, + "lr": 0.03, + "negative_goodness": 1.0258872706943087, + "pair_accuracy": 0.9407333333339956, + "positive_goodness": 4.0914868221706815, + "runtime_seconds": 11.489105224609375 + } + ], + "layer": 16 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8778899590174357, + "lr": 0.03, + "negative_goodness": 1.3020035125096638, + "pair_accuracy": 0.8818222222752041, + "positive_goodness": 3.3060080192989774, + "runtime_seconds": 11.411525249481201 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8345706110212538, + "lr": 0.03, + "negative_goodness": 1.280667839940389, + "pair_accuracy": 0.8831555555575423, + "positive_goodness": 3.411566069391039, + "runtime_seconds": 11.411093711853027 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8132698686493768, + "lr": 0.03, + "negative_goodness": 1.2700839863035414, + "pair_accuracy": 0.8968444444444444, + "positive_goodness": 3.5052186207241482, + "runtime_seconds": 11.414422273635864 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7849615915086534, + "lr": 0.03, + "negative_goodness": 1.2225899628533257, + "pair_accuracy": 0.8984000000019868, + "positive_goodness": 3.594915585433112, + "runtime_seconds": 11.40564227104187 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7518328544669681, + "lr": 0.03, + "negative_goodness": 1.206861057970259, + "pair_accuracy": 0.8994888888888889, + "positive_goodness": 3.6625433993021645, + "runtime_seconds": 11.407093524932861 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7541667163213094, + "lr": 0.03, + "negative_goodness": 1.1577165252685546, + "pair_accuracy": 0.9005999999894037, + "positive_goodness": 3.7698585171169703, + "runtime_seconds": 11.400676965713501 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.739497305562761, + "lr": 0.03, + "negative_goodness": 1.1787491742451985, + "pair_accuracy": 0.9093999999788073, + "positive_goodness": 3.726414663611518, + "runtime_seconds": 11.4019136428833 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6995644869698419, + "lr": 0.03, + "negative_goodness": 1.1724853363460965, + "pair_accuracy": 0.9265333333227369, + "positive_goodness": 3.790528178405762, + "runtime_seconds": 11.39671277999878 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7282243003421359, + "lr": 0.03, + "negative_goodness": 1.170282567469279, + "pair_accuracy": 0.9110222222225534, + "positive_goodness": 3.747770587878757, + "runtime_seconds": 11.395386457443237 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7370741773393419, + "lr": 0.03, + "negative_goodness": 1.1865092753940158, + "pair_accuracy": 0.9069111111111111, + "positive_goodness": 3.7261062078581917, + "runtime_seconds": 11.40120005607605 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7272635718133714, + "lr": 0.03, + "negative_goodness": 1.155601115820143, + "pair_accuracy": 0.9087333333015442, + "positive_goodness": 3.6798297815534804, + "runtime_seconds": 11.403172016143799 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7272491498099433, + "lr": 0.03, + "negative_goodness": 1.1531237418810527, + "pair_accuracy": 0.9120222222222222, + "positive_goodness": 3.7804766793568927, + "runtime_seconds": 11.405818939208984 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7229687028196123, + "lr": 0.03, + "negative_goodness": 1.1729744496451484, + "pair_accuracy": 0.9122, + "positive_goodness": 3.729553064303928, + "runtime_seconds": 11.407121419906616 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6956096220228407, + "lr": 0.03, + "negative_goodness": 1.1212622887505426, + "pair_accuracy": 0.9231111111111111, + "positive_goodness": 3.8179926508373683, + "runtime_seconds": 11.397711515426636 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6780408235443963, + "lr": 0.03, + "negative_goodness": 1.1109374502393934, + "pair_accuracy": 0.9240666666560703, + "positive_goodness": 3.839642442194621, + "runtime_seconds": 11.389739513397217 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7222260410732693, + "lr": 0.03, + "negative_goodness": 1.1353862975968254, + "pair_accuracy": 0.9069555555555555, + "positive_goodness": 3.752523799175686, + "runtime_seconds": 11.392278671264648 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6847725882636176, + "lr": 0.03, + "negative_goodness": 1.1022012262980143, + "pair_accuracy": 0.9186222222328186, + "positive_goodness": 3.7879389378441704, + "runtime_seconds": 11.39792013168335 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.67931686179373, + "lr": 0.03, + "negative_goodness": 1.1019170108583238, + "pair_accuracy": 0.9149555555555555, + "positive_goodness": 3.8497440105862086, + "runtime_seconds": 11.398510217666626 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.681632706366645, + "lr": 0.03, + "negative_goodness": 1.0949594888475207, + "pair_accuracy": 0.912, + "positive_goodness": 3.857762930891249, + "runtime_seconds": 11.4032461643219 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6807656131532457, + "lr": 0.03, + "negative_goodness": 1.113674265967475, + "pair_accuracy": 0.9278222222354677, + "positive_goodness": 3.8597234155654907, + "runtime_seconds": 11.397762298583984 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6568966510560778, + "lr": 0.03, + "negative_goodness": 1.052275086042616, + "pair_accuracy": 0.9224, + "positive_goodness": 3.998084701241387, + "runtime_seconds": 11.389761209487915 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6523776440726387, + "lr": 0.03, + "negative_goodness": 1.0585639647801717, + "pair_accuracy": 0.9251111111111111, + "positive_goodness": 3.988406601164076, + "runtime_seconds": 11.390177726745605 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6698606166097852, + "lr": 0.03, + "negative_goodness": 1.059729789437188, + "pair_accuracy": 0.9264888888888889, + "positive_goodness": 3.9499813256157768, + "runtime_seconds": 11.39331340789795 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6767855382124582, + "lr": 0.03, + "negative_goodness": 1.0962845183796353, + "pair_accuracy": 0.9200888888888888, + "positive_goodness": 3.866574421861437, + "runtime_seconds": 11.400814771652222 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6681559310436249, + "lr": 0.03, + "negative_goodness": 1.05962479212549, + "pair_accuracy": 0.9242666666560703, + "positive_goodness": 3.953779282803006, + "runtime_seconds": 11.394862651824951 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6580384875085619, + "lr": 0.03, + "negative_goodness": 1.0404926973342896, + "pair_accuracy": 0.9182444444444444, + "positive_goodness": 4.013242787594265, + "runtime_seconds": 11.400310754776001 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6556592156092326, + "lr": 0.03, + "negative_goodness": 1.0391947637399037, + "pair_accuracy": 0.9148, + "positive_goodness": 3.9430634727478027, + "runtime_seconds": 11.395452499389648 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6495145433266958, + "lr": 0.03, + "negative_goodness": 1.035141086875068, + "pair_accuracy": 0.9265111111111111, + "positive_goodness": 4.048542617119683, + "runtime_seconds": 11.390067100524902 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6545808230400085, + "lr": 0.03, + "negative_goodness": 1.03744416897032, + "pair_accuracy": 0.9250444444444444, + "positive_goodness": 3.989464785427517, + "runtime_seconds": 11.398654222488403 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6071680326355828, + "lr": 0.03, + "negative_goodness": 0.9928528082529704, + "pair_accuracy": 0.9376, + "positive_goodness": 4.1280806634691025, + "runtime_seconds": 11.394030809402466 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6564588803291321, + "lr": 0.03, + "negative_goodness": 1.048547625647651, + "pair_accuracy": 0.9268666666560703, + "positive_goodness": 4.011526395204332, + "runtime_seconds": 11.390095710754395 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6915162199762133, + "lr": 0.03, + "negative_goodness": 1.0527576429102157, + "pair_accuracy": 0.9129555555555555, + "positive_goodness": 3.9125674082862005, + "runtime_seconds": 11.376333236694336 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6647219253116183, + "lr": 0.03, + "negative_goodness": 1.0455589829126994, + "pair_accuracy": 0.9225111111005148, + "positive_goodness": 3.96108216764662, + "runtime_seconds": 11.376374006271362 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6611198013517592, + "lr": 0.03, + "negative_goodness": 1.0240171091927421, + "pair_accuracy": 0.9153777777671814, + "positive_goodness": 3.995133366394043, + "runtime_seconds": 11.379848003387451 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6286041328218248, + "lr": 0.03, + "negative_goodness": 0.9957966912375557, + "pair_accuracy": 0.9247999999682108, + "positive_goodness": 4.073972783025106, + "runtime_seconds": 11.38550853729248 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6168696037981245, + "lr": 0.03, + "negative_goodness": 0.9752029914220174, + "pair_accuracy": 0.9332, + "positive_goodness": 4.086082436879476, + "runtime_seconds": 11.394899129867554 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6445652841567994, + "lr": 0.03, + "negative_goodness": 1.0346120649761623, + "pair_accuracy": 0.932733333322737, + "positive_goodness": 4.030500244691637, + "runtime_seconds": 11.381865739822388 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6182633950975206, + "lr": 0.03, + "negative_goodness": 0.9857562998983596, + "pair_accuracy": 0.9240222221904331, + "positive_goodness": 4.130385513983833, + "runtime_seconds": 11.375680446624756 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6341926379521687, + "lr": 0.03, + "negative_goodness": 0.9965379685295953, + "pair_accuracy": 0.9281333333386316, + "positive_goodness": 4.07446285065545, + "runtime_seconds": 11.372761726379395 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6626359832975599, + "lr": 0.03, + "negative_goodness": 1.043187768236796, + "pair_accuracy": 0.9175555555767483, + "positive_goodness": 3.987830042690701, + "runtime_seconds": 11.381198644638062 + } + ], + "layer": 17 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0238712080425687, + "lr": 0.03, + "negative_goodness": 1.6267969138463338, + "pair_accuracy": 0.8635333333545261, + "positive_goodness": 3.4405414626227486, + "runtime_seconds": 11.458187818527222 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.875248497666253, + "lr": 0.03, + "negative_goodness": 1.4113246383243137, + "pair_accuracy": 0.8891555555449592, + "positive_goodness": 3.537755567932129, + "runtime_seconds": 11.458462476730347 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8570412212583753, + "lr": 0.03, + "negative_goodness": 1.3480560964372423, + "pair_accuracy": 0.8769333333121405, + "positive_goodness": 3.4812887993706596, + "runtime_seconds": 11.45902419090271 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7620544821951124, + "lr": 0.03, + "negative_goodness": 1.2164268571005927, + "pair_accuracy": 0.9001333333333333, + "positive_goodness": 3.7835476169162328, + "runtime_seconds": 11.461973190307617 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7821060390260485, + "lr": 0.03, + "negative_goodness": 1.2471443727493285, + "pair_accuracy": 0.898377777809567, + "positive_goodness": 3.6897428286658394, + "runtime_seconds": 11.467332363128662 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.699793931584888, + "lr": 0.03, + "negative_goodness": 1.1276142264684041, + "pair_accuracy": 0.9205777777777778, + "positive_goodness": 3.958661265903049, + "runtime_seconds": 11.458478689193726 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7200518340216743, + "lr": 0.03, + "negative_goodness": 1.1428078466203477, + "pair_accuracy": 0.9049555555343628, + "positive_goodness": 3.8757632691701254, + "runtime_seconds": 11.461958885192871 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7276201088110605, + "lr": 0.03, + "negative_goodness": 1.1310185886065165, + "pair_accuracy": 0.9052444444338481, + "positive_goodness": 3.8477619941711425, + "runtime_seconds": 11.465466976165771 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.697704225148095, + "lr": 0.03, + "negative_goodness": 1.0837657649993897, + "pair_accuracy": 0.913844444433848, + "positive_goodness": 3.9636478348202178, + "runtime_seconds": 11.462913751602173 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7546042550881704, + "lr": 0.03, + "negative_goodness": 1.155122189225091, + "pair_accuracy": 0.8957111111005147, + "positive_goodness": 3.7985451854281957, + "runtime_seconds": 11.46703815460205 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7112306126515071, + "lr": 0.03, + "negative_goodness": 1.0944936880879932, + "pair_accuracy": 0.9051333333333333, + "positive_goodness": 3.8808950395372177, + "runtime_seconds": 11.47171950340271 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7186666619751189, + "lr": 0.03, + "negative_goodness": 1.1041798067390918, + "pair_accuracy": 0.9001777777777777, + "positive_goodness": 3.8866955695258247, + "runtime_seconds": 11.470775604248047 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6876446948157416, + "lr": 0.03, + "negative_goodness": 1.068367762819926, + "pair_accuracy": 0.9173111111534966, + "positive_goodness": 3.989405051803589, + "runtime_seconds": 11.466109991073608 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7034430094507006, + "lr": 0.03, + "negative_goodness": 1.0808769000106389, + "pair_accuracy": 0.9096444444444445, + "positive_goodness": 3.9345895313686796, + "runtime_seconds": 11.45791244506836 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6381483037577735, + "lr": 0.03, + "negative_goodness": 1.007320471684138, + "pair_accuracy": 0.9239777777777778, + "positive_goodness": 4.173850257576836, + "runtime_seconds": 11.465054273605347 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7105574566311307, + "lr": 0.03, + "negative_goodness": 1.0828821138805813, + "pair_accuracy": 0.9065555555767483, + "positive_goodness": 3.9530222195519342, + "runtime_seconds": 11.468143939971924 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7203307428995768, + "lr": 0.03, + "negative_goodness": 1.087520958190494, + "pair_accuracy": 0.9079111111111111, + "positive_goodness": 3.8520314056396483, + "runtime_seconds": 11.473730325698853 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7014325802644094, + "lr": 0.03, + "negative_goodness": 1.0590627021047805, + "pair_accuracy": 0.9077777777777778, + "positive_goodness": 3.9206238221910263, + "runtime_seconds": 11.4680814743042 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6453215148184035, + "lr": 0.03, + "negative_goodness": 1.0081046334902446, + "pair_accuracy": 0.9273111111005148, + "positive_goodness": 4.116268181567722, + "runtime_seconds": 11.455830335617065 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7206089069578383, + "lr": 0.03, + "negative_goodness": 1.0759896638870239, + "pair_accuracy": 0.8996444444444445, + "positive_goodness": 3.9054997420416937, + "runtime_seconds": 11.456252098083496 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7251647478315565, + "lr": 0.03, + "negative_goodness": 1.0857200366550022, + "pair_accuracy": 0.8963333333545261, + "positive_goodness": 3.8997216314527723, + "runtime_seconds": 11.463079452514648 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.64600939956771, + "lr": 0.03, + "negative_goodness": 0.996404217698839, + "pair_accuracy": 0.9239111111005147, + "positive_goodness": 4.093870976935492, + "runtime_seconds": 11.468068599700928 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6732896330091689, + "lr": 0.03, + "negative_goodness": 1.0263761287265354, + "pair_accuracy": 0.9150444444364971, + "positive_goodness": 4.061544457499187, + "runtime_seconds": 11.471715688705444 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6253658033132553, + "lr": 0.03, + "negative_goodness": 0.9649615334365103, + "pair_accuracy": 0.9220444444444444, + "positive_goodness": 4.284620618777805, + "runtime_seconds": 11.467228889465332 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6809599262396494, + "lr": 0.03, + "negative_goodness": 1.0085183369636537, + "pair_accuracy": 0.9082444444444444, + "positive_goodness": 4.085452321624756, + "runtime_seconds": 11.463412761688232 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6498948894129859, + "lr": 0.03, + "negative_goodness": 0.996808154270384, + "pair_accuracy": 0.9242222222222222, + "positive_goodness": 4.116955759769016, + "runtime_seconds": 11.469215631484985 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6407717226717207, + "lr": 0.03, + "negative_goodness": 0.9865594613214334, + "pair_accuracy": 0.9189111111111111, + "positive_goodness": 4.156074464077419, + "runtime_seconds": 11.463778734207153 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6425734723885854, + "lr": 0.03, + "negative_goodness": 0.9739522951655918, + "pair_accuracy": 0.9158222222222222, + "positive_goodness": 4.183378729587131, + "runtime_seconds": 11.463738918304443 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6360524368392096, + "lr": 0.03, + "negative_goodness": 0.975266733402676, + "pair_accuracy": 0.9278888888782925, + "positive_goodness": 4.194047065395779, + "runtime_seconds": 11.457951545715332 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.620321017297109, + "lr": 0.03, + "negative_goodness": 0.974578369363149, + "pair_accuracy": 0.9376888888782925, + "positive_goodness": 4.238451406266954, + "runtime_seconds": 11.466377258300781 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6293363606452942, + "lr": 0.03, + "negative_goodness": 0.9650486266030206, + "pair_accuracy": 0.9294222222010294, + "positive_goodness": 4.265275078837077, + "runtime_seconds": 11.467689514160156 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6541497399859958, + "lr": 0.03, + "negative_goodness": 1.0048234191417693, + "pair_accuracy": 0.9252666666560703, + "positive_goodness": 4.077626754336887, + "runtime_seconds": 11.474833250045776 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.618113844596015, + "lr": 0.03, + "negative_goodness": 0.9215122717751397, + "pair_accuracy": 0.9207111111323039, + "positive_goodness": 4.346696318393284, + "runtime_seconds": 11.472180366516113 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6284509398354424, + "lr": 0.03, + "negative_goodness": 0.9420025441275702, + "pair_accuracy": 0.9208222222752042, + "positive_goodness": 4.258650866360134, + "runtime_seconds": 11.463392734527588 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6443040979517831, + "lr": 0.03, + "negative_goodness": 0.9455102918545405, + "pair_accuracy": 0.9206444444444445, + "positive_goodness": 4.154368501366509, + "runtime_seconds": 11.463207006454468 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5886359766642253, + "lr": 0.03, + "negative_goodness": 0.8900165515899658, + "pair_accuracy": 0.9290444444391462, + "positive_goodness": 4.40786255607605, + "runtime_seconds": 11.470798254013062 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6178384296841092, + "lr": 0.03, + "negative_goodness": 0.9052698461532592, + "pair_accuracy": 0.9202, + "positive_goodness": 4.3244637471093075, + "runtime_seconds": 11.46959662437439 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6243698041174147, + "lr": 0.03, + "negative_goodness": 0.9499153301450941, + "pair_accuracy": 0.9236666666560703, + "positive_goodness": 4.257938927290175, + "runtime_seconds": 11.4730806350708 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5716669845581055, + "lr": 0.03, + "negative_goodness": 0.8611269852956136, + "pair_accuracy": 0.9315333333121406, + "positive_goodness": 4.4412541074117025, + "runtime_seconds": 11.468679189682007 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6149952730178833, + "lr": 0.03, + "negative_goodness": 0.9392577539020115, + "pair_accuracy": 0.92824444448683, + "positive_goodness": 4.298384775543213, + "runtime_seconds": 11.469684600830078 + } + ], + "layer": 18 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8897968382729424, + "lr": 0.03, + "negative_goodness": 1.2942851863437228, + "pair_accuracy": 0.8700444444126553, + "positive_goodness": 3.281372258292304, + "runtime_seconds": 11.056148767471313 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7675300782945421, + "lr": 0.03, + "negative_goodness": 1.1725007422659133, + "pair_accuracy": 0.9020666666454739, + "positive_goodness": 3.640765602408515, + "runtime_seconds": 11.04533314704895 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7705299105538262, + "lr": 0.03, + "negative_goodness": 1.1633437631395127, + "pair_accuracy": 0.8955111111005147, + "positive_goodness": 3.6506315539889864, + "runtime_seconds": 11.040218830108643 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7747511502901713, + "lr": 0.03, + "negative_goodness": 1.14982639503479, + "pair_accuracy": 0.898200000052982, + "positive_goodness": 3.6608936979081896, + "runtime_seconds": 11.043161392211914 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8368180810610454, + "lr": 0.03, + "negative_goodness": 1.2289307603200277, + "pair_accuracy": 0.8712444444338481, + "positive_goodness": 3.4823982645670575, + "runtime_seconds": 11.029962301254272 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7660647264268663, + "lr": 0.03, + "negative_goodness": 1.1317598656548393, + "pair_accuracy": 0.8949555555555555, + "positive_goodness": 3.6811716128879124, + "runtime_seconds": 11.037769794464111 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.792335047589408, + "lr": 0.03, + "negative_goodness": 1.1791967719925776, + "pair_accuracy": 0.8906222222222222, + "positive_goodness": 3.653780731455485, + "runtime_seconds": 11.037715435028076 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7417601909743415, + "lr": 0.03, + "negative_goodness": 1.123517779858907, + "pair_accuracy": 0.9085777778095669, + "positive_goodness": 3.765036004384359, + "runtime_seconds": 11.035395860671997 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7576801489618089, + "lr": 0.03, + "negative_goodness": 1.1071360138363309, + "pair_accuracy": 0.895511111164093, + "positive_goodness": 3.7130664882236055, + "runtime_seconds": 11.036988496780396 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7656070228364733, + "lr": 0.03, + "negative_goodness": 1.1382857746971977, + "pair_accuracy": 0.8963999999894037, + "positive_goodness": 3.696016763856676, + "runtime_seconds": 11.029341459274292 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7372738844500648, + "lr": 0.03, + "negative_goodness": 1.0960620536698236, + "pair_accuracy": 0.9043111111111111, + "positive_goodness": 3.753188484446208, + "runtime_seconds": 11.028864860534668 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7130952417903477, + "lr": 0.03, + "negative_goodness": 1.0554641263167064, + "pair_accuracy": 0.9084666666666666, + "positive_goodness": 3.914613025961982, + "runtime_seconds": 11.025443315505981 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7321566489007738, + "lr": 0.03, + "negative_goodness": 1.090039578374227, + "pair_accuracy": 0.9058888888888889, + "positive_goodness": 3.8106720087687176, + "runtime_seconds": 11.02683424949646 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7190828419261509, + "lr": 0.03, + "negative_goodness": 1.076899893951416, + "pair_accuracy": 0.9095333333121406, + "positive_goodness": 3.792647679901123, + "runtime_seconds": 11.030236959457397 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7017752679877811, + "lr": 0.03, + "negative_goodness": 1.0617893258624607, + "pair_accuracy": 0.9214666666666667, + "positive_goodness": 3.8725761288960774, + "runtime_seconds": 11.028100728988647 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7254317340532939, + "lr": 0.03, + "negative_goodness": 1.095533413028717, + "pair_accuracy": 0.9063555555979411, + "positive_goodness": 3.773855685043335, + "runtime_seconds": 11.04168701171875 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7267078950352139, + "lr": 0.03, + "negative_goodness": 1.069862123521169, + "pair_accuracy": 0.9041777777777777, + "positive_goodness": 3.823130665418837, + "runtime_seconds": 11.034801959991455 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7108385338041517, + "lr": 0.03, + "negative_goodness": 1.0609655254152086, + "pair_accuracy": 0.9085111111534966, + "positive_goodness": 3.8717107067108154, + "runtime_seconds": 11.032256126403809 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7044290168868171, + "lr": 0.03, + "negative_goodness": 1.0682763629701402, + "pair_accuracy": 0.9091111111005147, + "positive_goodness": 3.8506938236660426, + "runtime_seconds": 11.028140544891357 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6998607492023045, + "lr": 0.03, + "negative_goodness": 1.0456291405995686, + "pair_accuracy": 0.9123999999894036, + "positive_goodness": 3.8799837927076553, + "runtime_seconds": 11.039151430130005 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6916552094035678, + "lr": 0.03, + "negative_goodness": 1.055586628871494, + "pair_accuracy": 0.9177555555449591, + "positive_goodness": 3.897475470055474, + "runtime_seconds": 11.030864477157593 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7160895680427551, + "lr": 0.03, + "negative_goodness": 1.0308590716997783, + "pair_accuracy": 0.9060222222010295, + "positive_goodness": 3.845046791712443, + "runtime_seconds": 11.033905982971191 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7325767435656654, + "lr": 0.03, + "negative_goodness": 1.0910205457925797, + "pair_accuracy": 0.8998666666666667, + "positive_goodness": 3.7684525178697372, + "runtime_seconds": 11.041051864624023 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6936178482373555, + "lr": 0.03, + "negative_goodness": 1.0210628179338244, + "pair_accuracy": 0.9148666666560703, + "positive_goodness": 3.903477737681071, + "runtime_seconds": 11.04029130935669 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6991065873463949, + "lr": 0.03, + "negative_goodness": 1.019480429787106, + "pair_accuracy": 0.9104444444470935, + "positive_goodness": 3.9596194022814433, + "runtime_seconds": 11.029513597488403 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7013771608511606, + "lr": 0.03, + "negative_goodness": 1.032268196217219, + "pair_accuracy": 0.9059999999788072, + "positive_goodness": 3.8859318993462457, + "runtime_seconds": 11.02440333366394 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6590976129849752, + "lr": 0.03, + "negative_goodness": 0.9778981248643663, + "pair_accuracy": 0.9193555555555556, + "positive_goodness": 4.0102544122484, + "runtime_seconds": 11.026361227035522 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6730045509232415, + "lr": 0.03, + "negative_goodness": 0.9915513085736168, + "pair_accuracy": 0.911, + "positive_goodness": 3.997576932356093, + "runtime_seconds": 11.033700942993164 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6645780202600691, + "lr": 0.03, + "negative_goodness": 0.980203039126926, + "pair_accuracy": 0.9176888888782925, + "positive_goodness": 3.9766530134412976, + "runtime_seconds": 11.040428161621094 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6978399725966984, + "lr": 0.03, + "negative_goodness": 1.0290309374279447, + "pair_accuracy": 0.9137333333333333, + "positive_goodness": 3.9406833396487766, + "runtime_seconds": 11.047670364379883 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6861910125732422, + "lr": 0.03, + "negative_goodness": 1.0010407320128547, + "pair_accuracy": 0.9090888888915379, + "positive_goodness": 3.9566620972739326, + "runtime_seconds": 11.041018724441528 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6719412999682957, + "lr": 0.03, + "negative_goodness": 0.9733510562366909, + "pair_accuracy": 0.9112666666772631, + "positive_goodness": 4.0392858127594, + "runtime_seconds": 11.04247760772705 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6777115481906467, + "lr": 0.03, + "negative_goodness": 1.0003335967063904, + "pair_accuracy": 0.9161111110952166, + "positive_goodness": 4.007618843862746, + "runtime_seconds": 11.04442286491394 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6503387535148196, + "lr": 0.03, + "negative_goodness": 0.9593006705813938, + "pair_accuracy": 0.9202, + "positive_goodness": 4.094335821617975, + "runtime_seconds": 11.039626359939575 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6382088255670335, + "lr": 0.03, + "negative_goodness": 0.9451066217356258, + "pair_accuracy": 0.9256888888888889, + "positive_goodness": 4.147031889258491, + "runtime_seconds": 11.04181456565857 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.665387974749671, + "lr": 0.03, + "negative_goodness": 0.9599508424229092, + "pair_accuracy": 0.9181777778095669, + "positive_goodness": 4.090465255228678, + "runtime_seconds": 11.044080257415771 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6114597426944309, + "lr": 0.03, + "negative_goodness": 0.916444284523858, + "pair_accuracy": 0.9398000000105964, + "positive_goodness": 4.254020530721876, + "runtime_seconds": 11.04451036453247 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6787692109028498, + "lr": 0.03, + "negative_goodness": 0.9924484639816814, + "pair_accuracy": 0.9142666666666667, + "positive_goodness": 4.032047976430257, + "runtime_seconds": 11.049831628799438 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.669404616800944, + "lr": 0.03, + "negative_goodness": 0.9578225483788384, + "pair_accuracy": 0.913866666677263, + "positive_goodness": 4.076801224772136, + "runtime_seconds": 11.041583776473999 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6231370561732186, + "lr": 0.03, + "negative_goodness": 0.8812808318773906, + "pair_accuracy": 0.9197111111111111, + "positive_goodness": 4.219706010860867, + "runtime_seconds": 11.04699993133545 + } + ], + "layer": 19 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0159315459887186, + "lr": 0.03, + "negative_goodness": 1.5060517064412435, + "pair_accuracy": 0.8396444444656372, + "positive_goodness": 3.219766240819295, + "runtime_seconds": 11.14686894416809 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8449754423353407, + "lr": 0.03, + "negative_goodness": 1.321000718413459, + "pair_accuracy": 0.8906888889047835, + "positive_goodness": 3.4788123643451265, + "runtime_seconds": 11.126720428466797 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7991632269965278, + "lr": 0.03, + "negative_goodness": 1.2459060908529493, + "pair_accuracy": 0.9014888888782925, + "positive_goodness": 3.5688275544484456, + "runtime_seconds": 11.119493007659912 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7600319583045112, + "lr": 0.03, + "negative_goodness": 1.2020499038802253, + "pair_accuracy": 0.9082888888676961, + "positive_goodness": 3.6954628224690755, + "runtime_seconds": 11.114173412322998 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7055248845418294, + "lr": 0.03, + "negative_goodness": 1.1064341090096368, + "pair_accuracy": 0.9226, + "positive_goodness": 3.879192309718662, + "runtime_seconds": 11.127914667129517 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7463480601522657, + "lr": 0.03, + "negative_goodness": 1.1620532824092442, + "pair_accuracy": 0.9048888888888889, + "positive_goodness": 3.765120003467136, + "runtime_seconds": 11.11577582359314 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.714597853480445, + "lr": 0.03, + "negative_goodness": 1.1142130030843946, + "pair_accuracy": 0.9070444444444444, + "positive_goodness": 3.932217236879137, + "runtime_seconds": 11.11468243598938 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7341278751691183, + "lr": 0.03, + "negative_goodness": 1.1118655537711248, + "pair_accuracy": 0.9101111111111111, + "positive_goodness": 3.7870368191189234, + "runtime_seconds": 11.108931303024292 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6994435772365994, + "lr": 0.03, + "negative_goodness": 1.092015451558431, + "pair_accuracy": 0.9140444444232517, + "positive_goodness": 3.935842131551107, + "runtime_seconds": 11.113586664199829 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6627329576386346, + "lr": 0.03, + "negative_goodness": 1.066448647901747, + "pair_accuracy": 0.9284444444444444, + "positive_goodness": 3.9944156928168404, + "runtime_seconds": 11.105719566345215 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6943741464243994, + "lr": 0.03, + "negative_goodness": 1.0632417104297214, + "pair_accuracy": 0.9185777777671814, + "positive_goodness": 3.879582884640164, + "runtime_seconds": 11.105685710906982 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6923339837286208, + "lr": 0.03, + "negative_goodness": 1.0589468596140543, + "pair_accuracy": 0.91164444448683, + "positive_goodness": 3.9890255343119305, + "runtime_seconds": 11.108553886413574 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6555828375922309, + "lr": 0.03, + "negative_goodness": 1.0206069918314615, + "pair_accuracy": 0.9192222222222223, + "positive_goodness": 4.122530371136135, + "runtime_seconds": 11.11194896697998 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6593441983699798, + "lr": 0.03, + "negative_goodness": 1.0120020165655348, + "pair_accuracy": 0.9237333333333333, + "positive_goodness": 4.1242109159681535, + "runtime_seconds": 11.113760471343994 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7054265214761098, + "lr": 0.03, + "negative_goodness": 1.042225859260559, + "pair_accuracy": 0.9061111111111111, + "positive_goodness": 3.967176022254096, + "runtime_seconds": 11.116923093795776 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.660067798781395, + "lr": 0.03, + "negative_goodness": 1.0177289138476053, + "pair_accuracy": 0.9218222222222222, + "positive_goodness": 4.16506443456014, + "runtime_seconds": 11.122292280197144 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6422319798151652, + "lr": 0.03, + "negative_goodness": 0.9938028288099501, + "pair_accuracy": 0.9260888888676961, + "positive_goodness": 4.13302086046007, + "runtime_seconds": 11.117006063461304 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6479314363691542, + "lr": 0.03, + "negative_goodness": 1.011535687552558, + "pair_accuracy": 0.9233555555661519, + "positive_goodness": 4.103052482922872, + "runtime_seconds": 11.112883567810059 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6761519215742747, + "lr": 0.03, + "negative_goodness": 1.0371086683750153, + "pair_accuracy": 0.9133333333333333, + "positive_goodness": 4.026422662438287, + "runtime_seconds": 11.117064714431763 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6706064676708645, + "lr": 0.03, + "negative_goodness": 1.0027934560139975, + "pair_accuracy": 0.9183777777777777, + "positive_goodness": 4.094995846472846, + "runtime_seconds": 11.12255859375 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6309142844729954, + "lr": 0.03, + "negative_goodness": 0.9597272156503466, + "pair_accuracy": 0.9256888888888889, + "positive_goodness": 4.2328829788208004, + "runtime_seconds": 11.126716375350952 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6551690208435058, + "lr": 0.03, + "negative_goodness": 1.007851409573025, + "pair_accuracy": 0.9230444444974264, + "positive_goodness": 4.09389004512363, + "runtime_seconds": 11.105561256408691 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6112751581139034, + "lr": 0.03, + "negative_goodness": 0.9295570630444421, + "pair_accuracy": 0.9275111111111111, + "positive_goodness": 4.321526034630669, + "runtime_seconds": 11.113544702529907 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6487738689263661, + "lr": 0.03, + "negative_goodness": 0.9682493840641445, + "pair_accuracy": 0.9241111111111111, + "positive_goodness": 4.17864088567098, + "runtime_seconds": 11.111833333969116 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6452777474032508, + "lr": 0.03, + "negative_goodness": 0.9571507864369286, + "pair_accuracy": 0.9118222222116258, + "positive_goodness": 4.2143416286044655, + "runtime_seconds": 11.114066362380981 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6568354656643338, + "lr": 0.03, + "negative_goodness": 0.9738039085388184, + "pair_accuracy": 0.9196444444523917, + "positive_goodness": 4.170342621570163, + "runtime_seconds": 11.111068964004517 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6527136901219686, + "lr": 0.03, + "negative_goodness": 0.9741771711561414, + "pair_accuracy": 0.912044444433848, + "positive_goodness": 4.171915949079725, + "runtime_seconds": 11.11466360092163 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6347881847381592, + "lr": 0.03, + "negative_goodness": 0.9752460170374976, + "pair_accuracy": 0.9236222222222222, + "positive_goodness": 4.195937968190511, + "runtime_seconds": 11.10994267463684 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6824342876222399, + "lr": 0.03, + "negative_goodness": 1.0067006530655755, + "pair_accuracy": 0.9112666666666667, + "positive_goodness": 4.015544774500529, + "runtime_seconds": 11.12522554397583 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6263354563977983, + "lr": 0.03, + "negative_goodness": 0.9457056217723423, + "pair_accuracy": 0.9197555555555555, + "positive_goodness": 4.227846555921767, + "runtime_seconds": 11.125265121459961 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6868328608089023, + "lr": 0.03, + "negative_goodness": 1.0034477077696058, + "pair_accuracy": 0.8999555555237664, + "positive_goodness": 4.065208229107327, + "runtime_seconds": 11.10265564918518 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6711669765260484, + "lr": 0.03, + "negative_goodness": 1.0045035263061524, + "pair_accuracy": 0.9144000000529819, + "positive_goodness": 4.12256383688185, + "runtime_seconds": 11.11836862564087 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.670402642774582, + "lr": 0.03, + "negative_goodness": 1.0086089979065789, + "pair_accuracy": 0.9106444444444445, + "positive_goodness": 4.092948599158393, + "runtime_seconds": 11.121069192886353 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6020092468632592, + "lr": 0.03, + "negative_goodness": 0.9069583732154635, + "pair_accuracy": 0.9305777777777777, + "positive_goodness": 4.354486865785387, + "runtime_seconds": 11.11877989768982 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6124957174619039, + "lr": 0.03, + "negative_goodness": 0.9275702701992459, + "pair_accuracy": 0.9275333333121406, + "positive_goodness": 4.3420462059021, + "runtime_seconds": 11.110337257385254 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6168937229050531, + "lr": 0.03, + "negative_goodness": 0.9384199421988593, + "pair_accuracy": 0.9298444444338481, + "positive_goodness": 4.293281845347087, + "runtime_seconds": 11.092948198318481 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6403769088851081, + "lr": 0.03, + "negative_goodness": 0.974670251862208, + "pair_accuracy": 0.9239333333121406, + "positive_goodness": 4.201837698491414, + "runtime_seconds": 11.095546245574951 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5918380961736043, + "lr": 0.03, + "negative_goodness": 0.8865425747129653, + "pair_accuracy": 0.9329333333545261, + "positive_goodness": 4.424838272603353, + "runtime_seconds": 11.108196258544922 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.588706819290585, + "lr": 0.03, + "negative_goodness": 0.86786878874037, + "pair_accuracy": 0.9379111111217074, + "positive_goodness": 4.401966753345065, + "runtime_seconds": 11.095087051391602 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6041920864078734, + "lr": 0.03, + "negative_goodness": 0.892962294514974, + "pair_accuracy": 0.9221555555555555, + "positive_goodness": 4.3659705052693685, + "runtime_seconds": 11.081132888793945 + } + ], + "layer": 20 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8881645433849759, + "lr": 0.03, + "negative_goodness": 1.269579215833876, + "pair_accuracy": 0.8776222221904331, + "positive_goodness": 3.333434263314141, + "runtime_seconds": 10.312065362930298 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8251499989297655, + "lr": 0.03, + "negative_goodness": 1.2463688258488972, + "pair_accuracy": 0.8870888888888889, + "positive_goodness": 3.500197876824273, + "runtime_seconds": 9.843517065048218 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7567361609829797, + "lr": 0.03, + "negative_goodness": 1.1509097148259482, + "pair_accuracy": 0.9057111111005147, + "positive_goodness": 3.6458424524095325, + "runtime_seconds": 10.075620412826538 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8118733232922024, + "lr": 0.03, + "negative_goodness": 1.1970054453531902, + "pair_accuracy": 0.8890888888888889, + "positive_goodness": 3.530623871485392, + "runtime_seconds": 9.840810060501099 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7094494288762411, + "lr": 0.03, + "negative_goodness": 1.091615876314375, + "pair_accuracy": 0.9137111111111111, + "positive_goodness": 3.7822360170152454, + "runtime_seconds": 9.933433532714844 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.737636995236079, + "lr": 0.03, + "negative_goodness": 1.1005836195045047, + "pair_accuracy": 0.9007111111111111, + "positive_goodness": 3.754209392038981, + "runtime_seconds": 9.745323419570923 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.66383554959827, + "lr": 0.03, + "negative_goodness": 1.0245686241573757, + "pair_accuracy": 0.9206666666560703, + "positive_goodness": 3.9872735278659395, + "runtime_seconds": 9.973944664001465 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7211032234218385, + "lr": 0.03, + "negative_goodness": 1.1051502022584279, + "pair_accuracy": 0.911, + "positive_goodness": 3.799056103769938, + "runtime_seconds": 9.799665212631226 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6551723930570814, + "lr": 0.03, + "negative_goodness": 0.9901889557732476, + "pair_accuracy": 0.9237999999788072, + "positive_goodness": 4.070426139068603, + "runtime_seconds": 9.90731167793274 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6513851806746589, + "lr": 0.03, + "negative_goodness": 0.9696218379762438, + "pair_accuracy": 0.9218222222222222, + "positive_goodness": 4.040347370486789, + "runtime_seconds": 9.788477659225464 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6826728337287903, + "lr": 0.03, + "negative_goodness": 1.0201771539264255, + "pair_accuracy": 0.9032000000317891, + "positive_goodness": 4.0159361585829, + "runtime_seconds": 10.077486038208008 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6373336482895745, + "lr": 0.03, + "negative_goodness": 0.9773968426598443, + "pair_accuracy": 0.9206888888888889, + "positive_goodness": 4.0911422974480525, + "runtime_seconds": 9.792579412460327 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6590163462850782, + "lr": 0.03, + "negative_goodness": 0.9703390027364095, + "pair_accuracy": 0.9179111111111111, + "positive_goodness": 4.084614022657607, + "runtime_seconds": 9.877023458480835 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7030290870030721, + "lr": 0.03, + "negative_goodness": 1.0501702172173395, + "pair_accuracy": 0.902755555566152, + "positive_goodness": 3.9091428763495553, + "runtime_seconds": 9.87649130821228 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6545703786002265, + "lr": 0.03, + "negative_goodness": 0.9942095385233561, + "pair_accuracy": 0.9223333333227369, + "positive_goodness": 4.0935355565388996, + "runtime_seconds": 10.081867694854736 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6470485781351726, + "lr": 0.03, + "negative_goodness": 0.9729014879014757, + "pair_accuracy": 0.9196, + "positive_goodness": 4.071464036856757, + "runtime_seconds": 9.757944583892822 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6655181950436698, + "lr": 0.03, + "negative_goodness": 0.9816463483201133, + "pair_accuracy": 0.9112666666666667, + "positive_goodness": 4.011623206159803, + "runtime_seconds": 9.837441444396973 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6067378798855676, + "lr": 0.03, + "negative_goodness": 0.9125529476589627, + "pair_accuracy": 0.9355777777777777, + "positive_goodness": 4.236775533718533, + "runtime_seconds": 9.752538442611694 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6194083744949764, + "lr": 0.03, + "negative_goodness": 0.9257384330961439, + "pair_accuracy": 0.9330444444444445, + "positive_goodness": 4.211429159800211, + "runtime_seconds": 10.057053089141846 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6614102448569403, + "lr": 0.03, + "negative_goodness": 0.9455141001701355, + "pair_accuracy": 0.9128, + "positive_goodness": 4.082400214915805, + "runtime_seconds": 9.738528966903687 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5988132469177246, + "lr": 0.03, + "negative_goodness": 0.9106998472743564, + "pair_accuracy": 0.9367333333757188, + "positive_goodness": 4.294793701553345, + "runtime_seconds": 9.808160305023193 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6158064560625288, + "lr": 0.03, + "negative_goodness": 0.8903891183561748, + "pair_accuracy": 0.9315111111111111, + "positive_goodness": 4.246721631028917, + "runtime_seconds": 9.79865312576294 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6242394544601441, + "lr": 0.03, + "negative_goodness": 0.9268351051754422, + "pair_accuracy": 0.9236888889418708, + "positive_goodness": 4.256469004143609, + "runtime_seconds": 9.982605457305908 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6351897318310208, + "lr": 0.03, + "negative_goodness": 0.9479790982246399, + "pair_accuracy": 0.9261999999894036, + "positive_goodness": 4.152758196894328, + "runtime_seconds": 9.60093379020691 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6335484645366669, + "lr": 0.03, + "negative_goodness": 0.9206142365879483, + "pair_accuracy": 0.9127333333333333, + "positive_goodness": 4.234188641018338, + "runtime_seconds": 9.842778205871582 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6730506138218774, + "lr": 0.03, + "negative_goodness": 0.9573538195186191, + "pair_accuracy": 0.9035111111111112, + "positive_goodness": 4.0553096385531955, + "runtime_seconds": 9.85113000869751 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.605177257876926, + "lr": 0.03, + "negative_goodness": 0.8855095486746895, + "pair_accuracy": 0.9246444444338481, + "positive_goodness": 4.27328407965766, + "runtime_seconds": 9.979038715362549 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.593786323457294, + "lr": 0.03, + "negative_goodness": 0.8948529388639662, + "pair_accuracy": 0.9280222222222222, + "positive_goodness": 4.325428230158488, + "runtime_seconds": 9.944522857666016 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6185544545385573, + "lr": 0.03, + "negative_goodness": 0.9003954335954454, + "pair_accuracy": 0.9169777777565851, + "positive_goodness": 4.256947866990831, + "runtime_seconds": 9.83584976196289 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.647206793196996, + "lr": 0.03, + "negative_goodness": 0.9099014339870877, + "pair_accuracy": 0.9146, + "positive_goodness": 4.149663179185655, + "runtime_seconds": 9.877631664276123 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.651788014072842, + "lr": 0.03, + "negative_goodness": 0.9375842256016201, + "pair_accuracy": 0.9124666666666666, + "positive_goodness": 4.153271003299289, + "runtime_seconds": 9.966643810272217 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6388278070449829, + "lr": 0.03, + "negative_goodness": 0.9616028508504232, + "pair_accuracy": 0.9228888888888889, + "positive_goodness": 4.218381388176812, + "runtime_seconds": 9.824944734573364 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6065880462010701, + "lr": 0.03, + "negative_goodness": 0.8712106365627713, + "pair_accuracy": 0.9253555555237664, + "positive_goodness": 4.373270506456163, + "runtime_seconds": 9.811578273773193 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6617401881959704, + "lr": 0.03, + "negative_goodness": 0.9324241824256049, + "pair_accuracy": 0.9125333333015442, + "positive_goodness": 4.145188413492838, + "runtime_seconds": 9.807898044586182 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6153364201015896, + "lr": 0.03, + "negative_goodness": 0.9204057542800903, + "pair_accuracy": 0.9231333333333334, + "positive_goodness": 4.28296179945204, + "runtime_seconds": 9.9091317653656 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6083721281157599, + "lr": 0.03, + "negative_goodness": 0.8796086328506469, + "pair_accuracy": 0.9222222222116259, + "positive_goodness": 4.380474408806695, + "runtime_seconds": 9.79871416091919 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6233050287803014, + "lr": 0.03, + "negative_goodness": 0.8871852931499481, + "pair_accuracy": 0.9220666666666667, + "positive_goodness": 4.321192449188232, + "runtime_seconds": 9.780346393585205 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6077679660373264, + "lr": 0.03, + "negative_goodness": 0.8918207466125488, + "pair_accuracy": 0.9280222222223878, + "positive_goodness": 4.3507827926635745, + "runtime_seconds": 9.987098693847656 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6356136862966749, + "lr": 0.03, + "negative_goodness": 0.8857557588577271, + "pair_accuracy": 0.9117111111005147, + "positive_goodness": 4.242736364237468, + "runtime_seconds": 9.863014459609985 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6385602628601922, + "lr": 0.03, + "negative_goodness": 0.9178850177552965, + "pair_accuracy": 0.9188888889312744, + "positive_goodness": 4.194714624192979, + "runtime_seconds": 9.799524307250977 + } + ], + "layer": 21 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1425297839694553, + "lr": 0.03, + "negative_goodness": 1.6748451942443847, + "pair_accuracy": 0.8050666666719648, + "positive_goodness": 3.056648136647542, + "runtime_seconds": 9.737308502197266 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8923790642102559, + "lr": 0.03, + "negative_goodness": 1.353847053941091, + "pair_accuracy": 0.8753999999682108, + "positive_goodness": 3.412654547712538, + "runtime_seconds": 10.01595401763916 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8129586520406935, + "lr": 0.03, + "negative_goodness": 1.2405240736431546, + "pair_accuracy": 0.8864222222434149, + "positive_goodness": 3.568287437354194, + "runtime_seconds": 9.900436639785767 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7955885203997294, + "lr": 0.03, + "negative_goodness": 1.1718998906665379, + "pair_accuracy": 0.8754444444338481, + "positive_goodness": 3.6992203032599553, + "runtime_seconds": 9.875757694244385 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7842299243291219, + "lr": 0.03, + "negative_goodness": 1.1420214914957683, + "pair_accuracy": 0.8883555555767483, + "positive_goodness": 3.7052800753487483, + "runtime_seconds": 9.778049230575562 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7482156053860982, + "lr": 0.03, + "negative_goodness": 1.1155448295593262, + "pair_accuracy": 0.9011333333333333, + "positive_goodness": 3.860564830440945, + "runtime_seconds": 10.082462549209595 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7292594421439701, + "lr": 0.03, + "negative_goodness": 1.0747983111593458, + "pair_accuracy": 0.9118222222116258, + "positive_goodness": 3.8404936949835884, + "runtime_seconds": 9.84516954421997 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6668918752564325, + "lr": 0.03, + "negative_goodness": 0.9887104051113129, + "pair_accuracy": 0.9212444444444444, + "positive_goodness": 4.065972469160292, + "runtime_seconds": 9.873403072357178 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7316255342271593, + "lr": 0.03, + "negative_goodness": 1.0677384557088216, + "pair_accuracy": 0.8988666666560703, + "positive_goodness": 3.8815508778466117, + "runtime_seconds": 9.858925580978394 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6984536246617635, + "lr": 0.03, + "negative_goodness": 1.0185694818072848, + "pair_accuracy": 0.9075555555449591, + "positive_goodness": 3.9474600248124863, + "runtime_seconds": 10.084466457366943 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7156036849233839, + "lr": 0.03, + "negative_goodness": 1.030807160419888, + "pair_accuracy": 0.8941555555343628, + "positive_goodness": 3.9409784742567275, + "runtime_seconds": 9.820594072341919 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7107813565889994, + "lr": 0.03, + "negative_goodness": 1.0124943130228254, + "pair_accuracy": 0.9002888888888889, + "positive_goodness": 3.966812789493137, + "runtime_seconds": 9.927628993988037 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6887333458476597, + "lr": 0.03, + "negative_goodness": 0.9942567683166927, + "pair_accuracy": 0.9038222222222222, + "positive_goodness": 4.000730838521322, + "runtime_seconds": 9.804759740829468 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7070580721934636, + "lr": 0.03, + "negative_goodness": 1.0125714521143172, + "pair_accuracy": 0.8972888888888889, + "positive_goodness": 3.961963419511583, + "runtime_seconds": 10.087706089019775 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6696270593484243, + "lr": 0.03, + "negative_goodness": 0.9724699235916138, + "pair_accuracy": 0.9126444444444445, + "positive_goodness": 4.102872494252523, + "runtime_seconds": 9.75637149810791 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6926164064831204, + "lr": 0.03, + "negative_goodness": 0.9940088387383355, + "pair_accuracy": 0.9073999999894036, + "positive_goodness": 4.006097166146173, + "runtime_seconds": 9.930743217468262 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6735704031414456, + "lr": 0.03, + "negative_goodness": 0.9642101926061842, + "pair_accuracy": 0.9095555555449591, + "positive_goodness": 4.126646055306328, + "runtime_seconds": 9.983771324157715 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7001136427985297, + "lr": 0.03, + "negative_goodness": 1.0184299933115641, + "pair_accuracy": 0.9091111110899184, + "positive_goodness": 3.98192630704244, + "runtime_seconds": 10.044289588928223 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6329325511932373, + "lr": 0.03, + "negative_goodness": 0.9195326830969917, + "pair_accuracy": 0.9212888889206781, + "positive_goodness": 4.282900132115682, + "runtime_seconds": 9.789300918579102 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6626369604799482, + "lr": 0.03, + "negative_goodness": 0.9646712840822008, + "pair_accuracy": 0.9145333333333333, + "positive_goodness": 4.173599193912082, + "runtime_seconds": 9.859226942062378 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6400040479448107, + "lr": 0.03, + "negative_goodness": 0.9264581188201905, + "pair_accuracy": 0.9164888888994853, + "positive_goodness": 4.220302565214369, + "runtime_seconds": 9.887630462646484 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6566174213621352, + "lr": 0.03, + "negative_goodness": 0.9352625430425008, + "pair_accuracy": 0.9078888889047835, + "positive_goodness": 4.191010115581088, + "runtime_seconds": 9.992117166519165 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6767474761539035, + "lr": 0.03, + "negative_goodness": 0.9675698185602823, + "pair_accuracy": 0.9021555555582047, + "positive_goodness": 4.05461943819258, + "runtime_seconds": 9.834648847579956 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6552515185727014, + "lr": 0.03, + "negative_goodness": 0.9461156769646538, + "pair_accuracy": 0.9080222222222222, + "positive_goodness": 4.176473136647543, + "runtime_seconds": 9.794219255447388 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6331706919352214, + "lr": 0.03, + "negative_goodness": 0.8990879618008931, + "pair_accuracy": 0.9164666666719649, + "positive_goodness": 4.312531542311774, + "runtime_seconds": 9.950608730316162 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6517698099454244, + "lr": 0.03, + "negative_goodness": 0.9325013540532854, + "pair_accuracy": 0.9100888888782925, + "positive_goodness": 4.1482281583574085, + "runtime_seconds": 9.931151151657104 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6668737206194136, + "lr": 0.03, + "negative_goodness": 0.9299625075075362, + "pair_accuracy": 0.904, + "positive_goodness": 4.2235485136244035, + "runtime_seconds": 9.797780752182007 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6803004922442966, + "lr": 0.03, + "negative_goodness": 0.9700965859942966, + "pair_accuracy": 0.9058222222328186, + "positive_goodness": 4.098268174489339, + "runtime_seconds": 9.77621865272522 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6687055692248874, + "lr": 0.03, + "negative_goodness": 0.9336308770073785, + "pair_accuracy": 0.9015333333015442, + "positive_goodness": 4.139167847972446, + "runtime_seconds": 9.908114433288574 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.674937947596444, + "lr": 0.03, + "negative_goodness": 0.9608933854950799, + "pair_accuracy": 0.9032444444444444, + "positive_goodness": 4.0590745944552955, + "runtime_seconds": 9.898412466049194 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6369926897419823, + "lr": 0.03, + "negative_goodness": 0.9234248402489557, + "pair_accuracy": 0.9160888888676961, + "positive_goodness": 4.234926638200548, + "runtime_seconds": 9.77702021598816 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6650424405972163, + "lr": 0.03, + "negative_goodness": 0.9294843626287248, + "pair_accuracy": 0.9093555555555556, + "positive_goodness": 4.2209758759392635, + "runtime_seconds": 9.73686170578003 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6328603158526951, + "lr": 0.03, + "negative_goodness": 0.9175892281638252, + "pair_accuracy": 0.9153777777459886, + "positive_goodness": 4.292430530378554, + "runtime_seconds": 9.927372455596924 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6425685404406654, + "lr": 0.03, + "negative_goodness": 0.8930337191316816, + "pair_accuracy": 0.9101555555555556, + "positive_goodness": 4.227550921800401, + "runtime_seconds": 9.905876874923706 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6800879379696316, + "lr": 0.03, + "negative_goodness": 0.945432575501336, + "pair_accuracy": 0.9017111110899183, + "positive_goodness": 4.141923625119527, + "runtime_seconds": 9.9472017288208 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6102805757866966, + "lr": 0.03, + "negative_goodness": 0.8663110853698519, + "pair_accuracy": 0.9226, + "positive_goodness": 4.4483737358093265, + "runtime_seconds": 9.674419164657593 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6813071437809203, + "lr": 0.03, + "negative_goodness": 0.9577986324681176, + "pair_accuracy": 0.8999111111111111, + "positive_goodness": 4.1521433356391055, + "runtime_seconds": 9.982878923416138 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6590344801372952, + "lr": 0.03, + "negative_goodness": 0.9222008040746053, + "pair_accuracy": 0.9090666667196485, + "positive_goodness": 4.2113091326819525, + "runtime_seconds": 9.844735383987427 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6278884483973185, + "lr": 0.03, + "negative_goodness": 0.8756357998636034, + "pair_accuracy": 0.9180444444603391, + "positive_goodness": 4.328324479293824, + "runtime_seconds": 9.86838984489441 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.663906315252516, + "lr": 0.03, + "negative_goodness": 0.924801326391432, + "pair_accuracy": 0.9035777777989705, + "positive_goodness": 4.162703207227919, + "runtime_seconds": 9.79780101776123 + } + ], + "layer": 22 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9063953370836046, + "lr": 0.03, + "negative_goodness": 1.2821892659929064, + "pair_accuracy": 0.8533111111429003, + "positive_goodness": 3.3038999427371554, + "runtime_seconds": 9.993561506271362 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8675342436260647, + "lr": 0.03, + "negative_goodness": 1.3051679518381754, + "pair_accuracy": 0.8644444444974263, + "positive_goodness": 3.301750230111016, + "runtime_seconds": 9.817720651626587 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7510025411817762, + "lr": 0.03, + "negative_goodness": 1.1451432731204563, + "pair_accuracy": 0.8891777778307597, + "positive_goodness": 3.7062560922410754, + "runtime_seconds": 9.872278451919556 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7809694588343302, + "lr": 0.03, + "negative_goodness": 1.1642433547125923, + "pair_accuracy": 0.8885777777671814, + "positive_goodness": 3.6940874348110624, + "runtime_seconds": 9.623603820800781 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7432291091707017, + "lr": 0.03, + "negative_goodness": 1.127787467151218, + "pair_accuracy": 0.8908444444550409, + "positive_goodness": 3.739205248514811, + "runtime_seconds": 10.004968166351318 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.73954883468416, + "lr": 0.03, + "negative_goodness": 1.0970384375254314, + "pair_accuracy": 0.9014, + "positive_goodness": 3.7801341794755725, + "runtime_seconds": 9.740286588668823 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.8149913377443949, + "lr": 0.03, + "negative_goodness": 1.1726405399428474, + "pair_accuracy": 0.8681555555555556, + "positive_goodness": 3.522812597571479, + "runtime_seconds": 9.822396755218506 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6922095374531216, + "lr": 0.03, + "negative_goodness": 1.038653727918201, + "pair_accuracy": 0.9144, + "positive_goodness": 3.9119723178863524, + "runtime_seconds": 9.660771131515503 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7214454894277784, + "lr": 0.03, + "negative_goodness": 1.0697889428668552, + "pair_accuracy": 0.8971555555449592, + "positive_goodness": 3.8330642506917316, + "runtime_seconds": 9.998628854751587 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7369461787170835, + "lr": 0.03, + "negative_goodness": 1.0502558211750455, + "pair_accuracy": 0.8898666666666667, + "positive_goodness": 3.8683310238308377, + "runtime_seconds": 9.808428764343262 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6853813572565715, + "lr": 0.03, + "negative_goodness": 1.052326403533088, + "pair_accuracy": 0.9115333333757188, + "positive_goodness": 3.954663879500495, + "runtime_seconds": 9.848608255386353 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6615131763988071, + "lr": 0.03, + "negative_goodness": 1.0297020576265123, + "pair_accuracy": 0.9232666666666667, + "positive_goodness": 3.9986368975321454, + "runtime_seconds": 9.737563371658325 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.619598372443517, + "lr": 0.03, + "negative_goodness": 0.9532510510232713, + "pair_accuracy": 0.9282444444444444, + "positive_goodness": 4.265007527245416, + "runtime_seconds": 9.994598388671875 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6639747494538625, + "lr": 0.03, + "negative_goodness": 0.9555062126053704, + "pair_accuracy": 0.9083777777777777, + "positive_goodness": 4.092525115542942, + "runtime_seconds": 9.749613046646118 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6478851926379734, + "lr": 0.03, + "negative_goodness": 0.9842791645579868, + "pair_accuracy": 0.9207777777777778, + "positive_goodness": 4.123988693110148, + "runtime_seconds": 9.865755796432495 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6974610544257693, + "lr": 0.03, + "negative_goodness": 1.0332402134391996, + "pair_accuracy": 0.9097111111111111, + "positive_goodness": 3.9367523871527776, + "runtime_seconds": 9.996083974838257 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7056979952017466, + "lr": 0.03, + "negative_goodness": 1.0399852906968858, + "pair_accuracy": 0.9128, + "positive_goodness": 3.9697699119567873, + "runtime_seconds": 10.002360105514526 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6517980801529355, + "lr": 0.03, + "negative_goodness": 0.9620999343077342, + "pair_accuracy": 0.9183111111111111, + "positive_goodness": 4.121725176493327, + "runtime_seconds": 10.005833387374878 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6881412785265181, + "lr": 0.03, + "negative_goodness": 0.9881109303050571, + "pair_accuracy": 0.9124666666560702, + "positive_goodness": 4.060151494428847, + "runtime_seconds": 9.995383501052856 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6716578653547499, + "lr": 0.03, + "negative_goodness": 0.9827705767446094, + "pair_accuracy": 0.9105111111111112, + "positive_goodness": 4.128630847337511, + "runtime_seconds": 9.995143175125122 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6772952109601763, + "lr": 0.03, + "negative_goodness": 1.011960527250502, + "pair_accuracy": 0.9088222222222222, + "positive_goodness": 4.0527343539767795, + "runtime_seconds": 9.992181301116943 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6212303226047092, + "lr": 0.03, + "negative_goodness": 0.9325257623248631, + "pair_accuracy": 0.9217777777777778, + "positive_goodness": 4.289334576310051, + "runtime_seconds": 9.991932392120361 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6340920580969917, + "lr": 0.03, + "negative_goodness": 0.9061255318535699, + "pair_accuracy": 0.9119999999788072, + "positive_goodness": 4.260370438808865, + "runtime_seconds": 9.995000839233398 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6758303497314453, + "lr": 0.03, + "negative_goodness": 0.9926710238986545, + "pair_accuracy": 0.9180666666666667, + "positive_goodness": 4.056352860260009, + "runtime_seconds": 10.000931978225708 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.611085856130388, + "lr": 0.03, + "negative_goodness": 0.8880085057576498, + "pair_accuracy": 0.9240000000529819, + "positive_goodness": 4.304467231411404, + "runtime_seconds": 9.996618747711182 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6557094589445326, + "lr": 0.03, + "negative_goodness": 0.964448283598158, + "pair_accuracy": 0.9146222222116258, + "positive_goodness": 4.180238928265042, + "runtime_seconds": 9.999067068099976 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6466414227061802, + "lr": 0.03, + "negative_goodness": 0.9475020941204495, + "pair_accuracy": 0.9143555555449592, + "positive_goodness": 4.146114269595676, + "runtime_seconds": 10.003045320510864 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6754566741784414, + "lr": 0.03, + "negative_goodness": 0.9868158641603257, + "pair_accuracy": 0.9164, + "positive_goodness": 4.118335230339898, + "runtime_seconds": 9.997300148010254 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6259289637565613, + "lr": 0.03, + "negative_goodness": 0.9233284832424588, + "pair_accuracy": 0.9174888888782925, + "positive_goodness": 4.29570474243164, + "runtime_seconds": 9.999748468399048 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5814369130002128, + "lr": 0.03, + "negative_goodness": 0.8577498827351464, + "pair_accuracy": 0.9371555555555555, + "positive_goodness": 4.485380961524116, + "runtime_seconds": 10.005926609039307 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6267607856326634, + "lr": 0.03, + "negative_goodness": 0.8818120198249817, + "pair_accuracy": 0.9150888889418708, + "positive_goodness": 4.398649579048157, + "runtime_seconds": 10.00106430053711 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6620328968418969, + "lr": 0.03, + "negative_goodness": 0.9225231530984243, + "pair_accuracy": 0.9044444444444445, + "positive_goodness": 4.1740453092787, + "runtime_seconds": 9.99000859260559 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6474589947938919, + "lr": 0.03, + "negative_goodness": 0.9178750205304887, + "pair_accuracy": 0.9060222222222222, + "positive_goodness": 4.321573487345377, + "runtime_seconds": 9.998370885848999 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6233022975073921, + "lr": 0.03, + "negative_goodness": 0.9029258880827162, + "pair_accuracy": 0.9207111111429003, + "positive_goodness": 4.4061433108011885, + "runtime_seconds": 9.991890668869019 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5783767602973514, + "lr": 0.03, + "negative_goodness": 0.8651227012899186, + "pair_accuracy": 0.9358222222222222, + "positive_goodness": 4.475177309672038, + "runtime_seconds": 9.99657154083252 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6251813918219672, + "lr": 0.03, + "negative_goodness": 0.8834655347188314, + "pair_accuracy": 0.9236888889100816, + "positive_goodness": 4.364158306672838, + "runtime_seconds": 10.003841161727905 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6537203983306885, + "lr": 0.03, + "negative_goodness": 0.9134703694449531, + "pair_accuracy": 0.9079777777779433, + "positive_goodness": 4.3280288617451985, + "runtime_seconds": 9.995168447494507 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6146174406475491, + "lr": 0.03, + "negative_goodness": 0.8828985183080037, + "pair_accuracy": 0.9216222222301695, + "positive_goodness": 4.359011995983124, + "runtime_seconds": 9.99098825454712 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6085645258161757, + "lr": 0.03, + "negative_goodness": 0.8688129183239407, + "pair_accuracy": 0.9174444444444444, + "positive_goodness": 4.401026074769762, + "runtime_seconds": 10.006775140762329 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6176732622888353, + "lr": 0.03, + "negative_goodness": 0.8648968572086758, + "pair_accuracy": 0.9181111110899184, + "positive_goodness": 4.425267076237996, + "runtime_seconds": 9.99769377708435 + } + ], + "layer": 23 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.1499169572724237, + "lr": 0.03, + "negative_goodness": 1.7150626821623909, + "pair_accuracy": 0.787733333322737, + "positive_goodness": 3.149105027219984, + "runtime_seconds": 10.017301321029663 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.862276980887519, + "lr": 0.03, + "negative_goodness": 1.3761818639119465, + "pair_accuracy": 0.8641333333121406, + "positive_goodness": 3.647079677073161, + "runtime_seconds": 10.028666734695435 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8208829814010197, + "lr": 0.03, + "negative_goodness": 1.2855739379141067, + "pair_accuracy": 0.8740888888888889, + "positive_goodness": 3.7101817275153266, + "runtime_seconds": 10.011423110961914 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7711763194825915, + "lr": 0.03, + "negative_goodness": 1.2318105571746827, + "pair_accuracy": 0.897355555566152, + "positive_goodness": 3.8375424824608695, + "runtime_seconds": 10.01513934135437 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7571718318568336, + "lr": 0.03, + "negative_goodness": 1.1855391702387068, + "pair_accuracy": 0.8918444444444444, + "positive_goodness": 3.792354956902398, + "runtime_seconds": 10.015815258026123 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7210684820810954, + "lr": 0.03, + "negative_goodness": 1.1440414271884494, + "pair_accuracy": 0.9085777777784401, + "positive_goodness": 3.932660065481398, + "runtime_seconds": 10.013560056686401 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7495236698998345, + "lr": 0.03, + "negative_goodness": 1.1299403498331706, + "pair_accuracy": 0.890511111164093, + "positive_goodness": 3.824982868385315, + "runtime_seconds": 10.018441915512085 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7080749800788032, + "lr": 0.03, + "negative_goodness": 1.071161287996504, + "pair_accuracy": 0.898533333322737, + "positive_goodness": 3.999514247555203, + "runtime_seconds": 10.00893235206604 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6739809307522244, + "lr": 0.03, + "negative_goodness": 1.0411093948364258, + "pair_accuracy": 0.9071333333386316, + "positive_goodness": 4.042894931284587, + "runtime_seconds": 10.016138553619385 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7129645613008075, + "lr": 0.03, + "negative_goodness": 1.0721798154989879, + "pair_accuracy": 0.9036666666666666, + "positive_goodness": 3.969536837005615, + "runtime_seconds": 10.027096033096313 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6188611870739195, + "lr": 0.03, + "negative_goodness": 0.9844074019696978, + "pair_accuracy": 0.9282, + "positive_goodness": 4.317660299428304, + "runtime_seconds": 10.066328525543213 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6714865581724379, + "lr": 0.03, + "negative_goodness": 1.0220262929704453, + "pair_accuracy": 0.9176000000158946, + "positive_goodness": 4.115154445054796, + "runtime_seconds": 10.012224674224854 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.664500237941742, + "lr": 0.03, + "negative_goodness": 1.0022587037086488, + "pair_accuracy": 0.9021555555237664, + "positive_goodness": 4.171603312386407, + "runtime_seconds": 10.020734310150146 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6692168541457918, + "lr": 0.03, + "negative_goodness": 0.9938885171201494, + "pair_accuracy": 0.9034, + "positive_goodness": 4.131118164401585, + "runtime_seconds": 10.012383222579956 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6614815886603461, + "lr": 0.03, + "negative_goodness": 0.986999926410781, + "pair_accuracy": 0.9047555555555555, + "positive_goodness": 4.1747631461249455, + "runtime_seconds": 10.009117364883423 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.63201781651179, + "lr": 0.03, + "negative_goodness": 0.942918321662479, + "pair_accuracy": 0.9135777777777778, + "positive_goodness": 4.246121618652344, + "runtime_seconds": 10.012680768966675 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6356402978579203, + "lr": 0.03, + "negative_goodness": 0.9277490626759, + "pair_accuracy": 0.9100444444444444, + "positive_goodness": 4.3026078637441, + "runtime_seconds": 10.016139268875122 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6638353772640229, + "lr": 0.03, + "negative_goodness": 1.0112152004241943, + "pair_accuracy": 0.9051111111005147, + "positive_goodness": 4.124814250013563, + "runtime_seconds": 10.009111642837524 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6266249609576331, + "lr": 0.03, + "negative_goodness": 0.9189503387451172, + "pair_accuracy": 0.9130444444444444, + "positive_goodness": 4.305652025943332, + "runtime_seconds": 10.04856538772583 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6190559244314829, + "lr": 0.03, + "negative_goodness": 0.9227825326389737, + "pair_accuracy": 0.9111777777777778, + "positive_goodness": 4.292675548214382, + "runtime_seconds": 10.066322088241577 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6399056768841214, + "lr": 0.03, + "negative_goodness": 0.9369923552195231, + "pair_accuracy": 0.9024888888994853, + "positive_goodness": 4.251150292883979, + "runtime_seconds": 10.010757446289062 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7033529266569349, + "lr": 0.03, + "negative_goodness": 1.0006875924746195, + "pair_accuracy": 0.8905999999788072, + "positive_goodness": 4.014221489164564, + "runtime_seconds": 10.012919187545776 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6180129198127323, + "lr": 0.03, + "negative_goodness": 0.8995996591885884, + "pair_accuracy": 0.9091777777777778, + "positive_goodness": 4.355004449886746, + "runtime_seconds": 10.017956018447876 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6342668400552538, + "lr": 0.03, + "negative_goodness": 0.9138778467390273, + "pair_accuracy": 0.9172444444444444, + "positive_goodness": 4.278461887359619, + "runtime_seconds": 10.013375520706177 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5925407177819146, + "lr": 0.03, + "negative_goodness": 0.8596929501427545, + "pair_accuracy": 0.9178222222752042, + "positive_goodness": 4.371769077661303, + "runtime_seconds": 10.019278049468994 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6452767585860358, + "lr": 0.03, + "negative_goodness": 0.9230077415678236, + "pair_accuracy": 0.9049333333333334, + "positive_goodness": 4.377670126046075, + "runtime_seconds": 10.013969659805298 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6654049845351113, + "lr": 0.03, + "negative_goodness": 0.9342033046404521, + "pair_accuracy": 0.8941555555555556, + "positive_goodness": 4.1539048148261175, + "runtime_seconds": 10.00963306427002 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6372471890343561, + "lr": 0.03, + "negative_goodness": 0.933024249235789, + "pair_accuracy": 0.9185333333227369, + "positive_goodness": 4.264212601555719, + "runtime_seconds": 10.002684354782104 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6337039084328545, + "lr": 0.03, + "negative_goodness": 0.9379842082235548, + "pair_accuracy": 0.9176888888782925, + "positive_goodness": 4.238111461003621, + "runtime_seconds": 10.023680448532104 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6275699707931942, + "lr": 0.03, + "negative_goodness": 0.8846522171974183, + "pair_accuracy": 0.9067111111111111, + "positive_goodness": 4.398529322391086, + "runtime_seconds": 10.04190444946289 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5715102721797095, + "lr": 0.03, + "negative_goodness": 0.8401175433794658, + "pair_accuracy": 0.9224, + "positive_goodness": 4.517007708316379, + "runtime_seconds": 10.053844928741455 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6423697421577241, + "lr": 0.03, + "negative_goodness": 0.8938210610734092, + "pair_accuracy": 0.9060222222222222, + "positive_goodness": 4.311148053148058, + "runtime_seconds": 10.009666919708252 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6320409386740791, + "lr": 0.03, + "negative_goodness": 0.8856709917068482, + "pair_accuracy": 0.9034444444444445, + "positive_goodness": 4.313630762524075, + "runtime_seconds": 10.049237251281738 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6177026893880632, + "lr": 0.03, + "negative_goodness": 0.8883920827335782, + "pair_accuracy": 0.9050666666666667, + "positive_goodness": 4.444708211686876, + "runtime_seconds": 10.078399181365967 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.585146470811632, + "lr": 0.03, + "negative_goodness": 0.8236493263138666, + "pair_accuracy": 0.9189555555449592, + "positive_goodness": 4.562109501012166, + "runtime_seconds": 10.060802221298218 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5994775459925333, + "lr": 0.03, + "negative_goodness": 0.8714770076857673, + "pair_accuracy": 0.9244444444974264, + "positive_goodness": 4.432496951686011, + "runtime_seconds": 10.011601686477661 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6660344969643487, + "lr": 0.03, + "negative_goodness": 0.8966664981842041, + "pair_accuracy": 0.899955555597941, + "positive_goodness": 4.264954697375828, + "runtime_seconds": 10.015611410140991 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6699272077401479, + "lr": 0.03, + "negative_goodness": 0.911514763185713, + "pair_accuracy": 0.8932, + "positive_goodness": 4.237340748680962, + "runtime_seconds": 10.010088205337524 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6289260950194465, + "lr": 0.03, + "negative_goodness": 0.8936687642945184, + "pair_accuracy": 0.9038222222010295, + "positive_goodness": 4.356372689988878, + "runtime_seconds": 10.015506267547607 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6410751819186741, + "lr": 0.03, + "negative_goodness": 0.8975582965321011, + "pair_accuracy": 0.9100222222275204, + "positive_goodness": 4.347843708377415, + "runtime_seconds": 10.014419555664062 + } + ], + "layer": 24 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9156956342273288, + "lr": 0.03, + "negative_goodness": 1.2744096091376411, + "pair_accuracy": 0.8460888889100816, + "positive_goodness": 3.241411830223931, + "runtime_seconds": 9.994781017303467 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7945889218542311, + "lr": 0.03, + "negative_goodness": 1.1657386093139648, + "pair_accuracy": 0.8785333333333334, + "positive_goodness": 3.633951196416219, + "runtime_seconds": 9.99396824836731 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8682658582157559, + "lr": 0.03, + "negative_goodness": 1.2649704164557987, + "pair_accuracy": 0.8489555555555556, + "positive_goodness": 3.4497206462436254, + "runtime_seconds": 9.997618675231934 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7966600628111098, + "lr": 0.03, + "negative_goodness": 1.201978729777866, + "pair_accuracy": 0.8780222221904331, + "positive_goodness": 3.627027499262492, + "runtime_seconds": 9.997895240783691 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8397631089263492, + "lr": 0.03, + "negative_goodness": 1.1884865045759414, + "pair_accuracy": 0.8517777777671814, + "positive_goodness": 3.491463515387641, + "runtime_seconds": 9.993839263916016 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.746947581799825, + "lr": 0.03, + "negative_goodness": 1.1076670710139804, + "pair_accuracy": 0.8816888888888889, + "positive_goodness": 3.81700332836575, + "runtime_seconds": 9.987244367599487 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.8099665078692966, + "lr": 0.03, + "negative_goodness": 1.1864122143851386, + "pair_accuracy": 0.8606444444656373, + "positive_goodness": 3.5627292899661596, + "runtime_seconds": 9.993288040161133 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7136135042455461, + "lr": 0.03, + "negative_goodness": 1.066845770380232, + "pair_accuracy": 0.8866666666666667, + "positive_goodness": 3.960549517144097, + "runtime_seconds": 9.988036155700684 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7614226719114515, + "lr": 0.03, + "negative_goodness": 1.1130996048397488, + "pair_accuracy": 0.8873777777777778, + "positive_goodness": 3.799475735918681, + "runtime_seconds": 9.989776611328125 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7824494532267252, + "lr": 0.03, + "negative_goodness": 1.1653124734454685, + "pair_accuracy": 0.8797777777777778, + "positive_goodness": 3.644248066965739, + "runtime_seconds": 9.994952917098999 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7705715397251977, + "lr": 0.03, + "negative_goodness": 1.143079610384835, + "pair_accuracy": 0.876533333322737, + "positive_goodness": 3.6588973770989313, + "runtime_seconds": 9.990703582763672 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7226027072482639, + "lr": 0.03, + "negative_goodness": 1.084085762723287, + "pair_accuracy": 0.8949111111323039, + "positive_goodness": 3.8704359272003175, + "runtime_seconds": 9.985678434371948 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.74963895939721, + "lr": 0.03, + "negative_goodness": 1.0908094732920328, + "pair_accuracy": 0.8821333333121406, + "positive_goodness": 3.769889411248101, + "runtime_seconds": 9.993733882904053 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7293800431039599, + "lr": 0.03, + "negative_goodness": 1.0752734311209784, + "pair_accuracy": 0.8870222222752041, + "positive_goodness": 3.9099451125250924, + "runtime_seconds": 9.99046540260315 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7291933175590303, + "lr": 0.03, + "negative_goodness": 1.0826821887466642, + "pair_accuracy": 0.8966222222222222, + "positive_goodness": 3.852228061336941, + "runtime_seconds": 9.995542764663696 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.756208097574446, + "lr": 0.03, + "negative_goodness": 1.1300659677505493, + "pair_accuracy": 0.8878222222646077, + "positive_goodness": 3.752500954225328, + "runtime_seconds": 9.968535661697388 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7332867321014405, + "lr": 0.03, + "negative_goodness": 1.076296139907837, + "pair_accuracy": 0.885777777778109, + "positive_goodness": 3.872284299045139, + "runtime_seconds": 10.034621238708496 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7632201564629872, + "lr": 0.03, + "negative_goodness": 1.0851504744476743, + "pair_accuracy": 0.8796888888888889, + "positive_goodness": 3.7430934202406143, + "runtime_seconds": 10.047853946685791 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7509747156937917, + "lr": 0.03, + "negative_goodness": 1.1061250494480133, + "pair_accuracy": 0.8835111111111111, + "positive_goodness": 3.7727191251118977, + "runtime_seconds": 10.051154375076294 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7318355027649138, + "lr": 0.03, + "negative_goodness": 1.0818484890143076, + "pair_accuracy": 0.8953333333333333, + "positive_goodness": 3.8571338484022353, + "runtime_seconds": 10.051645040512085 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7066634716272354, + "lr": 0.03, + "negative_goodness": 1.052059801647398, + "pair_accuracy": 0.8991111111111111, + "positive_goodness": 3.94099151441786, + "runtime_seconds": 10.057369709014893 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7384962366369036, + "lr": 0.03, + "negative_goodness": 1.0612862804306877, + "pair_accuracy": 0.8858888888888888, + "positive_goodness": 3.8849198235405815, + "runtime_seconds": 10.06177282333374 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7040973768870036, + "lr": 0.03, + "negative_goodness": 1.020586633321974, + "pair_accuracy": 0.8987777777671814, + "positive_goodness": 3.9788957151201036, + "runtime_seconds": 10.055988073348999 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6592743795606825, + "lr": 0.03, + "negative_goodness": 0.9730433664957683, + "pair_accuracy": 0.9001999999894036, + "positive_goodness": 4.187326781378852, + "runtime_seconds": 10.051180362701416 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6615950688680013, + "lr": 0.03, + "negative_goodness": 0.9563438813527425, + "pair_accuracy": 0.9043777777883741, + "positive_goodness": 4.179543604755402, + "runtime_seconds": 10.05319595336914 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7366856751547919, + "lr": 0.03, + "negative_goodness": 1.0567808694415621, + "pair_accuracy": 0.8915777777777778, + "positive_goodness": 3.952175377358331, + "runtime_seconds": 10.05028748512268 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7054878774272071, + "lr": 0.03, + "negative_goodness": 1.0331461829397413, + "pair_accuracy": 0.8852888888888889, + "positive_goodness": 4.0464912287818064, + "runtime_seconds": 10.052772283554077 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.7098597661336263, + "lr": 0.03, + "negative_goodness": 1.0270559885236952, + "pair_accuracy": 0.8951555555555556, + "positive_goodness": 3.9318908452351886, + "runtime_seconds": 10.057686805725098 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6555931070433723, + "lr": 0.03, + "negative_goodness": 0.9832627337667678, + "pair_accuracy": 0.9128888888676961, + "positive_goodness": 4.22407068256802, + "runtime_seconds": 10.048103094100952 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7677071920288934, + "lr": 0.03, + "negative_goodness": 1.1082024140463935, + "pair_accuracy": 0.8770444444444444, + "positive_goodness": 3.734865996551514, + "runtime_seconds": 10.043132543563843 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6949661350170772, + "lr": 0.03, + "negative_goodness": 0.9793019681771596, + "pair_accuracy": 0.8981555555555556, + "positive_goodness": 3.977219241163466, + "runtime_seconds": 10.056259393692017 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6775262031767103, + "lr": 0.03, + "negative_goodness": 0.9991601709365845, + "pair_accuracy": 0.8992666666772631, + "positive_goodness": 4.188407940038045, + "runtime_seconds": 10.046976566314697 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6673039017995198, + "lr": 0.03, + "negative_goodness": 0.9816924666934543, + "pair_accuracy": 0.9094444444447756, + "positive_goodness": 4.10413924530877, + "runtime_seconds": 10.054481506347656 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6545212643702825, + "lr": 0.03, + "negative_goodness": 0.9611306492196189, + "pair_accuracy": 0.9098, + "positive_goodness": 4.126805654652913, + "runtime_seconds": 10.057473421096802 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6828081083933513, + "lr": 0.03, + "negative_goodness": 1.0129584018071494, + "pair_accuracy": 0.90564444444461, + "positive_goodness": 4.124658046298557, + "runtime_seconds": 10.051753520965576 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6738656066046821, + "lr": 0.03, + "negative_goodness": 0.9749763679504394, + "pair_accuracy": 0.8985777777724796, + "positive_goodness": 4.046684522056579, + "runtime_seconds": 10.048254489898682 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6852923019621108, + "lr": 0.03, + "negative_goodness": 1.004231835132175, + "pair_accuracy": 0.9025333333757188, + "positive_goodness": 4.0579376986609565, + "runtime_seconds": 10.05697250366211 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.7142373358620537, + "lr": 0.03, + "negative_goodness": 1.0471296072853935, + "pair_accuracy": 0.8958666666560703, + "positive_goodness": 3.946293104786343, + "runtime_seconds": 10.04898715019226 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6599911453458998, + "lr": 0.03, + "negative_goodness": 0.9481218861473931, + "pair_accuracy": 0.9043333333121406, + "positive_goodness": 4.1878479051378035, + "runtime_seconds": 10.054591178894043 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.7247016898155212, + "lr": 0.03, + "negative_goodness": 1.011338179450565, + "pair_accuracy": 0.8912222222222222, + "positive_goodness": 3.9476418816460503, + "runtime_seconds": 10.056423425674438 + } + ], + "layer": 25 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0754077345954047, + "lr": 0.03, + "negative_goodness": 1.596887409930759, + "pair_accuracy": 0.810977777783076, + "positive_goodness": 3.1357420365651447, + "runtime_seconds": 10.076606750488281 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.9762479615529378, + "lr": 0.03, + "negative_goodness": 1.4439638203514946, + "pair_accuracy": 0.8263333333174387, + "positive_goodness": 3.2491509752909344, + "runtime_seconds": 10.0710768699646 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.9164666673660279, + "lr": 0.03, + "negative_goodness": 1.3571886390686034, + "pair_accuracy": 0.8392222222540113, + "positive_goodness": 3.367342446306017, + "runtime_seconds": 10.079352617263794 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8966873063776228, + "lr": 0.03, + "negative_goodness": 1.3267776253700256, + "pair_accuracy": 0.8410444444444445, + "positive_goodness": 3.4256441567314995, + "runtime_seconds": 10.07205843925476 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8861953901078966, + "lr": 0.03, + "negative_goodness": 1.3362455600738525, + "pair_accuracy": 0.8413999999682109, + "positive_goodness": 3.4328766171773273, + "runtime_seconds": 10.077913522720337 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.8172218341827393, + "lr": 0.03, + "negative_goodness": 1.2520480536672804, + "pair_accuracy": 0.867511111164093, + "positive_goodness": 3.653805304675632, + "runtime_seconds": 10.075621366500854 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.8077260306464301, + "lr": 0.03, + "negative_goodness": 1.2514317898432414, + "pair_accuracy": 0.8684000000423855, + "positive_goodness": 3.6675525063408747, + "runtime_seconds": 10.096473217010498 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7875268301433987, + "lr": 0.03, + "negative_goodness": 1.2261982879426745, + "pair_accuracy": 0.8811777778307597, + "positive_goodness": 3.673423330391778, + "runtime_seconds": 10.101830005645752 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7377033579720391, + "lr": 0.03, + "negative_goodness": 1.1584482105255127, + "pair_accuracy": 0.8978000000423855, + "positive_goodness": 3.8597183612823485, + "runtime_seconds": 10.044204950332642 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7154156240463256, + "lr": 0.03, + "negative_goodness": 1.115414048830668, + "pair_accuracy": 0.9035111111005147, + "positive_goodness": 4.053048053910997, + "runtime_seconds": 9.959750890731812 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7976701223426395, + "lr": 0.03, + "negative_goodness": 1.210196309651269, + "pair_accuracy": 0.8828222222116259, + "positive_goodness": 3.6833365005493164, + "runtime_seconds": 9.949445724487305 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7235005657619901, + "lr": 0.03, + "negative_goodness": 1.1155193075180054, + "pair_accuracy": 0.8973333333545261, + "positive_goodness": 3.963395411406623, + "runtime_seconds": 9.962594747543335 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7296312120755514, + "lr": 0.03, + "negative_goodness": 1.1331677461094327, + "pair_accuracy": 0.8908888888782925, + "positive_goodness": 3.917820081753201, + "runtime_seconds": 9.984773397445679 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7701314965142144, + "lr": 0.03, + "negative_goodness": 1.1441217078526815, + "pair_accuracy": 0.8783555556085375, + "positive_goodness": 3.8138339547475177, + "runtime_seconds": 9.987700700759888 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7311767669889662, + "lr": 0.03, + "negative_goodness": 1.1132369034449259, + "pair_accuracy": 0.8962888889100816, + "positive_goodness": 3.9018770863850913, + "runtime_seconds": 10.01404881477356 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7168961059040494, + "lr": 0.03, + "negative_goodness": 1.0772306450102065, + "pair_accuracy": 0.8943777777777778, + "positive_goodness": 4.018947581905789, + "runtime_seconds": 10.054666996002197 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7587659620390999, + "lr": 0.03, + "negative_goodness": 1.119686351384057, + "pair_accuracy": 0.8862666666666666, + "positive_goodness": 3.8785154051038955, + "runtime_seconds": 10.049308061599731 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6935367529021369, + "lr": 0.03, + "negative_goodness": 1.0563864450772604, + "pair_accuracy": 0.9101777777777778, + "positive_goodness": 4.00179469265408, + "runtime_seconds": 10.040748357772827 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6761993137306638, + "lr": 0.03, + "negative_goodness": 1.0326671551810371, + "pair_accuracy": 0.9130666666666667, + "positive_goodness": 4.094303120083279, + "runtime_seconds": 10.050718307495117 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.696706811009513, + "lr": 0.03, + "negative_goodness": 1.0560373175726996, + "pair_accuracy": 0.8994444444444445, + "positive_goodness": 4.065681282552084, + "runtime_seconds": 10.033474445343018 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6921081252415975, + "lr": 0.03, + "negative_goodness": 1.0280698344972399, + "pair_accuracy": 0.9017555555555555, + "positive_goodness": 4.09735645904541, + "runtime_seconds": 10.010486841201782 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7416076448016696, + "lr": 0.03, + "negative_goodness": 1.1050043071534899, + "pair_accuracy": 0.8888666666666667, + "positive_goodness": 3.917665804417928, + "runtime_seconds": 9.959796667098999 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7092351714875963, + "lr": 0.03, + "negative_goodness": 1.0465496017032199, + "pair_accuracy": 0.8946444444444445, + "positive_goodness": 3.945064657465617, + "runtime_seconds": 9.967967748641968 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6807079020182292, + "lr": 0.03, + "negative_goodness": 1.0224991474469503, + "pair_accuracy": 0.9042444444444444, + "positive_goodness": 4.167212455240885, + "runtime_seconds": 9.962337493896484 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.683357845211029, + "lr": 0.03, + "negative_goodness": 1.0228102578163147, + "pair_accuracy": 0.8981777777777777, + "positive_goodness": 4.122249009026421, + "runtime_seconds": 9.965726613998413 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6621526769081751, + "lr": 0.03, + "negative_goodness": 0.9899435514873929, + "pair_accuracy": 0.9129777777777778, + "positive_goodness": 4.182604258643257, + "runtime_seconds": 9.963778257369995 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6482936428070069, + "lr": 0.03, + "negative_goodness": 0.9618897447374132, + "pair_accuracy": 0.9163999999841054, + "positive_goodness": 4.23324228439331, + "runtime_seconds": 9.987390518188477 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6338101997163561, + "lr": 0.03, + "negative_goodness": 0.9734577690760294, + "pair_accuracy": 0.9259555555449591, + "positive_goodness": 4.228517674001058, + "runtime_seconds": 10.018527030944824 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6905010349379646, + "lr": 0.03, + "negative_goodness": 1.012852920182546, + "pair_accuracy": 0.9076666666560703, + "positive_goodness": 4.039442114173042, + "runtime_seconds": 10.025130987167358 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6970536393589444, + "lr": 0.03, + "negative_goodness": 0.9974601542790731, + "pair_accuracy": 0.8990444444444444, + "positive_goodness": 4.026280062950982, + "runtime_seconds": 10.01854920387268 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6632667289310031, + "lr": 0.03, + "negative_goodness": 0.987554546631707, + "pair_accuracy": 0.9079777778201633, + "positive_goodness": 4.158877842140198, + "runtime_seconds": 10.02120041847229 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7026220719867282, + "lr": 0.03, + "negative_goodness": 1.0051252111646864, + "pair_accuracy": 0.8960888888782925, + "positive_goodness": 4.086066511535645, + "runtime_seconds": 10.017961502075195 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6896990658230252, + "lr": 0.03, + "negative_goodness": 1.0289629948510064, + "pair_accuracy": 0.9053111111164093, + "positive_goodness": 4.065423633066813, + "runtime_seconds": 10.008915662765503 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6425413310156928, + "lr": 0.03, + "negative_goodness": 0.9222569998847113, + "pair_accuracy": 0.9095555555873447, + "positive_goodness": 4.343401405058967, + "runtime_seconds": 10.007014989852905 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6707297577646043, + "lr": 0.03, + "negative_goodness": 0.9647929321712918, + "pair_accuracy": 0.896711111079322, + "positive_goodness": 4.233567786958482, + "runtime_seconds": 10.01942229270935 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.7168790613280402, + "lr": 0.03, + "negative_goodness": 1.0180048845291139, + "pair_accuracy": 0.9004888888782925, + "positive_goodness": 4.039086233562894, + "runtime_seconds": 10.003113269805908 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6436714503420724, + "lr": 0.03, + "negative_goodness": 0.9414882581339942, + "pair_accuracy": 0.9114666666666666, + "positive_goodness": 4.239902316877577, + "runtime_seconds": 9.970606327056885 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6932852479086982, + "lr": 0.03, + "negative_goodness": 0.9963754517290327, + "pair_accuracy": 0.8969777777777778, + "positive_goodness": 4.077731266530355, + "runtime_seconds": 9.954761981964111 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.715259654712677, + "lr": 0.03, + "negative_goodness": 1.0097330750571356, + "pair_accuracy": 0.9001555555449592, + "positive_goodness": 4.026358721330431, + "runtime_seconds": 9.965373754501343 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6456105721685621, + "lr": 0.03, + "negative_goodness": 0.9618749086459478, + "pair_accuracy": 0.9229555555555555, + "positive_goodness": 4.123552245924208, + "runtime_seconds": 9.95861005783081 + } + ], + "layer": 26 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9973282545301649, + "lr": 0.03, + "negative_goodness": 1.3472452763663398, + "pair_accuracy": 0.8030444444550409, + "positive_goodness": 3.0183549074808758, + "runtime_seconds": 9.934853315353394 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8926615781784057, + "lr": 0.03, + "negative_goodness": 1.310264387427436, + "pair_accuracy": 0.8296888888888889, + "positive_goodness": 3.35326974673801, + "runtime_seconds": 9.955825805664062 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.87830744916598, + "lr": 0.03, + "negative_goodness": 1.2815399506198035, + "pair_accuracy": 0.8334666666666667, + "positive_goodness": 3.506083349439833, + "runtime_seconds": 9.939950942993164 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8059913368966845, + "lr": 0.03, + "negative_goodness": 1.1953766734653049, + "pair_accuracy": 0.8590888888888889, + "positive_goodness": 3.685178758494059, + "runtime_seconds": 9.934147596359253 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.831401661740409, + "lr": 0.03, + "negative_goodness": 1.2392694284545052, + "pair_accuracy": 0.8546444444444444, + "positive_goodness": 3.60773788011339, + "runtime_seconds": 9.974394798278809 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7831082214991252, + "lr": 0.03, + "negative_goodness": 1.1557285052405464, + "pair_accuracy": 0.8672666666666666, + "positive_goodness": 3.881923151652018, + "runtime_seconds": 10.045059204101562 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7646622186289893, + "lr": 0.03, + "negative_goodness": 1.1270515451643202, + "pair_accuracy": 0.8722222222222222, + "positive_goodness": 3.914533170912001, + "runtime_seconds": 10.05236291885376 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.8011969330575731, + "lr": 0.03, + "negative_goodness": 1.1617358395894368, + "pair_accuracy": 0.8568888888676961, + "positive_goodness": 3.7001441154903834, + "runtime_seconds": 10.037477970123291 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7904286113950941, + "lr": 0.03, + "negative_goodness": 1.173604451974233, + "pair_accuracy": 0.8557333333863152, + "positive_goodness": 3.7644097271813286, + "runtime_seconds": 10.028298139572144 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7464506815592448, + "lr": 0.03, + "negative_goodness": 1.1125031994077894, + "pair_accuracy": 0.8801555555767483, + "positive_goodness": 3.8770199554867215, + "runtime_seconds": 9.992729902267456 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7029205097304451, + "lr": 0.03, + "negative_goodness": 1.0501743556976317, + "pair_accuracy": 0.8897111111111111, + "positive_goodness": 4.079000473403931, + "runtime_seconds": 9.991872787475586 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7591919878853692, + "lr": 0.03, + "negative_goodness": 1.0962554896036785, + "pair_accuracy": 0.8797555555343628, + "positive_goodness": 3.845050106345283, + "runtime_seconds": 9.990530967712402 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7349691083484226, + "lr": 0.03, + "negative_goodness": 1.0974861016379462, + "pair_accuracy": 0.8804888888570997, + "positive_goodness": 3.94255365041097, + "runtime_seconds": 9.997494220733643 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7506375585397085, + "lr": 0.03, + "negative_goodness": 1.1095067710055246, + "pair_accuracy": 0.8786222222222222, + "positive_goodness": 4.006408699544271, + "runtime_seconds": 10.023826122283936 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7696398738861084, + "lr": 0.03, + "negative_goodness": 1.1149286233266196, + "pair_accuracy": 0.8768444444974264, + "positive_goodness": 3.833420631451077, + "runtime_seconds": 10.056577444076538 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7286696126725939, + "lr": 0.03, + "negative_goodness": 1.0653853897306653, + "pair_accuracy": 0.8852222222222222, + "positive_goodness": 3.90455148340861, + "runtime_seconds": 10.071783065795898 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6493422082265218, + "lr": 0.03, + "negative_goodness": 0.9631670120239257, + "pair_accuracy": 0.9090888888888888, + "positive_goodness": 4.297175278049045, + "runtime_seconds": 10.031126976013184 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7091110239426295, + "lr": 0.03, + "negative_goodness": 1.0446134230507744, + "pair_accuracy": 0.8925111111111111, + "positive_goodness": 3.9862861300998262, + "runtime_seconds": 9.98530125617981 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7133112908230888, + "lr": 0.03, + "negative_goodness": 1.0550913597981135, + "pair_accuracy": 0.8871555555555556, + "positive_goodness": 4.042543593766954, + "runtime_seconds": 9.989944696426392 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6957509308232201, + "lr": 0.03, + "negative_goodness": 1.0169714228285684, + "pair_accuracy": 0.8974222222222222, + "positive_goodness": 4.135588390519884, + "runtime_seconds": 9.990549087524414 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7001423362731933, + "lr": 0.03, + "negative_goodness": 1.0057772292878893, + "pair_accuracy": 0.8836666666878594, + "positive_goodness": 4.078019660674201, + "runtime_seconds": 9.995277166366577 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7085884099960327, + "lr": 0.03, + "negative_goodness": 1.0206549253463746, + "pair_accuracy": 0.8851111110899184, + "positive_goodness": 4.071334675216675, + "runtime_seconds": 9.99241590499878 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6862817740228441, + "lr": 0.03, + "negative_goodness": 1.0167168387518988, + "pair_accuracy": 0.9007111111323038, + "positive_goodness": 4.126021142790052, + "runtime_seconds": 10.004430532455444 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.7517608533329434, + "lr": 0.03, + "negative_goodness": 1.0722415704727173, + "pair_accuracy": 0.8798666666878594, + "positive_goodness": 3.8836316171858045, + "runtime_seconds": 10.032402276992798 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6668887093861898, + "lr": 0.03, + "negative_goodness": 0.9820835566414727, + "pair_accuracy": 0.9120000000052982, + "positive_goodness": 4.18927176844279, + "runtime_seconds": 10.169947862625122 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7153224961969588, + "lr": 0.03, + "negative_goodness": 1.0043122079637317, + "pair_accuracy": 0.8876666666666667, + "positive_goodness": 4.065902161492242, + "runtime_seconds": 10.17181944847107 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.7111277776930067, + "lr": 0.03, + "negative_goodness": 1.0064700892342462, + "pair_accuracy": 0.8786000000211928, + "positive_goodness": 4.051913648796082, + "runtime_seconds": 10.182847738265991 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6727970147132873, + "lr": 0.03, + "negative_goodness": 0.9645984424591064, + "pair_accuracy": 0.9009999999788072, + "positive_goodness": 4.256581723446316, + "runtime_seconds": 10.19404411315918 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6959864339828491, + "lr": 0.03, + "negative_goodness": 0.9728437945948707, + "pair_accuracy": 0.8917777778095669, + "positive_goodness": 4.262056596713596, + "runtime_seconds": 10.185352563858032 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6597436196433173, + "lr": 0.03, + "negative_goodness": 0.9170972577200995, + "pair_accuracy": 0.9016222222222222, + "positive_goodness": 4.397140443632337, + "runtime_seconds": 10.179966449737549 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6148202548133003, + "lr": 0.03, + "negative_goodness": 0.8819213944964939, + "pair_accuracy": 0.9091111111111111, + "positive_goodness": 4.524136940511068, + "runtime_seconds": 10.099060297012329 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.7074653270297581, + "lr": 0.03, + "negative_goodness": 0.9888508654064603, + "pair_accuracy": 0.8868666666719649, + "positive_goodness": 4.149769893179999, + "runtime_seconds": 10.049208402633667 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6836402069515652, + "lr": 0.03, + "negative_goodness": 0.9776879566616482, + "pair_accuracy": 0.8933777777777778, + "positive_goodness": 4.171838005362616, + "runtime_seconds": 10.042532444000244 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6670516364097595, + "lr": 0.03, + "negative_goodness": 0.9457303720262316, + "pair_accuracy": 0.9013333333333333, + "positive_goodness": 4.164082375420464, + "runtime_seconds": 10.04600977897644 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6738511717213524, + "lr": 0.03, + "negative_goodness": 0.9500945160441928, + "pair_accuracy": 0.9022222222222223, + "positive_goodness": 4.199338584899903, + "runtime_seconds": 10.045025110244751 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6990323189841376, + "lr": 0.03, + "negative_goodness": 0.9818230692227682, + "pair_accuracy": 0.8951111111534966, + "positive_goodness": 4.137388456090291, + "runtime_seconds": 10.050849914550781 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6543590897454156, + "lr": 0.03, + "negative_goodness": 0.9266872259351943, + "pair_accuracy": 0.8925777777777778, + "positive_goodness": 4.269241975487603, + "runtime_seconds": 10.054288864135742 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5992650219069587, + "lr": 0.03, + "negative_goodness": 0.850706814400355, + "pair_accuracy": 0.9219555555555555, + "positive_goodness": 4.595891338094075, + "runtime_seconds": 10.04828405380249 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6489611943350898, + "lr": 0.03, + "negative_goodness": 0.8984929085413615, + "pair_accuracy": 0.9016222222142749, + "positive_goodness": 4.4515726386176215, + "runtime_seconds": 10.053179740905762 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6220126320997874, + "lr": 0.03, + "negative_goodness": 0.8917177544911703, + "pair_accuracy": 0.9148, + "positive_goodness": 4.507846427408854, + "runtime_seconds": 10.047499656677246 + } + ], + "layer": 27 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9221804879506429, + "lr": 0.03, + "negative_goodness": 1.4421615493880378, + "pair_accuracy": 0.8548666666666667, + "positive_goodness": 3.690839225853814, + "runtime_seconds": 10.06937861442566 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8524957663642035, + "lr": 0.03, + "negative_goodness": 1.316926120270623, + "pair_accuracy": 0.8548888888676961, + "positive_goodness": 3.636083918041653, + "runtime_seconds": 10.066743850708008 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8428867513868544, + "lr": 0.03, + "negative_goodness": 1.333231357574463, + "pair_accuracy": 0.8659999999947018, + "positive_goodness": 3.5961456211513942, + "runtime_seconds": 10.075169801712036 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8268709253311157, + "lr": 0.03, + "negative_goodness": 1.2770334579043918, + "pair_accuracy": 0.8533777777883742, + "positive_goodness": 3.576441589567396, + "runtime_seconds": 10.049075365066528 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7722645816617542, + "lr": 0.03, + "negative_goodness": 1.2160701414320203, + "pair_accuracy": 0.8679333333333333, + "positive_goodness": 3.7907052485148114, + "runtime_seconds": 10.071882009506226 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7866826685799493, + "lr": 0.03, + "negative_goodness": 1.2062876518673367, + "pair_accuracy": 0.870022222243415, + "positive_goodness": 3.752792928144667, + "runtime_seconds": 10.076712369918823 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.8135524946636624, + "lr": 0.03, + "negative_goodness": 1.2496504679150051, + "pair_accuracy": 0.8594222222116258, + "positive_goodness": 3.6068329129960803, + "runtime_seconds": 10.037766933441162 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7690014966752794, + "lr": 0.03, + "negative_goodness": 1.1917794512006972, + "pair_accuracy": 0.8723111111376021, + "positive_goodness": 3.7192677597469754, + "runtime_seconds": 10.01506781578064 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.769010848792394, + "lr": 0.03, + "negative_goodness": 1.1731632952372233, + "pair_accuracy": 0.8718888888888889, + "positive_goodness": 3.756338835228814, + "runtime_seconds": 10.014165878295898 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7511619844648573, + "lr": 0.03, + "negative_goodness": 1.1535993487887912, + "pair_accuracy": 0.8752888888994853, + "positive_goodness": 3.7409136021931966, + "runtime_seconds": 10.039623737335205 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7241712444517348, + "lr": 0.03, + "negative_goodness": 1.1008472691429987, + "pair_accuracy": 0.8798222222116259, + "positive_goodness": 3.916466936916775, + "runtime_seconds": 10.07686972618103 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7416243732240465, + "lr": 0.03, + "negative_goodness": 1.1351803434160022, + "pair_accuracy": 0.8799333333333333, + "positive_goodness": 3.8154443522983126, + "runtime_seconds": 10.036656618118286 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7094451395246718, + "lr": 0.03, + "negative_goodness": 1.0685231173621283, + "pair_accuracy": 0.8808666666878594, + "positive_goodness": 3.989925425084432, + "runtime_seconds": 10.025899887084961 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7148241577360365, + "lr": 0.03, + "negative_goodness": 1.0845477354897393, + "pair_accuracy": 0.8849555555449592, + "positive_goodness": 3.911201128556993, + "runtime_seconds": 10.018788814544678 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6930087102916506, + "lr": 0.03, + "negative_goodness": 1.0636577247036827, + "pair_accuracy": 0.8943777777777778, + "positive_goodness": 3.9985579076978897, + "runtime_seconds": 10.02344799041748 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7004732805728913, + "lr": 0.03, + "negative_goodness": 1.073780974287457, + "pair_accuracy": 0.8944444444444445, + "positive_goodness": 3.968824179840088, + "runtime_seconds": 10.02070426940918 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7491039667977227, + "lr": 0.03, + "negative_goodness": 1.1161790109846328, + "pair_accuracy": 0.8762888888994853, + "positive_goodness": 3.869361102909512, + "runtime_seconds": 10.02143383026123 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7184235352410211, + "lr": 0.03, + "negative_goodness": 1.0503856811099583, + "pair_accuracy": 0.8832222222540114, + "positive_goodness": 3.949131775156657, + "runtime_seconds": 10.013886451721191 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7466217429425981, + "lr": 0.03, + "negative_goodness": 1.1106359362443288, + "pair_accuracy": 0.8788444444444444, + "positive_goodness": 3.8023085593329538, + "runtime_seconds": 10.01969027519226 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7227921074019538, + "lr": 0.03, + "negative_goodness": 1.0845147467295329, + "pair_accuracy": 0.8831555555449592, + "positive_goodness": 3.968156818856133, + "runtime_seconds": 10.013653755187988 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.699165432686276, + "lr": 0.03, + "negative_goodness": 1.0401889172448053, + "pair_accuracy": 0.8862666666666666, + "positive_goodness": 4.008302087148031, + "runtime_seconds": 10.014911413192749 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6821367959552341, + "lr": 0.03, + "negative_goodness": 1.0594229918162028, + "pair_accuracy": 0.8988222222328186, + "positive_goodness": 4.059271290545993, + "runtime_seconds": 10.012058019638062 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.7241336813555823, + "lr": 0.03, + "negative_goodness": 1.0665042506800757, + "pair_accuracy": 0.8767777777777778, + "positive_goodness": 3.8803978759765627, + "runtime_seconds": 10.013751029968262 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6835401442633735, + "lr": 0.03, + "negative_goodness": 1.0330341786702475, + "pair_accuracy": 0.8921333333121406, + "positive_goodness": 4.060235665935941, + "runtime_seconds": 10.014899969100952 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6719598876025942, + "lr": 0.03, + "negative_goodness": 1.046072091680103, + "pair_accuracy": 0.9048, + "positive_goodness": 4.033721761491564, + "runtime_seconds": 10.017635107040405 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7453496799257067, + "lr": 0.03, + "negative_goodness": 1.0913183999167548, + "pair_accuracy": 0.8807111111111111, + "positive_goodness": 3.9400767964257133, + "runtime_seconds": 10.032061338424683 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6893340722613864, + "lr": 0.03, + "negative_goodness": 1.034689483133952, + "pair_accuracy": 0.8918666666984558, + "positive_goodness": 3.954676326031155, + "runtime_seconds": 10.066969871520996 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6481950015227, + "lr": 0.03, + "negative_goodness": 0.9818008621056875, + "pair_accuracy": 0.9057777777777778, + "positive_goodness": 4.220318747202556, + "runtime_seconds": 10.069378852844238 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.7012042014439901, + "lr": 0.03, + "negative_goodness": 1.0581065533743965, + "pair_accuracy": 0.8952444444232517, + "positive_goodness": 4.010556592263116, + "runtime_seconds": 10.071995735168457 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.7076526320404477, + "lr": 0.03, + "negative_goodness": 1.017533501434326, + "pair_accuracy": 0.8768666666666667, + "positive_goodness": 3.9958628561231824, + "runtime_seconds": 10.074715852737427 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6721812032063802, + "lr": 0.03, + "negative_goodness": 1.0128212443033855, + "pair_accuracy": 0.8955999999788072, + "positive_goodness": 4.113391945605808, + "runtime_seconds": 10.08425235748291 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6736586957560645, + "lr": 0.03, + "negative_goodness": 0.9858647690667046, + "pair_accuracy": 0.8947999999894036, + "positive_goodness": 4.079405556318495, + "runtime_seconds": 10.074700355529785 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.656178367834621, + "lr": 0.03, + "negative_goodness": 0.9879680991278754, + "pair_accuracy": 0.8972444444444444, + "positive_goodness": 4.21054910413954, + "runtime_seconds": 10.078937530517578 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6918513781123691, + "lr": 0.03, + "negative_goodness": 1.0177796881887649, + "pair_accuracy": 0.8911777777671814, + "positive_goodness": 4.102686592186822, + "runtime_seconds": 10.075247287750244 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6537729209052192, + "lr": 0.03, + "negative_goodness": 0.9694091121249728, + "pair_accuracy": 0.905955555539661, + "positive_goodness": 4.172705106608073, + "runtime_seconds": 10.074471712112427 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6908776702457005, + "lr": 0.03, + "negative_goodness": 1.0198383930630155, + "pair_accuracy": 0.8917777777817514, + "positive_goodness": 4.117477531433106, + "runtime_seconds": 10.077322244644165 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6187315314531326, + "lr": 0.03, + "negative_goodness": 0.9207328838586807, + "pair_accuracy": 0.9149555555555555, + "positive_goodness": 4.323871966128879, + "runtime_seconds": 10.07469892501831 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.7011205619812012, + "lr": 0.03, + "negative_goodness": 1.0125919501066207, + "pair_accuracy": 0.8915333333333333, + "positive_goodness": 4.09265262383355, + "runtime_seconds": 10.082283735275269 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6941359952078925, + "lr": 0.03, + "negative_goodness": 1.0029119274987115, + "pair_accuracy": 0.8879333333439298, + "positive_goodness": 4.1199705115424266, + "runtime_seconds": 10.161360502243042 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6934937392897076, + "lr": 0.03, + "negative_goodness": 1.0356123130374484, + "pair_accuracy": 0.894, + "positive_goodness": 4.001264546457927, + "runtime_seconds": 10.283041715621948 + } + ], + "layer": 28 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8872664809650845, + "lr": 0.03, + "negative_goodness": 1.2610374111493428, + "pair_accuracy": 0.8587777777777778, + "positive_goodness": 3.3473619202507865, + "runtime_seconds": 10.075911283493042 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8152782010396321, + "lr": 0.03, + "negative_goodness": 1.2313219055599636, + "pair_accuracy": 0.875066666650772, + "positive_goodness": 3.6185823184543184, + "runtime_seconds": 10.052309274673462 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8432079123602974, + "lr": 0.03, + "negative_goodness": 1.2450663547092014, + "pair_accuracy": 0.8600444444974263, + "positive_goodness": 3.5660711201137967, + "runtime_seconds": 10.057555198669434 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7406409569528368, + "lr": 0.03, + "negative_goodness": 1.1286076908535427, + "pair_accuracy": 0.896022222190433, + "positive_goodness": 3.7959241757710775, + "runtime_seconds": 10.055273532867432 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7477033920500014, + "lr": 0.03, + "negative_goodness": 1.1399196376376681, + "pair_accuracy": 0.8910666666560703, + "positive_goodness": 3.823968726094564, + "runtime_seconds": 10.057621479034424 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6972613058514066, + "lr": 0.03, + "negative_goodness": 1.046972129758199, + "pair_accuracy": 0.9012222222752041, + "positive_goodness": 3.97553061565823, + "runtime_seconds": 10.048738479614258 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7318992391374376, + "lr": 0.03, + "negative_goodness": 1.0814495833714803, + "pair_accuracy": 0.8955555556085375, + "positive_goodness": 3.8686297049840292, + "runtime_seconds": 10.049051761627197 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.712668365462621, + "lr": 0.03, + "negative_goodness": 1.0769799249649048, + "pair_accuracy": 0.8945777777777778, + "positive_goodness": 3.9590901997036405, + "runtime_seconds": 10.056838989257812 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7350172864278157, + "lr": 0.03, + "negative_goodness": 1.0922421755896674, + "pair_accuracy": 0.887866666677263, + "positive_goodness": 3.8425066430833605, + "runtime_seconds": 10.053906679153442 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7106776964399549, + "lr": 0.03, + "negative_goodness": 1.03155348233117, + "pair_accuracy": 0.892977777783076, + "positive_goodness": 3.9699693670484755, + "runtime_seconds": 10.052066087722778 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7636669287469652, + "lr": 0.03, + "negative_goodness": 1.1068337872399223, + "pair_accuracy": 0.8829111111111111, + "positive_goodness": 3.7947679885016545, + "runtime_seconds": 10.052846431732178 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.693778589047326, + "lr": 0.03, + "negative_goodness": 1.0416959836112127, + "pair_accuracy": 0.9020444444974264, + "positive_goodness": 4.029539180628459, + "runtime_seconds": 10.053679704666138 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7032332011752659, + "lr": 0.03, + "negative_goodness": 1.0310190940009223, + "pair_accuracy": 0.8945999999788072, + "positive_goodness": 3.9878929563734267, + "runtime_seconds": 9.969465017318726 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6727694860246446, + "lr": 0.03, + "negative_goodness": 0.9637052684677971, + "pair_accuracy": 0.8945111111429003, + "positive_goodness": 4.1230434628380666, + "runtime_seconds": 9.936309814453125 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6612760377142164, + "lr": 0.03, + "negative_goodness": 0.9833089513142904, + "pair_accuracy": 0.901266666645474, + "positive_goodness": 4.1421237720913355, + "runtime_seconds": 9.935980081558228 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6925224894629585, + "lr": 0.03, + "negative_goodness": 1.0219133551067776, + "pair_accuracy": 0.8986666667196486, + "positive_goodness": 4.031834189690484, + "runtime_seconds": 9.943340539932251 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6518647311316597, + "lr": 0.03, + "negative_goodness": 0.9656548398971557, + "pair_accuracy": 0.9117777777936723, + "positive_goodness": 4.167345799827576, + "runtime_seconds": 9.936884641647339 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6708154081874423, + "lr": 0.03, + "negative_goodness": 0.9562301291041904, + "pair_accuracy": 0.8969111111217075, + "positive_goodness": 4.158211990102132, + "runtime_seconds": 9.9376859664917 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6889491943703757, + "lr": 0.03, + "negative_goodness": 1.0134726977825166, + "pair_accuracy": 0.9012444444444444, + "positive_goodness": 4.0857928572760684, + "runtime_seconds": 9.940669059753418 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6862767931169934, + "lr": 0.03, + "negative_goodness": 0.9972146312395732, + "pair_accuracy": 0.9030222222222222, + "positive_goodness": 4.0260133941226535, + "runtime_seconds": 9.935781002044678 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6661904729631212, + "lr": 0.03, + "negative_goodness": 0.9499649792777167, + "pair_accuracy": 0.9001555555343628, + "positive_goodness": 4.172589739693536, + "runtime_seconds": 9.93812894821167 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6448802200873693, + "lr": 0.03, + "negative_goodness": 0.9610180839750502, + "pair_accuracy": 0.9153333333333333, + "positive_goodness": 4.198410610029432, + "runtime_seconds": 9.941867351531982 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6505506902482775, + "lr": 0.03, + "negative_goodness": 0.9458975958294339, + "pair_accuracy": 0.9051555555555556, + "positive_goodness": 4.102512123976814, + "runtime_seconds": 9.941321611404419 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6318902108510335, + "lr": 0.03, + "negative_goodness": 0.906355880800883, + "pair_accuracy": 0.9145999999682108, + "positive_goodness": 4.299184376059638, + "runtime_seconds": 9.93293309211731 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6742010083410475, + "lr": 0.03, + "negative_goodness": 0.9879235899819269, + "pair_accuracy": 0.9091555555449592, + "positive_goodness": 4.020564132902357, + "runtime_seconds": 9.938151121139526 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6611903122425079, + "lr": 0.03, + "negative_goodness": 0.9511012084060245, + "pair_accuracy": 0.9121999999894036, + "positive_goodness": 4.063488042195638, + "runtime_seconds": 9.93993067741394 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6343254096137153, + "lr": 0.03, + "negative_goodness": 0.8884135166168213, + "pair_accuracy": 0.9000444444126553, + "positive_goodness": 4.282340312533909, + "runtime_seconds": 9.939478158950806 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6162593627505832, + "lr": 0.03, + "negative_goodness": 0.8893184426307679, + "pair_accuracy": 0.9152, + "positive_goodness": 4.370227366977268, + "runtime_seconds": 9.94476866722107 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6494569510512882, + "lr": 0.03, + "negative_goodness": 0.914290450472302, + "pair_accuracy": 0.9096222222222222, + "positive_goodness": 4.222837505086263, + "runtime_seconds": 9.943783521652222 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.649281451564365, + "lr": 0.03, + "negative_goodness": 0.9491903128465017, + "pair_accuracy": 0.9100444444444444, + "positive_goodness": 4.165049964904785, + "runtime_seconds": 9.912127494812012 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6267527609083388, + "lr": 0.03, + "negative_goodness": 0.8621317677074009, + "pair_accuracy": 0.9097999999788072, + "positive_goodness": 4.326508144675361, + "runtime_seconds": 9.938801050186157 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6182045806460911, + "lr": 0.03, + "negative_goodness": 0.9018237980312771, + "pair_accuracy": 0.9188444444451067, + "positive_goodness": 4.337988528357612, + "runtime_seconds": 9.93671202659607 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6067689634535047, + "lr": 0.03, + "negative_goodness": 0.8781712947951423, + "pair_accuracy": 0.9196666666669978, + "positive_goodness": 4.353846157073975, + "runtime_seconds": 9.943166017532349 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5989051156255933, + "lr": 0.03, + "negative_goodness": 0.8483096996943156, + "pair_accuracy": 0.9159999999788072, + "positive_goodness": 4.425803072060479, + "runtime_seconds": 9.944528341293335 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6357768700705634, + "lr": 0.03, + "negative_goodness": 0.9040650556988187, + "pair_accuracy": 0.9079777777777778, + "positive_goodness": 4.275713812849257, + "runtime_seconds": 9.946199655532837 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5554883244196573, + "lr": 0.03, + "negative_goodness": 0.8000754127926296, + "pair_accuracy": 0.9273555556085374, + "positive_goodness": 4.61751587562561, + "runtime_seconds": 10.01645565032959 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5930362655772103, + "lr": 0.03, + "negative_goodness": 0.8604083919299973, + "pair_accuracy": 0.9272222222222222, + "positive_goodness": 4.505759776051839, + "runtime_seconds": 10.05826997756958 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5665851357565985, + "lr": 0.03, + "negative_goodness": 0.8112014442549812, + "pair_accuracy": 0.9317333333333333, + "positive_goodness": 4.557649175601536, + "runtime_seconds": 10.055928945541382 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6486136707729764, + "lr": 0.03, + "negative_goodness": 0.9208026482052273, + "pair_accuracy": 0.9037555555237664, + "positive_goodness": 4.193940994856093, + "runtime_seconds": 10.060580968856812 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5825926074504852, + "lr": 0.03, + "negative_goodness": 0.8419774212943183, + "pair_accuracy": 0.9270666666666667, + "positive_goodness": 4.485696997748481, + "runtime_seconds": 10.058736562728882 + } + ], + "layer": 29 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0527203379419114, + "lr": 0.03, + "negative_goodness": 1.5877434749391344, + "pair_accuracy": 0.8324444444444444, + "positive_goodness": 3.182067077170478, + "runtime_seconds": 10.086768627166748 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.788087932756212, + "lr": 0.03, + "negative_goodness": 1.290195106463962, + "pair_accuracy": 0.9029111111111111, + "positive_goodness": 3.8272314534399245, + "runtime_seconds": 10.080604076385498 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8240290346198612, + "lr": 0.03, + "negative_goodness": 1.2883222598499722, + "pair_accuracy": 0.8768888888888889, + "positive_goodness": 3.6531016982184514, + "runtime_seconds": 10.078842878341675 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7540148819393582, + "lr": 0.03, + "negative_goodness": 1.201154007021586, + "pair_accuracy": 0.8926444444126553, + "positive_goodness": 3.7884464078267417, + "runtime_seconds": 10.080885648727417 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7304339881261189, + "lr": 0.03, + "negative_goodness": 1.11335860909356, + "pair_accuracy": 0.8918444444444444, + "positive_goodness": 3.898308421749539, + "runtime_seconds": 10.081835269927979 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7381379524336921, + "lr": 0.03, + "negative_goodness": 1.1443515233357748, + "pair_accuracy": 0.8901555555555556, + "positive_goodness": 3.868937818230523, + "runtime_seconds": 10.080406427383423 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6831766552024418, + "lr": 0.03, + "negative_goodness": 1.070170069196489, + "pair_accuracy": 0.9114222222222222, + "positive_goodness": 4.039746946546766, + "runtime_seconds": 10.088561534881592 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6987806320296394, + "lr": 0.03, + "negative_goodness": 1.0417220620473227, + "pair_accuracy": 0.8938, + "positive_goodness": 4.047304008483887, + "runtime_seconds": 10.078872442245483 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6953670828077528, + "lr": 0.03, + "negative_goodness": 1.0859488314098782, + "pair_accuracy": 0.8991555556085374, + "positive_goodness": 3.9399426977793377, + "runtime_seconds": 10.080844163894653 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6607558449851142, + "lr": 0.03, + "negative_goodness": 1.0119096382141113, + "pair_accuracy": 0.9087111111243565, + "positive_goodness": 4.119036094326443, + "runtime_seconds": 10.082869529724121 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6828879483540853, + "lr": 0.03, + "negative_goodness": 1.0088172818289862, + "pair_accuracy": 0.9062666666772631, + "positive_goodness": 4.0514768441942, + "runtime_seconds": 10.022312641143799 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6904014861795638, + "lr": 0.03, + "negative_goodness": 1.0369989606115553, + "pair_accuracy": 0.9006888888888889, + "positive_goodness": 4.070397160932753, + "runtime_seconds": 9.960794687271118 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6862743945651584, + "lr": 0.03, + "negative_goodness": 1.0161548536512586, + "pair_accuracy": 0.897888888894187, + "positive_goodness": 4.067146290249295, + "runtime_seconds": 9.970938920974731 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6702796202659607, + "lr": 0.03, + "negative_goodness": 1.0132985458532968, + "pair_accuracy": 0.9079333333333334, + "positive_goodness": 4.021163410780165, + "runtime_seconds": 9.962181568145752 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6462748288048639, + "lr": 0.03, + "negative_goodness": 0.9852922549989488, + "pair_accuracy": 0.9157111110899183, + "positive_goodness": 4.2022311364491785, + "runtime_seconds": 9.956971645355225 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.691898160847028, + "lr": 0.03, + "negative_goodness": 1.0194975871774885, + "pair_accuracy": 0.9000222222222222, + "positive_goodness": 4.008075190565322, + "runtime_seconds": 9.960087060928345 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6390584043184916, + "lr": 0.03, + "negative_goodness": 0.9593855672306485, + "pair_accuracy": 0.9139111111111111, + "positive_goodness": 4.211722837151421, + "runtime_seconds": 9.96293568611145 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6642331957923041, + "lr": 0.03, + "negative_goodness": 0.9854809447924296, + "pair_accuracy": 0.9004, + "positive_goodness": 4.086551359812419, + "runtime_seconds": 9.961516618728638 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6420701846864488, + "lr": 0.03, + "negative_goodness": 0.9589314679039849, + "pair_accuracy": 0.9092222222328186, + "positive_goodness": 4.215237318929036, + "runtime_seconds": 9.970555543899536 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6297289316495259, + "lr": 0.03, + "negative_goodness": 0.939714123556349, + "pair_accuracy": 0.9128888888994853, + "positive_goodness": 4.292680307261149, + "runtime_seconds": 9.966504096984863 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6529057266553243, + "lr": 0.03, + "negative_goodness": 0.9539199870321485, + "pair_accuracy": 0.9073555555555556, + "positive_goodness": 4.236390735456679, + "runtime_seconds": 9.963369369506836 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6424391766574647, + "lr": 0.03, + "negative_goodness": 0.9338950427002377, + "pair_accuracy": 0.9047333333333333, + "positive_goodness": 4.280757856326633, + "runtime_seconds": 9.961521625518799 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6510272367795309, + "lr": 0.03, + "negative_goodness": 0.9597343287997776, + "pair_accuracy": 0.9040888888782925, + "positive_goodness": 4.247062623935276, + "runtime_seconds": 9.964585781097412 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6357407645834817, + "lr": 0.03, + "negative_goodness": 0.9544500939183765, + "pair_accuracy": 0.9161333333333334, + "positive_goodness": 4.203912306806776, + "runtime_seconds": 9.959272384643555 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5999115397188398, + "lr": 0.03, + "negative_goodness": 0.9009284882757399, + "pair_accuracy": 0.9260222222222222, + "positive_goodness": 4.3876876272413465, + "runtime_seconds": 9.972989559173584 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6131733137263192, + "lr": 0.03, + "negative_goodness": 0.8923269851737552, + "pair_accuracy": 0.9164222222222222, + "positive_goodness": 4.413973066880968, + "runtime_seconds": 9.958343267440796 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6059297974904378, + "lr": 0.03, + "negative_goodness": 0.8803764810562134, + "pair_accuracy": 0.9136444444126554, + "positive_goodness": 4.485605068206787, + "runtime_seconds": 9.964277982711792 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.609437104458279, + "lr": 0.03, + "negative_goodness": 0.8947190713670519, + "pair_accuracy": 0.91804444444461, + "positive_goodness": 4.388487822376357, + "runtime_seconds": 9.967325925827026 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5720112884309557, + "lr": 0.03, + "negative_goodness": 0.8215835718366835, + "pair_accuracy": 0.9218000000317892, + "positive_goodness": 4.506159666654798, + "runtime_seconds": 9.961693525314331 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6171475888358222, + "lr": 0.03, + "negative_goodness": 0.8973235738118489, + "pair_accuracy": 0.9156222222540114, + "positive_goodness": 4.433674624252319, + "runtime_seconds": 9.965421199798584 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6026717464235094, + "lr": 0.03, + "negative_goodness": 0.8965384653303359, + "pair_accuracy": 0.9197333333121406, + "positive_goodness": 4.348265403027004, + "runtime_seconds": 9.969858407974243 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6182759827666813, + "lr": 0.03, + "negative_goodness": 0.9278526787757874, + "pair_accuracy": 0.9221999999894036, + "positive_goodness": 4.2701071266174315, + "runtime_seconds": 9.958916902542114 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6008725464926825, + "lr": 0.03, + "negative_goodness": 0.8593160827636719, + "pair_accuracy": 0.9128444444444445, + "positive_goodness": 4.439442906867133, + "runtime_seconds": 9.964844465255737 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5809445569144355, + "lr": 0.03, + "negative_goodness": 0.8483521804597642, + "pair_accuracy": 0.927800000026491, + "positive_goodness": 4.426817695257399, + "runtime_seconds": 9.984657049179077 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6144778468237982, + "lr": 0.03, + "negative_goodness": 0.8901861510806613, + "pair_accuracy": 0.9142888888782925, + "positive_goodness": 4.346625002585517, + "runtime_seconds": 9.974590539932251 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5873738614188301, + "lr": 0.03, + "negative_goodness": 0.8564273905012343, + "pair_accuracy": 0.9186000000211928, + "positive_goodness": 4.446309986495971, + "runtime_seconds": 10.073867797851562 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6248400505913628, + "lr": 0.03, + "negative_goodness": 0.9047472524642944, + "pair_accuracy": 0.9162666666984558, + "positive_goodness": 4.3347523162417945, + "runtime_seconds": 9.98305892944336 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5860988001081678, + "lr": 0.03, + "negative_goodness": 0.8431514192581177, + "pair_accuracy": 0.917688888920678, + "positive_goodness": 4.47760111495124, + "runtime_seconds": 9.959319114685059 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6870628238572015, + "lr": 0.03, + "negative_goodness": 0.9740370849821303, + "pair_accuracy": 0.9029111110899184, + "positive_goodness": 4.058611128913031, + "runtime_seconds": 9.96480679512024 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.612367590883043, + "lr": 0.03, + "negative_goodness": 0.8885421074761285, + "pair_accuracy": 0.9109777777698305, + "positive_goodness": 4.3208310052023995, + "runtime_seconds": 9.958999156951904 + } + ], + "layer": 30 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.895979928864373, + "lr": 0.03, + "negative_goodness": 1.290446430545383, + "pair_accuracy": 0.8529777777777777, + "positive_goodness": 3.392071725294325, + "runtime_seconds": 9.941535711288452 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8470980894088745, + "lr": 0.03, + "negative_goodness": 1.2547754196802776, + "pair_accuracy": 0.8701999999788073, + "positive_goodness": 3.4581240982055665, + "runtime_seconds": 9.938080310821533 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7766326047685411, + "lr": 0.03, + "negative_goodness": 1.165174969546, + "pair_accuracy": 0.8812666667196486, + "positive_goodness": 3.6795301303863526, + "runtime_seconds": 9.948394536972046 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8024202027638754, + "lr": 0.03, + "negative_goodness": 1.2075080683814154, + "pair_accuracy": 0.8821777777777777, + "positive_goodness": 3.602763991970486, + "runtime_seconds": 9.944293975830078 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7875595242288378, + "lr": 0.03, + "negative_goodness": 1.1723277770996094, + "pair_accuracy": 0.8847111110899184, + "positive_goodness": 3.674202441660563, + "runtime_seconds": 9.945285081863403 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7294254644870758, + "lr": 0.03, + "negative_goodness": 1.0897259763081868, + "pair_accuracy": 0.8928666666666667, + "positive_goodness": 3.882092630089654, + "runtime_seconds": 9.939208984375 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7210018215020497, + "lr": 0.03, + "negative_goodness": 1.0696194546805489, + "pair_accuracy": 0.8972, + "positive_goodness": 3.93049799202813, + "runtime_seconds": 9.945659399032593 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7103763287279341, + "lr": 0.03, + "negative_goodness": 1.0486638186666701, + "pair_accuracy": 0.9001111111111111, + "positive_goodness": 3.9809768812391493, + "runtime_seconds": 9.955312728881836 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7624806364589267, + "lr": 0.03, + "negative_goodness": 1.0983493452919855, + "pair_accuracy": 0.8854666666666666, + "positive_goodness": 3.771486375512017, + "runtime_seconds": 10.035012245178223 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.764994488313463, + "lr": 0.03, + "negative_goodness": 1.102196972380744, + "pair_accuracy": 0.8821111111270057, + "positive_goodness": 3.8084122876061333, + "runtime_seconds": 9.972667694091797 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6964509235805936, + "lr": 0.03, + "negative_goodness": 1.019561678949992, + "pair_accuracy": 0.9031555555555556, + "positive_goodness": 4.012372899670071, + "runtime_seconds": 9.958285570144653 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7235280456331041, + "lr": 0.03, + "negative_goodness": 1.0496018133693272, + "pair_accuracy": 0.8983555555555556, + "positive_goodness": 3.827729673936632, + "runtime_seconds": 10.00172209739685 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7117616896099515, + "lr": 0.03, + "negative_goodness": 1.0365675638198852, + "pair_accuracy": 0.8999999999682109, + "positive_goodness": 3.945915899149577, + "runtime_seconds": 9.968883275985718 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7558829819785224, + "lr": 0.03, + "negative_goodness": 1.0807425625695122, + "pair_accuracy": 0.8805555555449591, + "positive_goodness": 3.8103411047617595, + "runtime_seconds": 9.94870376586914 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6646051271014743, + "lr": 0.03, + "negative_goodness": 0.9695889597998725, + "pair_accuracy": 0.9057333333015442, + "positive_goodness": 4.15263760049608, + "runtime_seconds": 9.950842142105103 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6417580501132542, + "lr": 0.03, + "negative_goodness": 0.946171255938212, + "pair_accuracy": 0.9156222222328186, + "positive_goodness": 4.232892220391167, + "runtime_seconds": 9.964691162109375 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6534262130790287, + "lr": 0.03, + "negative_goodness": 0.9538364314052794, + "pair_accuracy": 0.9109777777777778, + "positive_goodness": 4.123261120266385, + "runtime_seconds": 9.942144632339478 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7089250459459093, + "lr": 0.03, + "negative_goodness": 1.001858676592509, + "pair_accuracy": 0.8968888888888888, + "positive_goodness": 3.9790655146280924, + "runtime_seconds": 9.947757720947266 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.64147910756535, + "lr": 0.03, + "negative_goodness": 0.9349143468962775, + "pair_accuracy": 0.9186222222222222, + "positive_goodness": 4.2275092376708985, + "runtime_seconds": 9.951383352279663 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.675972006114324, + "lr": 0.03, + "negative_goodness": 0.973912505891588, + "pair_accuracy": 0.8994, + "positive_goodness": 4.105897184668647, + "runtime_seconds": 10.008575201034546 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.7064764359050327, + "lr": 0.03, + "negative_goodness": 1.0160143802483876, + "pair_accuracy": 0.8959333333333334, + "positive_goodness": 3.9194104285346136, + "runtime_seconds": 10.005050659179688 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.7235590193324619, + "lr": 0.03, + "negative_goodness": 1.0357334026813507, + "pair_accuracy": 0.8940888888888889, + "positive_goodness": 3.942307022221883, + "runtime_seconds": 9.972609519958496 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6745021367232005, + "lr": 0.03, + "negative_goodness": 0.9642095009803772, + "pair_accuracy": 0.9085111111111112, + "positive_goodness": 4.0482156278822155, + "runtime_seconds": 9.94254207611084 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5992626527150472, + "lr": 0.03, + "negative_goodness": 0.8920467013676961, + "pair_accuracy": 0.9271111110952165, + "positive_goodness": 4.488479331673516, + "runtime_seconds": 9.952384233474731 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6746314722273085, + "lr": 0.03, + "negative_goodness": 0.9322276020156013, + "pair_accuracy": 0.9022222222275205, + "positive_goodness": 4.242767919158935, + "runtime_seconds": 9.95505166053772 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.695904409927792, + "lr": 0.03, + "negative_goodness": 0.9807526109695435, + "pair_accuracy": 0.8954666666666666, + "positive_goodness": 3.985301328955756, + "runtime_seconds": 9.9653959274292 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6856453708807627, + "lr": 0.03, + "negative_goodness": 0.9685449261718326, + "pair_accuracy": 0.8975777777777778, + "positive_goodness": 4.037819399939643, + "runtime_seconds": 10.07592487335205 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6405441376553641, + "lr": 0.03, + "negative_goodness": 0.9084001415782504, + "pair_accuracy": 0.9122222222222223, + "positive_goodness": 4.277704297044542, + "runtime_seconds": 10.052971363067627 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6213930712064107, + "lr": 0.03, + "negative_goodness": 0.891893326801724, + "pair_accuracy": 0.9242222222222222, + "positive_goodness": 4.37259070693122, + "runtime_seconds": 9.991830587387085 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6297228010177612, + "lr": 0.03, + "negative_goodness": 0.8961210288365682, + "pair_accuracy": 0.9161555555555555, + "positive_goodness": 4.262028603363037, + "runtime_seconds": 9.989911794662476 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.647338176314036, + "lr": 0.03, + "negative_goodness": 0.9046119753837586, + "pair_accuracy": 0.9082444444338481, + "positive_goodness": 4.188191352886624, + "runtime_seconds": 9.999000549316406 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6731981682035658, + "lr": 0.03, + "negative_goodness": 0.9188089425616794, + "pair_accuracy": 0.8985777777671814, + "positive_goodness": 4.1417329043918185, + "runtime_seconds": 9.993606328964233 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6508849210712645, + "lr": 0.03, + "negative_goodness": 0.9018299841668871, + "pair_accuracy": 0.9115333333333333, + "positive_goodness": 4.177507799699572, + "runtime_seconds": 9.968447208404541 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5885925085438622, + "lr": 0.03, + "negative_goodness": 0.8462221169842614, + "pair_accuracy": 0.9280444444444444, + "positive_goodness": 4.3928562408447265, + "runtime_seconds": 10.00088095664978 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.7056479803562165, + "lr": 0.03, + "negative_goodness": 0.9771130065282186, + "pair_accuracy": 0.8927777777671814, + "positive_goodness": 3.9954342680189345, + "runtime_seconds": 10.01268458366394 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6254214613596598, + "lr": 0.03, + "negative_goodness": 0.9008965121587118, + "pair_accuracy": 0.9207111111005147, + "positive_goodness": 4.31839161406623, + "runtime_seconds": 10.014852523803711 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5961019582536485, + "lr": 0.03, + "negative_goodness": 0.845542149564955, + "pair_accuracy": 0.9194666666693158, + "positive_goodness": 4.481413532087538, + "runtime_seconds": 10.023297548294067 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6583883216963874, + "lr": 0.03, + "negative_goodness": 0.8677720418188307, + "pair_accuracy": 0.9043999999894036, + "positive_goodness": 4.229232146199545, + "runtime_seconds": 10.024507999420166 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6117039626863268, + "lr": 0.03, + "negative_goodness": 0.8451396239704556, + "pair_accuracy": 0.9235111111005148, + "positive_goodness": 4.414740050760905, + "runtime_seconds": 10.030121803283691 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6207292570537991, + "lr": 0.03, + "negative_goodness": 0.8818078737470839, + "pair_accuracy": 0.913177777756585, + "positive_goodness": 4.316921987491184, + "runtime_seconds": 10.029073476791382 + } + ], + "layer": 31 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9571427828047011, + "lr": 0.03, + "negative_goodness": 1.5520686111874051, + "pair_accuracy": 0.8731777777618832, + "positive_goodness": 3.3990124696095783, + "runtime_seconds": 10.039581775665283 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8794643289989895, + "lr": 0.03, + "negative_goodness": 1.3608488307740954, + "pair_accuracy": 0.8567777778307597, + "positive_goodness": 3.4355942749023436, + "runtime_seconds": 10.038503885269165 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7641720732159085, + "lr": 0.03, + "negative_goodness": 1.2215954839070637, + "pair_accuracy": 0.8891333333863153, + "positive_goodness": 3.7938356208377413, + "runtime_seconds": 10.042763471603394 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7166664213074578, + "lr": 0.03, + "negative_goodness": 1.1188248928493925, + "pair_accuracy": 0.895022222190433, + "positive_goodness": 4.012185805384318, + "runtime_seconds": 10.04386854171753 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7254519999821981, + "lr": 0.03, + "negative_goodness": 1.1377262819714016, + "pair_accuracy": 0.8964222221904331, + "positive_goodness": 3.975534745279948, + "runtime_seconds": 10.04730772972107 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7029708911948734, + "lr": 0.03, + "negative_goodness": 1.0940044045448303, + "pair_accuracy": 0.8981111111005148, + "positive_goodness": 4.074843015289306, + "runtime_seconds": 10.037362337112427 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6801964372952779, + "lr": 0.03, + "negative_goodness": 1.0449840445624456, + "pair_accuracy": 0.8990666666348776, + "positive_goodness": 4.269941180123223, + "runtime_seconds": 10.050221920013428 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.675688521194458, + "lr": 0.03, + "negative_goodness": 1.0224669211493598, + "pair_accuracy": 0.8932444444444444, + "positive_goodness": 4.321503373887804, + "runtime_seconds": 10.03706669807434 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7042894311242633, + "lr": 0.03, + "negative_goodness": 1.063742868677775, + "pair_accuracy": 0.8948666666666667, + "positive_goodness": 4.044369224378798, + "runtime_seconds": 10.034702777862549 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6790701773643494, + "lr": 0.03, + "negative_goodness": 1.052576384968228, + "pair_accuracy": 0.9058666667090521, + "positive_goodness": 4.115242134475708, + "runtime_seconds": 10.036365032196045 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7306722278171115, + "lr": 0.03, + "negative_goodness": 1.0795858651796977, + "pair_accuracy": 0.8833999999788072, + "positive_goodness": 4.001572648281521, + "runtime_seconds": 10.039581537246704 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6491553808424207, + "lr": 0.03, + "negative_goodness": 1.018415724934472, + "pair_accuracy": 0.9097111111005147, + "positive_goodness": 4.222223449961344, + "runtime_seconds": 10.035427570343018 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6580566675610012, + "lr": 0.03, + "negative_goodness": 1.0083212010065714, + "pair_accuracy": 0.9056666666666666, + "positive_goodness": 4.186473304155138, + "runtime_seconds": 10.034303426742554 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6580777748531765, + "lr": 0.03, + "negative_goodness": 0.9913626222822401, + "pair_accuracy": 0.9039555555343628, + "positive_goodness": 4.220936089070638, + "runtime_seconds": 10.035549879074097 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6092214242405362, + "lr": 0.03, + "negative_goodness": 0.9306989562299517, + "pair_accuracy": 0.915, + "positive_goodness": 4.4361083973354765, + "runtime_seconds": 10.077055215835571 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6917043537775676, + "lr": 0.03, + "negative_goodness": 1.0348192432191636, + "pair_accuracy": 0.9019333333333334, + "positive_goodness": 4.154803136698405, + "runtime_seconds": 10.094709873199463 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6509540586895413, + "lr": 0.03, + "negative_goodness": 0.9689782274564107, + "pair_accuracy": 0.9058888889418708, + "positive_goodness": 4.230463677512275, + "runtime_seconds": 10.099379777908325 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.601288715145323, + "lr": 0.03, + "negative_goodness": 0.9354167592896355, + "pair_accuracy": 0.9126222222222222, + "positive_goodness": 4.431241020202637, + "runtime_seconds": 10.08792757987976 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6532455193678538, + "lr": 0.03, + "negative_goodness": 0.9694784748501248, + "pair_accuracy": 0.9087111111111111, + "positive_goodness": 4.2373405392964685, + "runtime_seconds": 10.116528034210205 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6524370477411482, + "lr": 0.03, + "negative_goodness": 0.9723046496338315, + "pair_accuracy": 0.9121111111111111, + "positive_goodness": 4.19987468761868, + "runtime_seconds": 10.109500885009766 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.654826272837321, + "lr": 0.03, + "negative_goodness": 0.967896542316013, + "pair_accuracy": 0.9018222222261959, + "positive_goodness": 4.268311320580376, + "runtime_seconds": 10.087544679641724 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.598096206665039, + "lr": 0.03, + "negative_goodness": 0.9158814609527588, + "pair_accuracy": 0.9203111111005148, + "positive_goodness": 4.437094774034288, + "runtime_seconds": 10.095343589782715 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6432597007115682, + "lr": 0.03, + "negative_goodness": 0.9645690663443671, + "pair_accuracy": 0.9102666666560703, + "positive_goodness": 4.157738592232598, + "runtime_seconds": 10.098874568939209 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6358317419581944, + "lr": 0.03, + "negative_goodness": 0.93814405930837, + "pair_accuracy": 0.910844444433848, + "positive_goodness": 4.301655565897624, + "runtime_seconds": 10.02159857749939 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.674162274424235, + "lr": 0.03, + "negative_goodness": 0.9654614650938246, + "pair_accuracy": 0.9006000000132455, + "positive_goodness": 4.140436628511217, + "runtime_seconds": 9.9755380153656 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.651431692843967, + "lr": 0.03, + "negative_goodness": 0.9781975006527371, + "pair_accuracy": 0.9090222222222222, + "positive_goodness": 4.114314656405979, + "runtime_seconds": 9.95893406867981 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6496234606822332, + "lr": 0.03, + "negative_goodness": 0.9579895099772348, + "pair_accuracy": 0.9081777777777778, + "positive_goodness": 4.185779586368137, + "runtime_seconds": 9.957132577896118 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6751617731306288, + "lr": 0.03, + "negative_goodness": 0.988876019583808, + "pair_accuracy": 0.8933111111111111, + "positive_goodness": 4.173401971435547, + "runtime_seconds": 9.958920001983643 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6539076081805759, + "lr": 0.03, + "negative_goodness": 0.9367621976799435, + "pair_accuracy": 0.8952666666666667, + "positive_goodness": 4.18576299235026, + "runtime_seconds": 9.980079650878906 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6136250913619995, + "lr": 0.03, + "negative_goodness": 0.904463730208079, + "pair_accuracy": 0.9129555555555555, + "positive_goodness": 4.353628818511963, + "runtime_seconds": 9.80945611000061 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6644755024592082, + "lr": 0.03, + "negative_goodness": 0.9664013893657261, + "pair_accuracy": 0.8946888889100817, + "positive_goodness": 4.170952260080973, + "runtime_seconds": 9.966344356536865 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6019549238204956, + "lr": 0.03, + "negative_goodness": 0.890285473314921, + "pair_accuracy": 0.9147555555555555, + "positive_goodness": 4.457985679372151, + "runtime_seconds": 9.960948944091797 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5802971987141503, + "lr": 0.03, + "negative_goodness": 0.8840256461673313, + "pair_accuracy": 0.9234888888782925, + "positive_goodness": 4.531624133385552, + "runtime_seconds": 9.960318326950073 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6517886258973016, + "lr": 0.03, + "negative_goodness": 0.9573378620995415, + "pair_accuracy": 0.9069111111111111, + "positive_goodness": 4.239572627936469, + "runtime_seconds": 9.980063915252686 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.668260863939921, + "lr": 0.03, + "negative_goodness": 0.949830303838518, + "pair_accuracy": 0.8990000000529819, + "positive_goodness": 4.130654037729899, + "runtime_seconds": 10.021864652633667 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5937122045252058, + "lr": 0.03, + "negative_goodness": 0.9155929442829556, + "pair_accuracy": 0.9256666666560703, + "positive_goodness": 4.405227042728, + "runtime_seconds": 10.015639781951904 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5416820426093207, + "lr": 0.03, + "negative_goodness": 0.8407078239864774, + "pair_accuracy": 0.9348666666348775, + "positive_goodness": 4.664542295159234, + "runtime_seconds": 10.019673109054565 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6586524071481493, + "lr": 0.03, + "negative_goodness": 0.9317456098344591, + "pair_accuracy": 0.8930666666560703, + "positive_goodness": 4.2250420800526935, + "runtime_seconds": 10.017554521560669 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6180063926272922, + "lr": 0.03, + "negative_goodness": 0.8802477085325453, + "pair_accuracy": 0.9085333333439297, + "positive_goodness": 4.428954905615913, + "runtime_seconds": 10.012910842895508 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5924858799987369, + "lr": 0.03, + "negative_goodness": 0.8800422372817993, + "pair_accuracy": 0.9206666666560703, + "positive_goodness": 4.419239108700222, + "runtime_seconds": 10.01402997970581 + } + ], + "layer": 32 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8775403936068217, + "lr": 0.03, + "negative_goodness": 1.2273791823175217, + "pair_accuracy": 0.8661333333015442, + "positive_goodness": 3.420658640586005, + "runtime_seconds": 9.992662906646729 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7927472955120934, + "lr": 0.03, + "negative_goodness": 1.17883691294988, + "pair_accuracy": 0.9036444444338481, + "positive_goodness": 3.6525686351352267, + "runtime_seconds": 9.990451574325562 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8065639284133911, + "lr": 0.03, + "negative_goodness": 1.1892847421858046, + "pair_accuracy": 0.8838000000529819, + "positive_goodness": 3.64062059334649, + "runtime_seconds": 9.99244236946106 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7591941640430027, + "lr": 0.03, + "negative_goodness": 1.122080318461524, + "pair_accuracy": 0.9033555555343628, + "positive_goodness": 3.811155273861355, + "runtime_seconds": 9.996013402938843 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7338217303435007, + "lr": 0.03, + "negative_goodness": 1.0963964272499085, + "pair_accuracy": 0.906044444433848, + "positive_goodness": 3.8497717044830324, + "runtime_seconds": 9.99138879776001 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7685676646444533, + "lr": 0.03, + "negative_goodness": 1.1290455292807684, + "pair_accuracy": 0.8805111111111111, + "positive_goodness": 3.7093546789381238, + "runtime_seconds": 9.993512868881226 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7316480138778686, + "lr": 0.03, + "negative_goodness": 1.078336469078064, + "pair_accuracy": 0.9008222222010295, + "positive_goodness": 3.7862786136627196, + "runtime_seconds": 10.005736827850342 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7039810042540232, + "lr": 0.03, + "negative_goodness": 1.0398155017852784, + "pair_accuracy": 0.9023333333333333, + "positive_goodness": 3.9362326671176486, + "runtime_seconds": 9.999104261398315 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6947509299437205, + "lr": 0.03, + "negative_goodness": 1.0034758990234798, + "pair_accuracy": 0.9054888888888889, + "positive_goodness": 3.9959831748962404, + "runtime_seconds": 9.996104001998901 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6988318108558654, + "lr": 0.03, + "negative_goodness": 0.9939229785919189, + "pair_accuracy": 0.9001111111111111, + "positive_goodness": 3.991137611728244, + "runtime_seconds": 9.99892807006836 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6722750585291121, + "lr": 0.03, + "negative_goodness": 0.9686307286752595, + "pair_accuracy": 0.9126, + "positive_goodness": 4.072514661492241, + "runtime_seconds": 9.991799592971802 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6405874509440528, + "lr": 0.03, + "negative_goodness": 0.923414528910319, + "pair_accuracy": 0.9093999999894036, + "positive_goodness": 4.171556795501709, + "runtime_seconds": 9.993040084838867 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6565637683444553, + "lr": 0.03, + "negative_goodness": 0.9368971870422363, + "pair_accuracy": 0.9031555555595292, + "positive_goodness": 4.184329573567708, + "runtime_seconds": 9.999590396881104 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6565613461070591, + "lr": 0.03, + "negative_goodness": 0.9381132279581493, + "pair_accuracy": 0.9147777777777778, + "positive_goodness": 4.133940472581651, + "runtime_seconds": 10.010251998901367 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6447041301091512, + "lr": 0.03, + "negative_goodness": 0.9203395280943977, + "pair_accuracy": 0.9095555555979411, + "positive_goodness": 4.2028245365142824, + "runtime_seconds": 10.011644840240479 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5567014875306023, + "lr": 0.03, + "negative_goodness": 0.8435284137937757, + "pair_accuracy": 0.9455333333015442, + "positive_goodness": 4.562089350255331, + "runtime_seconds": 9.989068984985352 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6305024337450663, + "lr": 0.03, + "negative_goodness": 0.8925985129674275, + "pair_accuracy": 0.9178888889418708, + "positive_goodness": 4.2576495532989505, + "runtime_seconds": 9.959112405776978 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6617442392243279, + "lr": 0.03, + "negative_goodness": 0.9226030493153466, + "pair_accuracy": 0.9122222222222223, + "positive_goodness": 4.119569735972086, + "runtime_seconds": 9.932730913162231 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5761878124660916, + "lr": 0.03, + "negative_goodness": 0.8072646861924065, + "pair_accuracy": 0.9253777777989706, + "positive_goodness": 4.532887953694662, + "runtime_seconds": 9.97112774848938 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6621815564685398, + "lr": 0.03, + "negative_goodness": 0.9145419657071432, + "pair_accuracy": 0.8994444444868299, + "positive_goodness": 4.210193179088169, + "runtime_seconds": 9.990675687789917 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5213090593655905, + "lr": 0.03, + "negative_goodness": 0.7797811721060011, + "pair_accuracy": 0.9454222222222223, + "positive_goodness": 4.8625082515292695, + "runtime_seconds": 9.994551420211792 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6229051810158623, + "lr": 0.03, + "negative_goodness": 0.8762820412635803, + "pair_accuracy": 0.9171555555608537, + "positive_goodness": 4.412780305322011, + "runtime_seconds": 9.99372410774231 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5779249543719821, + "lr": 0.03, + "negative_goodness": 0.829430890888638, + "pair_accuracy": 0.9247111110952165, + "positive_goodness": 4.504672818438212, + "runtime_seconds": 10.045057535171509 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.623989332347446, + "lr": 0.03, + "negative_goodness": 0.8743355153348711, + "pair_accuracy": 0.9118888889312744, + "positive_goodness": 4.255456501324971, + "runtime_seconds": 10.12296438217163 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6041246428410212, + "lr": 0.03, + "negative_goodness": 0.836143303835392, + "pair_accuracy": 0.9269333333333334, + "positive_goodness": 4.399119314829508, + "runtime_seconds": 10.123973608016968 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6392179711871677, + "lr": 0.03, + "negative_goodness": 0.8958638978110419, + "pair_accuracy": 0.9136, + "positive_goodness": 4.288931726752387, + "runtime_seconds": 10.13003158569336 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6190047213872274, + "lr": 0.03, + "negative_goodness": 0.8671034500969781, + "pair_accuracy": 0.9134, + "positive_goodness": 4.371512220170763, + "runtime_seconds": 10.13254427909851 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6143062329875099, + "lr": 0.03, + "negative_goodness": 0.8397880694601271, + "pair_accuracy": 0.9191111111111111, + "positive_goodness": 4.311558707682292, + "runtime_seconds": 10.139798402786255 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5951693809138404, + "lr": 0.03, + "negative_goodness": 0.8188633423169454, + "pair_accuracy": 0.9251777777777778, + "positive_goodness": 4.424571667395698, + "runtime_seconds": 10.138149499893188 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5824643005715476, + "lr": 0.03, + "negative_goodness": 0.8094125044557783, + "pair_accuracy": 0.9229555555555555, + "positive_goodness": 4.398123372395833, + "runtime_seconds": 10.01394772529602 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5527095999532275, + "lr": 0.03, + "negative_goodness": 0.7743318483485115, + "pair_accuracy": 0.9304888888888889, + "positive_goodness": 4.67305200076633, + "runtime_seconds": 10.004955768585205 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5671391041782167, + "lr": 0.03, + "negative_goodness": 0.7992114481422636, + "pair_accuracy": 0.926, + "positive_goodness": 4.566206705559624, + "runtime_seconds": 9.995394229888916 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6404328046586778, + "lr": 0.03, + "negative_goodness": 0.8786908142937554, + "pair_accuracy": 0.9092888888782925, + "positive_goodness": 4.303854938083225, + "runtime_seconds": 10.008742809295654 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5988216574350993, + "lr": 0.03, + "negative_goodness": 0.8177063032574123, + "pair_accuracy": 0.9145111110899183, + "positive_goodness": 4.455991916402181, + "runtime_seconds": 10.013883590698242 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5880609906276068, + "lr": 0.03, + "negative_goodness": 0.8075659817384349, + "pair_accuracy": 0.9291555555555555, + "positive_goodness": 4.484889797210693, + "runtime_seconds": 10.016686201095581 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5678715657578575, + "lr": 0.03, + "negative_goodness": 0.8059767995331022, + "pair_accuracy": 0.9290222222222222, + "positive_goodness": 4.506658736165365, + "runtime_seconds": 10.014960765838623 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.681596632639567, + "lr": 0.03, + "negative_goodness": 0.8925116480297512, + "pair_accuracy": 0.8839111111031638, + "positive_goodness": 4.253534789111879, + "runtime_seconds": 10.02083945274353 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6148919690026178, + "lr": 0.03, + "negative_goodness": 0.8630514293193817, + "pair_accuracy": 0.914666666677263, + "positive_goodness": 4.345549751027425, + "runtime_seconds": 10.017074584960938 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5865168914371066, + "lr": 0.03, + "negative_goodness": 0.83904118636979, + "pair_accuracy": 0.9225555555449592, + "positive_goodness": 4.435555789438883, + "runtime_seconds": 10.021092176437378 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5875494206640456, + "lr": 0.03, + "negative_goodness": 0.7921849565347036, + "pair_accuracy": 0.9165111111005148, + "positive_goodness": 4.527910473293728, + "runtime_seconds": 10.01183557510376 + } + ], + "layer": 33 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0322268606821696, + "lr": 0.03, + "negative_goodness": 1.499512560749054, + "pair_accuracy": 0.8447111111005148, + "positive_goodness": 3.4011474119822185, + "runtime_seconds": 10.042738676071167 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8970918490038978, + "lr": 0.03, + "negative_goodness": 1.2760134171062045, + "pair_accuracy": 0.8587777777777778, + "positive_goodness": 3.5121770126342775, + "runtime_seconds": 10.038506269454956 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.821084361606174, + "lr": 0.03, + "negative_goodness": 1.1945672932518854, + "pair_accuracy": 0.8815333333346579, + "positive_goodness": 3.6766391008377077, + "runtime_seconds": 10.044474124908447 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7742545225779216, + "lr": 0.03, + "negative_goodness": 1.11501746963395, + "pair_accuracy": 0.8850888888888889, + "positive_goodness": 3.8142459789699976, + "runtime_seconds": 10.042149305343628 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7330536900626289, + "lr": 0.03, + "negative_goodness": 1.0507697550879584, + "pair_accuracy": 0.8910000000211927, + "positive_goodness": 3.9470704854541356, + "runtime_seconds": 10.046587944030762 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7463804876115587, + "lr": 0.03, + "negative_goodness": 1.0578610524601406, + "pair_accuracy": 0.8950444444444444, + "positive_goodness": 3.971044548288981, + "runtime_seconds": 10.040633916854858 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7250130729887221, + "lr": 0.03, + "negative_goodness": 1.0129307129542033, + "pair_accuracy": 0.8891555555873447, + "positive_goodness": 4.006303492418925, + "runtime_seconds": 10.044931650161743 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.724167554834154, + "lr": 0.03, + "negative_goodness": 1.0106694728003607, + "pair_accuracy": 0.8889777777671813, + "positive_goodness": 4.047528644858466, + "runtime_seconds": 10.029603719711304 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6869968540297614, + "lr": 0.03, + "negative_goodness": 1.010615208530426, + "pair_accuracy": 0.9027555555343628, + "positive_goodness": 4.170037489022149, + "runtime_seconds": 10.028364658355713 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7400360490798951, + "lr": 0.03, + "negative_goodness": 1.0389742355452642, + "pair_accuracy": 0.8911333333333333, + "positive_goodness": 4.012480381435818, + "runtime_seconds": 10.053470373153687 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7250646805445353, + "lr": 0.03, + "negative_goodness": 1.0066181325276693, + "pair_accuracy": 0.8931555555449592, + "positive_goodness": 3.9962674436781143, + "runtime_seconds": 9.98809266090393 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6819559499475691, + "lr": 0.03, + "negative_goodness": 0.964628881406784, + "pair_accuracy": 0.8947777777777778, + "positive_goodness": 4.156859045155843, + "runtime_seconds": 9.977768659591675 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7263831901550293, + "lr": 0.03, + "negative_goodness": 1.033441501214769, + "pair_accuracy": 0.8859999999894036, + "positive_goodness": 3.9794387770758735, + "runtime_seconds": 9.986523389816284 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6749831137180329, + "lr": 0.03, + "negative_goodness": 0.937254129796558, + "pair_accuracy": 0.8951333333333333, + "positive_goodness": 4.261985037316216, + "runtime_seconds": 9.978404998779297 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6800421013090345, + "lr": 0.03, + "negative_goodness": 0.969499709447225, + "pair_accuracy": 0.8988000000317892, + "positive_goodness": 4.241831187820434, + "runtime_seconds": 9.977480411529541 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6704044833289252, + "lr": 0.03, + "negative_goodness": 0.9378520092858209, + "pair_accuracy": 0.9006888888994853, + "positive_goodness": 4.381905769115024, + "runtime_seconds": 9.980923652648926 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5907360054016113, + "lr": 0.03, + "negative_goodness": 0.841774925602807, + "pair_accuracy": 0.9210888888676961, + "positive_goodness": 4.626781287765503, + "runtime_seconds": 9.981258869171143 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6716638093736437, + "lr": 0.03, + "negative_goodness": 0.9219930821524726, + "pair_accuracy": 0.8962222222195732, + "positive_goodness": 4.284145861074659, + "runtime_seconds": 9.97743535041809 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6553468619081709, + "lr": 0.03, + "negative_goodness": 0.9103772512012058, + "pair_accuracy": 0.9050666666666667, + "positive_goodness": 4.281350135887994, + "runtime_seconds": 10.042561531066895 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7751317802535163, + "lr": 0.03, + "negative_goodness": 1.0692926647345224, + "pair_accuracy": 0.8762666666666666, + "positive_goodness": 3.870941742875841, + "runtime_seconds": 10.083948373794556 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6701943402608236, + "lr": 0.03, + "negative_goodness": 0.9398774054421319, + "pair_accuracy": 0.905244444460339, + "positive_goodness": 4.103929814910889, + "runtime_seconds": 10.091166496276855 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6580072066201104, + "lr": 0.03, + "negative_goodness": 0.9557242758644952, + "pair_accuracy": 0.9069555555449592, + "positive_goodness": 4.240184212409125, + "runtime_seconds": 10.102039575576782 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.65268775203228, + "lr": 0.03, + "negative_goodness": 0.9284731643093956, + "pair_accuracy": 0.9085555555555556, + "positive_goodness": 4.232778013695611, + "runtime_seconds": 10.093748092651367 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6822126950793796, + "lr": 0.03, + "negative_goodness": 0.9150874194463094, + "pair_accuracy": 0.8979555555555555, + "positive_goodness": 4.214970655737983, + "runtime_seconds": 10.086221933364868 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6270517845259772, + "lr": 0.03, + "negative_goodness": 0.8801493833753797, + "pair_accuracy": 0.9190888888782925, + "positive_goodness": 4.352150057898627, + "runtime_seconds": 10.006826877593994 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6593625307771894, + "lr": 0.03, + "negative_goodness": 0.9321091445181104, + "pair_accuracy": 0.9062888888888889, + "positive_goodness": 4.263657899051243, + "runtime_seconds": 9.97603702545166 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6528644825617472, + "lr": 0.03, + "negative_goodness": 0.8930055034849379, + "pair_accuracy": 0.8982888888782925, + "positive_goodness": 4.321861668226454, + "runtime_seconds": 9.975977420806885 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6180813258144591, + "lr": 0.03, + "negative_goodness": 0.884113206908438, + "pair_accuracy": 0.9173555555555556, + "positive_goodness": 4.3812693970574275, + "runtime_seconds": 9.98067855834961 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6221480687141419, + "lr": 0.03, + "negative_goodness": 0.8662332567638821, + "pair_accuracy": 0.9131777777459886, + "positive_goodness": 4.446895021650526, + "runtime_seconds": 9.97879147529602 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.669458787396219, + "lr": 0.03, + "negative_goodness": 0.9248237756874826, + "pair_accuracy": 0.8953555555555556, + "positive_goodness": 4.252143858422174, + "runtime_seconds": 9.970752239227295 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6689498403337266, + "lr": 0.03, + "negative_goodness": 0.9420995299975077, + "pair_accuracy": 0.9025777777777778, + "positive_goodness": 4.292258964962429, + "runtime_seconds": 10.005688190460205 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6483897400750054, + "lr": 0.03, + "negative_goodness": 0.8891202805572086, + "pair_accuracy": 0.8994666666666666, + "positive_goodness": 4.3786569729275175, + "runtime_seconds": 10.088218927383423 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5999346148702833, + "lr": 0.03, + "negative_goodness": 0.8756128525627984, + "pair_accuracy": 0.9239555555555555, + "positive_goodness": 4.55067520260281, + "runtime_seconds": 10.050219297409058 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6417808109283447, + "lr": 0.03, + "negative_goodness": 0.8748477415773603, + "pair_accuracy": 0.9001333333333333, + "positive_goodness": 4.370932582685683, + "runtime_seconds": 9.995544195175171 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6566176889869901, + "lr": 0.03, + "negative_goodness": 0.9097574879540338, + "pair_accuracy": 0.9069333333333334, + "positive_goodness": 4.286806278906928, + "runtime_seconds": 9.979860067367554 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6513466975238588, + "lr": 0.03, + "negative_goodness": 0.9046460872358746, + "pair_accuracy": 0.9078444444444445, + "positive_goodness": 4.274458298831516, + "runtime_seconds": 9.98018193244934 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6256502517912123, + "lr": 0.03, + "negative_goodness": 0.867761403327518, + "pair_accuracy": 0.9108888888782926, + "positive_goodness": 4.452331737857395, + "runtime_seconds": 9.979122877120972 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6088246240297953, + "lr": 0.03, + "negative_goodness": 0.8406873490015666, + "pair_accuracy": 0.9172666666666667, + "positive_goodness": 4.450889373694526, + "runtime_seconds": 9.974372625350952 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5812595694647895, + "lr": 0.03, + "negative_goodness": 0.8225127309838931, + "pair_accuracy": 0.9252666666666667, + "positive_goodness": 4.6902395175509985, + "runtime_seconds": 9.979822397232056 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.601476653043429, + "lr": 0.03, + "negative_goodness": 0.8278890450689528, + "pair_accuracy": 0.9160666666666667, + "positive_goodness": 4.519076377360026, + "runtime_seconds": 9.973611831665039 + } + ], + "layer": 34 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9515418414857653, + "lr": 0.03, + "negative_goodness": 1.2739862032572429, + "pair_accuracy": 0.8322666666878594, + "positive_goodness": 3.175459010865953, + "runtime_seconds": 9.998822927474976 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8477554979430305, + "lr": 0.03, + "negative_goodness": 1.1933743820402358, + "pair_accuracy": 0.8724666666454739, + "positive_goodness": 3.517106308492025, + "runtime_seconds": 10.022502899169922 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8686725905100504, + "lr": 0.03, + "negative_goodness": 1.2367680440478854, + "pair_accuracy": 0.8567333333227369, + "positive_goodness": 3.3895784996880427, + "runtime_seconds": 9.967468738555908 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7564294039620294, + "lr": 0.03, + "negative_goodness": 1.076971116341485, + "pair_accuracy": 0.8941333333121406, + "positive_goodness": 3.79045933848487, + "runtime_seconds": 9.968336343765259 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7898334192117056, + "lr": 0.03, + "negative_goodness": 1.1155597940338982, + "pair_accuracy": 0.8804222222222222, + "positive_goodness": 3.64250667008294, + "runtime_seconds": 9.962807178497314 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7427123117234972, + "lr": 0.03, + "negative_goodness": 1.0518072357707553, + "pair_accuracy": 0.8871333333863153, + "positive_goodness": 3.8428616831885445, + "runtime_seconds": 9.961266994476318 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7824915260314942, + "lr": 0.03, + "negative_goodness": 1.090081955104404, + "pair_accuracy": 0.8704666667196486, + "positive_goodness": 3.7575076669481065, + "runtime_seconds": 9.962836503982544 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7109308692508274, + "lr": 0.03, + "negative_goodness": 0.9943841414875454, + "pair_accuracy": 0.8987333333757188, + "positive_goodness": 3.9533771893395317, + "runtime_seconds": 9.960517406463623 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7338733688195547, + "lr": 0.03, + "negative_goodness": 1.0251271610048083, + "pair_accuracy": 0.8910222222222223, + "positive_goodness": 3.8738577471415203, + "runtime_seconds": 9.960889339447021 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6972229278988309, + "lr": 0.03, + "negative_goodness": 0.975767645072937, + "pair_accuracy": 0.9016666666878594, + "positive_goodness": 3.9942685209698148, + "runtime_seconds": 9.96416187286377 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6961530681980981, + "lr": 0.03, + "negative_goodness": 0.9708151338365343, + "pair_accuracy": 0.8929333333333334, + "positive_goodness": 4.083466359795464, + "runtime_seconds": 9.962813138961792 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7515587283028496, + "lr": 0.03, + "negative_goodness": 1.012211818684472, + "pair_accuracy": 0.8817333333333334, + "positive_goodness": 3.7750276686774358, + "runtime_seconds": 9.960017919540405 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7047059248394436, + "lr": 0.03, + "negative_goodness": 0.9801699511845906, + "pair_accuracy": 0.8866666667090521, + "positive_goodness": 4.001315868504842, + "runtime_seconds": 9.962721586227417 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7149894047631158, + "lr": 0.03, + "negative_goodness": 0.9620648279825846, + "pair_accuracy": 0.8903333333227369, + "positive_goodness": 4.053847910139296, + "runtime_seconds": 9.9607572555542 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6984740332921346, + "lr": 0.03, + "negative_goodness": 0.961666829723782, + "pair_accuracy": 0.8911555555449592, + "positive_goodness": 3.953339498562283, + "runtime_seconds": 9.957852602005005 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6864043823983934, + "lr": 0.03, + "negative_goodness": 0.9628466314103868, + "pair_accuracy": 0.8952666666666667, + "positive_goodness": 4.027035693571302, + "runtime_seconds": 9.964977025985718 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6641308839003245, + "lr": 0.03, + "negative_goodness": 0.9061342604504691, + "pair_accuracy": 0.9025333333333333, + "positive_goodness": 4.167397422620986, + "runtime_seconds": 9.96232295036316 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.7229331155141194, + "lr": 0.03, + "negative_goodness": 0.9940783415688409, + "pair_accuracy": 0.8843999999894037, + "positive_goodness": 3.9369881610446504, + "runtime_seconds": 9.959840536117554 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6856107444551256, + "lr": 0.03, + "negative_goodness": 0.9385235283957587, + "pair_accuracy": 0.8972000000026491, + "positive_goodness": 4.066670898416307, + "runtime_seconds": 9.960780382156372 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.7028258686277602, + "lr": 0.03, + "negative_goodness": 0.9692251867930094, + "pair_accuracy": 0.8971777777459886, + "positive_goodness": 3.9588773829989963, + "runtime_seconds": 9.959509134292603 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6859439936372969, + "lr": 0.03, + "negative_goodness": 0.9353966668658786, + "pair_accuracy": 0.8972666666666667, + "positive_goodness": 4.034217646450466, + "runtime_seconds": 9.966397285461426 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6387627098613315, + "lr": 0.03, + "negative_goodness": 0.871050050968594, + "pair_accuracy": 0.91, + "positive_goodness": 4.280765774705675, + "runtime_seconds": 9.961848020553589 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6693130444420708, + "lr": 0.03, + "negative_goodness": 0.8756975954691569, + "pair_accuracy": 0.9012, + "positive_goodness": 4.198388085444768, + "runtime_seconds": 9.968264102935791 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6847673743353949, + "lr": 0.03, + "negative_goodness": 0.9233788727336459, + "pair_accuracy": 0.8890888888994852, + "positive_goodness": 4.0239408730189, + "runtime_seconds": 9.960436344146729 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.69750763897366, + "lr": 0.03, + "negative_goodness": 0.9113589918772379, + "pair_accuracy": 0.8897555555873446, + "positive_goodness": 4.027187818527222, + "runtime_seconds": 9.956815242767334 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6550784800979826, + "lr": 0.03, + "negative_goodness": 0.8718901119656033, + "pair_accuracy": 0.8956666666666667, + "positive_goodness": 4.126782785373264, + "runtime_seconds": 9.959018230438232 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6625450910144383, + "lr": 0.03, + "negative_goodness": 0.8762359239896138, + "pair_accuracy": 0.8949777778201633, + "positive_goodness": 4.218327990976969, + "runtime_seconds": 9.988890409469604 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.666782505332099, + "lr": 0.03, + "negative_goodness": 0.8946670900132921, + "pair_accuracy": 0.8948222222222222, + "positive_goodness": 4.115743523809645, + "runtime_seconds": 10.021612882614136 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6447014591323005, + "lr": 0.03, + "negative_goodness": 0.8854921655866835, + "pair_accuracy": 0.9058666667196486, + "positive_goodness": 4.192705834579468, + "runtime_seconds": 10.019863843917847 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6496417700979444, + "lr": 0.03, + "negative_goodness": 0.8827225479443868, + "pair_accuracy": 0.9071777777883742, + "positive_goodness": 4.186931176503499, + "runtime_seconds": 9.97486686706543 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6622269511222839, + "lr": 0.03, + "negative_goodness": 0.8767299788130655, + "pair_accuracy": 0.8956888888888889, + "positive_goodness": 4.268105158699884, + "runtime_seconds": 9.96399474143982 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6740938622633617, + "lr": 0.03, + "negative_goodness": 0.8776231344646878, + "pair_accuracy": 0.9009777777777778, + "positive_goodness": 4.1574000651041665, + "runtime_seconds": 10.016211986541748 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6424284669876099, + "lr": 0.03, + "negative_goodness": 0.8576509076754252, + "pair_accuracy": 0.9002222222328186, + "positive_goodness": 4.19173925098843, + "runtime_seconds": 9.992271661758423 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6756472893555959, + "lr": 0.03, + "negative_goodness": 0.948481321700414, + "pair_accuracy": 0.893, + "positive_goodness": 4.102813733334012, + "runtime_seconds": 10.078316450119019 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6722494408289591, + "lr": 0.03, + "negative_goodness": 0.8997828509542677, + "pair_accuracy": 0.8982222222275205, + "positive_goodness": 4.138722323396471, + "runtime_seconds": 10.085516929626465 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6569785958078173, + "lr": 0.03, + "negative_goodness": 0.8853114342371623, + "pair_accuracy": 0.8984, + "positive_goodness": 4.15676932703654, + "runtime_seconds": 9.99568223953247 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.672388138071696, + "lr": 0.03, + "negative_goodness": 0.8971145927429199, + "pair_accuracy": 0.9032000000001655, + "positive_goodness": 4.239187364027235, + "runtime_seconds": 9.95421290397644 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6074937474568685, + "lr": 0.03, + "negative_goodness": 0.8203023203425938, + "pair_accuracy": 0.9132222222275204, + "positive_goodness": 4.460209921688504, + "runtime_seconds": 9.961001873016357 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6636783986621433, + "lr": 0.03, + "negative_goodness": 0.9038263324949476, + "pair_accuracy": 0.8988666666348776, + "positive_goodness": 4.155631894005669, + "runtime_seconds": 9.962318420410156 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6259292001962662, + "lr": 0.03, + "negative_goodness": 0.8563831973764632, + "pair_accuracy": 0.9043333333333333, + "positive_goodness": 4.329961453840467, + "runtime_seconds": 9.958215236663818 + } + ], + "layer": 35 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.017660606765747, + "lr": 0.03, + "negative_goodness": 1.4644456815507676, + "pair_accuracy": 0.8530222222116258, + "positive_goodness": 3.234465563498603, + "runtime_seconds": 9.985161781311035 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8601226961135864, + "lr": 0.03, + "negative_goodness": 1.2785305224524603, + "pair_accuracy": 0.8827111111111111, + "positive_goodness": 3.6159995277404784, + "runtime_seconds": 9.980775594711304 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8437308552847969, + "lr": 0.03, + "negative_goodness": 1.2370693484306337, + "pair_accuracy": 0.8788666666454739, + "positive_goodness": 3.584685969034831, + "runtime_seconds": 9.980294466018677 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7977903335995145, + "lr": 0.03, + "negative_goodness": 1.1735356215159098, + "pair_accuracy": 0.8827333333757189, + "positive_goodness": 3.7300540885925293, + "runtime_seconds": 9.981762409210205 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7896331175910102, + "lr": 0.03, + "negative_goodness": 1.1440556893666585, + "pair_accuracy": 0.8829555555555556, + "positive_goodness": 3.747936379114787, + "runtime_seconds": 9.981125593185425 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7867407500372993, + "lr": 0.03, + "negative_goodness": 1.1465790122561985, + "pair_accuracy": 0.8844222222228845, + "positive_goodness": 3.7279620622528924, + "runtime_seconds": 9.981381893157959 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7536214086956448, + "lr": 0.03, + "negative_goodness": 1.0929513443099128, + "pair_accuracy": 0.8854888888994853, + "positive_goodness": 3.9235703424453736, + "runtime_seconds": 9.99712347984314 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7235473349253336, + "lr": 0.03, + "negative_goodness": 1.062254759724935, + "pair_accuracy": 0.891466666677263, + "positive_goodness": 3.9118428479088676, + "runtime_seconds": 9.983838319778442 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7121923045900133, + "lr": 0.03, + "negative_goodness": 1.0394301169077556, + "pair_accuracy": 0.9007333333333334, + "positive_goodness": 4.011440042834812, + "runtime_seconds": 9.984777450561523 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7345801015218099, + "lr": 0.03, + "negative_goodness": 1.06815555973053, + "pair_accuracy": 0.8902, + "positive_goodness": 3.902160744476318, + "runtime_seconds": 9.982107400894165 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7159721934000651, + "lr": 0.03, + "negative_goodness": 1.030243331612481, + "pair_accuracy": 0.8950444444126553, + "positive_goodness": 4.132744580290052, + "runtime_seconds": 9.982451677322388 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7106224691020118, + "lr": 0.03, + "negative_goodness": 1.0333759019321866, + "pair_accuracy": 0.8966444444444445, + "positive_goodness": 4.023230458747016, + "runtime_seconds": 9.978997945785522 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7437363721635607, + "lr": 0.03, + "negative_goodness": 1.0348422218455209, + "pair_accuracy": 0.8833555555555556, + "positive_goodness": 3.8924326735602484, + "runtime_seconds": 9.98607325553894 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7478605555852255, + "lr": 0.03, + "negative_goodness": 1.0654895467970107, + "pair_accuracy": 0.8769999999682109, + "positive_goodness": 3.8985533212449814, + "runtime_seconds": 9.98323678970337 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6680435750643412, + "lr": 0.03, + "negative_goodness": 0.972067733669281, + "pair_accuracy": 0.9007111111005147, + "positive_goodness": 4.175576770443387, + "runtime_seconds": 10.025638103485107 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.7142782561884986, + "lr": 0.03, + "negative_goodness": 0.9884009260018667, + "pair_accuracy": 0.8887111111005147, + "positive_goodness": 4.025378960164388, + "runtime_seconds": 10.09997010231018 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.752757520225313, + "lr": 0.03, + "negative_goodness": 1.064283939743042, + "pair_accuracy": 0.8833333333333333, + "positive_goodness": 3.8733990881178113, + "runtime_seconds": 10.066629886627197 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6650405123551687, + "lr": 0.03, + "negative_goodness": 0.9572206263489194, + "pair_accuracy": 0.9022666666666667, + "positive_goodness": 4.1866550729539656, + "runtime_seconds": 10.052061796188354 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6629093949953715, + "lr": 0.03, + "negative_goodness": 0.9638626567204793, + "pair_accuracy": 0.9115333333121406, + "positive_goodness": 4.187503624767728, + "runtime_seconds": 10.056437730789185 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6741603254371219, + "lr": 0.03, + "negative_goodness": 0.9552570031907823, + "pair_accuracy": 0.8996444444444445, + "positive_goodness": 4.198576664818658, + "runtime_seconds": 9.997413396835327 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6901068243503571, + "lr": 0.03, + "negative_goodness": 0.9794173510551453, + "pair_accuracy": 0.895933333322737, + "positive_goodness": 4.102979101223416, + "runtime_seconds": 9.983061075210571 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6987583338154687, + "lr": 0.03, + "negative_goodness": 0.9787740860833062, + "pair_accuracy": 0.8899333333227369, + "positive_goodness": 4.080886443667942, + "runtime_seconds": 9.98528790473938 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6711122394561767, + "lr": 0.03, + "negative_goodness": 0.9439741090244717, + "pair_accuracy": 0.9002, + "positive_goodness": 4.182054471079509, + "runtime_seconds": 9.985221147537231 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6911050295008554, + "lr": 0.03, + "negative_goodness": 0.9580786972893609, + "pair_accuracy": 0.8917111111111111, + "positive_goodness": 4.032699739668105, + "runtime_seconds": 9.990108251571655 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6915610218471951, + "lr": 0.03, + "negative_goodness": 0.936229362127516, + "pair_accuracy": 0.8818222222222222, + "positive_goodness": 4.138957581498888, + "runtime_seconds": 9.99152660369873 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6483866043514676, + "lr": 0.03, + "negative_goodness": 0.9159507114569346, + "pair_accuracy": 0.9047555555555555, + "positive_goodness": 4.278015472242568, + "runtime_seconds": 10.004818439483643 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5900238190333048, + "lr": 0.03, + "negative_goodness": 0.8275630159166124, + "pair_accuracy": 0.9171111111323039, + "positive_goodness": 4.613340049425761, + "runtime_seconds": 10.04016399383545 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6264475736061732, + "lr": 0.03, + "negative_goodness": 0.8813688348399268, + "pair_accuracy": 0.9110888888888888, + "positive_goodness": 4.3638620231628416, + "runtime_seconds": 10.045512437820435 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6561417633454005, + "lr": 0.03, + "negative_goodness": 0.9196205754015181, + "pair_accuracy": 0.9048888888888889, + "positive_goodness": 4.171678367657131, + "runtime_seconds": 10.045051097869873 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6852696612119674, + "lr": 0.03, + "negative_goodness": 0.9520459472152922, + "pair_accuracy": 0.8909111111111111, + "positive_goodness": 4.145524906921387, + "runtime_seconds": 10.044321298599243 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.668940435968505, + "lr": 0.03, + "negative_goodness": 0.9308808668825361, + "pair_accuracy": 0.8977777777777778, + "positive_goodness": 4.245609780629476, + "runtime_seconds": 10.041504383087158 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6436759376949734, + "lr": 0.03, + "negative_goodness": 0.9039504076851739, + "pair_accuracy": 0.905155555597941, + "positive_goodness": 4.3162866705152725, + "runtime_seconds": 10.042577266693115 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6256821543163723, + "lr": 0.03, + "negative_goodness": 0.847578062333001, + "pair_accuracy": 0.9045333333545261, + "positive_goodness": 4.41184127368927, + "runtime_seconds": 10.039039611816406 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6328216609477997, + "lr": 0.03, + "negative_goodness": 0.8903494365268283, + "pair_accuracy": 0.9100444444444444, + "positive_goodness": 4.353873739284939, + "runtime_seconds": 10.039884805679321 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.7027934914218055, + "lr": 0.03, + "negative_goodness": 0.9632313057528602, + "pair_accuracy": 0.8819333333333333, + "positive_goodness": 4.052958532630073, + "runtime_seconds": 10.03977918624878 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6098917439672682, + "lr": 0.03, + "negative_goodness": 0.8584046775182088, + "pair_accuracy": 0.9149333333757188, + "positive_goodness": 4.430767816925049, + "runtime_seconds": 10.095376014709473 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.661377605316374, + "lr": 0.03, + "negative_goodness": 0.9161045271873474, + "pair_accuracy": 0.9068444444444445, + "positive_goodness": 4.238326773113674, + "runtime_seconds": 10.10683560371399 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6763242351531983, + "lr": 0.03, + "negative_goodness": 0.9187178114149306, + "pair_accuracy": 0.8973333333333333, + "positive_goodness": 4.194669361996651, + "runtime_seconds": 10.11292576789856 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5977383260832893, + "lr": 0.03, + "negative_goodness": 0.8452419548299578, + "pair_accuracy": 0.9165333333333333, + "positive_goodness": 4.410339419725206, + "runtime_seconds": 10.10292363166809 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6349437619527181, + "lr": 0.03, + "negative_goodness": 0.8593855927361382, + "pair_accuracy": 0.9026222222222222, + "positive_goodness": 4.297739414850871, + "runtime_seconds": 10.102056980133057 + } + ], + "layer": 36 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9844550242741903, + "lr": 0.03, + "negative_goodness": 1.2961218488057455, + "pair_accuracy": 0.8115555555449592, + "positive_goodness": 3.117497361755371, + "runtime_seconds": 10.09500527381897 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.775170346795188, + "lr": 0.03, + "negative_goodness": 1.101099989467197, + "pair_accuracy": 0.8892, + "positive_goodness": 3.792788373989529, + "runtime_seconds": 10.080813646316528 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8202306736628214, + "lr": 0.03, + "negative_goodness": 1.1390947163687812, + "pair_accuracy": 0.8702222222646078, + "positive_goodness": 3.6348761897193063, + "runtime_seconds": 10.082440614700317 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7937059325430128, + "lr": 0.03, + "negative_goodness": 1.122771711836921, + "pair_accuracy": 0.8789555555767483, + "positive_goodness": 3.6360692790137397, + "runtime_seconds": 10.073468446731567 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.736245004606247, + "lr": 0.03, + "negative_goodness": 1.0590656236754523, + "pair_accuracy": 0.8917111111111111, + "positive_goodness": 3.942247825495402, + "runtime_seconds": 10.073387622833252 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7540748082478841, + "lr": 0.03, + "negative_goodness": 1.0666477018780178, + "pair_accuracy": 0.8831111111005148, + "positive_goodness": 3.8389620517518783, + "runtime_seconds": 10.067903518676758 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7786173186937968, + "lr": 0.03, + "negative_goodness": 1.0896247003979154, + "pair_accuracy": 0.8778888888888889, + "positive_goodness": 3.7354992033216687, + "runtime_seconds": 10.078101873397827 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.738416267908944, + "lr": 0.03, + "negative_goodness": 1.0647863897535537, + "pair_accuracy": 0.8809777777777777, + "positive_goodness": 3.9465703553941514, + "runtime_seconds": 10.078062057495117 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.760775032403734, + "lr": 0.03, + "negative_goodness": 1.0692346486197577, + "pair_accuracy": 0.8784444444444445, + "positive_goodness": 3.8015551073710125, + "runtime_seconds": 10.080201625823975 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7227605537891388, + "lr": 0.03, + "negative_goodness": 1.0336865297820832, + "pair_accuracy": 0.8931111111111111, + "positive_goodness": 3.9597296447753907, + "runtime_seconds": 10.08073353767395 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7185703379127715, + "lr": 0.03, + "negative_goodness": 0.9736303386661741, + "pair_accuracy": 0.8793333333333333, + "positive_goodness": 3.9593011606004502, + "runtime_seconds": 10.078272819519043 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7177237571636835, + "lr": 0.03, + "negative_goodness": 1.0211690945439869, + "pair_accuracy": 0.8951333333333333, + "positive_goodness": 3.9728965526156954, + "runtime_seconds": 10.077175617218018 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6858191565831502, + "lr": 0.03, + "negative_goodness": 0.9592367655860053, + "pair_accuracy": 0.8968000000052981, + "positive_goodness": 4.07613529870775, + "runtime_seconds": 10.076665878295898 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7039333526823256, + "lr": 0.03, + "negative_goodness": 0.9892613338364495, + "pair_accuracy": 0.8942, + "positive_goodness": 3.987987514368693, + "runtime_seconds": 10.07770586013794 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.7187970116297404, + "lr": 0.03, + "negative_goodness": 1.0023443513234456, + "pair_accuracy": 0.8886666666666667, + "positive_goodness": 4.02138624382019, + "runtime_seconds": 10.07567572593689 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.692404626168145, + "lr": 0.03, + "negative_goodness": 0.9716952077441745, + "pair_accuracy": 0.8864888888888889, + "positive_goodness": 4.154300475565592, + "runtime_seconds": 10.080533266067505 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.7054903799798754, + "lr": 0.03, + "negative_goodness": 0.9783198663923476, + "pair_accuracy": 0.8942222222010294, + "positive_goodness": 4.043300562879774, + "runtime_seconds": 10.076720714569092 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.683643981552124, + "lr": 0.03, + "negative_goodness": 0.9706232560793558, + "pair_accuracy": 0.9000666666454739, + "positive_goodness": 4.14700607647366, + "runtime_seconds": 10.073118209838867 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.646786274072859, + "lr": 0.03, + "negative_goodness": 0.9041431274202135, + "pair_accuracy": 0.8979333333757189, + "positive_goodness": 4.27083523135715, + "runtime_seconds": 10.074635744094849 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6481765329122543, + "lr": 0.03, + "negative_goodness": 0.895178413216273, + "pair_accuracy": 0.9025777777777778, + "positive_goodness": 4.326637114122179, + "runtime_seconds": 10.07763671875 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6015416181908714, + "lr": 0.03, + "negative_goodness": 0.8386599930286407, + "pair_accuracy": 0.9182888888888889, + "positive_goodness": 4.518417228529189, + "runtime_seconds": 10.076727867126465 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.704261390127076, + "lr": 0.03, + "negative_goodness": 0.9791233874042828, + "pair_accuracy": 0.8847111111111111, + "positive_goodness": 4.122004435136583, + "runtime_seconds": 10.077731847763062 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6150173309750027, + "lr": 0.03, + "negative_goodness": 0.8682488751729329, + "pair_accuracy": 0.9103333333280351, + "positive_goodness": 4.395159215672811, + "runtime_seconds": 10.075066804885864 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6264795018513998, + "lr": 0.03, + "negative_goodness": 0.8583762802653843, + "pair_accuracy": 0.9093555555557211, + "positive_goodness": 4.459501206673516, + "runtime_seconds": 10.077122688293457 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.7300223679913415, + "lr": 0.03, + "negative_goodness": 0.9773748893472883, + "pair_accuracy": 0.8770888888888889, + "positive_goodness": 3.9280369705200195, + "runtime_seconds": 10.090771198272705 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.7108447946972317, + "lr": 0.03, + "negative_goodness": 0.9696070358700223, + "pair_accuracy": 0.8860444444338481, + "positive_goodness": 4.065411273744371, + "runtime_seconds": 10.106485843658447 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6517274273554484, + "lr": 0.03, + "negative_goodness": 0.8984224930975172, + "pair_accuracy": 0.9035999999682108, + "positive_goodness": 4.238310269546509, + "runtime_seconds": 10.085636615753174 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6240693633874257, + "lr": 0.03, + "negative_goodness": 0.8719130776405335, + "pair_accuracy": 0.9086666666666666, + "positive_goodness": 4.369689799033271, + "runtime_seconds": 10.082068920135498 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6332063141928779, + "lr": 0.03, + "negative_goodness": 0.8766779900868734, + "pair_accuracy": 0.9072444444656372, + "positive_goodness": 4.434643169932896, + "runtime_seconds": 10.073848247528076 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6396398240513271, + "lr": 0.03, + "negative_goodness": 0.8665754569583469, + "pair_accuracy": 0.8974222222010294, + "positive_goodness": 4.354673322762383, + "runtime_seconds": 10.073721647262573 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6428078939331903, + "lr": 0.03, + "negative_goodness": 0.8975953558074103, + "pair_accuracy": 0.9066, + "positive_goodness": 4.349167521158854, + "runtime_seconds": 10.07813310623169 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6470104046291775, + "lr": 0.03, + "negative_goodness": 0.8631999375237359, + "pair_accuracy": 0.9073333333545262, + "positive_goodness": 4.406757494693332, + "runtime_seconds": 10.078409194946289 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6695312079111735, + "lr": 0.03, + "negative_goodness": 0.8881729813257854, + "pair_accuracy": 0.9001333333333333, + "positive_goodness": 4.29093483505249, + "runtime_seconds": 10.075666666030884 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.641937988864051, + "lr": 0.03, + "negative_goodness": 0.8739150980419583, + "pair_accuracy": 0.9012444444338481, + "positive_goodness": 4.301036187405057, + "runtime_seconds": 10.084974765777588 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6356201938761605, + "lr": 0.03, + "negative_goodness": 0.8647303432981173, + "pair_accuracy": 0.9016222222222222, + "positive_goodness": 4.362098940658569, + "runtime_seconds": 10.074069738388062 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6373374134275648, + "lr": 0.03, + "negative_goodness": 0.8426532042821249, + "pair_accuracy": 0.9053555555873447, + "positive_goodness": 4.36538196404775, + "runtime_seconds": 10.066593647003174 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6880461799489127, + "lr": 0.03, + "negative_goodness": 0.9425799671093623, + "pair_accuracy": 0.8900666666666667, + "positive_goodness": 4.092791657087538, + "runtime_seconds": 9.95663046836853 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6070629447433684, + "lr": 0.03, + "negative_goodness": 0.7961986843956841, + "pair_accuracy": 0.9067333333333333, + "positive_goodness": 4.548892499118381, + "runtime_seconds": 9.952649116516113 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6576264483345879, + "lr": 0.03, + "negative_goodness": 0.8892843728383382, + "pair_accuracy": 0.8983111111005148, + "positive_goodness": 4.2401720470216535, + "runtime_seconds": 9.97258973121643 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6765905447006225, + "lr": 0.03, + "negative_goodness": 0.9072884622891744, + "pair_accuracy": 0.8926888889418708, + "positive_goodness": 4.219999381256104, + "runtime_seconds": 10.012591361999512 + } + ], + "layer": 37 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9603222092734442, + "lr": 0.03, + "negative_goodness": 1.3763688084496393, + "pair_accuracy": 0.8401333333333333, + "positive_goodness": 3.380863320414225, + "runtime_seconds": 10.14347219467163 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8186551518652174, + "lr": 0.03, + "negative_goodness": 1.189796886836158, + "pair_accuracy": 0.8730666666666667, + "positive_goodness": 3.7106136008368598, + "runtime_seconds": 10.051413774490356 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7765526193671757, + "lr": 0.03, + "negative_goodness": 1.1032818089273242, + "pair_accuracy": 0.8823555555555556, + "positive_goodness": 3.8231490734524196, + "runtime_seconds": 10.038787841796875 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.726358902668953, + "lr": 0.03, + "negative_goodness": 1.049431330400043, + "pair_accuracy": 0.8881777777777777, + "positive_goodness": 3.9083232632954914, + "runtime_seconds": 10.052400827407837 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7664889167785645, + "lr": 0.03, + "negative_goodness": 1.0765569445292156, + "pair_accuracy": 0.8766444444444444, + "positive_goodness": 3.8526254706912573, + "runtime_seconds": 10.04604434967041 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7264230482313369, + "lr": 0.03, + "negative_goodness": 1.0043345182842678, + "pair_accuracy": 0.8829333333598243, + "positive_goodness": 4.006991455205282, + "runtime_seconds": 10.049007654190063 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7324474462138282, + "lr": 0.03, + "negative_goodness": 1.0206763019773695, + "pair_accuracy": 0.8783333333333333, + "positive_goodness": 4.02994428346422, + "runtime_seconds": 10.077637672424316 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7740027501848009, + "lr": 0.03, + "negative_goodness": 1.0796982300228544, + "pair_accuracy": 0.8765777777883742, + "positive_goodness": 3.852670436032613, + "runtime_seconds": 10.048825979232788 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.684822843975491, + "lr": 0.03, + "negative_goodness": 0.9933329599062601, + "pair_accuracy": 0.9013777777459886, + "positive_goodness": 4.043669514550103, + "runtime_seconds": 10.04373049736023 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7079136656655206, + "lr": 0.03, + "negative_goodness": 0.9938655927658081, + "pair_accuracy": 0.8878000000052982, + "positive_goodness": 4.006203349664476, + "runtime_seconds": 10.044939517974854 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6901465763303969, + "lr": 0.03, + "negative_goodness": 0.9818168599658542, + "pair_accuracy": 0.8886888888888889, + "positive_goodness": 4.0917287497202555, + "runtime_seconds": 10.043960571289062 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.679492989593082, + "lr": 0.03, + "negative_goodness": 0.9498953950458102, + "pair_accuracy": 0.8956000000211928, + "positive_goodness": 4.14540946735806, + "runtime_seconds": 10.036707878112793 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7225157851325141, + "lr": 0.03, + "negative_goodness": 1.0210700191073947, + "pair_accuracy": 0.8863999999894037, + "positive_goodness": 3.9679475369135537, + "runtime_seconds": 10.036444664001465 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6548289209683736, + "lr": 0.03, + "negative_goodness": 0.9247718804253472, + "pair_accuracy": 0.8977999999894036, + "positive_goodness": 4.241158334181044, + "runtime_seconds": 10.043118476867676 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6639313312106663, + "lr": 0.03, + "negative_goodness": 0.9429128271950615, + "pair_accuracy": 0.9026888888782925, + "positive_goodness": 4.243620150015089, + "runtime_seconds": 10.04661250114441 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6523055692566766, + "lr": 0.03, + "negative_goodness": 0.9172833630137973, + "pair_accuracy": 0.9000666666560703, + "positive_goodness": 4.2721374098036025, + "runtime_seconds": 10.047792911529541 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6953094447029962, + "lr": 0.03, + "negative_goodness": 0.9502287826114231, + "pair_accuracy": 0.8923111111111112, + "positive_goodness": 4.172413816515605, + "runtime_seconds": 10.028233766555786 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6498709713511996, + "lr": 0.03, + "negative_goodness": 0.900939888938268, + "pair_accuracy": 0.8974666666666666, + "positive_goodness": 4.3250301063537595, + "runtime_seconds": 9.998133659362793 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.7011260701497396, + "lr": 0.03, + "negative_goodness": 0.9965636367797852, + "pair_accuracy": 0.889666666677263, + "positive_goodness": 4.058028882726034, + "runtime_seconds": 10.042960166931152 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6379665372159746, + "lr": 0.03, + "negative_goodness": 0.898592868844668, + "pair_accuracy": 0.9060888888888889, + "positive_goodness": 4.271121848551433, + "runtime_seconds": 10.04159140586853 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6813113847732544, + "lr": 0.03, + "negative_goodness": 0.9156777991082933, + "pair_accuracy": 0.8916666666666667, + "positive_goodness": 4.213101913324992, + "runtime_seconds": 10.03674602508545 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6890522903972202, + "lr": 0.03, + "negative_goodness": 0.9440437766393026, + "pair_accuracy": 0.8909555555449592, + "positive_goodness": 4.098720684602526, + "runtime_seconds": 10.04645299911499 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6659842575126224, + "lr": 0.03, + "negative_goodness": 0.916411987982856, + "pair_accuracy": 0.8995333333333333, + "positive_goodness": 4.159383911471897, + "runtime_seconds": 10.059457778930664 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6233117734379239, + "lr": 0.03, + "negative_goodness": 0.8855161915567186, + "pair_accuracy": 0.8984666666666666, + "positive_goodness": 4.372427256266276, + "runtime_seconds": 10.10793399810791 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6549536719852024, + "lr": 0.03, + "negative_goodness": 0.8968143885506524, + "pair_accuracy": 0.8942000000105964, + "positive_goodness": 4.325274636862013, + "runtime_seconds": 10.101380109786987 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6531773858944575, + "lr": 0.03, + "negative_goodness": 0.908243909561634, + "pair_accuracy": 0.9013111111111111, + "positive_goodness": 4.273779348670112, + "runtime_seconds": 10.096346616744995 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5795056484010485, + "lr": 0.03, + "negative_goodness": 0.8077118283801609, + "pair_accuracy": 0.9206444444444445, + "positive_goodness": 4.647581088002522, + "runtime_seconds": 10.097864627838135 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.645115561358134, + "lr": 0.03, + "negative_goodness": 0.8966607926474677, + "pair_accuracy": 0.8964888888782925, + "positive_goodness": 4.376681829537286, + "runtime_seconds": 10.09805965423584 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6478425573243035, + "lr": 0.03, + "negative_goodness": 0.9081459748797946, + "pair_accuracy": 0.9047111111111111, + "positive_goodness": 4.304067346360949, + "runtime_seconds": 10.099363803863525 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6333470203823514, + "lr": 0.03, + "negative_goodness": 0.8737166614373525, + "pair_accuracy": 0.9055111111111112, + "positive_goodness": 4.371160708363851, + "runtime_seconds": 10.105852842330933 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6169174257914225, + "lr": 0.03, + "negative_goodness": 0.8702309056811862, + "pair_accuracy": 0.9025777777777778, + "positive_goodness": 4.481205748078558, + "runtime_seconds": 10.100684404373169 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6039668794261085, + "lr": 0.03, + "negative_goodness": 0.8472762819502089, + "pair_accuracy": 0.9140888888888888, + "positive_goodness": 4.4040085904439294, + "runtime_seconds": 10.09962248802185 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.7012413679334852, + "lr": 0.03, + "negative_goodness": 0.9370677416059706, + "pair_accuracy": 0.8813333333333333, + "positive_goodness": 4.150819061851501, + "runtime_seconds": 10.097330093383789 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6109990913497078, + "lr": 0.03, + "negative_goodness": 0.8476930606630113, + "pair_accuracy": 0.9048, + "positive_goodness": 4.426769283548991, + "runtime_seconds": 10.101054430007935 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6286291126966477, + "lr": 0.03, + "negative_goodness": 0.8535242718021075, + "pair_accuracy": 0.9109777777777778, + "positive_goodness": 4.427066253323026, + "runtime_seconds": 10.099457025527954 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6319640471988254, + "lr": 0.03, + "negative_goodness": 0.8512851130697462, + "pair_accuracy": 0.9001777777459886, + "positive_goodness": 4.406204919603136, + "runtime_seconds": 10.103289127349854 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6122188353141149, + "lr": 0.03, + "negative_goodness": 0.8560016632265515, + "pair_accuracy": 0.9072888888888889, + "positive_goodness": 4.3973769316779245, + "runtime_seconds": 10.100402116775513 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6376579239739312, + "lr": 0.03, + "negative_goodness": 0.8405525158882141, + "pair_accuracy": 0.8957333333386315, + "positive_goodness": 4.425547610653771, + "runtime_seconds": 10.09932565689087 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6386647228346931, + "lr": 0.03, + "negative_goodness": 0.8890973055945502, + "pair_accuracy": 0.9009111111111111, + "positive_goodness": 4.315291705746121, + "runtime_seconds": 10.09532618522644 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6093856119791666, + "lr": 0.03, + "negative_goodness": 0.8533947391086154, + "pair_accuracy": 0.912444444460339, + "positive_goodness": 4.497864351654052, + "runtime_seconds": 10.09546422958374 + } + ], + "layer": 38 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9044709194819133, + "lr": 0.03, + "negative_goodness": 1.2142236764272054, + "pair_accuracy": 0.8399333333227369, + "positive_goodness": 3.263226277054681, + "runtime_seconds": 10.075417041778564 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7500944805781047, + "lr": 0.03, + "negative_goodness": 1.1147213100433349, + "pair_accuracy": 0.8832666666666666, + "positive_goodness": 3.779538465881348, + "runtime_seconds": 10.074279069900513 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7175201727337307, + "lr": 0.03, + "negative_goodness": 1.057327103614807, + "pair_accuracy": 0.8864222222222222, + "positive_goodness": 3.9051674144321016, + "runtime_seconds": 10.068827629089355 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7356099473529392, + "lr": 0.03, + "negative_goodness": 1.0773882938385009, + "pair_accuracy": 0.8791999999788073, + "positive_goodness": 3.9087134789360896, + "runtime_seconds": 10.074800491333008 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7139083543141683, + "lr": 0.03, + "negative_goodness": 1.0028735670513578, + "pair_accuracy": 0.8939111111112766, + "positive_goodness": 3.998214962217543, + "runtime_seconds": 10.068125247955322 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6531007410897149, + "lr": 0.03, + "negative_goodness": 0.9380608325322469, + "pair_accuracy": 0.9010888889100817, + "positive_goodness": 4.211117925262451, + "runtime_seconds": 10.064229249954224 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.5921370742056105, + "lr": 0.03, + "negative_goodness": 0.8659508483846983, + "pair_accuracy": 0.9195333333333333, + "positive_goodness": 4.529179920874702, + "runtime_seconds": 10.066184520721436 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6349477030860053, + "lr": 0.03, + "negative_goodness": 0.8860166179233128, + "pair_accuracy": 0.9026888888908757, + "positive_goodness": 4.426948488277859, + "runtime_seconds": 10.08021092414856 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6720575386497709, + "lr": 0.03, + "negative_goodness": 0.9205606120957268, + "pair_accuracy": 0.8906888888888889, + "positive_goodness": 4.191794294653999, + "runtime_seconds": 10.072965621948242 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6180054580476549, + "lr": 0.03, + "negative_goodness": 0.8503872280014886, + "pair_accuracy": 0.9050000000211927, + "positive_goodness": 4.442140350892808, + "runtime_seconds": 10.079155921936035 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.673424582473437, + "lr": 0.03, + "negative_goodness": 0.9333072367429733, + "pair_accuracy": 0.8975333333333333, + "positive_goodness": 4.253342711385091, + "runtime_seconds": 10.070779800415039 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6537574671109517, + "lr": 0.03, + "negative_goodness": 0.8937398676978218, + "pair_accuracy": 0.8957777777671814, + "positive_goodness": 4.2839088564131, + "runtime_seconds": 10.076182842254639 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6420846963670519, + "lr": 0.03, + "negative_goodness": 0.8919938735114203, + "pair_accuracy": 0.8991333333333333, + "positive_goodness": 4.301709136284722, + "runtime_seconds": 10.081150770187378 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6626575554953681, + "lr": 0.03, + "negative_goodness": 0.9246866110165914, + "pair_accuracy": 0.901866666677263, + "positive_goodness": 4.240515174272326, + "runtime_seconds": 10.081074476242065 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6216205331802368, + "lr": 0.03, + "negative_goodness": 0.8362076749483744, + "pair_accuracy": 0.9060666666454739, + "positive_goodness": 4.4526291638268365, + "runtime_seconds": 10.106997966766357 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6273270875506931, + "lr": 0.03, + "negative_goodness": 0.8512620506286621, + "pair_accuracy": 0.9058666666454739, + "positive_goodness": 4.473657467969258, + "runtime_seconds": 10.108585357666016 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6161921378029718, + "lr": 0.03, + "negative_goodness": 0.838517136669159, + "pair_accuracy": 0.9095111111005147, + "positive_goodness": 4.422744445673625, + "runtime_seconds": 10.080228090286255 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6179622064855363, + "lr": 0.03, + "negative_goodness": 0.8459719989299774, + "pair_accuracy": 0.9082222222222223, + "positive_goodness": 4.443491106075711, + "runtime_seconds": 10.082779169082642 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6154576017432742, + "lr": 0.03, + "negative_goodness": 0.8693399793836806, + "pair_accuracy": 0.9091333333333333, + "positive_goodness": 4.452842464955648, + "runtime_seconds": 10.04371190071106 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6293964669333564, + "lr": 0.03, + "negative_goodness": 0.8258719447241889, + "pair_accuracy": 0.9017111111217075, + "positive_goodness": 4.465951396963331, + "runtime_seconds": 10.01569128036499 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6387009578069051, + "lr": 0.03, + "negative_goodness": 0.8521813700252109, + "pair_accuracy": 0.8935555555555555, + "positive_goodness": 4.402424097569783, + "runtime_seconds": 10.01470685005188 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5901265096876356, + "lr": 0.03, + "negative_goodness": 0.8230535802629259, + "pair_accuracy": 0.913, + "positive_goodness": 4.587548886998494, + "runtime_seconds": 10.027304649353027 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6109180419021183, + "lr": 0.03, + "negative_goodness": 0.8479657842715581, + "pair_accuracy": 0.9101777777777778, + "positive_goodness": 4.505268793826633, + "runtime_seconds": 10.01566195487976 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5694042356067234, + "lr": 0.03, + "negative_goodness": 0.7766684695243835, + "pair_accuracy": 0.9135333333465788, + "positive_goodness": 4.729237869411045, + "runtime_seconds": 10.018244504928589 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5963392577992546, + "lr": 0.03, + "negative_goodness": 0.8076098109351264, + "pair_accuracy": 0.9136666666666666, + "positive_goodness": 4.543729345533583, + "runtime_seconds": 10.022573232650757 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5538205151822831, + "lr": 0.03, + "negative_goodness": 0.7715649987406201, + "pair_accuracy": 0.9234666666666667, + "positive_goodness": 4.7359119082980685, + "runtime_seconds": 10.021238327026367 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6437126606199477, + "lr": 0.03, + "negative_goodness": 0.8717399771690368, + "pair_accuracy": 0.9056222222222222, + "positive_goodness": 4.4195051621331105, + "runtime_seconds": 10.014024496078491 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5598264515770807, + "lr": 0.03, + "negative_goodness": 0.7859159027099609, + "pair_accuracy": 0.9222666666666667, + "positive_goodness": 4.6620259649488665, + "runtime_seconds": 10.028610706329346 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5725173719406128, + "lr": 0.03, + "negative_goodness": 0.7662812419997321, + "pair_accuracy": 0.9167333333598243, + "positive_goodness": 4.743723135587905, + "runtime_seconds": 10.007451057434082 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5641133482641644, + "lr": 0.03, + "negative_goodness": 0.7733078421354294, + "pair_accuracy": 0.9197555555555555, + "positive_goodness": 4.734938589986165, + "runtime_seconds": 10.016318798065186 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5257834778255887, + "lr": 0.03, + "negative_goodness": 0.709512573422326, + "pair_accuracy": 0.9336666666666666, + "positive_goodness": 5.0064747824774845, + "runtime_seconds": 10.011372566223145 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5979486792299482, + "lr": 0.03, + "negative_goodness": 0.8153031711366442, + "pair_accuracy": 0.9081333333333333, + "positive_goodness": 4.5843277150472, + "runtime_seconds": 10.009856700897217 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5627534814622667, + "lr": 0.03, + "negative_goodness": 0.7784830200705264, + "pair_accuracy": 0.9273555555555556, + "positive_goodness": 4.639844144270155, + "runtime_seconds": 10.008188247680664 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5481490387916564, + "lr": 0.03, + "negative_goodness": 0.7339183986081017, + "pair_accuracy": 0.9234666666560702, + "positive_goodness": 4.850538806194729, + "runtime_seconds": 10.010855197906494 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5377916672097313, + "lr": 0.03, + "negative_goodness": 0.7296721399837071, + "pair_accuracy": 0.9281555555555555, + "positive_goodness": 4.926957164679633, + "runtime_seconds": 10.007832765579224 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6018742252137926, + "lr": 0.03, + "negative_goodness": 0.7822212873246934, + "pair_accuracy": 0.9069555555502574, + "positive_goodness": 4.542578429582384, + "runtime_seconds": 10.0096755027771 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5796442643377516, + "lr": 0.03, + "negative_goodness": 0.7808812333742777, + "pair_accuracy": 0.9144222222222222, + "positive_goodness": 4.621278313106961, + "runtime_seconds": 9.99123740196228 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5635497565746307, + "lr": 0.03, + "negative_goodness": 0.737340135257774, + "pair_accuracy": 0.9225777777777778, + "positive_goodness": 4.776759235551622, + "runtime_seconds": 9.959260702133179 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5525763244628906, + "lr": 0.03, + "negative_goodness": 0.7300878989272648, + "pair_accuracy": 0.918422222216924, + "positive_goodness": 4.733958689414131, + "runtime_seconds": 10.015881061553955 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5871695651902092, + "lr": 0.03, + "negative_goodness": 0.7796169621043735, + "pair_accuracy": 0.9088666666666667, + "positive_goodness": 4.73535216700236, + "runtime_seconds": 10.072513580322266 + } + ], + "layer": 39 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.0027902308252123, + "lr": 0.03, + "negative_goodness": 1.5153584428787232, + "pair_accuracy": 0.8267999999894036, + "positive_goodness": 3.29705428153144, + "runtime_seconds": 10.090317010879517 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7738685701582166, + "lr": 0.03, + "negative_goodness": 1.2019021007961697, + "pair_accuracy": 0.8797333333545261, + "positive_goodness": 3.7911420841640897, + "runtime_seconds": 10.085747718811035 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.698990989361869, + "lr": 0.03, + "negative_goodness": 1.0704554426087274, + "pair_accuracy": 0.8896888888888889, + "positive_goodness": 4.095352726999919, + "runtime_seconds": 10.07345461845398 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6809368813832601, + "lr": 0.03, + "negative_goodness": 1.0340294476403131, + "pair_accuracy": 0.889333333322737, + "positive_goodness": 4.257007818476359, + "runtime_seconds": 9.976145029067993 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6729929353210661, + "lr": 0.03, + "negative_goodness": 1.0237215215947892, + "pair_accuracy": 0.9012888888888889, + "positive_goodness": 4.218907311672634, + "runtime_seconds": 9.976564407348633 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6808227183977763, + "lr": 0.03, + "negative_goodness": 0.9954688462787205, + "pair_accuracy": 0.8873111111111112, + "positive_goodness": 4.207010585191515, + "runtime_seconds": 10.016380310058594 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6913516261206732, + "lr": 0.03, + "negative_goodness": 1.0464060576544867, + "pair_accuracy": 0.8908222222222222, + "positive_goodness": 4.188519542418586, + "runtime_seconds": 10.06538438796997 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.767975661081738, + "lr": 0.03, + "negative_goodness": 1.1001295874595642, + "pair_accuracy": 0.8646666666666667, + "positive_goodness": 3.8551598964267306, + "runtime_seconds": 10.089754104614258 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.578876721599367, + "lr": 0.03, + "negative_goodness": 0.8944320587264167, + "pair_accuracy": 0.9261111111111111, + "positive_goodness": 4.575780179172092, + "runtime_seconds": 10.05726408958435 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6549000314076742, + "lr": 0.03, + "negative_goodness": 0.9600286405139499, + "pair_accuracy": 0.9068222222328186, + "positive_goodness": 4.394900127262539, + "runtime_seconds": 10.033554553985596 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6759350988944371, + "lr": 0.03, + "negative_goodness": 0.9620548123558362, + "pair_accuracy": 0.8880444444444444, + "positive_goodness": 4.306739748721653, + "runtime_seconds": 10.02891492843628 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6444714936150445, + "lr": 0.03, + "negative_goodness": 0.9109411119461059, + "pair_accuracy": 0.902777777756585, + "positive_goodness": 4.426352488242255, + "runtime_seconds": 10.041491746902466 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6779732542249891, + "lr": 0.03, + "negative_goodness": 0.9582045177035862, + "pair_accuracy": 0.8934444444974263, + "positive_goodness": 4.169119452200995, + "runtime_seconds": 10.036263227462769 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6152724397023519, + "lr": 0.03, + "negative_goodness": 0.91256888800727, + "pair_accuracy": 0.9102222222540114, + "positive_goodness": 4.584412598461575, + "runtime_seconds": 10.038360834121704 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.64356365714603, + "lr": 0.03, + "negative_goodness": 0.9066061575783624, + "pair_accuracy": 0.9028444444762336, + "positive_goodness": 4.4423869728512235, + "runtime_seconds": 10.029078006744385 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6253506800545586, + "lr": 0.03, + "negative_goodness": 0.9159292721642388, + "pair_accuracy": 0.9118888888570997, + "positive_goodness": 4.428233952755399, + "runtime_seconds": 10.034005403518677 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.598731938764784, + "lr": 0.03, + "negative_goodness": 0.8900439156214396, + "pair_accuracy": 0.9135111111111112, + "positive_goodness": 4.529793977271186, + "runtime_seconds": 10.032622337341309 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6315020548264185, + "lr": 0.03, + "negative_goodness": 0.9147757451746199, + "pair_accuracy": 0.9094444444444445, + "positive_goodness": 4.433125136057535, + "runtime_seconds": 10.033443212509155 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5877526648892296, + "lr": 0.03, + "negative_goodness": 0.8593690444946289, + "pair_accuracy": 0.9195333333333333, + "positive_goodness": 4.588804868231879, + "runtime_seconds": 10.038478374481201 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6038979951646593, + "lr": 0.03, + "negative_goodness": 0.8577457665549384, + "pair_accuracy": 0.908177777756585, + "positive_goodness": 4.606211064571805, + "runtime_seconds": 10.035969018936157 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6008145780245463, + "lr": 0.03, + "negative_goodness": 0.8582283403820462, + "pair_accuracy": 0.9145333333333333, + "positive_goodness": 4.553166742706299, + "runtime_seconds": 10.036993026733398 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5802692399395837, + "lr": 0.03, + "negative_goodness": 0.8229504777007632, + "pair_accuracy": 0.9144888888888889, + "positive_goodness": 4.723898253292508, + "runtime_seconds": 10.040794372558594 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.596472993320889, + "lr": 0.03, + "negative_goodness": 0.8744417613983154, + "pair_accuracy": 0.9187555555449591, + "positive_goodness": 4.605642348861695, + "runtime_seconds": 10.03654408454895 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5818802404986487, + "lr": 0.03, + "negative_goodness": 0.844024964984258, + "pair_accuracy": 0.9186444444444445, + "positive_goodness": 4.587659560733371, + "runtime_seconds": 10.037172555923462 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5852548425224092, + "lr": 0.03, + "negative_goodness": 0.834198618472947, + "pair_accuracy": 0.9223777777777777, + "positive_goodness": 4.6672416628519695, + "runtime_seconds": 10.038603782653809 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5491242509179645, + "lr": 0.03, + "negative_goodness": 0.789697386182679, + "pair_accuracy": 0.9240666666666667, + "positive_goodness": 4.8397904406229655, + "runtime_seconds": 10.047062158584595 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5837403767903646, + "lr": 0.03, + "negative_goodness": 0.8092917204168107, + "pair_accuracy": 0.9174444444444444, + "positive_goodness": 4.675492954169379, + "runtime_seconds": 10.031696796417236 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6679705109702216, + "lr": 0.03, + "negative_goodness": 0.8871947802649603, + "pair_accuracy": 0.8952222222381168, + "positive_goodness": 4.408279705047607, + "runtime_seconds": 10.033767700195312 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5696497958289253, + "lr": 0.03, + "negative_goodness": 0.8114976280000474, + "pair_accuracy": 0.9177555555582046, + "positive_goodness": 4.792171216011047, + "runtime_seconds": 10.03218674659729 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5674437318060134, + "lr": 0.03, + "negative_goodness": 0.8032631411658393, + "pair_accuracy": 0.9186000000105964, + "positive_goodness": 4.760330782657199, + "runtime_seconds": 10.003559589385986 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5667968853871027, + "lr": 0.03, + "negative_goodness": 0.8226558421134949, + "pair_accuracy": 0.9253333333333333, + "positive_goodness": 4.651089817555746, + "runtime_seconds": 9.979890823364258 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6049371695730421, + "lr": 0.03, + "negative_goodness": 0.8505319831106398, + "pair_accuracy": 0.9049111111429002, + "positive_goodness": 4.641781528303358, + "runtime_seconds": 9.978708744049072 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5897386172533036, + "lr": 0.03, + "negative_goodness": 0.778912195136812, + "pair_accuracy": 0.9136, + "positive_goodness": 4.670616476864285, + "runtime_seconds": 9.979605436325073 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.522963986963696, + "lr": 0.03, + "negative_goodness": 0.7559430969450209, + "pair_accuracy": 0.9336444444444445, + "positive_goodness": 4.956953589630127, + "runtime_seconds": 9.979033946990967 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.569443716041247, + "lr": 0.03, + "negative_goodness": 0.7820498005125258, + "pair_accuracy": 0.9172666666666667, + "positive_goodness": 4.740655473921034, + "runtime_seconds": 9.975197076797485 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5548781802813212, + "lr": 0.03, + "negative_goodness": 0.7988120058748457, + "pair_accuracy": 0.9293333333333333, + "positive_goodness": 4.767505611928304, + "runtime_seconds": 9.976511716842651 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6126131960895327, + "lr": 0.03, + "negative_goodness": 0.8368411929660373, + "pair_accuracy": 0.9090222222222222, + "positive_goodness": 4.63910174975925, + "runtime_seconds": 9.977298259735107 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5653397281222873, + "lr": 0.03, + "negative_goodness": 0.7941440729326672, + "pair_accuracy": 0.9250666666666667, + "positive_goodness": 4.800498033057319, + "runtime_seconds": 9.98648738861084 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5515986431545682, + "lr": 0.03, + "negative_goodness": 0.76908803358078, + "pair_accuracy": 0.9202222222328186, + "positive_goodness": 4.845120483610365, + "runtime_seconds": 9.97733759880066 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5532063944710626, + "lr": 0.03, + "negative_goodness": 0.7416361449347602, + "pair_accuracy": 0.9177111111429003, + "positive_goodness": 4.903250675498115, + "runtime_seconds": 9.97734260559082 + } + ], + "layer": 40 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8410653310563829, + "lr": 0.03, + "negative_goodness": 1.1795543827056885, + "pair_accuracy": 0.8598444444444444, + "positive_goodness": 3.5608075041453042, + "runtime_seconds": 9.956555128097534 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7161834541426765, + "lr": 0.03, + "negative_goodness": 1.0943142999119229, + "pair_accuracy": 0.8975777777671814, + "positive_goodness": 3.968211346435547, + "runtime_seconds": 9.959474086761475 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7173308681276109, + "lr": 0.03, + "negative_goodness": 1.0747483500586617, + "pair_accuracy": 0.8885777778201633, + "positive_goodness": 3.9395175884246827, + "runtime_seconds": 9.948861837387085 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6683649241712358, + "lr": 0.03, + "negative_goodness": 1.0184696601337857, + "pair_accuracy": 0.8982888888782925, + "positive_goodness": 4.145121842447916, + "runtime_seconds": 9.959560871124268 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6530723746379217, + "lr": 0.03, + "negative_goodness": 0.9814650938140022, + "pair_accuracy": 0.9019777777777778, + "positive_goodness": 4.361980755021837, + "runtime_seconds": 9.95118236541748 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6415567388110691, + "lr": 0.03, + "negative_goodness": 0.9484892399046156, + "pair_accuracy": 0.9028222222116259, + "positive_goodness": 4.394668313641018, + "runtime_seconds": 9.952096462249756 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6131181784523858, + "lr": 0.03, + "negative_goodness": 0.9277969763014051, + "pair_accuracy": 0.9118, + "positive_goodness": 4.482808672163221, + "runtime_seconds": 9.953466892242432 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6317904665152232, + "lr": 0.03, + "negative_goodness": 0.9186702896542019, + "pair_accuracy": 0.9000222222222222, + "positive_goodness": 4.428046324920654, + "runtime_seconds": 9.965519666671753 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6125078812864092, + "lr": 0.03, + "negative_goodness": 0.8989775419553121, + "pair_accuracy": 0.9088222222222222, + "positive_goodness": 4.56401384379069, + "runtime_seconds": 9.950330018997192 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6570642395655314, + "lr": 0.03, + "negative_goodness": 0.9314143106460572, + "pair_accuracy": 0.902044444433848, + "positive_goodness": 4.4486942111969, + "runtime_seconds": 9.991445064544678 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.581409693230523, + "lr": 0.03, + "negative_goodness": 0.8743475409931607, + "pair_accuracy": 0.9179333333333334, + "positive_goodness": 4.663845074399313, + "runtime_seconds": 10.013336181640625 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.5654696006324557, + "lr": 0.03, + "negative_goodness": 0.8517795768419901, + "pair_accuracy": 0.9297111111111112, + "positive_goodness": 4.7337277431911895, + "runtime_seconds": 10.00649905204773 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5405684370332294, + "lr": 0.03, + "negative_goodness": 0.8017586196899414, + "pair_accuracy": 0.9317333333333333, + "positive_goodness": 4.858577959696452, + "runtime_seconds": 10.005365133285522 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6625536021974352, + "lr": 0.03, + "negative_goodness": 0.9032483280287849, + "pair_accuracy": 0.8957111111217075, + "positive_goodness": 4.398903069559733, + "runtime_seconds": 10.012637376785278 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5877065873172548, + "lr": 0.03, + "negative_goodness": 0.8510560491932763, + "pair_accuracy": 0.9154222222222222, + "positive_goodness": 4.650851553853353, + "runtime_seconds": 10.01427698135376 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6012389905293782, + "lr": 0.03, + "negative_goodness": 0.8804303578694661, + "pair_accuracy": 0.9222000000211927, + "positive_goodness": 4.582653355238173, + "runtime_seconds": 10.013288497924805 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6214829784287347, + "lr": 0.03, + "negative_goodness": 0.8693826074573728, + "pair_accuracy": 0.9053999999894036, + "positive_goodness": 4.552217796240913, + "runtime_seconds": 10.003208875656128 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5808002692275577, + "lr": 0.03, + "negative_goodness": 0.8532018755061759, + "pair_accuracy": 0.9252, + "positive_goodness": 4.631188228437636, + "runtime_seconds": 10.013558626174927 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5182080402268304, + "lr": 0.03, + "negative_goodness": 0.7532157998349932, + "pair_accuracy": 0.9357777777671814, + "positive_goodness": 5.070812141757541, + "runtime_seconds": 10.015659809112549 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.57666421684689, + "lr": 0.03, + "negative_goodness": 0.7954330261177487, + "pair_accuracy": 0.9156444444444445, + "positive_goodness": 4.793293437957764, + "runtime_seconds": 10.00693130493164 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6426753140767415, + "lr": 0.03, + "negative_goodness": 0.8610120055675506, + "pair_accuracy": 0.8975111111005147, + "positive_goodness": 4.536409318330553, + "runtime_seconds": 10.004117727279663 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6266093982272678, + "lr": 0.03, + "negative_goodness": 0.8823354832490286, + "pair_accuracy": 0.9010222222752041, + "positive_goodness": 4.419415385860867, + "runtime_seconds": 10.020032167434692 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6209464531368679, + "lr": 0.03, + "negative_goodness": 0.8715314350446065, + "pair_accuracy": 0.9125333333545261, + "positive_goodness": 4.479748315472073, + "runtime_seconds": 10.012541055679321 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6057916189405653, + "lr": 0.03, + "negative_goodness": 0.8567101050747765, + "pair_accuracy": 0.9115999999947018, + "positive_goodness": 4.544769990497166, + "runtime_seconds": 10.010693311691284 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5784205848058065, + "lr": 0.03, + "negative_goodness": 0.8356187997182211, + "pair_accuracy": 0.9188888888888889, + "positive_goodness": 4.7406698468526205, + "runtime_seconds": 10.00938630104065 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5592046421872245, + "lr": 0.03, + "negative_goodness": 0.7860097009923723, + "pair_accuracy": 0.9248, + "positive_goodness": 4.824309388478597, + "runtime_seconds": 10.01241683959961 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5569418246693081, + "lr": 0.03, + "negative_goodness": 0.7999911514282226, + "pair_accuracy": 0.9227333333863152, + "positive_goodness": 4.78462001347012, + "runtime_seconds": 10.011291980743408 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.4903130758073595, + "lr": 0.03, + "negative_goodness": 0.6959752661493089, + "pair_accuracy": 0.9355555555502574, + "positive_goodness": 5.26946835539076, + "runtime_seconds": 10.020337104797363 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5081647860633003, + "lr": 0.03, + "negative_goodness": 0.717801903072993, + "pair_accuracy": 0.9390000000105964, + "positive_goodness": 5.182534972084893, + "runtime_seconds": 10.011043071746826 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5289079062223434, + "lr": 0.03, + "negative_goodness": 0.7335750140501394, + "pair_accuracy": 0.9277111111111112, + "positive_goodness": 5.094796359507243, + "runtime_seconds": 10.019993543624878 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5715150426228841, + "lr": 0.03, + "negative_goodness": 0.7814284144904878, + "pair_accuracy": 0.907266666645474, + "positive_goodness": 4.9163242045508495, + "runtime_seconds": 10.015094518661499 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5254661731084188, + "lr": 0.03, + "negative_goodness": 0.6967168069879214, + "pair_accuracy": 0.9305555555767483, + "positive_goodness": 5.107960652605692, + "runtime_seconds": 10.018385887145996 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5610442151970334, + "lr": 0.03, + "negative_goodness": 0.7697857877810796, + "pair_accuracy": 0.9205555555555556, + "positive_goodness": 4.897938238440619, + "runtime_seconds": 10.008886814117432 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5436553597344292, + "lr": 0.03, + "negative_goodness": 0.7372141180621253, + "pair_accuracy": 0.9205555555558866, + "positive_goodness": 4.957036865848965, + "runtime_seconds": 10.014634370803833 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5247894151793586, + "lr": 0.03, + "negative_goodness": 0.7005205404890908, + "pair_accuracy": 0.9194444444444444, + "positive_goodness": 4.994695706770155, + "runtime_seconds": 9.992326974868774 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5909891812139088, + "lr": 0.03, + "negative_goodness": 0.7917696531467967, + "pair_accuracy": 0.9231111111111111, + "positive_goodness": 4.654295126914978, + "runtime_seconds": 10.015367031097412 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.4934218196153641, + "lr": 0.03, + "negative_goodness": 0.709395600605011, + "pair_accuracy": 0.9430666666666667, + "positive_goodness": 5.137511679840088, + "runtime_seconds": 10.007577657699585 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5280820315043131, + "lr": 0.03, + "negative_goodness": 0.7224501370218065, + "pair_accuracy": 0.9350444444444445, + "positive_goodness": 5.069763858003086, + "runtime_seconds": 10.013018608093262 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5271580532100466, + "lr": 0.03, + "negative_goodness": 0.6954125517828597, + "pair_accuracy": 0.9325555555555556, + "positive_goodness": 4.954243400149875, + "runtime_seconds": 10.00736951828003 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5793266274134318, + "lr": 0.03, + "negative_goodness": 0.7774267202112409, + "pair_accuracy": 0.9131333333227369, + "positive_goodness": 4.770529960208469, + "runtime_seconds": 10.017265796661377 + } + ], + "layer": 41 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.887427431148953, + "lr": 0.03, + "negative_goodness": 1.44226821269989, + "pair_accuracy": 0.8746888888888888, + "positive_goodness": 3.684422670279609, + "runtime_seconds": 10.035711288452148 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7198868732505375, + "lr": 0.03, + "negative_goodness": 1.1592497011608547, + "pair_accuracy": 0.9015111111111112, + "positive_goodness": 4.157158517795139, + "runtime_seconds": 10.040237426757812 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.6887618357552422, + "lr": 0.03, + "negative_goodness": 1.090113634109497, + "pair_accuracy": 0.9079777777989705, + "positive_goodness": 4.32568443514506, + "runtime_seconds": 10.030649900436401 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.6281640774197048, + "lr": 0.03, + "negative_goodness": 0.9982983204099867, + "pair_accuracy": 0.9254666666772631, + "positive_goodness": 4.429920133421156, + "runtime_seconds": 10.028395414352417 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7100401199446784, + "lr": 0.03, + "negative_goodness": 1.0622451689084371, + "pair_accuracy": 0.8925777777777778, + "positive_goodness": 4.128354220665826, + "runtime_seconds": 10.032638788223267 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6126116057501899, + "lr": 0.03, + "negative_goodness": 0.9584788209067451, + "pair_accuracy": 0.9204666667196486, + "positive_goodness": 4.52043928811815, + "runtime_seconds": 10.036487340927124 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.636044217300415, + "lr": 0.03, + "negative_goodness": 0.9692021562364367, + "pair_accuracy": 0.913, + "positive_goodness": 4.421305053244697, + "runtime_seconds": 10.035073518753052 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6361694331910875, + "lr": 0.03, + "negative_goodness": 0.9736818713082207, + "pair_accuracy": 0.9195333333227369, + "positive_goodness": 4.39525979669359, + "runtime_seconds": 10.034910917282104 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6201076570828755, + "lr": 0.03, + "negative_goodness": 0.9090611814459165, + "pair_accuracy": 0.9117333333333333, + "positive_goodness": 4.482860867733425, + "runtime_seconds": 10.033456087112427 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.5989608407974243, + "lr": 0.03, + "negative_goodness": 0.888242478524314, + "pair_accuracy": 0.9188, + "positive_goodness": 4.590633271959093, + "runtime_seconds": 10.032320976257324 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6170380057838228, + "lr": 0.03, + "negative_goodness": 0.9110174528386857, + "pair_accuracy": 0.9170888888888888, + "positive_goodness": 4.493333211686876, + "runtime_seconds": 10.029995918273926 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6055446593178643, + "lr": 0.03, + "negative_goodness": 0.9146360765033298, + "pair_accuracy": 0.9168222222381168, + "positive_goodness": 4.4985945256127255, + "runtime_seconds": 10.03518295288086 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6120567949295044, + "lr": 0.03, + "negative_goodness": 0.9123534190495809, + "pair_accuracy": 0.9192000000423856, + "positive_goodness": 4.509400270122952, + "runtime_seconds": 10.037546634674072 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5998949293136596, + "lr": 0.03, + "negative_goodness": 0.862089235581292, + "pair_accuracy": 0.9202000000211927, + "positive_goodness": 4.658831291876899, + "runtime_seconds": 10.032188177108765 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6371172179698944, + "lr": 0.03, + "negative_goodness": 0.9202970922735002, + "pair_accuracy": 0.9048222222222222, + "positive_goodness": 4.444242334577773, + "runtime_seconds": 10.03398084640503 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6151183974372016, + "lr": 0.03, + "negative_goodness": 0.894359079890781, + "pair_accuracy": 0.9164222222228845, + "positive_goodness": 4.544664575513204, + "runtime_seconds": 10.046368598937988 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6033443718406889, + "lr": 0.03, + "negative_goodness": 0.8496738063971202, + "pair_accuracy": 0.9135333333333333, + "positive_goodness": 4.609004012976753, + "runtime_seconds": 10.032380104064941 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5949251663578882, + "lr": 0.03, + "negative_goodness": 0.857126892346806, + "pair_accuracy": 0.9154666666666667, + "positive_goodness": 4.610677971310086, + "runtime_seconds": 10.045564889907837 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5337196722348531, + "lr": 0.03, + "negative_goodness": 0.7789395709143745, + "pair_accuracy": 0.9327777777671814, + "positive_goodness": 4.8381458056132, + "runtime_seconds": 9.999112844467163 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6118525693496069, + "lr": 0.03, + "negative_goodness": 0.8499740318086412, + "pair_accuracy": 0.9081777777777778, + "positive_goodness": 4.614898559485542, + "runtime_seconds": 9.964803457260132 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.564508903143141, + "lr": 0.03, + "negative_goodness": 0.8319906045807732, + "pair_accuracy": 0.9261777777989706, + "positive_goodness": 4.754343665356107, + "runtime_seconds": 10.040196180343628 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5302245810694165, + "lr": 0.03, + "negative_goodness": 0.7677643251286612, + "pair_accuracy": 0.9367333333333333, + "positive_goodness": 4.908162389967177, + "runtime_seconds": 10.10184383392334 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5871393154144288, + "lr": 0.03, + "negative_goodness": 0.7959679781172011, + "pair_accuracy": 0.9143333333227369, + "positive_goodness": 4.729840787845188, + "runtime_seconds": 10.098759412765503 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.48883789384100174, + "lr": 0.03, + "negative_goodness": 0.6983352086173163, + "pair_accuracy": 0.9472888888888888, + "positive_goodness": 5.0975899211459685, + "runtime_seconds": 10.100484609603882 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5462885640038384, + "lr": 0.03, + "negative_goodness": 0.7785916179127164, + "pair_accuracy": 0.9277555555555556, + "positive_goodness": 4.848670811780294, + "runtime_seconds": 10.105517148971558 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5169623890982734, + "lr": 0.03, + "negative_goodness": 0.7476172382990519, + "pair_accuracy": 0.9391111111005147, + "positive_goodness": 4.926960428110759, + "runtime_seconds": 10.095045804977417 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.51503655834198, + "lr": 0.03, + "negative_goodness": 0.704371912961536, + "pair_accuracy": 0.9391777777777778, + "positive_goodness": 5.126395542907715, + "runtime_seconds": 10.091853380203247 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5332332018242942, + "lr": 0.03, + "negative_goodness": 0.7308871531301074, + "pair_accuracy": 0.9266222222222222, + "positive_goodness": 5.0522072494294905, + "runtime_seconds": 10.104447364807129 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5464543623606364, + "lr": 0.03, + "negative_goodness": 0.7525234091758728, + "pair_accuracy": 0.9230666666668322, + "positive_goodness": 4.868062211449941, + "runtime_seconds": 10.093045473098755 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.607209702067905, + "lr": 0.03, + "negative_goodness": 0.8159460751745435, + "pair_accuracy": 0.9134222222328185, + "positive_goodness": 4.590862219662137, + "runtime_seconds": 10.098877429962158 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5025972669442494, + "lr": 0.03, + "negative_goodness": 0.7190810737291972, + "pair_accuracy": 0.9392444444444444, + "positive_goodness": 5.0960326885647245, + "runtime_seconds": 10.105468988418579 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.55843774878184, + "lr": 0.03, + "negative_goodness": 0.7848195170826382, + "pair_accuracy": 0.9214444444126553, + "positive_goodness": 4.868257752143013, + "runtime_seconds": 10.101815462112427 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5830221158769395, + "lr": 0.03, + "negative_goodness": 0.7907123623212179, + "pair_accuracy": 0.9182444444444444, + "positive_goodness": 4.726389292441474, + "runtime_seconds": 10.091635942459106 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5618393383873833, + "lr": 0.03, + "negative_goodness": 0.7933928614616395, + "pair_accuracy": 0.9281333333333334, + "positive_goodness": 4.8487280723571775, + "runtime_seconds": 10.09752631187439 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5464466075923707, + "lr": 0.03, + "negative_goodness": 0.7444698321938514, + "pair_accuracy": 0.9225111111111111, + "positive_goodness": 4.904703261841668, + "runtime_seconds": 10.09352731704712 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5270763546917173, + "lr": 0.03, + "negative_goodness": 0.7114512509822846, + "pair_accuracy": 0.9279333333333334, + "positive_goodness": 5.078466355726454, + "runtime_seconds": 10.047986507415771 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5294582007937961, + "lr": 0.03, + "negative_goodness": 0.7308565706570943, + "pair_accuracy": 0.9304888888888889, + "positive_goodness": 5.0052073592291935, + "runtime_seconds": 10.094313859939575 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5385455329312219, + "lr": 0.03, + "negative_goodness": 0.7294101170063019, + "pair_accuracy": 0.9225333333333333, + "positive_goodness": 4.958216426086426, + "runtime_seconds": 10.111279010772705 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.548955748377906, + "lr": 0.03, + "negative_goodness": 0.7590114922205607, + "pair_accuracy": 0.9352666666560703, + "positive_goodness": 4.849333688269721, + "runtime_seconds": 10.095786333084106 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.4998128714720408, + "lr": 0.03, + "negative_goodness": 0.6560110652128855, + "pair_accuracy": 0.9348444444444445, + "positive_goodness": 5.181431488545735, + "runtime_seconds": 10.21178126335144 + } + ], + "layer": 42 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8179889550659392, + "lr": 0.03, + "negative_goodness": 1.1462313527637058, + "pair_accuracy": 0.8672444444444445, + "positive_goodness": 3.647605067867703, + "runtime_seconds": 10.078670740127563 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7888289489322239, + "lr": 0.03, + "negative_goodness": 1.17002388159434, + "pair_accuracy": 0.890355555566152, + "positive_goodness": 3.753625008837382, + "runtime_seconds": 10.080565214157104 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7311965201907688, + "lr": 0.03, + "negative_goodness": 1.0644949744330512, + "pair_accuracy": 0.88844444448683, + "positive_goodness": 4.009015218904283, + "runtime_seconds": 10.080774784088135 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7094172963884142, + "lr": 0.03, + "negative_goodness": 1.0358666591856214, + "pair_accuracy": 0.8926222222222222, + "positive_goodness": 4.181194626935323, + "runtime_seconds": 10.090564250946045 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7163241175916459, + "lr": 0.03, + "negative_goodness": 1.0420260126007928, + "pair_accuracy": 0.8942444444444444, + "positive_goodness": 4.0782277452256945, + "runtime_seconds": 10.079569578170776 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6850003015730116, + "lr": 0.03, + "negative_goodness": 0.9829570609940423, + "pair_accuracy": 0.8998444444444444, + "positive_goodness": 4.150488767009311, + "runtime_seconds": 10.097111225128174 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6631348495483398, + "lr": 0.03, + "negative_goodness": 0.9855849769062466, + "pair_accuracy": 0.9106888888888889, + "positive_goodness": 4.24558134197659, + "runtime_seconds": 10.108320474624634 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7133040745205349, + "lr": 0.03, + "negative_goodness": 1.042136016262902, + "pair_accuracy": 0.8941333333386315, + "positive_goodness": 4.0741916755888195, + "runtime_seconds": 10.079923868179321 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6198467900011274, + "lr": 0.03, + "negative_goodness": 0.9451769719282787, + "pair_accuracy": 0.9271555555555555, + "positive_goodness": 4.39307469168769, + "runtime_seconds": 10.057103157043457 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7051939112451341, + "lr": 0.03, + "negative_goodness": 0.9893922040979067, + "pair_accuracy": 0.8908000000423856, + "positive_goodness": 4.216238576846653, + "runtime_seconds": 10.046756744384766 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6125929774390326, + "lr": 0.03, + "negative_goodness": 0.926178847834799, + "pair_accuracy": 0.9251777777459886, + "positive_goodness": 4.487662139765422, + "runtime_seconds": 10.02883529663086 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6108690706570943, + "lr": 0.03, + "negative_goodness": 0.9003370409184032, + "pair_accuracy": 0.9194444444232517, + "positive_goodness": 4.515347172546386, + "runtime_seconds": 10.027448177337646 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6249602450900608, + "lr": 0.03, + "negative_goodness": 0.9054474636210336, + "pair_accuracy": 0.9129111111058129, + "positive_goodness": 4.415955683030022, + "runtime_seconds": 10.023277044296265 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.674576816897922, + "lr": 0.03, + "negative_goodness": 0.9572825779199601, + "pair_accuracy": 0.8995777777777778, + "positive_goodness": 4.211818576526642, + "runtime_seconds": 10.019367456436157 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5770266006893582, + "lr": 0.03, + "negative_goodness": 0.8496399642070135, + "pair_accuracy": 0.9309111111111111, + "positive_goodness": 4.685720014063517, + "runtime_seconds": 10.039445638656616 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6194497462060716, + "lr": 0.03, + "negative_goodness": 0.8725013122240702, + "pair_accuracy": 0.909, + "positive_goodness": 4.4603203575981984, + "runtime_seconds": 10.019356727600098 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6122360109594133, + "lr": 0.03, + "negative_goodness": 0.8700629636102253, + "pair_accuracy": 0.9115555555555556, + "positive_goodness": 4.487841257900662, + "runtime_seconds": 10.021483421325684 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5529936537000868, + "lr": 0.03, + "negative_goodness": 0.8050271710740196, + "pair_accuracy": 0.9335555555661519, + "positive_goodness": 4.795033293787639, + "runtime_seconds": 10.0138840675354 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6195982169045342, + "lr": 0.03, + "negative_goodness": 0.8818307485063871, + "pair_accuracy": 0.9186444444974263, + "positive_goodness": 4.502899579577976, + "runtime_seconds": 10.00690245628357 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6485145110607147, + "lr": 0.03, + "negative_goodness": 0.9246666772815916, + "pair_accuracy": 0.9008222222222222, + "positive_goodness": 4.318816244930691, + "runtime_seconds": 10.01583194732666 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.619359090730879, + "lr": 0.03, + "negative_goodness": 0.8712964790840944, + "pair_accuracy": 0.9093555555555556, + "positive_goodness": 4.436643394809299, + "runtime_seconds": 10.003798961639404 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5833159824212392, + "lr": 0.03, + "negative_goodness": 0.8382012695524428, + "pair_accuracy": 0.9244222222222223, + "positive_goodness": 4.619830987124973, + "runtime_seconds": 10.016803979873657 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5784212253517574, + "lr": 0.03, + "negative_goodness": 0.8240616565002336, + "pair_accuracy": 0.9239555555555555, + "positive_goodness": 4.697486588880751, + "runtime_seconds": 9.988183498382568 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5901523152298398, + "lr": 0.03, + "negative_goodness": 0.8219476839971211, + "pair_accuracy": 0.9197111111111111, + "positive_goodness": 4.673067325592041, + "runtime_seconds": 10.024077892303467 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5398997092247009, + "lr": 0.03, + "negative_goodness": 0.7798298650821051, + "pair_accuracy": 0.9372444444232517, + "positive_goodness": 4.904782491811116, + "runtime_seconds": 10.014393091201782 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6071134946611193, + "lr": 0.03, + "negative_goodness": 0.8148499997854233, + "pair_accuracy": 0.9126444444444445, + "positive_goodness": 4.5933036802079945, + "runtime_seconds": 10.004595041275024 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5953148916244507, + "lr": 0.03, + "negative_goodness": 0.854289832828525, + "pair_accuracy": 0.9193777777777777, + "positive_goodness": 4.58754751264784, + "runtime_seconds": 10.040163516998291 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5548017180548774, + "lr": 0.03, + "negative_goodness": 0.7616561644236247, + "pair_accuracy": 0.9268888888888889, + "positive_goodness": 4.85307009540134, + "runtime_seconds": 10.007947206497192 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5529823384231991, + "lr": 0.03, + "negative_goodness": 0.7937288137111399, + "pair_accuracy": 0.9308888888888889, + "positive_goodness": 4.822380030653211, + "runtime_seconds": 10.013336420059204 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5409012065039741, + "lr": 0.03, + "negative_goodness": 0.7554583052370283, + "pair_accuracy": 0.926755555566152, + "positive_goodness": 4.852115006510417, + "runtime_seconds": 10.027529954910278 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.48813076962365043, + "lr": 0.03, + "negative_goodness": 0.6914338598661952, + "pair_accuracy": 0.9435111111323039, + "positive_goodness": 5.31680590209961, + "runtime_seconds": 10.062520027160645 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5636477982415093, + "lr": 0.03, + "negative_goodness": 0.7639542230857743, + "pair_accuracy": 0.9230666667090522, + "positive_goodness": 4.900458595403036, + "runtime_seconds": 10.079661130905151 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5240934502336714, + "lr": 0.03, + "negative_goodness": 0.7632961557249228, + "pair_accuracy": 0.9348222222222222, + "positive_goodness": 4.944244030337863, + "runtime_seconds": 10.048044204711914 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5664046451409658, + "lr": 0.03, + "negative_goodness": 0.7694805906097094, + "pair_accuracy": 0.9256444444444445, + "positive_goodness": 4.774511598587036, + "runtime_seconds": 9.968890905380249 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5958663291083441, + "lr": 0.03, + "negative_goodness": 0.8000580290900337, + "pair_accuracy": 0.9116444444444445, + "positive_goodness": 4.667495197974311, + "runtime_seconds": 9.946710348129272 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6243872205522325, + "lr": 0.03, + "negative_goodness": 0.8130516233576669, + "pair_accuracy": 0.9001111111111111, + "positive_goodness": 4.599979700915019, + "runtime_seconds": 10.062336921691895 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5686484395345052, + "lr": 0.03, + "negative_goodness": 0.8040521120170753, + "pair_accuracy": 0.9209111111217075, + "positive_goodness": 4.830848184288873, + "runtime_seconds": 10.077436447143555 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5611477889855703, + "lr": 0.03, + "negative_goodness": 0.7703935966081089, + "pair_accuracy": 0.9200666666666667, + "positive_goodness": 4.7988058352152505, + "runtime_seconds": 10.057904243469238 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.47971095938682556, + "lr": 0.03, + "negative_goodness": 0.6959808581087324, + "pair_accuracy": 0.9444222222222223, + "positive_goodness": 5.237649252743191, + "runtime_seconds": 10.088381052017212 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5063298049237993, + "lr": 0.03, + "negative_goodness": 0.6921519322143661, + "pair_accuracy": 0.9306444444444445, + "positive_goodness": 5.303913725280761, + "runtime_seconds": 10.06984281539917 + } + ], + "layer": 43 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9346767871432834, + "lr": 0.03, + "negative_goodness": 1.4683877006106907, + "pair_accuracy": 0.8670888889418708, + "positive_goodness": 3.6196975021362303, + "runtime_seconds": 10.091290712356567 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7984214011298285, + "lr": 0.03, + "negative_goodness": 1.2692857558780246, + "pair_accuracy": 0.8813555555555556, + "positive_goodness": 3.8025891786787245, + "runtime_seconds": 10.099543809890747 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8064818390051524, + "lr": 0.03, + "negative_goodness": 1.2206233721945021, + "pair_accuracy": 0.8678444444338481, + "positive_goodness": 3.7279663624233668, + "runtime_seconds": 10.080089807510376 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.718959611553616, + "lr": 0.03, + "negative_goodness": 1.1247085643132528, + "pair_accuracy": 0.8929777777777778, + "positive_goodness": 4.019505131191678, + "runtime_seconds": 10.101378202438354 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7011025118721856, + "lr": 0.03, + "negative_goodness": 1.1038812456342908, + "pair_accuracy": 0.8982444444974264, + "positive_goodness": 4.078388762495253, + "runtime_seconds": 10.084319829940796 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6952027022043864, + "lr": 0.03, + "negative_goodness": 1.0906676138983833, + "pair_accuracy": 0.9033333333651224, + "positive_goodness": 4.102948246553209, + "runtime_seconds": 10.08133316040039 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7069954497655233, + "lr": 0.03, + "negative_goodness": 1.0763319070392185, + "pair_accuracy": 0.890688888920678, + "positive_goodness": 4.0887845291985405, + "runtime_seconds": 10.084752559661865 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6445511186732186, + "lr": 0.03, + "negative_goodness": 0.9953670365863376, + "pair_accuracy": 0.9188888888888889, + "positive_goodness": 4.27519487991333, + "runtime_seconds": 10.071909427642822 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6423035937839084, + "lr": 0.03, + "negative_goodness": 1.0009610660129122, + "pair_accuracy": 0.9162666666673289, + "positive_goodness": 4.325644279988607, + "runtime_seconds": 10.09606122970581 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6418837658087413, + "lr": 0.03, + "negative_goodness": 0.9662708445707957, + "pair_accuracy": 0.8958222222222222, + "positive_goodness": 4.375575034840901, + "runtime_seconds": 10.08718752861023 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6194209810468886, + "lr": 0.03, + "negative_goodness": 0.9145867816925048, + "pair_accuracy": 0.912111111164093, + "positive_goodness": 4.48890809457567, + "runtime_seconds": 10.102477312088013 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7192885091198815, + "lr": 0.03, + "negative_goodness": 1.0595584604263306, + "pair_accuracy": 0.889533333322737, + "positive_goodness": 4.052926141357422, + "runtime_seconds": 10.067776679992676 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.674453549838066, + "lr": 0.03, + "negative_goodness": 0.9858992713729541, + "pair_accuracy": 0.8907777777777778, + "positive_goodness": 4.234053226894803, + "runtime_seconds": 10.10163950920105 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6197646417935689, + "lr": 0.03, + "negative_goodness": 0.9538008056004842, + "pair_accuracy": 0.9123555555979411, + "positive_goodness": 4.50951521542867, + "runtime_seconds": 10.078099250793457 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6421668072382609, + "lr": 0.03, + "negative_goodness": 0.9412178508122762, + "pair_accuracy": 0.9112, + "positive_goodness": 4.308680074649387, + "runtime_seconds": 10.074577331542969 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6517625487221612, + "lr": 0.03, + "negative_goodness": 0.9659093055327733, + "pair_accuracy": 0.9077555555555555, + "positive_goodness": 4.3450387087504065, + "runtime_seconds": 10.104088544845581 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5889569138420953, + "lr": 0.03, + "negative_goodness": 0.8709987914191352, + "pair_accuracy": 0.9243777777779434, + "positive_goodness": 4.580334243901571, + "runtime_seconds": 10.085711479187012 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6447001052114698, + "lr": 0.03, + "negative_goodness": 0.9291367623011271, + "pair_accuracy": 0.9019777778201633, + "positive_goodness": 4.389428783374362, + "runtime_seconds": 10.088613748550415 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6360152425554063, + "lr": 0.03, + "negative_goodness": 0.9470372845649719, + "pair_accuracy": 0.9125333333863153, + "positive_goodness": 4.3662574136734005, + "runtime_seconds": 10.093218803405762 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6394881881925795, + "lr": 0.03, + "negative_goodness": 0.9053104804356893, + "pair_accuracy": 0.8970666666719649, + "positive_goodness": 4.403017366960314, + "runtime_seconds": 10.104112386703491 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6330157224443224, + "lr": 0.03, + "negative_goodness": 0.9091789373291863, + "pair_accuracy": 0.9067111111111111, + "positive_goodness": 4.315316286129422, + "runtime_seconds": 10.076805114746094 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5771090676519606, + "lr": 0.03, + "negative_goodness": 0.8537562822553847, + "pair_accuracy": 0.9179777777671814, + "positive_goodness": 4.64185483432346, + "runtime_seconds": 10.091705560684204 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6114524514728122, + "lr": 0.03, + "negative_goodness": 0.8928262943267822, + "pair_accuracy": 0.9125777777777778, + "positive_goodness": 4.445650251727634, + "runtime_seconds": 10.066736459732056 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6020866572883394, + "lr": 0.03, + "negative_goodness": 0.8851158845530616, + "pair_accuracy": 0.919, + "positive_goodness": 4.569046004740398, + "runtime_seconds": 9.97428560256958 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5901178986867269, + "lr": 0.03, + "negative_goodness": 0.8410444557083978, + "pair_accuracy": 0.9153333333015442, + "positive_goodness": 4.559525248591105, + "runtime_seconds": 9.97572946548462 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6149011706087324, + "lr": 0.03, + "negative_goodness": 0.8855366744836172, + "pair_accuracy": 0.9102666666666667, + "positive_goodness": 4.490979231770833, + "runtime_seconds": 9.97760558128357 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5795924818462795, + "lr": 0.03, + "negative_goodness": 0.8267556357383729, + "pair_accuracy": 0.9205333333333333, + "positive_goodness": 4.674566679551866, + "runtime_seconds": 9.975169897079468 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5976186845408545, + "lr": 0.03, + "negative_goodness": 0.8727001194741991, + "pair_accuracy": 0.9151555555555556, + "positive_goodness": 4.649753768836128, + "runtime_seconds": 10.071118116378784 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6266046964433458, + "lr": 0.03, + "negative_goodness": 0.8811416229354011, + "pair_accuracy": 0.8956000000105964, + "positive_goodness": 4.515528265719944, + "runtime_seconds": 10.08766770362854 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5980989153226217, + "lr": 0.03, + "negative_goodness": 0.8413431424352857, + "pair_accuracy": 0.9086666666666666, + "positive_goodness": 4.639497564909194, + "runtime_seconds": 10.086966276168823 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5716602661980523, + "lr": 0.03, + "negative_goodness": 0.8121969635645548, + "pair_accuracy": 0.9206888889153798, + "positive_goodness": 4.73554785639445, + "runtime_seconds": 10.09742283821106 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6139340599060059, + "lr": 0.03, + "negative_goodness": 0.8556454732577006, + "pair_accuracy": 0.9051555555237664, + "positive_goodness": 4.5838341565026175, + "runtime_seconds": 10.07446837425232 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6212346639633178, + "lr": 0.03, + "negative_goodness": 0.8631563958591885, + "pair_accuracy": 0.9063999999788073, + "positive_goodness": 4.522479294331869, + "runtime_seconds": 10.084010124206543 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5459674128426446, + "lr": 0.03, + "negative_goodness": 0.7910254267904494, + "pair_accuracy": 0.9283111111058129, + "positive_goodness": 4.829119348589579, + "runtime_seconds": 10.079168796539307 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5197063999096553, + "lr": 0.03, + "negative_goodness": 0.7588317914326985, + "pair_accuracy": 0.9364, + "positive_goodness": 4.944094464535183, + "runtime_seconds": 10.086157321929932 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5546120406044854, + "lr": 0.03, + "negative_goodness": 0.7926846154636807, + "pair_accuracy": 0.9253111111429002, + "positive_goodness": 4.834729881668091, + "runtime_seconds": 10.089683294296265 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5906173386891683, + "lr": 0.03, + "negative_goodness": 0.8251942931599087, + "pair_accuracy": 0.9140000000423856, + "positive_goodness": 4.766774418873257, + "runtime_seconds": 10.085829019546509 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5788779208183289, + "lr": 0.03, + "negative_goodness": 0.8228042661428452, + "pair_accuracy": 0.9167555555555555, + "positive_goodness": 4.743212650807698, + "runtime_seconds": 10.08182954788208 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5214588431040446, + "lr": 0.03, + "negative_goodness": 0.7705103333155314, + "pair_accuracy": 0.9315111111114422, + "positive_goodness": 4.915619294754664, + "runtime_seconds": 10.081504583358765 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5371790157000224, + "lr": 0.03, + "negative_goodness": 0.7305568634616004, + "pair_accuracy": 0.9247777777777778, + "positive_goodness": 4.94543761783176, + "runtime_seconds": 10.079449653625488 + } + ], + "layer": 44 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8860680801762475, + "lr": 0.03, + "negative_goodness": 1.2736660833994549, + "pair_accuracy": 0.8462888888888889, + "positive_goodness": 3.3971416694641112, + "runtime_seconds": 10.060720443725586 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8095235125859579, + "lr": 0.03, + "negative_goodness": 1.233468624581231, + "pair_accuracy": 0.8758222222222222, + "positive_goodness": 3.7086089919196237, + "runtime_seconds": 10.065419673919678 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7687490080515543, + "lr": 0.03, + "negative_goodness": 1.162976470041275, + "pair_accuracy": 0.8832666666666666, + "positive_goodness": 3.8769055029975044, + "runtime_seconds": 10.064117193222046 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7229626228862338, + "lr": 0.03, + "negative_goodness": 1.1182381855540806, + "pair_accuracy": 0.8976666666666666, + "positive_goodness": 4.020081284756131, + "runtime_seconds": 10.083614110946655 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6494950989299351, + "lr": 0.03, + "negative_goodness": 1.0161853020773994, + "pair_accuracy": 0.9169777777777778, + "positive_goodness": 4.331090111456977, + "runtime_seconds": 10.08237075805664 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6622287333170573, + "lr": 0.03, + "negative_goodness": 0.9956000843207041, + "pair_accuracy": 0.90608888888922, + "positive_goodness": 4.316962286165025, + "runtime_seconds": 10.073353290557861 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6605287237591214, + "lr": 0.03, + "negative_goodness": 0.9793508658250173, + "pair_accuracy": 0.9118222221904331, + "positive_goodness": 4.336842023976644, + "runtime_seconds": 10.078052997589111 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6779779950459798, + "lr": 0.03, + "negative_goodness": 1.0014136907842424, + "pair_accuracy": 0.9029111111217075, + "positive_goodness": 4.195911247253418, + "runtime_seconds": 10.073413372039795 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6665464745362599, + "lr": 0.03, + "negative_goodness": 0.9857515238642692, + "pair_accuracy": 0.9168222222222222, + "positive_goodness": 4.2191411058213975, + "runtime_seconds": 10.073228120803833 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6668456158532037, + "lr": 0.03, + "negative_goodness": 0.9762346671210395, + "pair_accuracy": 0.9050666666719649, + "positive_goodness": 4.249607761319479, + "runtime_seconds": 10.080596923828125 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.652231997900539, + "lr": 0.03, + "negative_goodness": 0.9727108365376791, + "pair_accuracy": 0.9067555555555555, + "positive_goodness": 4.163765992312961, + "runtime_seconds": 10.071112155914307 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6449920952267117, + "lr": 0.03, + "negative_goodness": 0.9448702925019794, + "pair_accuracy": 0.9083777777777777, + "positive_goodness": 4.365559017033047, + "runtime_seconds": 10.041404962539673 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6530934161557091, + "lr": 0.03, + "negative_goodness": 0.9513328678819868, + "pair_accuracy": 0.9089111111111111, + "positive_goodness": 4.198437461725871, + "runtime_seconds": 10.014373064041138 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6232893709500631, + "lr": 0.03, + "negative_goodness": 0.9129342718124389, + "pair_accuracy": 0.9213777777777777, + "positive_goodness": 4.314717015520731, + "runtime_seconds": 10.010612964630127 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6439603883319431, + "lr": 0.03, + "negative_goodness": 0.9288154378361172, + "pair_accuracy": 0.9072222222010294, + "positive_goodness": 4.276440311219957, + "runtime_seconds": 10.014437437057495 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.5989538674036662, + "lr": 0.03, + "negative_goodness": 0.878886164744695, + "pair_accuracy": 0.9187111111323039, + "positive_goodness": 4.446385312949286, + "runtime_seconds": 10.021391868591309 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.632533130258984, + "lr": 0.03, + "negative_goodness": 0.9198600718074375, + "pair_accuracy": 0.9076, + "positive_goodness": 4.381458644104004, + "runtime_seconds": 10.020251750946045 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6147734961085849, + "lr": 0.03, + "negative_goodness": 0.8783927221126027, + "pair_accuracy": 0.9153111111111111, + "positive_goodness": 4.496247257741293, + "runtime_seconds": 10.017763614654541 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5514801948865254, + "lr": 0.03, + "negative_goodness": 0.8103742051071591, + "pair_accuracy": 0.9284222222222223, + "positive_goodness": 4.74955685128106, + "runtime_seconds": 10.017934799194336 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5974581809997559, + "lr": 0.03, + "negative_goodness": 0.8816708873695798, + "pair_accuracy": 0.9243777777989706, + "positive_goodness": 4.451486898040772, + "runtime_seconds": 10.017053604125977 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5984091165516111, + "lr": 0.03, + "negative_goodness": 0.81706239964962, + "pair_accuracy": 0.9126666666666666, + "positive_goodness": 4.510995239766439, + "runtime_seconds": 10.022727727890015 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6138833817164103, + "lr": 0.03, + "negative_goodness": 0.8652015167898602, + "pair_accuracy": 0.9155111111534966, + "positive_goodness": 4.467153117964003, + "runtime_seconds": 10.02014422416687 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5428742558346854, + "lr": 0.03, + "negative_goodness": 0.8026842041068607, + "pair_accuracy": 0.9358444444444445, + "positive_goodness": 4.633457951524522, + "runtime_seconds": 10.016863107681274 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.603065804221895, + "lr": 0.03, + "negative_goodness": 0.8473362385590871, + "pair_accuracy": 0.9150888888888888, + "positive_goodness": 4.468854666222466, + "runtime_seconds": 10.016139268875122 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6315376769304275, + "lr": 0.03, + "negative_goodness": 0.8683059722052681, + "pair_accuracy": 0.9031333333333333, + "positive_goodness": 4.310899039204916, + "runtime_seconds": 10.070836305618286 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6244244272232056, + "lr": 0.03, + "negative_goodness": 0.846631499979231, + "pair_accuracy": 0.9038444444444445, + "positive_goodness": 4.475329614596896, + "runtime_seconds": 10.067122459411621 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5937769415325589, + "lr": 0.03, + "negative_goodness": 0.8156543301847246, + "pair_accuracy": 0.9166222222540114, + "positive_goodness": 4.626213286421034, + "runtime_seconds": 10.077249765396118 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5497895376549826, + "lr": 0.03, + "negative_goodness": 0.7524658863888847, + "pair_accuracy": 0.9216, + "positive_goodness": 4.725271642812093, + "runtime_seconds": 10.076430082321167 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6042610728369818, + "lr": 0.03, + "negative_goodness": 0.8191545137670305, + "pair_accuracy": 0.911511111164093, + "positive_goodness": 4.53518788172404, + "runtime_seconds": 10.074847936630249 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5834533440007104, + "lr": 0.03, + "negative_goodness": 0.7903278296099768, + "pair_accuracy": 0.9145555555555556, + "positive_goodness": 4.6226212761773, + "runtime_seconds": 10.074146032333374 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6031155740790897, + "lr": 0.03, + "negative_goodness": 0.8240025059315893, + "pair_accuracy": 0.9159777777777778, + "positive_goodness": 4.492649683888753, + "runtime_seconds": 10.074518203735352 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5991065583811865, + "lr": 0.03, + "negative_goodness": 0.8319234763145447, + "pair_accuracy": 0.9165555555555556, + "positive_goodness": 4.497758159806994, + "runtime_seconds": 10.071101903915405 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5854352151234945, + "lr": 0.03, + "negative_goodness": 0.8198851407051087, + "pair_accuracy": 0.9229333333333334, + "positive_goodness": 4.590788896343443, + "runtime_seconds": 10.076202869415283 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5971673323419359, + "lr": 0.03, + "negative_goodness": 0.8254820499128765, + "pair_accuracy": 0.91564444448683, + "positive_goodness": 4.454475086720785, + "runtime_seconds": 10.072973012924194 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5771418919669258, + "lr": 0.03, + "negative_goodness": 0.7824590207126405, + "pair_accuracy": 0.910022222190433, + "positive_goodness": 4.749134894476996, + "runtime_seconds": 10.073547840118408 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.49575501692030166, + "lr": 0.03, + "negative_goodness": 0.6981429692678981, + "pair_accuracy": 0.9416, + "positive_goodness": 5.054091406334771, + "runtime_seconds": 10.07804560661316 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5156575779279073, + "lr": 0.03, + "negative_goodness": 0.710158234433333, + "pair_accuracy": 0.9334666666666667, + "positive_goodness": 5.095560399288601, + "runtime_seconds": 10.076952695846558 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5720213703579373, + "lr": 0.03, + "negative_goodness": 0.7591535099214978, + "pair_accuracy": 0.9216, + "positive_goodness": 4.6921911375257706, + "runtime_seconds": 10.073920965194702 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5793892247941759, + "lr": 0.03, + "negative_goodness": 0.7919173444430033, + "pair_accuracy": 0.9214444444444444, + "positive_goodness": 4.731639198790656, + "runtime_seconds": 10.07814884185791 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5550637698252996, + "lr": 0.03, + "negative_goodness": 0.7281312873668141, + "pair_accuracy": 0.9237111111111111, + "positive_goodness": 4.745413439390394, + "runtime_seconds": 10.079649686813354 + } + ], + "layer": 45 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9454643867916531, + "lr": 0.03, + "negative_goodness": 1.4746423621707492, + "pair_accuracy": 0.8562, + "positive_goodness": 3.5603540918986, + "runtime_seconds": 10.108945846557617 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8569478264914618, + "lr": 0.03, + "negative_goodness": 1.3262245496114096, + "pair_accuracy": 0.8653333333757188, + "positive_goodness": 3.5025284813351103, + "runtime_seconds": 10.10463547706604 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7708378267579609, + "lr": 0.03, + "negative_goodness": 1.1813692196263208, + "pair_accuracy": 0.8741555555555556, + "positive_goodness": 3.7858461574978297, + "runtime_seconds": 10.109174728393555 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7957794940577613, + "lr": 0.03, + "negative_goodness": 1.2369137157970005, + "pair_accuracy": 0.8657333333333334, + "positive_goodness": 3.6545450304667155, + "runtime_seconds": 10.10103464126587 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7209135388904148, + "lr": 0.03, + "negative_goodness": 1.125935685157776, + "pair_accuracy": 0.8940222222063277, + "positive_goodness": 3.999766543706258, + "runtime_seconds": 10.10451340675354 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7439595644579994, + "lr": 0.03, + "negative_goodness": 1.128408426560296, + "pair_accuracy": 0.8830888888888889, + "positive_goodness": 3.8039565533955892, + "runtime_seconds": 10.106029987335205 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6593189979341295, + "lr": 0.03, + "negative_goodness": 1.0352950593418546, + "pair_accuracy": 0.9099777777777778, + "positive_goodness": 4.140175774214002, + "runtime_seconds": 10.112250804901123 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7371553056610956, + "lr": 0.03, + "negative_goodness": 1.0770899088541668, + "pair_accuracy": 0.877155555566152, + "positive_goodness": 3.9596248834398056, + "runtime_seconds": 10.10774827003479 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.730634809647666, + "lr": 0.03, + "negative_goodness": 1.0826971373028225, + "pair_accuracy": 0.8820222222222223, + "positive_goodness": 3.9058528959062366, + "runtime_seconds": 10.106631994247437 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7113840471400155, + "lr": 0.03, + "negative_goodness": 1.087222335765097, + "pair_accuracy": 0.8826444444444445, + "positive_goodness": 3.928593242136637, + "runtime_seconds": 10.061813116073608 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6855294843249851, + "lr": 0.03, + "negative_goodness": 1.0160150193108453, + "pair_accuracy": 0.8945333333439297, + "positive_goodness": 4.075185091612074, + "runtime_seconds": 10.091916799545288 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6847959185812208, + "lr": 0.03, + "negative_goodness": 1.007796247927348, + "pair_accuracy": 0.8895555555343628, + "positive_goodness": 4.129701912646824, + "runtime_seconds": 10.108227014541626 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6835379494137234, + "lr": 0.03, + "negative_goodness": 1.022489470354716, + "pair_accuracy": 0.9013777777777778, + "positive_goodness": 4.06874282629225, + "runtime_seconds": 10.116569995880127 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6643015308856964, + "lr": 0.03, + "negative_goodness": 0.9892259294191996, + "pair_accuracy": 0.9011777777777777, + "positive_goodness": 4.129249195946588, + "runtime_seconds": 10.099916458129883 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6216559585412343, + "lr": 0.03, + "negative_goodness": 0.9327085357666015, + "pair_accuracy": 0.9072888888888889, + "positive_goodness": 4.323409793006049, + "runtime_seconds": 10.101758480072021 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6319279933532079, + "lr": 0.03, + "negative_goodness": 0.9427832969215181, + "pair_accuracy": 0.9053333333333333, + "positive_goodness": 4.347706460995144, + "runtime_seconds": 10.09842848777771 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6136130317158169, + "lr": 0.03, + "negative_goodness": 0.9310224804772271, + "pair_accuracy": 0.9183555555449592, + "positive_goodness": 4.427854076131185, + "runtime_seconds": 10.100592374801636 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6258697824690077, + "lr": 0.03, + "negative_goodness": 0.9254955580181545, + "pair_accuracy": 0.9151777777883742, + "positive_goodness": 4.311740976927015, + "runtime_seconds": 10.101432800292969 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6645634132941564, + "lr": 0.03, + "negative_goodness": 0.9393790666050381, + "pair_accuracy": 0.8924444444444445, + "positive_goodness": 4.103482025739882, + "runtime_seconds": 10.108880996704102 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5537030080000559, + "lr": 0.03, + "negative_goodness": 0.8481375500652525, + "pair_accuracy": 0.9295111111111111, + "positive_goodness": 4.651333989037408, + "runtime_seconds": 10.097234725952148 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.665707439215978, + "lr": 0.03, + "negative_goodness": 0.9784765799681345, + "pair_accuracy": 0.9081111111111111, + "positive_goodness": 4.2193978803846575, + "runtime_seconds": 10.078247547149658 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6195325894461737, + "lr": 0.03, + "negative_goodness": 0.8936072930230035, + "pair_accuracy": 0.9092222222010294, + "positive_goodness": 4.468801927057902, + "runtime_seconds": 10.03861141204834 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6370962489340041, + "lr": 0.03, + "negative_goodness": 0.9065140623516507, + "pair_accuracy": 0.9001111111005148, + "positive_goodness": 4.31263516070048, + "runtime_seconds": 10.096463680267334 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6166958722856309, + "lr": 0.03, + "negative_goodness": 0.9247695491154989, + "pair_accuracy": 0.9148222222010295, + "positive_goodness": 4.316009868918525, + "runtime_seconds": 10.104164600372314 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6277931456300947, + "lr": 0.03, + "negative_goodness": 0.8973218126614888, + "pair_accuracy": 0.9070888888888888, + "positive_goodness": 4.391159927283393, + "runtime_seconds": 10.10986590385437 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6653899891906314, + "lr": 0.03, + "negative_goodness": 0.9562301820331149, + "pair_accuracy": 0.8937111111111111, + "positive_goodness": 4.207376086849637, + "runtime_seconds": 10.10010051727295 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.591415165731642, + "lr": 0.03, + "negative_goodness": 0.8574602201461792, + "pair_accuracy": 0.915155555539661, + "positive_goodness": 4.5137336347791885, + "runtime_seconds": 10.098332405090332 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5791812678654988, + "lr": 0.03, + "negative_goodness": 0.8567208623250325, + "pair_accuracy": 0.919755555566152, + "positive_goodness": 4.5566861535390215, + "runtime_seconds": 10.09746766090393 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5901881730927362, + "lr": 0.03, + "negative_goodness": 0.8497645950317383, + "pair_accuracy": 0.9144666666454739, + "positive_goodness": 4.515974321026272, + "runtime_seconds": 10.093210458755493 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5899934298568301, + "lr": 0.03, + "negative_goodness": 0.8721463739103741, + "pair_accuracy": 0.9192888888888889, + "positive_goodness": 4.491561012352838, + "runtime_seconds": 10.097766637802124 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6054775657018026, + "lr": 0.03, + "negative_goodness": 0.8830433646096124, + "pair_accuracy": 0.9215333333545261, + "positive_goodness": 4.391409263653225, + "runtime_seconds": 10.10280156135559 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5556231348673503, + "lr": 0.03, + "negative_goodness": 0.8225973289648691, + "pair_accuracy": 0.9317555555555556, + "positive_goodness": 4.694046157667372, + "runtime_seconds": 10.086665868759155 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6633031949149237, + "lr": 0.03, + "negative_goodness": 0.9104946001264784, + "pair_accuracy": 0.8954666666666666, + "positive_goodness": 4.254643029573229, + "runtime_seconds": 10.037577629089355 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5863006171968248, + "lr": 0.03, + "negative_goodness": 0.8465041871176826, + "pair_accuracy": 0.9224222222116258, + "positive_goodness": 4.566744807688395, + "runtime_seconds": 10.042097806930542 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.603749256128735, + "lr": 0.03, + "negative_goodness": 0.8566013425747554, + "pair_accuracy": 0.9193111111111111, + "positive_goodness": 4.463840265994602, + "runtime_seconds": 10.03553032875061 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5705523841910892, + "lr": 0.03, + "negative_goodness": 0.8092608760224448, + "pair_accuracy": 0.9237333333333333, + "positive_goodness": 4.671305928632948, + "runtime_seconds": 10.040732145309448 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5695340238014857, + "lr": 0.03, + "negative_goodness": 0.8179248960071139, + "pair_accuracy": 0.9200666666666667, + "positive_goodness": 4.619156283315022, + "runtime_seconds": 10.044881582260132 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5504846266428629, + "lr": 0.03, + "negative_goodness": 0.8013754982524448, + "pair_accuracy": 0.9284222222116258, + "positive_goodness": 4.7047497393714055, + "runtime_seconds": 10.030721426010132 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.588724856821696, + "lr": 0.03, + "negative_goodness": 0.8454264588356019, + "pair_accuracy": 0.920577777809567, + "positive_goodness": 4.5581686618593, + "runtime_seconds": 10.020333290100098 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6513858989026812, + "lr": 0.03, + "negative_goodness": 0.885727674876319, + "pair_accuracy": 0.8982222222222223, + "positive_goodness": 4.248626653628879, + "runtime_seconds": 10.030713558197021 + } + ], + "layer": 46 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.7509889979124069, + "lr": 0.03, + "negative_goodness": 1.0585817709287009, + "pair_accuracy": 0.8746, + "positive_goodness": 3.889799285888672, + "runtime_seconds": 10.01156234741211 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.805835777229733, + "lr": 0.03, + "negative_goodness": 1.1566733598391215, + "pair_accuracy": 0.8537333333333333, + "positive_goodness": 3.773800124698215, + "runtime_seconds": 10.020140886306763 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8233789085812039, + "lr": 0.03, + "negative_goodness": 1.1767417003631593, + "pair_accuracy": 0.856977777809567, + "positive_goodness": 3.6628360942840574, + "runtime_seconds": 10.020422458648682 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7660000596576266, + "lr": 0.03, + "negative_goodness": 1.1699699289639791, + "pair_accuracy": 0.8808222222222222, + "positive_goodness": 3.791829153781467, + "runtime_seconds": 10.026766300201416 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.733036866484748, + "lr": 0.03, + "negative_goodness": 1.1111212184906005, + "pair_accuracy": 0.8952444444285499, + "positive_goodness": 3.955499699783325, + "runtime_seconds": 10.018225193023682 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6950301309373643, + "lr": 0.03, + "negative_goodness": 1.0478989976882935, + "pair_accuracy": 0.8980222222752041, + "positive_goodness": 3.9704496649424237, + "runtime_seconds": 10.019359111785889 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6737767413934072, + "lr": 0.03, + "negative_goodness": 1.0201217968198988, + "pair_accuracy": 0.8932222222222223, + "positive_goodness": 4.122900023481581, + "runtime_seconds": 10.023486852645874 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7090286201159159, + "lr": 0.03, + "negative_goodness": 1.0028425963189866, + "pair_accuracy": 0.8883777777671814, + "positive_goodness": 4.046813424216376, + "runtime_seconds": 10.01295018196106 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.698904631529914, + "lr": 0.03, + "negative_goodness": 1.0300119663556417, + "pair_accuracy": 0.8983111111376021, + "positive_goodness": 4.047114983876546, + "runtime_seconds": 10.022091388702393 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6327006008360121, + "lr": 0.03, + "negative_goodness": 0.9656412847783831, + "pair_accuracy": 0.917666666677263, + "positive_goodness": 4.353110056474474, + "runtime_seconds": 10.079277515411377 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6553725686603122, + "lr": 0.03, + "negative_goodness": 0.9768850217130449, + "pair_accuracy": 0.9101111111323039, + "positive_goodness": 4.227533372158474, + "runtime_seconds": 10.074438571929932 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.679126697921753, + "lr": 0.03, + "negative_goodness": 0.9706092124256823, + "pair_accuracy": 0.8967555555555555, + "positive_goodness": 4.282143689176771, + "runtime_seconds": 10.076709032058716 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.7141995176315308, + "lr": 0.03, + "negative_goodness": 0.9933788674354553, + "pair_accuracy": 0.8783111111005147, + "positive_goodness": 4.032003035693698, + "runtime_seconds": 10.076459884643555 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6885530319690705, + "lr": 0.03, + "negative_goodness": 0.9936960050847795, + "pair_accuracy": 0.8924444444444445, + "positive_goodness": 4.205625550842285, + "runtime_seconds": 10.009275436401367 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.625801300740242, + "lr": 0.03, + "negative_goodness": 0.911117371527354, + "pair_accuracy": 0.9086666666666666, + "positive_goodness": 4.4476646374172635, + "runtime_seconds": 9.960929155349731 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6413860049618615, + "lr": 0.03, + "negative_goodness": 0.9431101871437496, + "pair_accuracy": 0.9081555555555556, + "positive_goodness": 4.303457567511664, + "runtime_seconds": 9.957911252975464 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6813163195557065, + "lr": 0.03, + "negative_goodness": 0.9335349500762091, + "pair_accuracy": 0.8948888888888888, + "positive_goodness": 4.27055273352729, + "runtime_seconds": 9.953481912612915 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6426040394041274, + "lr": 0.03, + "negative_goodness": 0.9227099067714479, + "pair_accuracy": 0.8995555555449591, + "positive_goodness": 4.323266861258613, + "runtime_seconds": 9.962931632995605 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6426955445633994, + "lr": 0.03, + "negative_goodness": 0.9086218942602475, + "pair_accuracy": 0.9028222222222222, + "positive_goodness": 4.299542156304254, + "runtime_seconds": 9.96480107307434 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6422112154271867, + "lr": 0.03, + "negative_goodness": 0.9245647778378593, + "pair_accuracy": 0.9056888888888889, + "positive_goodness": 4.284907625622219, + "runtime_seconds": 9.977610111236572 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6167599059210883, + "lr": 0.03, + "negative_goodness": 0.891384081308047, + "pair_accuracy": 0.9105555555237664, + "positive_goodness": 4.472415610800849, + "runtime_seconds": 10.016610145568848 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.590854419154591, + "lr": 0.03, + "negative_goodness": 0.83902163330449, + "pair_accuracy": 0.9134, + "positive_goodness": 4.675784610493978, + "runtime_seconds": 10.025578498840332 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6227432424015469, + "lr": 0.03, + "negative_goodness": 0.9045049454821481, + "pair_accuracy": 0.9148666666666667, + "positive_goodness": 4.490438167911106, + "runtime_seconds": 10.020029306411743 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6655056238121456, + "lr": 0.03, + "negative_goodness": 0.9305003513839509, + "pair_accuracy": 0.9023555555555556, + "positive_goodness": 4.254048021104601, + "runtime_seconds": 10.015880823135376 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6547113120979733, + "lr": 0.03, + "negative_goodness": 0.9159140956418382, + "pair_accuracy": 0.9064222222222222, + "positive_goodness": 4.243242253367106, + "runtime_seconds": 10.02000117301941 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6153375278870264, + "lr": 0.03, + "negative_goodness": 0.8695896824532084, + "pair_accuracy": 0.9142, + "positive_goodness": 4.365218082088894, + "runtime_seconds": 10.020765781402588 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5798363998916414, + "lr": 0.03, + "negative_goodness": 0.827073824069235, + "pair_accuracy": 0.9248444444444445, + "positive_goodness": 4.6469362833235, + "runtime_seconds": 10.022949695587158 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6115806965695487, + "lr": 0.03, + "negative_goodness": 0.8784756493727366, + "pair_accuracy": 0.9127111111111111, + "positive_goodness": 4.41048894627889, + "runtime_seconds": 10.028647422790527 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6435682909859551, + "lr": 0.03, + "negative_goodness": 0.8962843156019846, + "pair_accuracy": 0.9065777777777778, + "positive_goodness": 4.349741572613186, + "runtime_seconds": 10.02534818649292 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5558024951166577, + "lr": 0.03, + "negative_goodness": 0.8091614491151439, + "pair_accuracy": 0.9355555555555556, + "positive_goodness": 4.635438447825114, + "runtime_seconds": 10.026383638381958 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6162388800938924, + "lr": 0.03, + "negative_goodness": 0.8810901050262981, + "pair_accuracy": 0.9132222222222223, + "positive_goodness": 4.511766043429905, + "runtime_seconds": 10.028766632080078 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6683057920244005, + "lr": 0.03, + "negative_goodness": 0.899460025106536, + "pair_accuracy": 0.8871333333757189, + "positive_goodness": 4.332178110037909, + "runtime_seconds": 10.023855924606323 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5917661720514298, + "lr": 0.03, + "negative_goodness": 0.8356529841396544, + "pair_accuracy": 0.9124888888888889, + "positive_goodness": 4.588539519585503, + "runtime_seconds": 10.024022102355957 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6573368574566312, + "lr": 0.03, + "negative_goodness": 0.9000139622820749, + "pair_accuracy": 0.8997333333333334, + "positive_goodness": 4.3485191574732465, + "runtime_seconds": 10.023800611495972 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5992025223626031, + "lr": 0.03, + "negative_goodness": 0.8345301794634925, + "pair_accuracy": 0.9160222222116259, + "positive_goodness": 4.588890724097358, + "runtime_seconds": 10.017826557159424 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5687634851270251, + "lr": 0.03, + "negative_goodness": 0.8191899930212233, + "pair_accuracy": 0.9224, + "positive_goodness": 4.6923967918396, + "runtime_seconds": 10.016484022140503 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5970765185011758, + "lr": 0.03, + "negative_goodness": 0.8329105189800262, + "pair_accuracy": 0.9171555555555555, + "positive_goodness": 4.635415122307672, + "runtime_seconds": 10.018012046813965 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6230689638747109, + "lr": 0.03, + "negative_goodness": 0.8441730618423886, + "pair_accuracy": 0.9024, + "positive_goodness": 4.445145313559638, + "runtime_seconds": 10.020829439163208 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6308746575726403, + "lr": 0.03, + "negative_goodness": 0.8820682648367352, + "pair_accuracy": 0.9069555555555555, + "positive_goodness": 4.445134696282281, + "runtime_seconds": 10.021762609481812 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5658772347768147, + "lr": 0.03, + "negative_goodness": 0.7782644309918085, + "pair_accuracy": 0.9239111111164093, + "positive_goodness": 4.685565070597331, + "runtime_seconds": 10.017639636993408 + } + ], + "layer": 47 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9556645269711812, + "lr": 0.03, + "negative_goodness": 1.5106836212370132, + "pair_accuracy": 0.8516888888888889, + "positive_goodness": 3.503890670606825, + "runtime_seconds": 10.037766695022583 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7897744193024105, + "lr": 0.03, + "negative_goodness": 1.2570382828288609, + "pair_accuracy": 0.8853999999894037, + "positive_goodness": 3.8313620577494305, + "runtime_seconds": 10.033180952072144 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8092127690739102, + "lr": 0.03, + "negative_goodness": 1.2716897482024299, + "pair_accuracy": 0.8856, + "positive_goodness": 3.7531701243930393, + "runtime_seconds": 10.03589415550232 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7166552744812436, + "lr": 0.03, + "negative_goodness": 1.154467751121521, + "pair_accuracy": 0.9090666666560703, + "positive_goodness": 3.983179187605116, + "runtime_seconds": 10.041009664535522 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.71950387573772, + "lr": 0.03, + "negative_goodness": 1.1148903985473844, + "pair_accuracy": 0.8930222222222223, + "positive_goodness": 3.9918710152520074, + "runtime_seconds": 10.04572319984436 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7119473086886936, + "lr": 0.03, + "negative_goodness": 1.098910070207384, + "pair_accuracy": 0.8982222222116258, + "positive_goodness": 3.9417206552505495, + "runtime_seconds": 10.029724597930908 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.724884518898858, + "lr": 0.03, + "negative_goodness": 1.1078229277292888, + "pair_accuracy": 0.9015333333253861, + "positive_goodness": 3.9670638195461696, + "runtime_seconds": 10.03264045715332 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7001290787378947, + "lr": 0.03, + "negative_goodness": 1.0707695884492663, + "pair_accuracy": 0.9003999999788073, + "positive_goodness": 4.007806697463989, + "runtime_seconds": 10.03383207321167 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6933284962972005, + "lr": 0.03, + "negative_goodness": 1.0437080457687378, + "pair_accuracy": 0.8991777777777777, + "positive_goodness": 4.030168920135498, + "runtime_seconds": 10.038951873779297 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6423406274689568, + "lr": 0.03, + "negative_goodness": 0.9893807917224037, + "pair_accuracy": 0.9096444444444445, + "positive_goodness": 4.178182397715251, + "runtime_seconds": 10.047543287277222 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6239662424431907, + "lr": 0.03, + "negative_goodness": 0.9718499404827754, + "pair_accuracy": 0.9102, + "positive_goodness": 4.29018596818712, + "runtime_seconds": 10.03997540473938 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.642972979508506, + "lr": 0.03, + "negative_goodness": 0.9690125368542142, + "pair_accuracy": 0.9061777777777777, + "positive_goodness": 4.192606391482883, + "runtime_seconds": 10.039491176605225 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6508428538587359, + "lr": 0.03, + "negative_goodness": 0.982193595541848, + "pair_accuracy": 0.9087333333333333, + "positive_goodness": 4.263385879516601, + "runtime_seconds": 10.045611381530762 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6597611940013037, + "lr": 0.03, + "negative_goodness": 0.9658930505275727, + "pair_accuracy": 0.9048, + "positive_goodness": 4.167958129119873, + "runtime_seconds": 10.038634777069092 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.627809990506702, + "lr": 0.03, + "negative_goodness": 0.9328276389016046, + "pair_accuracy": 0.9144888888888889, + "positive_goodness": 4.326137974802653, + "runtime_seconds": 10.034473419189453 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6330893417358399, + "lr": 0.03, + "negative_goodness": 0.9576519890467325, + "pair_accuracy": 0.9041555555568801, + "positive_goodness": 4.196567265023125, + "runtime_seconds": 10.037775993347168 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5838285841359032, + "lr": 0.03, + "negative_goodness": 0.8905726720849673, + "pair_accuracy": 0.9212666666666667, + "positive_goodness": 4.468698119778104, + "runtime_seconds": 10.03335976600647 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6446422501617007, + "lr": 0.03, + "negative_goodness": 0.9465318074120416, + "pair_accuracy": 0.905044444433848, + "positive_goodness": 4.3502023292541505, + "runtime_seconds": 10.035930871963501 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6439551240391201, + "lr": 0.03, + "negative_goodness": 0.9405104278988309, + "pair_accuracy": 0.9085777777777778, + "positive_goodness": 4.232613704299927, + "runtime_seconds": 10.040035486221313 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5802828068945143, + "lr": 0.03, + "negative_goodness": 0.8887045730087492, + "pair_accuracy": 0.9228666666666666, + "positive_goodness": 4.568433416408963, + "runtime_seconds": 10.038237571716309 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5320251569509507, + "lr": 0.03, + "negative_goodness": 0.7831897882699966, + "pair_accuracy": 0.9276888888888889, + "positive_goodness": 4.742801617177327, + "runtime_seconds": 10.036604404449463 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6396921509265899, + "lr": 0.03, + "negative_goodness": 0.8795291486369239, + "pair_accuracy": 0.9038888888888889, + "positive_goodness": 4.333175610521105, + "runtime_seconds": 10.035812139511108 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6022535765118069, + "lr": 0.03, + "negative_goodness": 0.8818741007275052, + "pair_accuracy": 0.9131333333757189, + "positive_goodness": 4.42444113752577, + "runtime_seconds": 10.03903579711914 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.551956500837538, + "lr": 0.03, + "negative_goodness": 0.81453553437127, + "pair_accuracy": 0.9326666666668323, + "positive_goodness": 4.7533246831681994, + "runtime_seconds": 10.040955543518066 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5986338709195455, + "lr": 0.03, + "negative_goodness": 0.8386497699313693, + "pair_accuracy": 0.9159777777671814, + "positive_goodness": 4.438725885518392, + "runtime_seconds": 10.040961027145386 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5671873942693074, + "lr": 0.03, + "negative_goodness": 0.808270259337955, + "pair_accuracy": 0.9130222222010295, + "positive_goodness": 4.672552778498332, + "runtime_seconds": 10.046934366226196 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5702879034625159, + "lr": 0.03, + "negative_goodness": 0.8369321902063158, + "pair_accuracy": 0.9313333333333333, + "positive_goodness": 4.558348125457764, + "runtime_seconds": 10.035788297653198 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5483188186221652, + "lr": 0.03, + "negative_goodness": 0.7896654685974122, + "pair_accuracy": 0.9268888888888889, + "positive_goodness": 4.71298992606269, + "runtime_seconds": 10.045343160629272 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6332948939032025, + "lr": 0.03, + "negative_goodness": 0.8649042991518975, + "pair_accuracy": 0.8965555555555556, + "positive_goodness": 4.363830300140381, + "runtime_seconds": 10.045077800750732 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.557010791460673, + "lr": 0.03, + "negative_goodness": 0.8074925434324477, + "pair_accuracy": 0.9258444444656372, + "positive_goodness": 4.615432792790731, + "runtime_seconds": 10.042617082595825 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.66138798116578, + "lr": 0.03, + "negative_goodness": 0.914602144294315, + "pair_accuracy": 0.8883333333333333, + "positive_goodness": 4.263681051211887, + "runtime_seconds": 10.040305852890015 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6230243718571133, + "lr": 0.03, + "negative_goodness": 0.8955985754556126, + "pair_accuracy": 0.9088444444444445, + "positive_goodness": 4.342459860992432, + "runtime_seconds": 10.045342683792114 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5706709362559849, + "lr": 0.03, + "negative_goodness": 0.7847752315574222, + "pair_accuracy": 0.9241111111111111, + "positive_goodness": 4.627227408684624, + "runtime_seconds": 10.040027141571045 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5756832020521164, + "lr": 0.03, + "negative_goodness": 0.8249718679812219, + "pair_accuracy": 0.9248444444444445, + "positive_goodness": 4.508928437635634, + "runtime_seconds": 10.038841009140015 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6083501094288296, + "lr": 0.03, + "negative_goodness": 0.8465094998783536, + "pair_accuracy": 0.9142666666772631, + "positive_goodness": 4.388291749106513, + "runtime_seconds": 10.04573678970337 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5218964057074653, + "lr": 0.03, + "negative_goodness": 0.75968275358412, + "pair_accuracy": 0.938377777756585, + "positive_goodness": 4.807339183341132, + "runtime_seconds": 10.032671451568604 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.558681070833736, + "lr": 0.03, + "negative_goodness": 0.7735006480163998, + "pair_accuracy": 0.9180666666666667, + "positive_goodness": 4.793060501268175, + "runtime_seconds": 9.98029637336731 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5921967450883654, + "lr": 0.03, + "negative_goodness": 0.7963791693899367, + "pair_accuracy": 0.9128444444447755, + "positive_goodness": 4.581961151568095, + "runtime_seconds": 9.97819185256958 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5068888984282811, + "lr": 0.03, + "negative_goodness": 0.7634167850560613, + "pair_accuracy": 0.9434222222222223, + "positive_goodness": 4.86082954864502, + "runtime_seconds": 9.978646516799927 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5496408563507927, + "lr": 0.03, + "negative_goodness": 0.7480275532934401, + "pair_accuracy": 0.9206444444126552, + "positive_goodness": 4.749715635257297, + "runtime_seconds": 9.978944540023804 + } + ], + "layer": 48 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8608405230416192, + "lr": 0.03, + "negative_goodness": 1.2208640418052674, + "pair_accuracy": 0.865311111079322, + "positive_goodness": 3.4254670569525825, + "runtime_seconds": 9.967078685760498 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.7547199245452881, + "lr": 0.03, + "negative_goodness": 1.1085793360392253, + "pair_accuracy": 0.8982666666666667, + "positive_goodness": 3.985507176589966, + "runtime_seconds": 10.00393009185791 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.6852944338189231, + "lr": 0.03, + "negative_goodness": 1.0554584245363872, + "pair_accuracy": 0.9123111111111111, + "positive_goodness": 4.120144678243001, + "runtime_seconds": 10.00311803817749 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.692991244761149, + "lr": 0.03, + "negative_goodness": 1.0249668288283877, + "pair_accuracy": 0.9030666666984558, + "positive_goodness": 4.077464967346192, + "runtime_seconds": 10.021154880523682 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6572486126052008, + "lr": 0.03, + "negative_goodness": 0.9845253782484267, + "pair_accuracy": 0.9027777777671814, + "positive_goodness": 4.33303185128106, + "runtime_seconds": 10.017723321914673 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6768978919770983, + "lr": 0.03, + "negative_goodness": 0.9661752522150675, + "pair_accuracy": 0.9086666666666666, + "positive_goodness": 4.260853667619493, + "runtime_seconds": 10.017683744430542 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.615332351620992, + "lr": 0.03, + "negative_goodness": 0.920110888698366, + "pair_accuracy": 0.9184666667090522, + "positive_goodness": 4.380157560306126, + "runtime_seconds": 10.07411789894104 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6298575479295518, + "lr": 0.03, + "negative_goodness": 0.9308792130258349, + "pair_accuracy": 0.9196888888835907, + "positive_goodness": 4.324788413365682, + "runtime_seconds": 10.072332859039307 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6420775249057346, + "lr": 0.03, + "negative_goodness": 0.9294133249388801, + "pair_accuracy": 0.9048444444550409, + "positive_goodness": 4.263689312129551, + "runtime_seconds": 10.068147420883179 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.615127854903539, + "lr": 0.03, + "negative_goodness": 0.8874823488261965, + "pair_accuracy": 0.9154666666666667, + "positive_goodness": 4.3442761461893715, + "runtime_seconds": 10.078190803527832 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6172329311688741, + "lr": 0.03, + "negative_goodness": 0.8940726155042649, + "pair_accuracy": 0.9151333333333334, + "positive_goodness": 4.469020071495904, + "runtime_seconds": 10.071424961090088 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6081538464943568, + "lr": 0.03, + "negative_goodness": 0.8801004032770793, + "pair_accuracy": 0.9146222222222222, + "positive_goodness": 4.378121089935303, + "runtime_seconds": 10.075969696044922 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6438125316725837, + "lr": 0.03, + "negative_goodness": 0.8964498304658466, + "pair_accuracy": 0.9054, + "positive_goodness": 4.340443732028537, + "runtime_seconds": 10.022891759872437 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6045488888634576, + "lr": 0.03, + "negative_goodness": 0.8625966466453341, + "pair_accuracy": 0.9225777777777778, + "positive_goodness": 4.437092317369249, + "runtime_seconds": 10.01377010345459 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.5972528516133626, + "lr": 0.03, + "negative_goodness": 0.8429029758234818, + "pair_accuracy": 0.9138666666666667, + "positive_goodness": 4.654232831319173, + "runtime_seconds": 10.006586790084839 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6254441573805279, + "lr": 0.03, + "negative_goodness": 0.8729054786278142, + "pair_accuracy": 0.9135111111111112, + "positive_goodness": 4.469083426539103, + "runtime_seconds": 10.021023511886597 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.5278818462716208, + "lr": 0.03, + "negative_goodness": 0.7416779708054331, + "pair_accuracy": 0.9348444444444445, + "positive_goodness": 4.800078827412923, + "runtime_seconds": 10.013746738433838 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6619796014255948, + "lr": 0.03, + "negative_goodness": 0.8896621986471944, + "pair_accuracy": 0.8926666666666667, + "positive_goodness": 4.322796659257677, + "runtime_seconds": 10.021897315979004 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6461988386789957, + "lr": 0.03, + "negative_goodness": 0.9148671193361282, + "pair_accuracy": 0.9057777777459887, + "positive_goodness": 4.23751984293196, + "runtime_seconds": 10.024096965789795 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.5259282250404358, + "lr": 0.03, + "negative_goodness": 0.7667646512470312, + "pair_accuracy": 0.933, + "positive_goodness": 4.86921011505127, + "runtime_seconds": 10.021299839019775 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5774568408038881, + "lr": 0.03, + "negative_goodness": 0.7854883380893204, + "pair_accuracy": 0.9189555555555555, + "positive_goodness": 4.695069144863552, + "runtime_seconds": 10.006296873092651 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6327136436515384, + "lr": 0.03, + "negative_goodness": 0.8715298809104496, + "pair_accuracy": 0.9046444444444445, + "positive_goodness": 4.410276686520047, + "runtime_seconds": 10.01503038406372 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5876503408749898, + "lr": 0.03, + "negative_goodness": 0.8386949739813805, + "pair_accuracy": 0.9167111111111111, + "positive_goodness": 4.699977326583863, + "runtime_seconds": 10.016330003738403 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5743087044186063, + "lr": 0.03, + "negative_goodness": 0.7983445225477218, + "pair_accuracy": 0.9244222222010294, + "positive_goodness": 4.657649431843228, + "runtime_seconds": 10.02325177192688 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6131763659265306, + "lr": 0.03, + "negative_goodness": 0.8315994108065963, + "pair_accuracy": 0.9086444444444445, + "positive_goodness": 4.39181091325548, + "runtime_seconds": 10.02388882637024 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5832107017093234, + "lr": 0.03, + "negative_goodness": 0.8070058507813348, + "pair_accuracy": 0.9198888889418708, + "positive_goodness": 4.587465270487468, + "runtime_seconds": 10.025866746902466 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5353840025530922, + "lr": 0.03, + "negative_goodness": 0.7450218870533837, + "pair_accuracy": 0.9305555555555556, + "positive_goodness": 5.033314386410184, + "runtime_seconds": 10.024152040481567 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5203048346148597, + "lr": 0.03, + "negative_goodness": 0.7428832002454334, + "pair_accuracy": 0.9401333333333334, + "positive_goodness": 4.995968236202664, + "runtime_seconds": 10.029066562652588 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6154123129500283, + "lr": 0.03, + "negative_goodness": 0.7926810667554538, + "pair_accuracy": 0.9013111111111111, + "positive_goodness": 4.552289917755127, + "runtime_seconds": 10.00430178642273 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5378996343188815, + "lr": 0.03, + "negative_goodness": 0.7414344246016609, + "pair_accuracy": 0.928222222243415, + "positive_goodness": 4.9378727868609955, + "runtime_seconds": 9.96708369255066 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5255037312401666, + "lr": 0.03, + "negative_goodness": 0.7255147974239455, + "pair_accuracy": 0.9341111111111111, + "positive_goodness": 4.899665665033129, + "runtime_seconds": 9.96572494506836 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6081364333179262, + "lr": 0.03, + "negative_goodness": 0.80558235751258, + "pair_accuracy": 0.9043555555555556, + "positive_goodness": 4.725248436313206, + "runtime_seconds": 9.967331171035767 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5960440228833093, + "lr": 0.03, + "negative_goodness": 0.7886643053134282, + "pair_accuracy": 0.9190222222222222, + "positive_goodness": 4.623406435818143, + "runtime_seconds": 9.961615324020386 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5157386973487006, + "lr": 0.03, + "negative_goodness": 0.7333225656251112, + "pair_accuracy": 0.9347111111270057, + "positive_goodness": 5.067700661087036, + "runtime_seconds": 9.965617656707764 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5855571704361174, + "lr": 0.03, + "negative_goodness": 0.766868519586987, + "pair_accuracy": 0.9195555555555556, + "positive_goodness": 4.728741132948134, + "runtime_seconds": 9.958759307861328 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5805177969190809, + "lr": 0.03, + "negative_goodness": 0.7867199313309458, + "pair_accuracy": 0.9191777777777778, + "positive_goodness": 4.692445377180311, + "runtime_seconds": 9.963146448135376 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5038030570136176, + "lr": 0.03, + "negative_goodness": 0.7075142765031921, + "pair_accuracy": 0.9438888888888889, + "positive_goodness": 4.942223031616211, + "runtime_seconds": 9.96493673324585 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5685872324069341, + "lr": 0.03, + "negative_goodness": 0.7444680289480421, + "pair_accuracy": 0.9134, + "positive_goodness": 4.908901737509834, + "runtime_seconds": 9.964640140533447 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.551952613374922, + "lr": 0.03, + "negative_goodness": 0.7280040501773357, + "pair_accuracy": 0.9210888889418708, + "positive_goodness": 4.954042401038276, + "runtime_seconds": 9.959681034088135 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5395793052037557, + "lr": 0.03, + "negative_goodness": 0.7257482261247105, + "pair_accuracy": 0.930311111137602, + "positive_goodness": 4.908160511165195, + "runtime_seconds": 9.958554029464722 + } + ], + "layer": 49 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9417036821047465, + "lr": 0.03, + "negative_goodness": 1.472839511680603, + "pair_accuracy": 0.8558222222328186, + "positive_goodness": 3.6374190672132705, + "runtime_seconds": 9.988226652145386 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.741615688641866, + "lr": 0.03, + "negative_goodness": 1.173699757385254, + "pair_accuracy": 0.902933333322737, + "positive_goodness": 4.018752478451199, + "runtime_seconds": 9.988186120986938 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7491230641523997, + "lr": 0.03, + "negative_goodness": 1.1386994646284316, + "pair_accuracy": 0.8881555555555556, + "positive_goodness": 4.04879218266805, + "runtime_seconds": 9.992087841033936 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7314277631547715, + "lr": 0.03, + "negative_goodness": 1.0939732081943088, + "pair_accuracy": 0.9019333333465788, + "positive_goodness": 4.033838463359409, + "runtime_seconds": 10.00942611694336 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.6815637051476373, + "lr": 0.03, + "negative_goodness": 1.0201830136193168, + "pair_accuracy": 0.9009111111111111, + "positive_goodness": 4.226503768030803, + "runtime_seconds": 9.999874114990234 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6854506649017335, + "lr": 0.03, + "negative_goodness": 1.0355264951494005, + "pair_accuracy": 0.9094000000052982, + "positive_goodness": 4.206494112904867, + "runtime_seconds": 10.008931875228882 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7062542958259582, + "lr": 0.03, + "negative_goodness": 1.0521224022547404, + "pair_accuracy": 0.9060888888676961, + "positive_goodness": 4.00074118715922, + "runtime_seconds": 10.012451887130737 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6609805555555556, + "lr": 0.03, + "negative_goodness": 0.9728611498249902, + "pair_accuracy": 0.8991555555555556, + "positive_goodness": 4.2190863035413955, + "runtime_seconds": 10.009950637817383 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6052597746637133, + "lr": 0.03, + "negative_goodness": 0.9317849480628967, + "pair_accuracy": 0.917377777756585, + "positive_goodness": 4.398246662224664, + "runtime_seconds": 10.038393497467041 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6786376272757848, + "lr": 0.03, + "negative_goodness": 0.9839559029791091, + "pair_accuracy": 0.9002444444444444, + "positive_goodness": 4.147732072194417, + "runtime_seconds": 10.06421184539795 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6206770835240683, + "lr": 0.03, + "negative_goodness": 0.9208910596423678, + "pair_accuracy": 0.9122222222752041, + "positive_goodness": 4.401299571990966, + "runtime_seconds": 10.040464401245117 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.646733862929874, + "lr": 0.03, + "negative_goodness": 0.9323406349182128, + "pair_accuracy": 0.9061555555767483, + "positive_goodness": 4.360781724294027, + "runtime_seconds": 10.050564050674438 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.5639505513085259, + "lr": 0.03, + "negative_goodness": 0.8436890657001072, + "pair_accuracy": 0.9298666666666666, + "positive_goodness": 4.67078637392256, + "runtime_seconds": 10.076878309249878 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.5593360664288203, + "lr": 0.03, + "negative_goodness": 0.824041066424052, + "pair_accuracy": 0.9322888888888888, + "positive_goodness": 4.67922130440606, + "runtime_seconds": 10.03655481338501 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.606739276793268, + "lr": 0.03, + "negative_goodness": 0.8725128966172536, + "pair_accuracy": 0.9174, + "positive_goodness": 4.459211054907905, + "runtime_seconds": 10.006713628768921 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.600708917678727, + "lr": 0.03, + "negative_goodness": 0.8456813607030444, + "pair_accuracy": 0.9163555555555556, + "positive_goodness": 4.567625704871284, + "runtime_seconds": 9.989904403686523 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6459993882285224, + "lr": 0.03, + "negative_goodness": 0.9007037235260009, + "pair_accuracy": 0.9058666666454739, + "positive_goodness": 4.367943588468764, + "runtime_seconds": 10.002551555633545 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.5583876171218024, + "lr": 0.03, + "negative_goodness": 0.8222946037928264, + "pair_accuracy": 0.9282888888676961, + "positive_goodness": 4.723868201149835, + "runtime_seconds": 10.058602094650269 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.5569208331108093, + "lr": 0.03, + "negative_goodness": 0.798187047041787, + "pair_accuracy": 0.9254222222010294, + "positive_goodness": 4.790486860317654, + "runtime_seconds": 10.06980037689209 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.650234633702702, + "lr": 0.03, + "negative_goodness": 0.863260536453459, + "pair_accuracy": 0.8929777777777778, + "positive_goodness": 4.308837576124403, + "runtime_seconds": 10.068115711212158 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.5183581573486328, + "lr": 0.03, + "negative_goodness": 0.7757959654914008, + "pair_accuracy": 0.9387555555555556, + "positive_goodness": 5.020672807481554, + "runtime_seconds": 10.0659658908844 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.5910618380970425, + "lr": 0.03, + "negative_goodness": 0.8475605146408081, + "pair_accuracy": 0.9246444444444445, + "positive_goodness": 4.576875430128309, + "runtime_seconds": 10.088855028152466 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.610367955160141, + "lr": 0.03, + "negative_goodness": 0.8323906297895644, + "pair_accuracy": 0.9023777777777777, + "positive_goodness": 4.565001979827881, + "runtime_seconds": 10.070333003997803 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.5436448327382406, + "lr": 0.03, + "negative_goodness": 0.757046722539266, + "pair_accuracy": 0.9256888888835907, + "positive_goodness": 4.8093398515701296, + "runtime_seconds": 10.075289249420166 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.5423519087632497, + "lr": 0.03, + "negative_goodness": 0.7400590353647868, + "pair_accuracy": 0.9283777777777777, + "positive_goodness": 4.89999448682997, + "runtime_seconds": 10.076465129852295 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.5914755151112875, + "lr": 0.03, + "negative_goodness": 0.8227116187731425, + "pair_accuracy": 0.9181333333439298, + "positive_goodness": 4.591641405614217, + "runtime_seconds": 10.072831630706787 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5640851142883301, + "lr": 0.03, + "negative_goodness": 0.771505237642924, + "pair_accuracy": 0.9194666666640176, + "positive_goodness": 4.85873080751631, + "runtime_seconds": 10.066692590713501 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6184421784586377, + "lr": 0.03, + "negative_goodness": 0.8269205969783995, + "pair_accuracy": 0.9096222222222222, + "positive_goodness": 4.4850337091234, + "runtime_seconds": 10.075336217880249 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.572581745783488, + "lr": 0.03, + "negative_goodness": 0.7949862522655063, + "pair_accuracy": 0.9179555555555555, + "positive_goodness": 4.622385264460246, + "runtime_seconds": 10.062690734863281 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.5761285832087198, + "lr": 0.03, + "negative_goodness": 0.8141102801216973, + "pair_accuracy": 0.9159777777777778, + "positive_goodness": 4.688251463402642, + "runtime_seconds": 10.070357322692871 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5997178022940953, + "lr": 0.03, + "negative_goodness": 0.8259070762395859, + "pair_accuracy": 0.9182444444444444, + "positive_goodness": 4.621902669355604, + "runtime_seconds": 10.070482730865479 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5994257957670424, + "lr": 0.03, + "negative_goodness": 0.8105767558627659, + "pair_accuracy": 0.9043777777777777, + "positive_goodness": 4.528519570075141, + "runtime_seconds": 10.064057111740112 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5293665364371406, + "lr": 0.03, + "negative_goodness": 0.7204504658699036, + "pair_accuracy": 0.9235555555555556, + "positive_goodness": 5.011023088497586, + "runtime_seconds": 10.067633867263794 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5538254436651866, + "lr": 0.03, + "negative_goodness": 0.7695837338990635, + "pair_accuracy": 0.9274222222222223, + "positive_goodness": 5.0185204643249515, + "runtime_seconds": 10.065929889678955 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5663096979141236, + "lr": 0.03, + "negative_goodness": 0.7761912986861335, + "pair_accuracy": 0.9148222222275204, + "positive_goodness": 4.919061868667603, + "runtime_seconds": 10.075566291809082 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5695985800213283, + "lr": 0.03, + "negative_goodness": 0.752494679345025, + "pair_accuracy": 0.9134, + "positive_goodness": 4.81699115808275, + "runtime_seconds": 10.081359624862671 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.5886774830288357, + "lr": 0.03, + "negative_goodness": 0.7689340820709865, + "pair_accuracy": 0.9161555555555555, + "positive_goodness": 4.581326762390137, + "runtime_seconds": 10.073734283447266 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5840071121692657, + "lr": 0.03, + "negative_goodness": 0.8005987316237556, + "pair_accuracy": 0.9154222222222222, + "positive_goodness": 4.627646752760145, + "runtime_seconds": 10.0725576877594 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5640161395178901, + "lr": 0.03, + "negative_goodness": 0.7558196410179138, + "pair_accuracy": 0.9205777777797646, + "positive_goodness": 4.757303883806864, + "runtime_seconds": 10.065823554992676 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5807559632937114, + "lr": 0.03, + "negative_goodness": 0.7790652207957374, + "pair_accuracy": 0.9202444444444444, + "positive_goodness": 4.596275162124634, + "runtime_seconds": 10.082281112670898 + } + ], + "layer": 50 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.8916916262096829, + "lr": 0.03, + "negative_goodness": 1.2235718869527181, + "pair_accuracy": 0.8423555555555555, + "positive_goodness": 3.277624616071913, + "runtime_seconds": 10.094029903411865 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8801621314154731, + "lr": 0.03, + "negative_goodness": 1.2874601439158122, + "pair_accuracy": 0.8439111111112767, + "positive_goodness": 3.46462568458981, + "runtime_seconds": 10.091124296188354 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7782271981053882, + "lr": 0.03, + "negative_goodness": 1.1636875904983945, + "pair_accuracy": 0.8726, + "positive_goodness": 3.7533600045945907, + "runtime_seconds": 10.044451236724854 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7099221853256226, + "lr": 0.03, + "negative_goodness": 1.055033711804284, + "pair_accuracy": 0.8970666666560703, + "positive_goodness": 4.118368091773987, + "runtime_seconds": 10.047720193862915 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.8068508183638254, + "lr": 0.03, + "negative_goodness": 1.1543930247015424, + "pair_accuracy": 0.8626666666666667, + "positive_goodness": 3.8044279781765407, + "runtime_seconds": 10.042433023452759 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7437988771862454, + "lr": 0.03, + "negative_goodness": 1.1057939317491319, + "pair_accuracy": 0.8823111111217075, + "positive_goodness": 3.9048104088677302, + "runtime_seconds": 10.050527334213257 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7126632631116443, + "lr": 0.03, + "negative_goodness": 1.0569900338994132, + "pair_accuracy": 0.8988222222222222, + "positive_goodness": 4.100460506947836, + "runtime_seconds": 10.040066242218018 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7053868012958103, + "lr": 0.03, + "negative_goodness": 1.024080309979121, + "pair_accuracy": 0.9015777778201634, + "positive_goodness": 4.075339190059238, + "runtime_seconds": 10.04748249053955 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6931406771977743, + "lr": 0.03, + "negative_goodness": 1.0425356752501593, + "pair_accuracy": 0.9011777777671814, + "positive_goodness": 3.9995582905663385, + "runtime_seconds": 10.039661169052124 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6920218367682563, + "lr": 0.03, + "negative_goodness": 0.998586137787501, + "pair_accuracy": 0.8944, + "positive_goodness": 4.125454496977064, + "runtime_seconds": 10.076419591903687 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7179563438097636, + "lr": 0.03, + "negative_goodness": 1.0376373873021867, + "pair_accuracy": 0.8908, + "positive_goodness": 4.030581722365485, + "runtime_seconds": 10.083015203475952 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6577165043751398, + "lr": 0.03, + "negative_goodness": 0.9520927327156067, + "pair_accuracy": 0.9063111111111111, + "positive_goodness": 4.303022258504232, + "runtime_seconds": 10.029913425445557 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6593450240347121, + "lr": 0.03, + "negative_goodness": 0.9611270493189494, + "pair_accuracy": 0.9068666666666667, + "positive_goodness": 4.263143110487196, + "runtime_seconds": 9.795397996902466 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6598168652004666, + "lr": 0.03, + "negative_goodness": 0.9624421834415859, + "pair_accuracy": 0.9056666666507721, + "positive_goodness": 4.264582021713257, + "runtime_seconds": 9.987699031829834 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6861370947731866, + "lr": 0.03, + "negative_goodness": 0.9768257255660163, + "pair_accuracy": 0.8928, + "positive_goodness": 4.148165478346083, + "runtime_seconds": 9.9945068359375 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6395842588106792, + "lr": 0.03, + "negative_goodness": 0.9463875416013929, + "pair_accuracy": 0.9119555555873446, + "positive_goodness": 4.304518228997124, + "runtime_seconds": 9.990309000015259 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6264479235755073, + "lr": 0.03, + "negative_goodness": 0.9241142172124651, + "pair_accuracy": 0.9155555555635029, + "positive_goodness": 4.385504115147061, + "runtime_seconds": 9.990079879760742 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6472865338616901, + "lr": 0.03, + "negative_goodness": 0.9396644218285879, + "pair_accuracy": 0.9119333333333334, + "positive_goodness": 4.361950411563449, + "runtime_seconds": 9.994667768478394 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6599823800298903, + "lr": 0.03, + "negative_goodness": 0.8997915664858288, + "pair_accuracy": 0.8968, + "positive_goodness": 4.3186571550157336, + "runtime_seconds": 9.988005876541138 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6195765741348267, + "lr": 0.03, + "negative_goodness": 0.8919681346416474, + "pair_accuracy": 0.9130888888888888, + "positive_goodness": 4.404558177778456, + "runtime_seconds": 9.987842082977295 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6563851802984874, + "lr": 0.03, + "negative_goodness": 0.9373459913624658, + "pair_accuracy": 0.9037333333333334, + "positive_goodness": 4.2269506585015195, + "runtime_seconds": 9.987286567687988 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6119672870477041, + "lr": 0.03, + "negative_goodness": 0.8608438099304835, + "pair_accuracy": 0.9146888888888889, + "positive_goodness": 4.493078649308947, + "runtime_seconds": 10.004509210586548 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6938771210617489, + "lr": 0.03, + "negative_goodness": 0.982613749730587, + "pair_accuracy": 0.8922444444444444, + "positive_goodness": 4.127942349497477, + "runtime_seconds": 10.022331476211548 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6152054488446977, + "lr": 0.03, + "negative_goodness": 0.9069699532932706, + "pair_accuracy": 0.9212, + "positive_goodness": 4.390490487331814, + "runtime_seconds": 10.049613237380981 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6033343175040351, + "lr": 0.03, + "negative_goodness": 0.8740071310308245, + "pair_accuracy": 0.9234222222116258, + "positive_goodness": 4.5692776895311145, + "runtime_seconds": 10.045437574386597 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6055138310803307, + "lr": 0.03, + "negative_goodness": 0.8379862177451451, + "pair_accuracy": 0.9206444444338481, + "positive_goodness": 4.5478488578796386, + "runtime_seconds": 10.044795274734497 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.5896425144195556, + "lr": 0.03, + "negative_goodness": 0.8378793379677667, + "pair_accuracy": 0.9139777777989705, + "positive_goodness": 4.592958962843153, + "runtime_seconds": 10.050657272338867 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6157212405628628, + "lr": 0.03, + "negative_goodness": 0.8552991679588954, + "pair_accuracy": 0.9097777777777778, + "positive_goodness": 4.50022475827535, + "runtime_seconds": 10.048592805862427 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.6480584205839369, + "lr": 0.03, + "negative_goodness": 0.8610575610346264, + "pair_accuracy": 0.9038222222222222, + "positive_goodness": 4.287657062996758, + "runtime_seconds": 10.046785593032837 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6681474072403378, + "lr": 0.03, + "negative_goodness": 0.9362824473089641, + "pair_accuracy": 0.8977555555555555, + "positive_goodness": 4.234708596462673, + "runtime_seconds": 10.047510147094727 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.629091486231486, + "lr": 0.03, + "negative_goodness": 0.880902905008528, + "pair_accuracy": 0.9112, + "positive_goodness": 4.411175779872471, + "runtime_seconds": 10.036160707473755 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5875038864188724, + "lr": 0.03, + "negative_goodness": 0.8212243908272849, + "pair_accuracy": 0.9152, + "positive_goodness": 4.607126539018419, + "runtime_seconds": 10.022600173950195 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6294390913009643, + "lr": 0.03, + "negative_goodness": 0.8555995503399108, + "pair_accuracy": 0.9126, + "positive_goodness": 4.347152785831027, + "runtime_seconds": 10.026370525360107 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6741595644844903, + "lr": 0.03, + "negative_goodness": 0.915696722290251, + "pair_accuracy": 0.8945777778307596, + "positive_goodness": 4.243603065003289, + "runtime_seconds": 9.843211889266968 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.5785855993853675, + "lr": 0.03, + "negative_goodness": 0.8352880256679323, + "pair_accuracy": 0.9282444444444444, + "positive_goodness": 4.516455561743842, + "runtime_seconds": 10.00782036781311 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5963580963558621, + "lr": 0.03, + "negative_goodness": 0.8372634417374929, + "pair_accuracy": 0.9156, + "positive_goodness": 4.65563289135827, + "runtime_seconds": 10.071914196014404 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6413373706181844, + "lr": 0.03, + "negative_goodness": 0.9062503250731362, + "pair_accuracy": 0.9087777777883742, + "positive_goodness": 4.304126962365045, + "runtime_seconds": 10.047698020935059 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5715888849894206, + "lr": 0.03, + "negative_goodness": 0.8214134670628441, + "pair_accuracy": 0.9300666666348775, + "positive_goodness": 4.62971503507826, + "runtime_seconds": 10.047661066055298 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6094175758573744, + "lr": 0.03, + "negative_goodness": 0.8267737293508318, + "pair_accuracy": 0.9082666666666667, + "positive_goodness": 4.546088009558783, + "runtime_seconds": 10.050795555114746 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5436757520675659, + "lr": 0.03, + "negative_goodness": 0.7642697678142124, + "pair_accuracy": 0.9368888888888889, + "positive_goodness": 4.897004393810696, + "runtime_seconds": 10.051849603652954 + } + ], + "layer": 51 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9557259690072801, + "lr": 0.03, + "negative_goodness": 1.4865344363424513, + "pair_accuracy": 0.8629555555555556, + "positive_goodness": 3.6184927385965984, + "runtime_seconds": 10.06769847869873 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8219555794768864, + "lr": 0.03, + "negative_goodness": 1.2903196554819742, + "pair_accuracy": 0.8778, + "positive_goodness": 3.858036119121975, + "runtime_seconds": 10.08444356918335 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.8830483508427938, + "lr": 0.03, + "negative_goodness": 1.3619267497168648, + "pair_accuracy": 0.8547777777989706, + "positive_goodness": 3.436230291112264, + "runtime_seconds": 10.061748266220093 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.8066341383245256, + "lr": 0.03, + "negative_goodness": 1.2473411675665114, + "pair_accuracy": 0.8755555555555555, + "positive_goodness": 3.7521911625332303, + "runtime_seconds": 10.072533369064331 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7798412339740329, + "lr": 0.03, + "negative_goodness": 1.2198930560006036, + "pair_accuracy": 0.8795111110899183, + "positive_goodness": 3.805536384624905, + "runtime_seconds": 10.068535327911377 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7680840172873603, + "lr": 0.03, + "negative_goodness": 1.1675408064312405, + "pair_accuracy": 0.871177777809567, + "positive_goodness": 3.859105820613437, + "runtime_seconds": 10.07304072380066 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.762499171813329, + "lr": 0.03, + "negative_goodness": 1.161603455066681, + "pair_accuracy": 0.8853555555343628, + "positive_goodness": 3.8439432194179957, + "runtime_seconds": 10.098217010498047 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7248866796917386, + "lr": 0.03, + "negative_goodness": 1.1079583627065024, + "pair_accuracy": 0.8889111111429002, + "positive_goodness": 3.9824595818837483, + "runtime_seconds": 10.119882106781006 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7290160112169054, + "lr": 0.03, + "negative_goodness": 1.1077965009901258, + "pair_accuracy": 0.8833111111164093, + "positive_goodness": 3.983990436108907, + "runtime_seconds": 10.115929365158081 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.7179092936648263, + "lr": 0.03, + "negative_goodness": 1.0821924878994624, + "pair_accuracy": 0.8916222222222222, + "positive_goodness": 3.9307518244425457, + "runtime_seconds": 10.122286319732666 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.7187846850713094, + "lr": 0.03, + "negative_goodness": 1.0966253268559774, + "pair_accuracy": 0.8872888888676961, + "positive_goodness": 3.9600891078101266, + "runtime_seconds": 10.12340497970581 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.7410257040871514, + "lr": 0.03, + "negative_goodness": 1.1070935099283854, + "pair_accuracy": 0.8802000000211928, + "positive_goodness": 3.936323635228475, + "runtime_seconds": 10.125559329986572 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6704885511504279, + "lr": 0.03, + "negative_goodness": 1.0207052899254694, + "pair_accuracy": 0.8971555555343628, + "positive_goodness": 4.303736881807115, + "runtime_seconds": 10.124876022338867 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.7239309618843927, + "lr": 0.03, + "negative_goodness": 1.0713069507598878, + "pair_accuracy": 0.8848222222646077, + "positive_goodness": 4.064977707884046, + "runtime_seconds": 10.123961687088013 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6675378434711032, + "lr": 0.03, + "negative_goodness": 0.9911460291544596, + "pair_accuracy": 0.8984666666984558, + "positive_goodness": 4.222072166188558, + "runtime_seconds": 10.103669881820679 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6409906976434919, + "lr": 0.03, + "negative_goodness": 1.0104387816005282, + "pair_accuracy": 0.9171555555555555, + "positive_goodness": 4.3052408507453075, + "runtime_seconds": 10.071081161499023 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6289795566876729, + "lr": 0.03, + "negative_goodness": 0.9423096930821737, + "pair_accuracy": 0.9099333333121405, + "positive_goodness": 4.516633827590942, + "runtime_seconds": 10.069024801254272 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.644540837987264, + "lr": 0.03, + "negative_goodness": 0.969155046166314, + "pair_accuracy": 0.9044222222752041, + "positive_goodness": 4.443670421685113, + "runtime_seconds": 10.071126461029053 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6823875209013621, + "lr": 0.03, + "negative_goodness": 0.9484774690945943, + "pair_accuracy": 0.8947777777777778, + "positive_goodness": 4.194467809549967, + "runtime_seconds": 10.065601825714111 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.665967687564426, + "lr": 0.03, + "negative_goodness": 0.9677410857306586, + "pair_accuracy": 0.8983555555237664, + "positive_goodness": 4.2335990572611495, + "runtime_seconds": 10.068625211715698 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6723186161677043, + "lr": 0.03, + "negative_goodness": 0.9574134540769789, + "pair_accuracy": 0.8908222222222222, + "positive_goodness": 4.13111358426412, + "runtime_seconds": 10.065691947937012 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6419587172879113, + "lr": 0.03, + "negative_goodness": 0.9564501709434721, + "pair_accuracy": 0.9109777777777778, + "positive_goodness": 4.351354719967312, + "runtime_seconds": 10.06331491470337 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6620844048579534, + "lr": 0.03, + "negative_goodness": 0.9669577015240988, + "pair_accuracy": 0.9060666666666667, + "positive_goodness": 4.241286279551188, + "runtime_seconds": 10.064308404922485 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6998578227096134, + "lr": 0.03, + "negative_goodness": 0.9880886892504163, + "pair_accuracy": 0.8863555555555556, + "positive_goodness": 4.049694043562147, + "runtime_seconds": 10.065243482589722 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6523337262736426, + "lr": 0.03, + "negative_goodness": 0.9519903821627299, + "pair_accuracy": 0.9040888888888889, + "positive_goodness": 4.24152026570638, + "runtime_seconds": 10.008589267730713 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6552094273090363, + "lr": 0.03, + "negative_goodness": 0.9475867868794335, + "pair_accuracy": 0.9087555555555555, + "positive_goodness": 4.2027712844848635, + "runtime_seconds": 10.00507664680481 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6009650703509649, + "lr": 0.03, + "negative_goodness": 0.8850744945804278, + "pair_accuracy": 0.9152222222222223, + "positive_goodness": 4.567711513264974, + "runtime_seconds": 10.00866985321045 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.6103424462238948, + "lr": 0.03, + "negative_goodness": 0.8861770941191249, + "pair_accuracy": 0.9131333333333334, + "positive_goodness": 4.534850696902805, + "runtime_seconds": 10.013022422790527 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.661388168504503, + "lr": 0.03, + "negative_goodness": 0.935618517197503, + "pair_accuracy": 0.8955777777989705, + "positive_goodness": 4.2640014832814535, + "runtime_seconds": 10.008787631988525 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6529282795164321, + "lr": 0.03, + "negative_goodness": 0.9317127757602268, + "pair_accuracy": 0.8959333333386316, + "positive_goodness": 4.302178327793545, + "runtime_seconds": 10.017077684402466 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6171348922305637, + "lr": 0.03, + "negative_goodness": 0.9063283411661783, + "pair_accuracy": 0.9099111111243566, + "positive_goodness": 4.383739192867279, + "runtime_seconds": 10.025922775268555 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6007845366107093, + "lr": 0.03, + "negative_goodness": 0.8810880542649163, + "pair_accuracy": 0.9222666666666667, + "positive_goodness": 4.483666107855902, + "runtime_seconds": 10.060091018676758 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5970280952692032, + "lr": 0.03, + "negative_goodness": 0.8551419611030154, + "pair_accuracy": 0.9170222222222222, + "positive_goodness": 4.601159931352403, + "runtime_seconds": 10.068086385726929 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.6164924780925115, + "lr": 0.03, + "negative_goodness": 0.8833399078104232, + "pair_accuracy": 0.9114666666666666, + "positive_goodness": 4.425194469536676, + "runtime_seconds": 10.031185865402222 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6498852258788215, + "lr": 0.03, + "negative_goodness": 0.9199689781930711, + "pair_accuracy": 0.9081555555555556, + "positive_goodness": 4.396459900241428, + "runtime_seconds": 10.010058164596558 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.6231957963413662, + "lr": 0.03, + "negative_goodness": 0.8923180164337158, + "pair_accuracy": 0.9083555555767483, + "positive_goodness": 4.419305969492594, + "runtime_seconds": 9.960160493850708 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6129528957472907, + "lr": 0.03, + "negative_goodness": 0.8618695213529799, + "pair_accuracy": 0.9099333333333334, + "positive_goodness": 4.603142702441746, + "runtime_seconds": 10.01974630355835 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.6263078727510241, + "lr": 0.03, + "negative_goodness": 0.8587322796503702, + "pair_accuracy": 0.9062666667090522, + "positive_goodness": 4.451288690121968, + "runtime_seconds": 10.018493175506592 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5635419694264729, + "lr": 0.03, + "negative_goodness": 0.8421482093916999, + "pair_accuracy": 0.9344222222381168, + "positive_goodness": 4.75666315313975, + "runtime_seconds": 10.007421016693115 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.6361039624876447, + "lr": 0.03, + "negative_goodness": 0.8626828548444642, + "pair_accuracy": 0.8913555555555556, + "positive_goodness": 4.4521284100850425, + "runtime_seconds": 10.08686375617981 + } + ], + "layer": 52 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9718790128919813, + "lr": 0.03, + "negative_goodness": 1.3510441356235081, + "pair_accuracy": 0.8151333333757188, + "positive_goodness": 3.0526271159278022, + "runtime_seconds": 10.092494249343872 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8496109322865804, + "lr": 0.03, + "negative_goodness": 1.259851583035787, + "pair_accuracy": 0.8513555555820466, + "positive_goodness": 3.5341094347847832, + "runtime_seconds": 9.994556188583374 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7416288950602213, + "lr": 0.03, + "negative_goodness": 1.1296082976871067, + "pair_accuracy": 0.883577777783076, + "positive_goodness": 3.957219272486369, + "runtime_seconds": 9.933800458908081 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7248642477300432, + "lr": 0.03, + "negative_goodness": 1.076265779765447, + "pair_accuracy": 0.8945777777777778, + "positive_goodness": 4.004423327975803, + "runtime_seconds": 10.100605726242065 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.790148381084866, + "lr": 0.03, + "negative_goodness": 1.1577402498986986, + "pair_accuracy": 0.8792666666666666, + "positive_goodness": 3.796155741712782, + "runtime_seconds": 10.011046171188354 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.7412220695442624, + "lr": 0.03, + "negative_goodness": 1.0655189945353403, + "pair_accuracy": 0.8893555555555556, + "positive_goodness": 3.8866528477138944, + "runtime_seconds": 9.981118202209473 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.7463502488560146, + "lr": 0.03, + "negative_goodness": 1.083470902411143, + "pair_accuracy": 0.8758666667196485, + "positive_goodness": 3.8914403356340195, + "runtime_seconds": 10.079550504684448 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.7480585777759552, + "lr": 0.03, + "negative_goodness": 1.0943528555181292, + "pair_accuracy": 0.8838222222222222, + "positive_goodness": 3.8668270648532443, + "runtime_seconds": 9.621582984924316 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.7394980255974664, + "lr": 0.03, + "negative_goodness": 1.0837025665654076, + "pair_accuracy": 0.8933333333121406, + "positive_goodness": 3.974250686921014, + "runtime_seconds": 9.993419408798218 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6718798721207513, + "lr": 0.03, + "negative_goodness": 0.9840706965446472, + "pair_accuracy": 0.9109333333015441, + "positive_goodness": 4.096597917599148, + "runtime_seconds": 9.846485614776611 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6862703913794623, + "lr": 0.03, + "negative_goodness": 0.9949673471132914, + "pair_accuracy": 0.9034222222228845, + "positive_goodness": 4.118851169925266, + "runtime_seconds": 10.065385818481445 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6531623085154428, + "lr": 0.03, + "negative_goodness": 0.9820671804322136, + "pair_accuracy": 0.9109777777777778, + "positive_goodness": 4.324530500369602, + "runtime_seconds": 9.684788465499878 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6543766511069403, + "lr": 0.03, + "negative_goodness": 0.9269744169606103, + "pair_accuracy": 0.9136888888941871, + "positive_goodness": 4.196119406572977, + "runtime_seconds": 9.848071575164795 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6766125249756707, + "lr": 0.03, + "negative_goodness": 0.9378234143257141, + "pair_accuracy": 0.8993111111217075, + "positive_goodness": 4.174975962363349, + "runtime_seconds": 9.697723388671875 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.6721136437601514, + "lr": 0.03, + "negative_goodness": 0.961543494436476, + "pair_accuracy": 0.9054666666666666, + "positive_goodness": 4.1395423909505205, + "runtime_seconds": 9.953272104263306 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6194746271981133, + "lr": 0.03, + "negative_goodness": 0.8648158974647522, + "pair_accuracy": 0.9108888888676961, + "positive_goodness": 4.43980187763638, + "runtime_seconds": 9.598949670791626 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.6352555414835612, + "lr": 0.03, + "negative_goodness": 0.9068665282673306, + "pair_accuracy": 0.912, + "positive_goodness": 4.389910551140043, + "runtime_seconds": 9.710328817367554 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6874113742881351, + "lr": 0.03, + "negative_goodness": 0.9548589927302467, + "pair_accuracy": 0.8892444444338481, + "positive_goodness": 4.078043294991387, + "runtime_seconds": 9.708877325057983 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.6376496504677667, + "lr": 0.03, + "negative_goodness": 0.9139197806199392, + "pair_accuracy": 0.9061555556085374, + "positive_goodness": 4.366563694636027, + "runtime_seconds": 10.010497808456421 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.631622070672777, + "lr": 0.03, + "negative_goodness": 0.9079685488541921, + "pair_accuracy": 0.9131777778095669, + "positive_goodness": 4.282988567903307, + "runtime_seconds": 9.704167366027832 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6150699621412489, + "lr": 0.03, + "negative_goodness": 0.8679924266550276, + "pair_accuracy": 0.9196444444232517, + "positive_goodness": 4.523173342386881, + "runtime_seconds": 9.852821350097656 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6355326179292466, + "lr": 0.03, + "negative_goodness": 0.8804945720222261, + "pair_accuracy": 0.9052666667090522, + "positive_goodness": 4.334465966415405, + "runtime_seconds": 9.841196537017822 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.6250092512660557, + "lr": 0.03, + "negative_goodness": 0.8794384945074717, + "pair_accuracy": 0.9124888888782925, + "positive_goodness": 4.422526327641805, + "runtime_seconds": 9.936088800430298 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6323445023086336, + "lr": 0.03, + "negative_goodness": 0.8904897312985526, + "pair_accuracy": 0.9059777777777778, + "positive_goodness": 4.39959243036906, + "runtime_seconds": 9.66298508644104 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6345611913151211, + "lr": 0.03, + "negative_goodness": 0.8671521504349179, + "pair_accuracy": 0.9061777777777777, + "positive_goodness": 4.359532369147407, + "runtime_seconds": 9.750511884689331 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6710732631789313, + "lr": 0.03, + "negative_goodness": 0.9364320003509522, + "pair_accuracy": 0.8891555555661519, + "positive_goodness": 4.224419231075711, + "runtime_seconds": 9.790742874145508 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6294779200659858, + "lr": 0.03, + "negative_goodness": 0.8965282597091463, + "pair_accuracy": 0.9159333333121406, + "positive_goodness": 4.394501925553216, + "runtime_seconds": 9.906727313995361 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5926749333911472, + "lr": 0.03, + "negative_goodness": 0.8322979968547821, + "pair_accuracy": 0.9233333333121406, + "positive_goodness": 4.523051480314467, + "runtime_seconds": 9.759467124938965 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.5953530943234762, + "lr": 0.03, + "negative_goodness": 0.8487500261598163, + "pair_accuracy": 0.9191999999894036, + "positive_goodness": 4.580622627894083, + "runtime_seconds": 9.668816804885864 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6216759985393948, + "lr": 0.03, + "negative_goodness": 0.8617629915449354, + "pair_accuracy": 0.9055777777777778, + "positive_goodness": 4.456602893659803, + "runtime_seconds": 9.79399299621582 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.6270539381557041, + "lr": 0.03, + "negative_goodness": 0.8674475597990884, + "pair_accuracy": 0.9052888888570998, + "positive_goodness": 4.4408058378007675, + "runtime_seconds": 9.865944147109985 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.6497213345315721, + "lr": 0.03, + "negative_goodness": 0.8860253891626994, + "pair_accuracy": 0.9066222222540113, + "positive_goodness": 4.264841134219699, + "runtime_seconds": 9.57270097732544 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.6478628385755751, + "lr": 0.03, + "negative_goodness": 0.8761018545044793, + "pair_accuracy": 0.8985111110899183, + "positive_goodness": 4.298050107023451, + "runtime_seconds": 9.696243047714233 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5726414273579915, + "lr": 0.03, + "negative_goodness": 0.7792402010891173, + "pair_accuracy": 0.9261333333333334, + "positive_goodness": 4.5869896021525065, + "runtime_seconds": 9.783829927444458 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6024058533615536, + "lr": 0.03, + "negative_goodness": 0.8513342796802521, + "pair_accuracy": 0.9102444444444444, + "positive_goodness": 4.477421525828044, + "runtime_seconds": 9.8998384475708 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5758690766652426, + "lr": 0.03, + "negative_goodness": 0.80571090531879, + "pair_accuracy": 0.9204888888941871, + "positive_goodness": 4.6717776040395105, + "runtime_seconds": 9.829283475875854 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6449668455574248, + "lr": 0.03, + "negative_goodness": 0.8706805886454052, + "pair_accuracy": 0.9014666666666666, + "positive_goodness": 4.356467363823785, + "runtime_seconds": 9.959268808364868 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5819087991714478, + "lr": 0.03, + "negative_goodness": 0.8065519719998042, + "pair_accuracy": 0.9193333333333333, + "positive_goodness": 4.750033147515191, + "runtime_seconds": 9.90379285812378 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.6394201956775454, + "lr": 0.03, + "negative_goodness": 0.8755332006778982, + "pair_accuracy": 0.9075777777777778, + "positive_goodness": 4.348800137159559, + "runtime_seconds": 9.989676237106323 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.5755850804355409, + "lr": 0.03, + "negative_goodness": 0.7802389313850138, + "pair_accuracy": 0.9215777777777778, + "positive_goodness": 4.6606920705159505, + "runtime_seconds": 9.804670572280884 + } + ], + "layer": 53 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 0.9807074271572961, + "lr": 0.03, + "negative_goodness": 1.4999519066386753, + "pair_accuracy": 0.8416666666666667, + "positive_goodness": 3.3484102884080675, + "runtime_seconds": 10.036877632141113 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 0.8130361848513286, + "lr": 0.03, + "negative_goodness": 1.2160580439885458, + "pair_accuracy": 0.8684222222222222, + "positive_goodness": 3.8970202912648517, + "runtime_seconds": 10.032211780548096 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 0.7887228790309694, + "lr": 0.03, + "negative_goodness": 1.1878752432982127, + "pair_accuracy": 0.8789111111111111, + "positive_goodness": 3.7453713707817924, + "runtime_seconds": 9.981094360351562 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 0.7168530240588717, + "lr": 0.03, + "negative_goodness": 1.0708329482184515, + "pair_accuracy": 0.891977777756585, + "positive_goodness": 4.147188596428765, + "runtime_seconds": 10.006554365158081 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 0.7424158500247532, + "lr": 0.03, + "negative_goodness": 1.1288277846548291, + "pair_accuracy": 0.8802444444338481, + "positive_goodness": 3.9260529023064508, + "runtime_seconds": 9.976880311965942 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 0.6924821253140767, + "lr": 0.03, + "negative_goodness": 1.0509499920845031, + "pair_accuracy": 0.8957111110899183, + "positive_goodness": 4.0846704053243, + "runtime_seconds": 9.977164268493652 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 0.6563846886899736, + "lr": 0.03, + "negative_goodness": 1.013778870042165, + "pair_accuracy": 0.9031777777777777, + "positive_goodness": 4.276834740532769, + "runtime_seconds": 9.76147723197937 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 0.6660584641138713, + "lr": 0.03, + "negative_goodness": 0.9720643954171074, + "pair_accuracy": 0.8950666666560703, + "positive_goodness": 4.268892664040459, + "runtime_seconds": 10.009721040725708 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 0.6510847702556186, + "lr": 0.03, + "negative_goodness": 0.985142549112108, + "pair_accuracy": 0.9080888889312744, + "positive_goodness": 4.35552965804206, + "runtime_seconds": 10.043035984039307 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 0.6921646865102979, + "lr": 0.03, + "negative_goodness": 1.0182068000157674, + "pair_accuracy": 0.8993111111111111, + "positive_goodness": 4.154021824391683, + "runtime_seconds": 9.945164442062378 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 0.6463984684626262, + "lr": 0.03, + "negative_goodness": 0.9608935300191244, + "pair_accuracy": 0.9027111111217075, + "positive_goodness": 4.395462994257609, + "runtime_seconds": 10.02266526222229 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 0.6677164799054464, + "lr": 0.03, + "negative_goodness": 0.9661255883905623, + "pair_accuracy": 0.9020222222222222, + "positive_goodness": 4.2232778763665095, + "runtime_seconds": 10.079427719116211 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 0.6703425888697306, + "lr": 0.03, + "negative_goodness": 0.9930842946158515, + "pair_accuracy": 0.8971111111058129, + "positive_goodness": 4.268312313312954, + "runtime_seconds": 10.118689060211182 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 0.6907692275895013, + "lr": 0.03, + "negative_goodness": 0.9726520744747585, + "pair_accuracy": 0.887511111079322, + "positive_goodness": 4.133011679712931, + "runtime_seconds": 10.136014223098755 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 0.609856798670027, + "lr": 0.03, + "negative_goodness": 0.9066604468133714, + "pair_accuracy": 0.9106222222646078, + "positive_goodness": 4.461973866907756, + "runtime_seconds": 10.119157075881958 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 0.6754365342722999, + "lr": 0.03, + "negative_goodness": 0.963351062933604, + "pair_accuracy": 0.9008222222222222, + "positive_goodness": 4.197352243126764, + "runtime_seconds": 10.1236572265625 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 0.639821836010615, + "lr": 0.03, + "negative_goodness": 0.9253360828611585, + "pair_accuracy": 0.9049777777777778, + "positive_goodness": 4.257168584272597, + "runtime_seconds": 10.134371757507324 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 0.6100702373080783, + "lr": 0.03, + "negative_goodness": 0.9000901824951172, + "pair_accuracy": 0.9121333333346578, + "positive_goodness": 4.497454149542914, + "runtime_seconds": 10.121986150741577 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 0.599796954758962, + "lr": 0.03, + "negative_goodness": 0.8847956112649705, + "pair_accuracy": 0.9184666666666667, + "positive_goodness": 4.4942666529337565, + "runtime_seconds": 10.113982915878296 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 0.6181610445234511, + "lr": 0.03, + "negative_goodness": 0.9079151033507453, + "pair_accuracy": 0.9115555555555556, + "positive_goodness": 4.436755153274536, + "runtime_seconds": 10.089767932891846 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 0.6077179144647387, + "lr": 0.03, + "negative_goodness": 0.893466213575999, + "pair_accuracy": 0.9162666666666667, + "positive_goodness": 4.438487059699164, + "runtime_seconds": 10.034184217453003 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 0.6458089850213793, + "lr": 0.03, + "negative_goodness": 0.9158478715260824, + "pair_accuracy": 0.9024666666560703, + "positive_goodness": 4.42747243147956, + "runtime_seconds": 10.119109392166138 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 0.5897651215129428, + "lr": 0.03, + "negative_goodness": 0.8748973689609104, + "pair_accuracy": 0.9209555555555555, + "positive_goodness": 4.479646542443169, + "runtime_seconds": 10.10434365272522 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 0.6350981562296549, + "lr": 0.03, + "negative_goodness": 0.9059914469718933, + "pair_accuracy": 0.9061999999841054, + "positive_goodness": 4.439764390458001, + "runtime_seconds": 10.010063409805298 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 0.6265923896577623, + "lr": 0.03, + "negative_goodness": 0.904791773739126, + "pair_accuracy": 0.9106, + "positive_goodness": 4.50167994817098, + "runtime_seconds": 10.041465044021606 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 0.6176078755802579, + "lr": 0.03, + "negative_goodness": 0.8897638054529826, + "pair_accuracy": 0.9160666667090522, + "positive_goodness": 4.42020427585178, + "runtime_seconds": 10.058037281036377 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 0.6441897880739635, + "lr": 0.03, + "negative_goodness": 0.8915969023452864, + "pair_accuracy": 0.8964666666666666, + "positive_goodness": 4.461563881598579, + "runtime_seconds": 9.988579273223877 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 0.5800856943130493, + "lr": 0.03, + "negative_goodness": 0.8301542021221585, + "pair_accuracy": 0.9161777777618832, + "positive_goodness": 4.775342433887058, + "runtime_seconds": 9.989960670471191 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 0.681165795411004, + "lr": 0.03, + "negative_goodness": 0.9199062249713473, + "pair_accuracy": 0.8937111111217075, + "positive_goodness": 4.277552214474148, + "runtime_seconds": 9.844154596328735 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 0.6098415383709802, + "lr": 0.03, + "negative_goodness": 0.8951451687574387, + "pair_accuracy": 0.9175555555555556, + "positive_goodness": 4.484138892873128, + "runtime_seconds": 10.031460285186768 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 0.5783637369632721, + "lr": 0.03, + "negative_goodness": 0.831114099544949, + "pair_accuracy": 0.9209555555555555, + "positive_goodness": 4.624740789922078, + "runtime_seconds": 10.014838218688965 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 0.5879253962410821, + "lr": 0.03, + "negative_goodness": 0.8038611711608039, + "pair_accuracy": 0.9078666666878594, + "positive_goodness": 4.663435381825765, + "runtime_seconds": 10.053936243057251 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 0.5620518515904744, + "lr": 0.03, + "negative_goodness": 0.7963704710218642, + "pair_accuracy": 0.9213333333757189, + "positive_goodness": 4.885161902597216, + "runtime_seconds": 10.016268491744995 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 0.5540232646094428, + "lr": 0.03, + "negative_goodness": 0.773684991285536, + "pair_accuracy": 0.9259111111217075, + "positive_goodness": 4.782418795098199, + "runtime_seconds": 10.031703233718872 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 0.6310140514161852, + "lr": 0.03, + "negative_goodness": 0.8799674922943115, + "pair_accuracy": 0.9025111111270057, + "positive_goodness": 4.507643565898471, + "runtime_seconds": 10.019968271255493 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 0.5956684907727772, + "lr": 0.03, + "negative_goodness": 0.8232087631702423, + "pair_accuracy": 0.9093111111111111, + "positive_goodness": 4.547465909025404, + "runtime_seconds": 10.034471988677979 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 0.6168594803651174, + "lr": 0.03, + "negative_goodness": 0.8669228552374575, + "pair_accuracy": 0.9097111111111111, + "positive_goodness": 4.560904757859972, + "runtime_seconds": 10.078142404556274 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 0.5976247637006972, + "lr": 0.03, + "negative_goodness": 0.816377407140202, + "pair_accuracy": 0.9116888888888889, + "positive_goodness": 4.622243157789442, + "runtime_seconds": 10.065387964248657 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 0.5976096093707615, + "lr": 0.03, + "negative_goodness": 0.8610959091795816, + "pair_accuracy": 0.9171111111111111, + "positive_goodness": 4.6870799832661945, + "runtime_seconds": 10.02001142501831 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 0.579174010292689, + "lr": 0.03, + "negative_goodness": 0.7979912240107854, + "pair_accuracy": 0.9161777777777778, + "positive_goodness": 4.678039666324191, + "runtime_seconds": 10.03408169746399 + } + ], + "layer": 54 + }, + { + "epochs": [ + { + "epoch": 1, + "examples": 45000, + "loss": 1.184197072177463, + "lr": 0.03, + "negative_goodness": 1.3534939605395, + "pair_accuracy": 0.8492444444444445, + "positive_goodness": 2.7945538678275215, + "runtime_seconds": 9.775387525558472 + }, + { + "epoch": 2, + "examples": 45000, + "loss": 1.1423126959694756, + "lr": 0.03, + "negative_goodness": 1.4092236820644803, + "pair_accuracy": 0.8524000000158946, + "positive_goodness": 2.940797165468004, + "runtime_seconds": 9.778346300125122 + }, + { + "epoch": 3, + "examples": 45000, + "loss": 1.1151776327874925, + "lr": 0.03, + "negative_goodness": 1.3359610687679715, + "pair_accuracy": 0.877, + "positive_goodness": 2.823753111775716, + "runtime_seconds": 9.774840593338013 + }, + { + "epoch": 4, + "examples": 45000, + "loss": 1.0267702130423653, + "lr": 0.03, + "negative_goodness": 1.236902877553304, + "pair_accuracy": 0.9089777777777778, + "positive_goodness": 3.254718428717719, + "runtime_seconds": 9.770963668823242 + }, + { + "epoch": 5, + "examples": 45000, + "loss": 1.0638780898412068, + "lr": 0.03, + "negative_goodness": 1.3004277753432592, + "pair_accuracy": 0.8862222222222222, + "positive_goodness": 3.1546159934149847, + "runtime_seconds": 9.771470069885254 + }, + { + "epoch": 6, + "examples": 45000, + "loss": 1.1263346552530924, + "lr": 0.03, + "negative_goodness": 1.3717342495388456, + "pair_accuracy": 0.8730222222116258, + "positive_goodness": 2.7517613819758098, + "runtime_seconds": 9.776463508605957 + }, + { + "epoch": 7, + "examples": 45000, + "loss": 1.0981660309473673, + "lr": 0.03, + "negative_goodness": 1.366488700580597, + "pair_accuracy": 0.8813111110899183, + "positive_goodness": 3.1017440055847167, + "runtime_seconds": 9.77240538597107 + }, + { + "epoch": 8, + "examples": 45000, + "loss": 1.0905553121990628, + "lr": 0.03, + "negative_goodness": 1.329091675037808, + "pair_accuracy": 0.8784444444444445, + "positive_goodness": 2.8984737918006047, + "runtime_seconds": 9.780826330184937 + }, + { + "epoch": 9, + "examples": 45000, + "loss": 1.093832874086168, + "lr": 0.03, + "negative_goodness": 1.3685844909667968, + "pair_accuracy": 0.873711111079322, + "positive_goodness": 2.945703710746765, + "runtime_seconds": 9.761581897735596 + }, + { + "epoch": 10, + "examples": 45000, + "loss": 1.1031283352533976, + "lr": 0.03, + "negative_goodness": 1.3643838286293877, + "pair_accuracy": 0.8654444444444445, + "positive_goodness": 2.8902267471313476, + "runtime_seconds": 9.762607097625732 + }, + { + "epoch": 11, + "examples": 45000, + "loss": 1.116275408914354, + "lr": 0.03, + "negative_goodness": 1.4166171488867865, + "pair_accuracy": 0.8648000000529819, + "positive_goodness": 2.7553805559370255, + "runtime_seconds": 9.799038648605347 + }, + { + "epoch": 12, + "examples": 45000, + "loss": 1.1229927423477173, + "lr": 0.03, + "negative_goodness": 1.4264613360087077, + "pair_accuracy": 0.8569111110899184, + "positive_goodness": 2.7742156440311008, + "runtime_seconds": 9.71019721031189 + }, + { + "epoch": 13, + "examples": 45000, + "loss": 1.106313990105523, + "lr": 0.03, + "negative_goodness": 1.372574581638972, + "pair_accuracy": 0.8615111111640931, + "positive_goodness": 2.7277152480655245, + "runtime_seconds": 9.671103715896606 + }, + { + "epoch": 14, + "examples": 45000, + "loss": 1.0745124628278944, + "lr": 0.03, + "negative_goodness": 1.3363565123240153, + "pair_accuracy": 0.8901777777698305, + "positive_goodness": 2.8931194282955595, + "runtime_seconds": 9.812143564224243 + }, + { + "epoch": 15, + "examples": 45000, + "loss": 1.106962527900272, + "lr": 0.03, + "negative_goodness": 1.4091783302307128, + "pair_accuracy": 0.8826222222116259, + "positive_goodness": 2.8020219415452745, + "runtime_seconds": 9.758358240127563 + }, + { + "epoch": 16, + "examples": 45000, + "loss": 1.0264011932346555, + "lr": 0.03, + "negative_goodness": 1.2492899574915568, + "pair_accuracy": 0.9042222222222223, + "positive_goodness": 3.021982100168864, + "runtime_seconds": 9.837210893630981 + }, + { + "epoch": 17, + "examples": 45000, + "loss": 1.08896471886105, + "lr": 0.03, + "negative_goodness": 1.3336359154807196, + "pair_accuracy": 0.8936000000317892, + "positive_goodness": 2.866050891918606, + "runtime_seconds": 9.832108736038208 + }, + { + "epoch": 18, + "examples": 45000, + "loss": 1.0629276325649686, + "lr": 0.03, + "negative_goodness": 1.2971882881376477, + "pair_accuracy": 0.8883777777671814, + "positive_goodness": 2.921773898612128, + "runtime_seconds": 9.847082138061523 + }, + { + "epoch": 19, + "examples": 45000, + "loss": 1.129939404466417, + "lr": 0.03, + "negative_goodness": 1.4075821552912393, + "pair_accuracy": 0.8699555556085374, + "positive_goodness": 2.8492314466052586, + "runtime_seconds": 9.82695746421814 + }, + { + "epoch": 20, + "examples": 45000, + "loss": 1.1219772945404052, + "lr": 0.03, + "negative_goodness": 1.3969622328440348, + "pair_accuracy": 0.8547111111217075, + "positive_goodness": 2.720562565867106, + "runtime_seconds": 9.850557088851929 + }, + { + "epoch": 21, + "examples": 45000, + "loss": 1.0430937149047852, + "lr": 0.03, + "negative_goodness": 1.2596888225237528, + "pair_accuracy": 0.8816888888570997, + "positive_goodness": 3.067155018488566, + "runtime_seconds": 9.849125862121582 + }, + { + "epoch": 22, + "examples": 45000, + "loss": 1.129721160252889, + "lr": 0.03, + "negative_goodness": 1.4573959629482693, + "pair_accuracy": 0.8826, + "positive_goodness": 2.72557011557685, + "runtime_seconds": 9.838758707046509 + }, + { + "epoch": 23, + "examples": 45000, + "loss": 1.0721709324412876, + "lr": 0.03, + "negative_goodness": 1.3606446496327718, + "pair_accuracy": 0.8961111111111111, + "positive_goodness": 2.8980208649105497, + "runtime_seconds": 9.839472532272339 + }, + { + "epoch": 24, + "examples": 45000, + "loss": 1.0590342574649387, + "lr": 0.03, + "negative_goodness": 1.2955707064734565, + "pair_accuracy": 0.8896666666666667, + "positive_goodness": 2.903395784801907, + "runtime_seconds": 9.84665560722351 + }, + { + "epoch": 25, + "examples": 45000, + "loss": 1.0687768292321098, + "lr": 0.03, + "negative_goodness": 1.3674302232530382, + "pair_accuracy": 0.8854888888888889, + "positive_goodness": 2.9371806815253363, + "runtime_seconds": 9.83845591545105 + }, + { + "epoch": 26, + "examples": 45000, + "loss": 1.0857497111320495, + "lr": 0.03, + "negative_goodness": 1.353075526216295, + "pair_accuracy": 0.8896888888888889, + "positive_goodness": 2.8680795538160537, + "runtime_seconds": 9.846652746200562 + }, + { + "epoch": 27, + "examples": 45000, + "loss": 1.0320108869128757, + "lr": 0.03, + "negative_goodness": 1.2962803907606337, + "pair_accuracy": 0.8926444444762336, + "positive_goodness": 2.9321901667488945, + "runtime_seconds": 9.841848611831665 + }, + { + "epoch": 28, + "examples": 45000, + "loss": 1.0658618065039316, + "lr": 0.03, + "negative_goodness": 1.3228056475109524, + "pair_accuracy": 0.9113777777777777, + "positive_goodness": 3.0528482829623753, + "runtime_seconds": 9.842699527740479 + }, + { + "epoch": 29, + "examples": 45000, + "loss": 1.107486102273729, + "lr": 0.03, + "negative_goodness": 1.363236610115899, + "pair_accuracy": 0.878000000052982, + "positive_goodness": 2.7865187793307835, + "runtime_seconds": 9.847919940948486 + }, + { + "epoch": 30, + "examples": 45000, + "loss": 1.0292572712580363, + "lr": 0.03, + "negative_goodness": 1.3128881837738884, + "pair_accuracy": 0.9016888888888889, + "positive_goodness": 3.026307892311944, + "runtime_seconds": 9.845136642456055 + }, + { + "epoch": 31, + "examples": 45000, + "loss": 1.1028423228581747, + "lr": 0.03, + "negative_goodness": 1.3455672025044758, + "pair_accuracy": 0.8734222222222222, + "positive_goodness": 2.753966315460205, + "runtime_seconds": 9.842769622802734 + }, + { + "epoch": 32, + "examples": 45000, + "loss": 1.0995811727735731, + "lr": 0.03, + "negative_goodness": 1.3632474960327148, + "pair_accuracy": 0.8721333333757189, + "positive_goodness": 2.7637847128232322, + "runtime_seconds": 9.84635066986084 + }, + { + "epoch": 33, + "examples": 45000, + "loss": 1.0494107941097683, + "lr": 0.03, + "negative_goodness": 1.2981311967637803, + "pair_accuracy": 0.8899777777777778, + "positive_goodness": 2.9735177730242413, + "runtime_seconds": 9.846891164779663 + }, + { + "epoch": 34, + "examples": 45000, + "loss": 1.029825587187873, + "lr": 0.03, + "negative_goodness": 1.3036056589444478, + "pair_accuracy": 0.8936444444444445, + "positive_goodness": 3.099489004908668, + "runtime_seconds": 9.83812165260315 + }, + { + "epoch": 35, + "examples": 45000, + "loss": 1.0700099790361193, + "lr": 0.03, + "negative_goodness": 1.35077005343967, + "pair_accuracy": 0.8826000000211928, + "positive_goodness": 2.874786649915907, + "runtime_seconds": 9.84872579574585 + }, + { + "epoch": 36, + "examples": 45000, + "loss": 1.0482548462020027, + "lr": 0.03, + "negative_goodness": 1.2838380948490566, + "pair_accuracy": 0.9033555555555556, + "positive_goodness": 2.984557406234741, + "runtime_seconds": 9.840322494506836 + }, + { + "epoch": 37, + "examples": 45000, + "loss": 1.0628106483565436, + "lr": 0.03, + "negative_goodness": 1.34119630590015, + "pair_accuracy": 0.8997555555555555, + "positive_goodness": 2.9668071258544924, + "runtime_seconds": 9.852153301239014 + }, + { + "epoch": 38, + "examples": 45000, + "loss": 1.0685276712311638, + "lr": 0.03, + "negative_goodness": 1.3287960688273113, + "pair_accuracy": 0.8903111111005148, + "positive_goodness": 2.873085006629096, + "runtime_seconds": 9.85128116607666 + }, + { + "epoch": 39, + "examples": 45000, + "loss": 1.0555061865912543, + "lr": 0.03, + "negative_goodness": 1.2827911155700684, + "pair_accuracy": 0.8979999999788072, + "positive_goodness": 3.080144876819187, + "runtime_seconds": 9.853239059448242 + }, + { + "epoch": 40, + "examples": 45000, + "loss": 1.0406131548351711, + "lr": 0.03, + "negative_goodness": 1.3308772464964125, + "pair_accuracy": 0.909844444433848, + "positive_goodness": 3.0667766004456416, + "runtime_seconds": 9.84464406967163 + } + ], + "layer": 55 + } + ], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 23673.216548919678, + "work": { + "candidate_label_evaluation_presentations": 50000, + "completed_global_epochs": 2240, + "feedforward_example_passes": 201600000, + "forward_macs_per_example": 125485696, + "forward_parameter_count": 853018, + "local_target_backward_example_evaluations": 201600000, + "local_vjp_example_evaluations": 0, + "logical_task_loss_queries": 0, + "num_trainable_layers": 56, + "ordinary_training_examples": 100800000, + "ordinary_validation_examples": 5000, + "relaxation_example_passes": 0, + "training_example_presentations": 201600000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json.manifest.json new file mode 100644 index 0000000..ab3281a --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json.manifest.json @@ -0,0 +1,130 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "ff", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "40", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.03", + "--output_lr", + "0.03", + "--lr_schedule", + "constant", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--ff_threshold", + "2.0", + "--ff_score_from_layer", + "1" + ], + "completed_unix_time": 1786404648.5199583, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 23686.061631202698, + "epochs": 40, + "experiment_name": "resnet-r2-ff-d56-lr0p03", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-b520421c-b120-70f0-eae8-1ff79ad97bf5", + "total_memory_mib": 49140 + }, + "lr_schedule": "constant", + "method": "ff", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-ff-d56-lr0p03.json", + "output_exists": true, + "output_sha256": "9b7724301a9ea9ca9b938355594c03d1f5d06bd7e77410c5b27ecac4344445b7", + "rate": 0.03, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json b/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json new file mode 100644 index 0000000..1bbd577 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json @@ -0,0 +1,1565 @@ +{ + "architecture": { + "base_width": 16, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 269722, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 100, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.003, + "lr_gamma": 0.1, + "lr_milestones": "60,90", + "lr_schedule": "pepita", + "max_steps": 0, + "method": "pepita", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json", + "output_lr": 0.003, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.2540517320844864, + "train_seconds": 5.846177339553833, + "validation": { + "accuracy": 0.2154, + "examples": 5000, + "loss": 2.17969853515625 + }, + "validation_seconds": 0.12939953804016113 + }, + { + "epoch": 2, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.170380486128065, + "train_seconds": 5.499620676040649, + "validation": { + "accuracy": 0.2038, + "examples": 5000, + "loss": 2.134754833984375 + }, + "validation_seconds": 0.10496997833251953 + }, + { + "epoch": 3, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1359980241563585, + "train_seconds": 5.497474431991577, + "validation": { + "accuracy": 0.225, + "examples": 5000, + "loss": 2.09927607421875 + }, + "validation_seconds": 0.10551118850708008 + }, + { + "epoch": 4, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1131656904008653, + "train_seconds": 5.495712995529175, + "validation": { + "accuracy": 0.2192, + "examples": 5000, + "loss": 2.096121337890625 + }, + "validation_seconds": 0.10497379302978516 + }, + { + "epoch": 5, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.096841954845852, + "train_seconds": 5.500979900360107, + "validation": { + "accuracy": 0.2506, + "examples": 5000, + "loss": 2.0687662109375 + }, + "validation_seconds": 0.10506653785705566 + }, + { + "epoch": 6, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.08288049314287, + "train_seconds": 5.502379894256592, + "validation": { + "accuracy": 0.2552, + "examples": 5000, + "loss": 2.0530152099609373 + }, + "validation_seconds": 0.10521101951599121 + }, + { + "epoch": 7, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.073017496617635, + "train_seconds": 5.497807741165161, + "validation": { + "accuracy": 0.2366, + "examples": 5000, + "loss": 2.05641904296875 + }, + "validation_seconds": 0.10515880584716797 + }, + { + "epoch": 8, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0659207894643146, + "train_seconds": 5.5049827098846436, + "validation": { + "accuracy": 0.257, + "examples": 5000, + "loss": 2.03794013671875 + }, + "validation_seconds": 0.10519862174987793 + }, + { + "epoch": 9, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0593762793223065, + "train_seconds": 5.4969775676727295, + "validation": { + "accuracy": 0.259, + "examples": 5000, + "loss": 2.0241550048828123 + }, + "validation_seconds": 0.10501766204833984 + }, + { + "epoch": 10, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0526318064371747, + "train_seconds": 5.5005998611450195, + "validation": { + "accuracy": 0.2634, + "examples": 5000, + "loss": 2.0274732177734376 + }, + "validation_seconds": 0.10511279106140137 + }, + { + "epoch": 11, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.047295862748888, + "train_seconds": 5.514113903045654, + "validation": { + "accuracy": 0.2494, + "examples": 5000, + "loss": 2.034632568359375 + }, + "validation_seconds": 0.10521459579467773 + }, + { + "epoch": 12, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.046622137429979, + "train_seconds": 5.512402534484863, + "validation": { + "accuracy": 0.2686, + "examples": 5000, + "loss": 2.011696142578125 + }, + "validation_seconds": 0.10520362854003906 + }, + { + "epoch": 13, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0403066360897486, + "train_seconds": 5.509901523590088, + "validation": { + "accuracy": 0.258, + "examples": 5000, + "loss": 2.012770654296875 + }, + "validation_seconds": 0.10524630546569824 + }, + { + "epoch": 14, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0382033768547907, + "train_seconds": 5.508563995361328, + "validation": { + "accuracy": 0.2696, + "examples": 5000, + "loss": 2.0040715087890626 + }, + "validation_seconds": 0.10525345802307129 + }, + { + "epoch": 15, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0371743593004017, + "train_seconds": 5.505054712295532, + "validation": { + "accuracy": 0.2662, + "examples": 5000, + "loss": 2.0049404296875 + }, + "validation_seconds": 0.1051628589630127 + }, + { + "epoch": 16, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0296005568610296, + "train_seconds": 5.501487493515015, + "validation": { + "accuracy": 0.2676, + "examples": 5000, + "loss": 1.9989622802734375 + }, + "validation_seconds": 0.10553455352783203 + }, + { + "epoch": 17, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.028386363093058, + "train_seconds": 5.505746603012085, + "validation": { + "accuracy": 0.2682, + "examples": 5000, + "loss": 2.0050867919921873 + }, + "validation_seconds": 0.10517692565917969 + }, + { + "epoch": 18, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0280178286234536, + "train_seconds": 5.502475023269653, + "validation": { + "accuracy": 0.2634, + "examples": 5000, + "loss": 2.004136962890625 + }, + "validation_seconds": 0.10525631904602051 + }, + { + "epoch": 19, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.022250784280565, + "train_seconds": 5.5056681632995605, + "validation": { + "accuracy": 0.2744, + "examples": 5000, + "loss": 1.9959945556640626 + }, + "validation_seconds": 0.10529446601867676 + }, + { + "epoch": 20, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0207669247521296, + "train_seconds": 5.503343820571899, + "validation": { + "accuracy": 0.272, + "examples": 5000, + "loss": 1.990040185546875 + }, + "validation_seconds": 0.10525321960449219 + }, + { + "epoch": 21, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0204448586357966, + "train_seconds": 5.503366708755493, + "validation": { + "accuracy": 0.2566, + "examples": 5000, + "loss": 2.01042177734375 + }, + "validation_seconds": 0.10516881942749023 + }, + { + "epoch": 22, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0200392368740507, + "train_seconds": 5.5079567432403564, + "validation": { + "accuracy": 0.2672, + "examples": 5000, + "loss": 1.998391650390625 + }, + "validation_seconds": 0.10535097122192383 + }, + { + "epoch": 23, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0179887710571287, + "train_seconds": 5.508881568908691, + "validation": { + "accuracy": 0.2746, + "examples": 5000, + "loss": 1.9914981689453124 + }, + "validation_seconds": 0.1053166389465332 + }, + { + "epoch": 24, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.012541169950697, + "train_seconds": 5.511257648468018, + "validation": { + "accuracy": 0.2688, + "examples": 5000, + "loss": 1.9919697998046875 + }, + "validation_seconds": 0.10536813735961914 + }, + { + "epoch": 25, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0146305094189114, + "train_seconds": 5.506036281585693, + "validation": { + "accuracy": 0.2658, + "examples": 5000, + "loss": 1.9949030029296875 + }, + "validation_seconds": 0.10530686378479004 + }, + { + "epoch": 26, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.010540577167935, + "train_seconds": 5.503469944000244, + "validation": { + "accuracy": 0.2656, + "examples": 5000, + "loss": 1.995817578125 + }, + "validation_seconds": 0.10513448715209961 + }, + { + "epoch": 27, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.011019580438402, + "train_seconds": 5.504186153411865, + "validation": { + "accuracy": 0.2672, + "examples": 5000, + "loss": 1.9920932861328124 + }, + "validation_seconds": 0.10550999641418457 + }, + { + "epoch": 28, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0107565660264757, + "train_seconds": 5.501965761184692, + "validation": { + "accuracy": 0.2562, + "examples": 5000, + "loss": 2.00138916015625 + }, + "validation_seconds": 0.10556864738464355 + }, + { + "epoch": 29, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0093017225901284, + "train_seconds": 5.500823736190796, + "validation": { + "accuracy": 0.2646, + "examples": 5000, + "loss": 1.9926986083984375 + }, + "validation_seconds": 0.1052248477935791 + }, + { + "epoch": 30, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0063924302630953, + "train_seconds": 5.504383563995361, + "validation": { + "accuracy": 0.2624, + "examples": 5000, + "loss": 1.9840763671875 + }, + "validation_seconds": 0.10538268089294434 + }, + { + "epoch": 31, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.005855017513699, + "train_seconds": 5.50493311882019, + "validation": { + "accuracy": 0.266, + "examples": 5000, + "loss": 1.9905255859375 + }, + "validation_seconds": 0.1052553653717041 + }, + { + "epoch": 32, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0042439441257054, + "train_seconds": 5.5027971267700195, + "validation": { + "accuracy": 0.2678, + "examples": 5000, + "loss": 1.970855615234375 + }, + "validation_seconds": 0.10536980628967285 + }, + { + "epoch": 33, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0055182597054375, + "train_seconds": 5.503118991851807, + "validation": { + "accuracy": 0.2644, + "examples": 5000, + "loss": 1.999814013671875 + }, + "validation_seconds": 0.1053767204284668 + }, + { + "epoch": 34, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0039610786014133, + "train_seconds": 5.502718448638916, + "validation": { + "accuracy": 0.2558, + "examples": 5000, + "loss": 1.9934460693359375 + }, + "validation_seconds": 0.10529065132141113 + }, + { + "epoch": 35, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.002173393376668, + "train_seconds": 5.504491090774536, + "validation": { + "accuracy": 0.2718, + "examples": 5000, + "loss": 1.983355615234375 + }, + "validation_seconds": 0.10515451431274414 + }, + { + "epoch": 36, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.999710160975986, + "train_seconds": 5.503596782684326, + "validation": { + "accuracy": 0.2654, + "examples": 5000, + "loss": 1.9914493896484375 + }, + "validation_seconds": 0.10546326637268066 + }, + { + "epoch": 37, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.002970977168613, + "train_seconds": 5.50464129447937, + "validation": { + "accuracy": 0.2574, + "examples": 5000, + "loss": 1.9867867919921876 + }, + "validation_seconds": 0.10530948638916016 + }, + { + "epoch": 38, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0020395337846546, + "train_seconds": 5.5078301429748535, + "validation": { + "accuracy": 0.2706, + "examples": 5000, + "loss": 1.9826528076171874 + }, + "validation_seconds": 0.10545110702514648 + }, + { + "epoch": 39, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9987340057584975, + "train_seconds": 5.504313707351685, + "validation": { + "accuracy": 0.2714, + "examples": 5000, + "loss": 1.9860088134765626 + }, + "validation_seconds": 0.10541486740112305 + }, + { + "epoch": 40, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9994566903856066, + "train_seconds": 5.507040977478027, + "validation": { + "accuracy": 0.2682, + "examples": 5000, + "loss": 1.9782498291015624 + }, + "validation_seconds": 0.10532188415527344 + }, + { + "epoch": 41, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0013496548334757, + "train_seconds": 5.507043361663818, + "validation": { + "accuracy": 0.2496, + "examples": 5000, + "loss": 1.9953440185546876 + }, + "validation_seconds": 0.10550904273986816 + }, + { + "epoch": 42, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9970213481055366, + "train_seconds": 5.501050710678101, + "validation": { + "accuracy": 0.2582, + "examples": 5000, + "loss": 1.980520703125 + }, + "validation_seconds": 0.10538554191589355 + }, + { + "epoch": 43, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0023596636454264, + "train_seconds": 5.504518508911133, + "validation": { + "accuracy": 0.276, + "examples": 5000, + "loss": 1.9724589599609375 + }, + "validation_seconds": 0.1053304672241211 + }, + { + "epoch": 44, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9987859845691256, + "train_seconds": 5.504904270172119, + "validation": { + "accuracy": 0.267, + "examples": 5000, + "loss": 1.991623046875 + }, + "validation_seconds": 0.10524177551269531 + }, + { + "epoch": 45, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0013457901000975, + "train_seconds": 5.505333662033081, + "validation": { + "accuracy": 0.2622, + "examples": 5000, + "loss": 1.987554638671875 + }, + "validation_seconds": 0.10548973083496094 + }, + { + "epoch": 46, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9991893960740832, + "train_seconds": 5.507464647293091, + "validation": { + "accuracy": 0.2706, + "examples": 5000, + "loss": 1.9903315673828126 + }, + "validation_seconds": 0.10526585578918457 + }, + { + "epoch": 47, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0000709929996066, + "train_seconds": 5.5044050216674805, + "validation": { + "accuracy": 0.2728, + "examples": 5000, + "loss": 1.9653149658203124 + }, + "validation_seconds": 0.10531759262084961 + }, + { + "epoch": 48, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0002532496558296, + "train_seconds": 5.506645917892456, + "validation": { + "accuracy": 0.2778, + "examples": 5000, + "loss": 1.9617325439453126 + }, + "validation_seconds": 0.10522890090942383 + }, + { + "epoch": 49, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.000072623655531, + "train_seconds": 5.511117935180664, + "validation": { + "accuracy": 0.2652, + "examples": 5000, + "loss": 1.982528076171875 + }, + "validation_seconds": 0.10529708862304688 + }, + { + "epoch": 50, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9980969663831922, + "train_seconds": 5.5062291622161865, + "validation": { + "accuracy": 0.2728, + "examples": 5000, + "loss": 1.97162822265625 + }, + "validation_seconds": 0.1052713394165039 + }, + { + "epoch": 51, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.994790125126309, + "train_seconds": 5.508324146270752, + "validation": { + "accuracy": 0.2596, + "examples": 5000, + "loss": 1.9888105224609376 + }, + "validation_seconds": 0.10519123077392578 + }, + { + "epoch": 52, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9956360507965087, + "train_seconds": 5.5053040981292725, + "validation": { + "accuracy": 0.2598, + "examples": 5000, + "loss": 1.987352001953125 + }, + "validation_seconds": 0.10519528388977051 + }, + { + "epoch": 53, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.999503876071506, + "train_seconds": 5.503100395202637, + "validation": { + "accuracy": 0.2622, + "examples": 5000, + "loss": 1.9869154541015626 + }, + "validation_seconds": 0.10521578788757324 + }, + { + "epoch": 54, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9960996131049262, + "train_seconds": 5.509793996810913, + "validation": { + "accuracy": 0.2582, + "examples": 5000, + "loss": 1.99108720703125 + }, + "validation_seconds": 0.10523152351379395 + }, + { + "epoch": 55, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9974913773854575, + "train_seconds": 5.506419658660889, + "validation": { + "accuracy": 0.2644, + "examples": 5000, + "loss": 1.9845252685546875 + }, + "validation_seconds": 0.10523486137390137 + }, + { + "epoch": 56, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9962633374955918, + "train_seconds": 5.5040602684021, + "validation": { + "accuracy": 0.263, + "examples": 5000, + "loss": 1.9842421875 + }, + "validation_seconds": 0.10512018203735352 + }, + { + "epoch": 57, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9980382384406197, + "train_seconds": 5.508415460586548, + "validation": { + "accuracy": 0.2684, + "examples": 5000, + "loss": 1.9884821533203125 + }, + "validation_seconds": 0.10535335540771484 + }, + { + "epoch": 58, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9994783101823594, + "train_seconds": 5.509284257888794, + "validation": { + "accuracy": 0.2598, + "examples": 5000, + "loss": 1.9920580078125 + }, + "validation_seconds": 0.10525155067443848 + }, + { + "epoch": 59, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.994873679860433, + "train_seconds": 5.507220506668091, + "validation": { + "accuracy": 0.2646, + "examples": 5000, + "loss": 1.984413818359375 + }, + "validation_seconds": 0.10529899597167969 + }, + { + "epoch": 60, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 1.9963021400875516, + "train_seconds": 5.505999326705933, + "validation": { + "accuracy": 0.2654, + "examples": 5000, + "loss": 1.97801142578125 + }, + "validation_seconds": 0.10530734062194824 + }, + { + "epoch": 61, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9900015704684788, + "train_seconds": 5.507628679275513, + "validation": { + "accuracy": 0.2764, + "examples": 5000, + "loss": 1.9735885986328126 + }, + "validation_seconds": 0.10532093048095703 + }, + { + "epoch": 62, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9889358832465278, + "train_seconds": 5.505821466445923, + "validation": { + "accuracy": 0.2746, + "examples": 5000, + "loss": 1.9731400634765626 + }, + "validation_seconds": 0.10542631149291992 + }, + { + "epoch": 63, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9910470107184517, + "train_seconds": 5.504168510437012, + "validation": { + "accuracy": 0.2706, + "examples": 5000, + "loss": 1.9786871826171875 + }, + "validation_seconds": 0.10542702674865723 + }, + { + "epoch": 64, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9888048695670233, + "train_seconds": 5.503101110458374, + "validation": { + "accuracy": 0.2722, + "examples": 5000, + "loss": 1.9800283447265625 + }, + "validation_seconds": 0.10519576072692871 + }, + { + "epoch": 65, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9899456718656752, + "train_seconds": 5.503441572189331, + "validation": { + "accuracy": 0.2718, + "examples": 5000, + "loss": 1.9744058837890626 + }, + "validation_seconds": 0.10547184944152832 + }, + { + "epoch": 66, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9914917253070408, + "train_seconds": 5.5037009716033936, + "validation": { + "accuracy": 0.2768, + "examples": 5000, + "loss": 1.970571484375 + }, + "validation_seconds": 0.10534429550170898 + }, + { + "epoch": 67, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9919023951212564, + "train_seconds": 5.503073453903198, + "validation": { + "accuracy": 0.2702, + "examples": 5000, + "loss": 1.9756346435546874 + }, + "validation_seconds": 0.10526371002197266 + }, + { + "epoch": 68, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9882253373675876, + "train_seconds": 5.506524085998535, + "validation": { + "accuracy": 0.2782, + "examples": 5000, + "loss": 1.9718321044921876 + }, + "validation_seconds": 0.10534501075744629 + }, + { + "epoch": 69, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9925236448076036, + "train_seconds": 5.504960536956787, + "validation": { + "accuracy": 0.2716, + "examples": 5000, + "loss": 1.9783636962890625 + }, + "validation_seconds": 0.10545825958251953 + }, + { + "epoch": 70, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9897801370620727, + "train_seconds": 5.505310773849487, + "validation": { + "accuracy": 0.269, + "examples": 5000, + "loss": 1.9767344482421876 + }, + "validation_seconds": 0.10528755187988281 + }, + { + "epoch": 71, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9880535870022245, + "train_seconds": 5.503842830657959, + "validation": { + "accuracy": 0.2676, + "examples": 5000, + "loss": 1.9813978271484376 + }, + "validation_seconds": 0.10532116889953613 + }, + { + "epoch": 72, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9871454840554132, + "train_seconds": 5.511318206787109, + "validation": { + "accuracy": 0.2758, + "examples": 5000, + "loss": 1.97433564453125 + }, + "validation_seconds": 0.10527610778808594 + }, + { + "epoch": 73, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9925464500427246, + "train_seconds": 5.5087480545043945, + "validation": { + "accuracy": 0.2724, + "examples": 5000, + "loss": 1.978366552734375 + }, + "validation_seconds": 0.10521245002746582 + }, + { + "epoch": 74, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9864715037875704, + "train_seconds": 5.508193492889404, + "validation": { + "accuracy": 0.2686, + "examples": 5000, + "loss": 1.9754033935546875 + }, + "validation_seconds": 0.10522222518920898 + }, + { + "epoch": 75, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9897771712409125, + "train_seconds": 5.50649881362915, + "validation": { + "accuracy": 0.2734, + "examples": 5000, + "loss": 1.974899462890625 + }, + "validation_seconds": 0.10533857345581055 + }, + { + "epoch": 76, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9886419744915433, + "train_seconds": 5.505634069442749, + "validation": { + "accuracy": 0.2692, + "examples": 5000, + "loss": 1.977445703125 + }, + "validation_seconds": 0.10537195205688477 + }, + { + "epoch": 77, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9885343943066067, + "train_seconds": 5.503732204437256, + "validation": { + "accuracy": 0.2682, + "examples": 5000, + "loss": 1.9802877685546876 + }, + "validation_seconds": 0.10513567924499512 + }, + { + "epoch": 78, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.993202943250868, + "train_seconds": 5.506843566894531, + "validation": { + "accuracy": 0.2686, + "examples": 5000, + "loss": 1.9763575439453125 + }, + "validation_seconds": 0.10518097877502441 + }, + { + "epoch": 79, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9885233533859252, + "train_seconds": 5.502397775650024, + "validation": { + "accuracy": 0.2736, + "examples": 5000, + "loss": 1.971972607421875 + }, + "validation_seconds": 0.10559296607971191 + }, + { + "epoch": 80, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9916276458104452, + "train_seconds": 5.503775358200073, + "validation": { + "accuracy": 0.2702, + "examples": 5000, + "loss": 1.983391943359375 + }, + "validation_seconds": 0.10515046119689941 + }, + { + "epoch": 81, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9912612048678928, + "train_seconds": 5.502410888671875, + "validation": { + "accuracy": 0.2722, + "examples": 5000, + "loss": 1.9768909912109376 + }, + "validation_seconds": 0.10518693923950195 + }, + { + "epoch": 82, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9933317521413167, + "train_seconds": 5.505985498428345, + "validation": { + "accuracy": 0.2738, + "examples": 5000, + "loss": 1.975612646484375 + }, + "validation_seconds": 0.10521507263183594 + }, + { + "epoch": 83, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.992124314159817, + "train_seconds": 5.503434896469116, + "validation": { + "accuracy": 0.2694, + "examples": 5000, + "loss": 1.9838414794921875 + }, + "validation_seconds": 0.10535001754760742 + }, + { + "epoch": 84, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9881560234917535, + "train_seconds": 5.5040600299835205, + "validation": { + "accuracy": 0.2718, + "examples": 5000, + "loss": 1.981240087890625 + }, + "validation_seconds": 0.10520529747009277 + }, + { + "epoch": 85, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9894730068630642, + "train_seconds": 5.501953840255737, + "validation": { + "accuracy": 0.2672, + "examples": 5000, + "loss": 1.9822439208984375 + }, + "validation_seconds": 0.10531735420227051 + }, + { + "epoch": 86, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9889087203979492, + "train_seconds": 5.50821590423584, + "validation": { + "accuracy": 0.2714, + "examples": 5000, + "loss": 1.9796527099609376 + }, + "validation_seconds": 0.10519123077392578 + }, + { + "epoch": 87, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9899914241366916, + "train_seconds": 5.50803017616272, + "validation": { + "accuracy": 0.27, + "examples": 5000, + "loss": 1.981403369140625 + }, + "validation_seconds": 0.1051938533782959 + }, + { + "epoch": 88, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.989610216055976, + "train_seconds": 5.505961179733276, + "validation": { + "accuracy": 0.2696, + "examples": 5000, + "loss": 1.976666259765625 + }, + "validation_seconds": 0.10532689094543457 + }, + { + "epoch": 89, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.9914015740712483, + "train_seconds": 5.505902290344238, + "validation": { + "accuracy": 0.2742, + "examples": 5000, + "loss": 1.968590283203125 + }, + "validation_seconds": 0.10518217086791992 + }, + { + "epoch": 90, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 1.991297045856052, + "train_seconds": 5.503058910369873, + "validation": { + "accuracy": 0.2686, + "examples": 5000, + "loss": 1.97153271484375 + }, + "validation_seconds": 0.10520219802856445 + }, + { + "epoch": 91, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9905024444156223, + "train_seconds": 5.507503986358643, + "validation": { + "accuracy": 0.2682, + "examples": 5000, + "loss": 1.9812239013671875 + }, + "validation_seconds": 0.10525727272033691 + }, + { + "epoch": 92, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9875384762658013, + "train_seconds": 5.504276514053345, + "validation": { + "accuracy": 0.271, + "examples": 5000, + "loss": 1.9784233642578124 + }, + "validation_seconds": 0.10520076751708984 + }, + { + "epoch": 93, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9886907052993774, + "train_seconds": 5.507157325744629, + "validation": { + "accuracy": 0.2704, + "examples": 5000, + "loss": 1.9788807373046875 + }, + "validation_seconds": 0.10531425476074219 + }, + { + "epoch": 94, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9844931491851807, + "train_seconds": 5.5087127685546875, + "validation": { + "accuracy": 0.2722, + "examples": 5000, + "loss": 1.9788918701171876 + }, + "validation_seconds": 0.10535192489624023 + }, + { + "epoch": 95, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9873190053304037, + "train_seconds": 5.506003379821777, + "validation": { + "accuracy": 0.2748, + "examples": 5000, + "loss": 1.9743620849609376 + }, + "validation_seconds": 0.10539937019348145 + }, + { + "epoch": 96, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.984961536640591, + "train_seconds": 5.514376163482666, + "validation": { + "accuracy": 0.2724, + "examples": 5000, + "loss": 1.9778021240234376 + }, + "validation_seconds": 0.10521578788757324 + }, + { + "epoch": 97, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9883242013295492, + "train_seconds": 5.5068089962005615, + "validation": { + "accuracy": 0.2696, + "examples": 5000, + "loss": 1.9805218017578126 + }, + "validation_seconds": 0.10537600517272949 + }, + { + "epoch": 98, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9905134502834745, + "train_seconds": 5.501478672027588, + "validation": { + "accuracy": 0.272, + "examples": 5000, + "loss": 1.977109130859375 + }, + "validation_seconds": 0.10539984703063965 + }, + { + "epoch": 99, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9883133581373427, + "train_seconds": 5.503832578659058, + "validation": { + "accuracy": 0.2712, + "examples": 5000, + "loss": 1.97808974609375 + }, + "validation_seconds": 0.10522580146789551 + }, + { + "epoch": 100, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 1.9903361009385852, + "train_seconds": 5.5033111572265625, + "validation": { + "accuracy": 0.2726, + "examples": 5000, + "loss": 1.9752171875 + }, + "validation_seconds": 0.10524916648864746 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.2726, + "evaluation_seconds": 0.10563945770263672, + "examples": 5000, + "finite": true, + "loss": 1.9752171875 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2008731648, + "peak_memory_reserved_bytes": 2543845376, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 561.5238916873932, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 100, + "feedforward_example_passes": 14005000, + "forward_macs_per_example": 40551040, + "forward_parameter_count": 269722, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 0, + "logical_task_loss_queries": 0, + "num_trainable_layers": 20, + "ordinary_training_examples": 4500000, + "ordinary_validation_examples": 505000, + "relaxation_example_passes": 0, + "training_example_presentations": 9000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json.manifest.json new file mode 100644 index 0000000..7264144 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json.manifest.json @@ -0,0 +1,128 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "pepita", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "100", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.003", + "--output_lr", + "0.003", + "--lr_schedule", + "pepita", + "--lr_milestones", + "60,90", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--pepita_projection_scale", + "0.05" + ], + "completed_unix_time": 1787031147.2250133, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 565.1650466918945, + "epochs": 100, + "experiment_name": "resnet-r2-pepita-d20-lr0p003", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "pepita", + "method": "pepita", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d20-lr0p003.json", + "output_exists": true, + "output_sha256": "b9120fa19a170ae650e432067cbf3df388be3f1606f18489180e646b65ac70e2", + "rate": 0.003, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json b/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json new file mode 100644 index 0000000..4a0e2d3 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json @@ -0,0 +1,1565 @@ +{ + "architecture": { + "base_width": 16, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 464154, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 100, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.003, + "lr_gamma": 0.1, + "lr_milestones": "60,90", + "lr_schedule": "pepita", + "max_steps": 0, + "method": "pepita", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json", + "output_lr": 0.003, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.2907867482503255, + "train_seconds": 9.360012769699097, + "validation": { + "accuracy": 0.181, + "examples": 5000, + "loss": 2.190810546875 + }, + "validation_seconds": 0.9673621654510498 + }, + { + "epoch": 2, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1893652502695717, + "train_seconds": 8.754016399383545, + "validation": { + "accuracy": 0.1966, + "examples": 5000, + "loss": 2.148013818359375 + }, + "validation_seconds": 0.169417142868042 + }, + { + "epoch": 3, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.161732895490858, + "train_seconds": 8.748339891433716, + "validation": { + "accuracy": 0.204, + "examples": 5000, + "loss": 2.12567041015625 + }, + "validation_seconds": 0.1697547435760498 + }, + { + "epoch": 4, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.146513925764296, + "train_seconds": 8.752492904663086, + "validation": { + "accuracy": 0.2226, + "examples": 5000, + "loss": 2.121567333984375 + }, + "validation_seconds": 0.17016267776489258 + }, + { + "epoch": 5, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1360204981486004, + "train_seconds": 8.763130903244019, + "validation": { + "accuracy": 0.235, + "examples": 5000, + "loss": 2.091363623046875 + }, + "validation_seconds": 0.1700141429901123 + }, + { + "epoch": 6, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1298021606021456, + "train_seconds": 8.769134998321533, + "validation": { + "accuracy": 0.2434, + "examples": 5000, + "loss": 2.0821609375 + }, + "validation_seconds": 0.17003583908081055 + }, + { + "epoch": 7, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.123042589187622, + "train_seconds": 8.775540590286255, + "validation": { + "accuracy": 0.2244, + "examples": 5000, + "loss": 2.086983447265625 + }, + "validation_seconds": 0.16979694366455078 + }, + { + "epoch": 8, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.118494323179457, + "train_seconds": 8.769744873046875, + "validation": { + "accuracy": 0.2466, + "examples": 5000, + "loss": 2.068868017578125 + }, + "validation_seconds": 0.17055869102478027 + }, + { + "epoch": 9, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1129601199256047, + "train_seconds": 8.778553485870361, + "validation": { + "accuracy": 0.2376, + "examples": 5000, + "loss": 2.071684130859375 + }, + "validation_seconds": 0.17051935195922852 + }, + { + "epoch": 10, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1110865395439995, + "train_seconds": 8.786123275756836, + "validation": { + "accuracy": 0.2464, + "examples": 5000, + "loss": 2.063233203125 + }, + "validation_seconds": 0.17072343826293945 + }, + { + "epoch": 11, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1061982627020943, + "train_seconds": 8.78468656539917, + "validation": { + "accuracy": 0.2348, + "examples": 5000, + "loss": 2.0644167724609375 + }, + "validation_seconds": 0.17047452926635742 + }, + { + "epoch": 12, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.107127010981242, + "train_seconds": 8.783576250076294, + "validation": { + "accuracy": 0.25, + "examples": 5000, + "loss": 2.0543634765625 + }, + "validation_seconds": 0.17060136795043945 + }, + { + "epoch": 13, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1037176490783693, + "train_seconds": 8.785626411437988, + "validation": { + "accuracy": 0.231, + "examples": 5000, + "loss": 2.064813330078125 + }, + "validation_seconds": 0.17042326927185059 + }, + { + "epoch": 14, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.099651092529297, + "train_seconds": 8.749484539031982, + "validation": { + "accuracy": 0.2542, + "examples": 5000, + "loss": 2.048787548828125 + }, + "validation_seconds": 0.17104148864746094 + }, + { + "epoch": 15, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1024867318047415, + "train_seconds": 8.785282135009766, + "validation": { + "accuracy": 0.263, + "examples": 5000, + "loss": 2.0486366455078127 + }, + "validation_seconds": 0.17087507247924805 + }, + { + "epoch": 16, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.096451896031698, + "train_seconds": 8.786890745162964, + "validation": { + "accuracy": 0.2558, + "examples": 5000, + "loss": 2.0489378662109377 + }, + "validation_seconds": 0.17048311233520508 + }, + { + "epoch": 17, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0968072445339625, + "train_seconds": 8.784305572509766, + "validation": { + "accuracy": 0.2578, + "examples": 5000, + "loss": 2.034791357421875 + }, + "validation_seconds": 0.17049431800842285 + }, + { + "epoch": 18, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0995056786007353, + "train_seconds": 8.785536050796509, + "validation": { + "accuracy": 0.2574, + "examples": 5000, + "loss": 2.04157490234375 + }, + "validation_seconds": 0.17114782333374023 + }, + { + "epoch": 19, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0956084409925673, + "train_seconds": 8.790813207626343, + "validation": { + "accuracy": 0.256, + "examples": 5000, + "loss": 2.0414225830078125 + }, + "validation_seconds": 0.17094159126281738 + }, + { + "epoch": 20, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.093731904051039, + "train_seconds": 8.78307318687439, + "validation": { + "accuracy": 0.2562, + "examples": 5000, + "loss": 2.0405240966796874 + }, + "validation_seconds": 0.17032456398010254 + }, + { + "epoch": 21, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0907077075534395, + "train_seconds": 8.78516411781311, + "validation": { + "accuracy": 0.252, + "examples": 5000, + "loss": 2.054002294921875 + }, + "validation_seconds": 0.17098021507263184 + }, + { + "epoch": 22, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0961777904086643, + "train_seconds": 8.780037879943848, + "validation": { + "accuracy": 0.264, + "examples": 5000, + "loss": 2.04380771484375 + }, + "validation_seconds": 0.17042255401611328 + }, + { + "epoch": 23, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.093424896197849, + "train_seconds": 8.782354831695557, + "validation": { + "accuracy": 0.2602, + "examples": 5000, + "loss": 2.0328253173828124 + }, + "validation_seconds": 0.17090868949890137 + }, + { + "epoch": 24, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.089472702874078, + "train_seconds": 8.781430721282959, + "validation": { + "accuracy": 0.2324, + "examples": 5000, + "loss": 2.0593703125 + }, + "validation_seconds": 0.17047786712646484 + }, + { + "epoch": 25, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0928057125091555, + "train_seconds": 8.784738302230835, + "validation": { + "accuracy": 0.2614, + "examples": 5000, + "loss": 2.0374564697265627 + }, + "validation_seconds": 0.17136812210083008 + }, + { + "epoch": 26, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0903255541059704, + "train_seconds": 8.7819344997406, + "validation": { + "accuracy": 0.2682, + "examples": 5000, + "loss": 2.030443994140625 + }, + "validation_seconds": 0.17069506645202637 + }, + { + "epoch": 27, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0905855367448596, + "train_seconds": 8.782570362091064, + "validation": { + "accuracy": 0.2552, + "examples": 5000, + "loss": 2.038783154296875 + }, + "validation_seconds": 0.17075705528259277 + }, + { + "epoch": 28, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0896595900429618, + "train_seconds": 8.781562089920044, + "validation": { + "accuracy": 0.24, + "examples": 5000, + "loss": 2.05001591796875 + }, + "validation_seconds": 0.17116379737854004 + }, + { + "epoch": 29, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.088815062628852, + "train_seconds": 8.783803462982178, + "validation": { + "accuracy": 0.2532, + "examples": 5000, + "loss": 2.038257421875 + }, + "validation_seconds": 0.17058849334716797 + }, + { + "epoch": 30, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0916393721262616, + "train_seconds": 8.789171695709229, + "validation": { + "accuracy": 0.2564, + "examples": 5000, + "loss": 2.0344188232421874 + }, + "validation_seconds": 0.1707472801208496 + }, + { + "epoch": 31, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0937151625739205, + "train_seconds": 8.783771753311157, + "validation": { + "accuracy": 0.264, + "examples": 5000, + "loss": 2.0369871826171875 + }, + "validation_seconds": 0.1707751750946045 + }, + { + "epoch": 32, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0915020135243734, + "train_seconds": 8.784525871276855, + "validation": { + "accuracy": 0.2614, + "examples": 5000, + "loss": 2.0325482177734373 + }, + "validation_seconds": 0.1706712245941162 + }, + { + "epoch": 33, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.089619413333469, + "train_seconds": 8.782733917236328, + "validation": { + "accuracy": 0.2568, + "examples": 5000, + "loss": 2.0395154541015623 + }, + "validation_seconds": 0.1710975170135498 + }, + { + "epoch": 34, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.09081546113756, + "train_seconds": 8.780365705490112, + "validation": { + "accuracy": 0.248, + "examples": 5000, + "loss": 2.03733984375 + }, + "validation_seconds": 0.17079472541809082 + }, + { + "epoch": 35, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.088437536536323, + "train_seconds": 8.78589916229248, + "validation": { + "accuracy": 0.258, + "examples": 5000, + "loss": 2.0394189697265626 + }, + "validation_seconds": 0.17055249214172363 + }, + { + "epoch": 36, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0895603517744274, + "train_seconds": 8.778956413269043, + "validation": { + "accuracy": 0.2592, + "examples": 5000, + "loss": 2.0462224609375 + }, + "validation_seconds": 0.17077350616455078 + }, + { + "epoch": 37, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.091470686488681, + "train_seconds": 8.784783363342285, + "validation": { + "accuracy": 0.2478, + "examples": 5000, + "loss": 2.0410172119140624 + }, + "validation_seconds": 0.1713721752166748 + }, + { + "epoch": 38, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.088535167863634, + "train_seconds": 8.789288759231567, + "validation": { + "accuracy": 0.256, + "examples": 5000, + "loss": 2.0440474853515624 + }, + "validation_seconds": 0.17089271545410156 + }, + { + "epoch": 39, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.088415269978841, + "train_seconds": 8.792225360870361, + "validation": { + "accuracy": 0.2636, + "examples": 5000, + "loss": 2.0371565673828127 + }, + "validation_seconds": 0.17076539993286133 + }, + { + "epoch": 40, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0889872140248618, + "train_seconds": 8.789459705352783, + "validation": { + "accuracy": 0.2554, + "examples": 5000, + "loss": 2.0380864013671873 + }, + "validation_seconds": 0.17058658599853516 + }, + { + "epoch": 41, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0890390762329103, + "train_seconds": 8.786790609359741, + "validation": { + "accuracy": 0.231, + "examples": 5000, + "loss": 2.0498098876953126 + }, + "validation_seconds": 0.17036080360412598 + }, + { + "epoch": 42, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.08899807349311, + "train_seconds": 8.78852891921997, + "validation": { + "accuracy": 0.2488, + "examples": 5000, + "loss": 2.037922509765625 + }, + "validation_seconds": 0.17090797424316406 + }, + { + "epoch": 43, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0921546830071343, + "train_seconds": 8.785106897354126, + "validation": { + "accuracy": 0.2548, + "examples": 5000, + "loss": 2.02869716796875 + }, + "validation_seconds": 0.17080354690551758 + }, + { + "epoch": 44, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.092005069584317, + "train_seconds": 8.787235975265503, + "validation": { + "accuracy": 0.2488, + "examples": 5000, + "loss": 2.049959033203125 + }, + "validation_seconds": 0.17099785804748535 + }, + { + "epoch": 45, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0909548716227215, + "train_seconds": 8.785946130752563, + "validation": { + "accuracy": 0.2476, + "examples": 5000, + "loss": 2.048082666015625 + }, + "validation_seconds": 0.1708078384399414 + }, + { + "epoch": 46, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0915795023176407, + "train_seconds": 8.780594825744629, + "validation": { + "accuracy": 0.2412, + "examples": 5000, + "loss": 2.0391725341796874 + }, + "validation_seconds": 0.17049098014831543 + }, + { + "epoch": 47, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0890224663416546, + "train_seconds": 8.786250352859497, + "validation": { + "accuracy": 0.2608, + "examples": 5000, + "loss": 2.0262034912109375 + }, + "validation_seconds": 0.17060184478759766 + }, + { + "epoch": 48, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0947676996866864, + "train_seconds": 8.783189058303833, + "validation": { + "accuracy": 0.2566, + "examples": 5000, + "loss": 2.039441650390625 + }, + "validation_seconds": 0.17084026336669922 + }, + { + "epoch": 49, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.095011107550727, + "train_seconds": 8.784818172454834, + "validation": { + "accuracy": 0.2562, + "examples": 5000, + "loss": 2.0363927734375 + }, + "validation_seconds": 0.17116999626159668 + }, + { + "epoch": 50, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0891846113416883, + "train_seconds": 8.784857749938965, + "validation": { + "accuracy": 0.2578, + "examples": 5000, + "loss": 2.0316771484375 + }, + "validation_seconds": 0.17063379287719727 + }, + { + "epoch": 51, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0929476915359495, + "train_seconds": 8.783042430877686, + "validation": { + "accuracy": 0.2558, + "examples": 5000, + "loss": 2.035330224609375 + }, + "validation_seconds": 0.17069602012634277 + }, + { + "epoch": 52, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.088564098273383, + "train_seconds": 8.782349824905396, + "validation": { + "accuracy": 0.2532, + "examples": 5000, + "loss": 2.04925380859375 + }, + "validation_seconds": 0.17074131965637207 + }, + { + "epoch": 53, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0938075734880237, + "train_seconds": 8.777364492416382, + "validation": { + "accuracy": 0.2616, + "examples": 5000, + "loss": 2.029446142578125 + }, + "validation_seconds": 0.1706075668334961 + }, + { + "epoch": 54, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0936699249479505, + "train_seconds": 8.779646873474121, + "validation": { + "accuracy": 0.2338, + "examples": 5000, + "loss": 2.0565670166015626 + }, + "validation_seconds": 0.1709299087524414 + }, + { + "epoch": 55, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.094998370022244, + "train_seconds": 8.779564380645752, + "validation": { + "accuracy": 0.2512, + "examples": 5000, + "loss": 2.0400283447265624 + }, + "validation_seconds": 0.1704554557800293 + }, + { + "epoch": 56, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0949601395924886, + "train_seconds": 8.783570289611816, + "validation": { + "accuracy": 0.2318, + "examples": 5000, + "loss": 2.05338671875 + }, + "validation_seconds": 0.17037272453308105 + }, + { + "epoch": 57, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0950200890858968, + "train_seconds": 8.780302047729492, + "validation": { + "accuracy": 0.238, + "examples": 5000, + "loss": 2.0578427001953123 + }, + "validation_seconds": 0.17057228088378906 + }, + { + "epoch": 58, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0979971645143296, + "train_seconds": 8.781599044799805, + "validation": { + "accuracy": 0.2442, + "examples": 5000, + "loss": 2.053140576171875 + }, + "validation_seconds": 0.17058086395263672 + }, + { + "epoch": 59, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0933812481774225, + "train_seconds": 8.783581972122192, + "validation": { + "accuracy": 0.2446, + "examples": 5000, + "loss": 2.039527001953125 + }, + "validation_seconds": 0.17032074928283691 + }, + { + "epoch": 60, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.0970389923095705, + "train_seconds": 8.782969951629639, + "validation": { + "accuracy": 0.2502, + "examples": 5000, + "loss": 2.035926953125 + }, + "validation_seconds": 0.17087292671203613 + }, + { + "epoch": 61, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.082931641260783, + "train_seconds": 8.786428213119507, + "validation": { + "accuracy": 0.2638, + "examples": 5000, + "loss": 2.027929052734375 + }, + "validation_seconds": 0.17071866989135742 + }, + { + "epoch": 62, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0817930608537463, + "train_seconds": 8.783912420272827, + "validation": { + "accuracy": 0.261, + "examples": 5000, + "loss": 2.02908349609375 + }, + "validation_seconds": 0.17039942741394043 + }, + { + "epoch": 63, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.082688669925266, + "train_seconds": 8.776290655136108, + "validation": { + "accuracy": 0.2552, + "examples": 5000, + "loss": 2.0279300537109375 + }, + "validation_seconds": 0.17133569717407227 + }, + { + "epoch": 64, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.079216177495321, + "train_seconds": 8.777651309967041, + "validation": { + "accuracy": 0.256, + "examples": 5000, + "loss": 2.03120791015625 + }, + "validation_seconds": 0.17051076889038086 + }, + { + "epoch": 65, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.080554505962796, + "train_seconds": 8.78281545639038, + "validation": { + "accuracy": 0.2608, + "examples": 5000, + "loss": 2.02845283203125 + }, + "validation_seconds": 0.1705918312072754 + }, + { + "epoch": 66, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.082145662413703, + "train_seconds": 8.77703309059143, + "validation": { + "accuracy": 0.2616, + "examples": 5000, + "loss": 2.034113623046875 + }, + "validation_seconds": 0.17079496383666992 + }, + { + "epoch": 67, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.085471007792155, + "train_seconds": 8.780488729476929, + "validation": { + "accuracy": 0.2596, + "examples": 5000, + "loss": 2.02885205078125 + }, + "validation_seconds": 0.17075657844543457 + }, + { + "epoch": 68, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.082207935248481, + "train_seconds": 8.777437210083008, + "validation": { + "accuracy": 0.2646, + "examples": 5000, + "loss": 2.02878359375 + }, + "validation_seconds": 0.17030858993530273 + }, + { + "epoch": 69, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0829909870147705, + "train_seconds": 8.775349855422974, + "validation": { + "accuracy": 0.2608, + "examples": 5000, + "loss": 2.0296310302734377 + }, + "validation_seconds": 0.17086458206176758 + }, + { + "epoch": 70, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.087993453131782, + "train_seconds": 8.779688596725464, + "validation": { + "accuracy": 0.2556, + "examples": 5000, + "loss": 2.0325456298828124 + }, + "validation_seconds": 0.1709308624267578 + }, + { + "epoch": 71, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0826953997717963, + "train_seconds": 8.779837608337402, + "validation": { + "accuracy": 0.2628, + "examples": 5000, + "loss": 2.03113203125 + }, + "validation_seconds": 0.17030596733093262 + }, + { + "epoch": 72, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0797708891126843, + "train_seconds": 8.777223110198975, + "validation": { + "accuracy": 0.2604, + "examples": 5000, + "loss": 2.028413623046875 + }, + "validation_seconds": 0.17042040824890137 + }, + { + "epoch": 73, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.082765061314901, + "train_seconds": 8.776772260665894, + "validation": { + "accuracy": 0.262, + "examples": 5000, + "loss": 2.0308223388671873 + }, + "validation_seconds": 0.17062807083129883 + }, + { + "epoch": 74, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0817778525034587, + "train_seconds": 8.77632451057434, + "validation": { + "accuracy": 0.2552, + "examples": 5000, + "loss": 2.030654541015625 + }, + "validation_seconds": 0.17026042938232422 + }, + { + "epoch": 75, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0830021292792424, + "train_seconds": 8.779340982437134, + "validation": { + "accuracy": 0.261, + "examples": 5000, + "loss": 2.0272352783203127 + }, + "validation_seconds": 0.17082738876342773 + }, + { + "epoch": 76, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0825558311886256, + "train_seconds": 8.77163314819336, + "validation": { + "accuracy": 0.2564, + "examples": 5000, + "loss": 2.027567041015625 + }, + "validation_seconds": 0.17051172256469727 + }, + { + "epoch": 77, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0800666684468587, + "train_seconds": 8.77224326133728, + "validation": { + "accuracy": 0.2624, + "examples": 5000, + "loss": 2.0338967041015623 + }, + "validation_seconds": 0.17057394981384277 + }, + { + "epoch": 78, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0857824336157904, + "train_seconds": 8.738943576812744, + "validation": { + "accuracy": 0.2544, + "examples": 5000, + "loss": 2.028851806640625 + }, + "validation_seconds": 0.17060256004333496 + }, + { + "epoch": 79, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.084057655037774, + "train_seconds": 8.777712106704712, + "validation": { + "accuracy": 0.2604, + "examples": 5000, + "loss": 2.0281642822265624 + }, + "validation_seconds": 0.17107605934143066 + }, + { + "epoch": 80, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0811874683803984, + "train_seconds": 8.779063940048218, + "validation": { + "accuracy": 0.253, + "examples": 5000, + "loss": 2.0347160400390627 + }, + "validation_seconds": 0.17075872421264648 + }, + { + "epoch": 81, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0836294103834363, + "train_seconds": 8.772765398025513, + "validation": { + "accuracy": 0.2602, + "examples": 5000, + "loss": 2.0277829833984375 + }, + "validation_seconds": 0.1703035831451416 + }, + { + "epoch": 82, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0816474262237548, + "train_seconds": 8.77937364578247, + "validation": { + "accuracy": 0.2562, + "examples": 5000, + "loss": 2.032101416015625 + }, + "validation_seconds": 0.17008304595947266 + }, + { + "epoch": 83, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.079528014585707, + "train_seconds": 8.772222995758057, + "validation": { + "accuracy": 0.2612, + "examples": 5000, + "loss": 2.0306248291015625 + }, + "validation_seconds": 0.17063570022583008 + }, + { + "epoch": 84, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.086056492869059, + "train_seconds": 8.776498079299927, + "validation": { + "accuracy": 0.2606, + "examples": 5000, + "loss": 2.02940244140625 + }, + "validation_seconds": 0.17087340354919434 + }, + { + "epoch": 85, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.084182815806071, + "train_seconds": 8.77548599243164, + "validation": { + "accuracy": 0.2572, + "examples": 5000, + "loss": 2.033000146484375 + }, + "validation_seconds": 0.1705179214477539 + }, + { + "epoch": 86, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.082096115875244, + "train_seconds": 8.770426034927368, + "validation": { + "accuracy": 0.2636, + "examples": 5000, + "loss": 2.0297025390625 + }, + "validation_seconds": 0.17105412483215332 + }, + { + "epoch": 87, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.079115822007921, + "train_seconds": 8.773576021194458, + "validation": { + "accuracy": 0.2562, + "examples": 5000, + "loss": 2.03552451171875 + }, + "validation_seconds": 0.1705305576324463 + }, + { + "epoch": 88, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0829235925462513, + "train_seconds": 8.77567744255066, + "validation": { + "accuracy": 0.2576, + "examples": 5000, + "loss": 2.0320380126953124 + }, + "validation_seconds": 0.17091631889343262 + }, + { + "epoch": 89, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.0760149488237167, + "train_seconds": 8.781469583511353, + "validation": { + "accuracy": 0.2592, + "examples": 5000, + "loss": 2.0298075439453127 + }, + "validation_seconds": 0.17047929763793945 + }, + { + "epoch": 90, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.081927407285902, + "train_seconds": 8.773695468902588, + "validation": { + "accuracy": 0.2602, + "examples": 5000, + "loss": 2.0261443359375 + }, + "validation_seconds": 0.17052292823791504 + }, + { + "epoch": 91, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.0803893885294595, + "train_seconds": 8.772611379623413, + "validation": { + "accuracy": 0.258, + "examples": 5000, + "loss": 2.0315213623046877 + }, + "validation_seconds": 0.1705324649810791 + }, + { + "epoch": 92, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.082957981236776, + "train_seconds": 8.775248765945435, + "validation": { + "accuracy": 0.2594, + "examples": 5000, + "loss": 2.03054775390625 + }, + "validation_seconds": 0.17024850845336914 + }, + { + "epoch": 93, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.0807154247283934, + "train_seconds": 8.77206015586853, + "validation": { + "accuracy": 0.2604, + "examples": 5000, + "loss": 2.029746728515625 + }, + "validation_seconds": 0.17039227485656738 + }, + { + "epoch": 94, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.08182001953125, + "train_seconds": 8.77450442314148, + "validation": { + "accuracy": 0.2586, + "examples": 5000, + "loss": 2.0316868408203126 + }, + "validation_seconds": 0.1709284782409668 + }, + { + "epoch": 95, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.077259977383084, + "train_seconds": 8.733667135238647, + "validation": { + "accuracy": 0.2606, + "examples": 5000, + "loss": 2.028377978515625 + }, + "validation_seconds": 0.17060089111328125 + }, + { + "epoch": 96, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.0803276078542075, + "train_seconds": 8.77278447151184, + "validation": { + "accuracy": 0.2576, + "examples": 5000, + "loss": 2.0310527099609375 + }, + "validation_seconds": 0.17107558250427246 + }, + { + "epoch": 97, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.0836526218838163, + "train_seconds": 8.760288715362549, + "validation": { + "accuracy": 0.257, + "examples": 5000, + "loss": 2.0310808349609375 + }, + "validation_seconds": 0.17074060440063477 + }, + { + "epoch": 98, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.0804748294406465, + "train_seconds": 8.81708836555481, + "validation": { + "accuracy": 0.2586, + "examples": 5000, + "loss": 2.0306814208984374 + }, + "validation_seconds": 0.17048335075378418 + }, + { + "epoch": 99, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.0814546067555746, + "train_seconds": 8.773549795150757, + "validation": { + "accuracy": 0.2598, + "examples": 5000, + "loss": 2.028793017578125 + }, + "validation_seconds": 0.17067503929138184 + }, + { + "epoch": 100, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.081185230784946, + "train_seconds": 8.774254560470581, + "validation": { + "accuracy": 0.2618, + "examples": 5000, + "loss": 2.029934326171875 + }, + "validation_seconds": 0.17056846618652344 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.2618, + "evaluation_seconds": 0.17047953605651855, + "examples": 5000, + "finite": true, + "loss": 2.029934326171875 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2322423808, + "peak_memory_reserved_bytes": 3162505216, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 896.4814369678497, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 100, + "feedforward_example_passes": 14005000, + "forward_macs_per_example": 68862592, + "forward_parameter_count": 464154, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 0, + "logical_task_loss_queries": 0, + "num_trainable_layers": 32, + "ordinary_training_examples": 4500000, + "ordinary_validation_examples": 505000, + "relaxation_example_passes": 0, + "training_example_presentations": 9000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json.manifest.json new file mode 100644 index 0000000..15f5dc6 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json.manifest.json @@ -0,0 +1,128 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "pepita", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "100", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.003", + "--output_lr", + "0.003", + "--lr_schedule", + "pepita", + "--lr_milestones", + "60,90", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--pepita_projection_scale", + "0.05" + ], + "completed_unix_time": 1787023816.5468557, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 909.2571918964386, + "epochs": 100, + "experiment_name": "resnet-r2-pepita-d32-lr0p003", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "pepita", + "method": "pepita", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d32-lr0p003.json", + "output_exists": true, + "output_sha256": "461ff5f189160d80f86da8972f3742e260ebfb4acc720aab961f403f047ba4f1", + "rate": 0.003, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json b/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json new file mode 100644 index 0000000..08c2291 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json @@ -0,0 +1,1565 @@ +{ + "architecture": { + "base_width": 16, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "forward_parameter_count": 853018, + "normalization": "batchnorm", + "residual_scale": 1.0 + }, + "args": { + "augment_train": 1, + "batch_size": 128, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "dp_alpha": 0.0, + "dp_beta": 0.1, + "dp_inference_passes": 16, + "ep_beta": 0.5, + "ep_dt": 0.5, + "ep_free_steps": 20, + "ep_nudge_steps": 4, + "epochs": 100, + "eval_every": 1, + "eval_split": "validation", + "feedback_seed": 1729, + "ff_score_from_layer": 1, + "ff_threshold": 2.0, + "loader_seed": 0, + "lr": 0.003, + "lr_gamma": 0.1, + "lr_milestones": "60,90", + "lr_schedule": "pepita", + "max_steps": 0, + "method": "pepita", + "momentum": 0.9, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json", + "output_lr": 0.003, + "pepita_projection_scale": 0.05, + "seed": 0, + "split_seed": 2027, + "train_limit": 0, + "val_examples": 5000, + "weight_decay": 0.0001, + "width": 16 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.3475769726223414, + "train_seconds": 16.284608602523804, + "validation": { + "accuracy": 0.1734, + "examples": 5000, + "loss": 2.24925087890625 + }, + "validation_seconds": 0.31821584701538086 + }, + { + "epoch": 2, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.2237265483432345, + "train_seconds": 15.681286573410034, + "validation": { + "accuracy": 0.157, + "examples": 5000, + "loss": 2.225540087890625 + }, + "validation_seconds": 0.2987985610961914 + }, + { + "epoch": 3, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.2015728338453506, + "train_seconds": 15.6518235206604, + "validation": { + "accuracy": 0.2018, + "examples": 5000, + "loss": 2.17788203125 + }, + "validation_seconds": 0.29958510398864746 + }, + { + "epoch": 4, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1849158473968506, + "train_seconds": 15.668159246444702, + "validation": { + "accuracy": 0.1936, + "examples": 5000, + "loss": 2.168885498046875 + }, + "validation_seconds": 0.29885411262512207 + }, + { + "epoch": 5, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1747074312422012, + "train_seconds": 15.680987358093262, + "validation": { + "accuracy": 0.1752, + "examples": 5000, + "loss": 2.192222607421875 + }, + "validation_seconds": 0.29928016662597656 + }, + { + "epoch": 6, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1689188510047064, + "train_seconds": 15.674059391021729, + "validation": { + "accuracy": 0.2234, + "examples": 5000, + "loss": 2.135821533203125 + }, + "validation_seconds": 0.2990145683288574 + }, + { + "epoch": 7, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.164094039789836, + "train_seconds": 15.639453887939453, + "validation": { + "accuracy": 0.2108, + "examples": 5000, + "loss": 2.135561669921875 + }, + "validation_seconds": 0.2990138530731201 + }, + { + "epoch": 8, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.163451834954156, + "train_seconds": 15.729496479034424, + "validation": { + "accuracy": 0.1988, + "examples": 5000, + "loss": 2.18637685546875 + }, + "validation_seconds": 0.29883646965026855 + }, + { + "epoch": 9, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1566063519795735, + "train_seconds": 15.675715208053589, + "validation": { + "accuracy": 0.2076, + "examples": 5000, + "loss": 2.162432666015625 + }, + "validation_seconds": 0.2995271682739258 + }, + { + "epoch": 10, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.161393331866794, + "train_seconds": 15.64734959602356, + "validation": { + "accuracy": 0.2054, + "examples": 5000, + "loss": 2.134214453125 + }, + "validation_seconds": 0.2993965148925781 + }, + { + "epoch": 11, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.157396332973904, + "train_seconds": 15.720870018005371, + "validation": { + "accuracy": 0.2262, + "examples": 5000, + "loss": 2.12485478515625 + }, + "validation_seconds": 0.2997002601623535 + }, + { + "epoch": 12, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1606002636379666, + "train_seconds": 15.66697907447815, + "validation": { + "accuracy": 0.226, + "examples": 5000, + "loss": 2.121274462890625 + }, + "validation_seconds": 0.2992849349975586 + }, + { + "epoch": 13, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1571049402448867, + "train_seconds": 15.68711233139038, + "validation": { + "accuracy": 0.216, + "examples": 5000, + "loss": 2.131235693359375 + }, + "validation_seconds": 0.2990860939025879 + }, + { + "epoch": 14, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1594723821428086, + "train_seconds": 15.663890361785889, + "validation": { + "accuracy": 0.2184, + "examples": 5000, + "loss": 2.133489697265625 + }, + "validation_seconds": 0.29947352409362793 + }, + { + "epoch": 15, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.156481715689765, + "train_seconds": 15.68401551246643, + "validation": { + "accuracy": 0.2246, + "examples": 5000, + "loss": 2.123956591796875 + }, + "validation_seconds": 0.29934096336364746 + }, + { + "epoch": 16, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1623080538855657, + "train_seconds": 15.691046476364136, + "validation": { + "accuracy": 0.2108, + "examples": 5000, + "loss": 2.1551609375 + }, + "validation_seconds": 0.2992236614227295 + }, + { + "epoch": 17, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1583885418362088, + "train_seconds": 15.685045003890991, + "validation": { + "accuracy": 0.2354, + "examples": 5000, + "loss": 2.107221484375 + }, + "validation_seconds": 0.29936671257019043 + }, + { + "epoch": 18, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.163076057688395, + "train_seconds": 15.706880807876587, + "validation": { + "accuracy": 0.2152, + "examples": 5000, + "loss": 2.1394583984375 + }, + "validation_seconds": 0.29956936836242676 + }, + { + "epoch": 19, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1610479665544298, + "train_seconds": 15.694411993026733, + "validation": { + "accuracy": 0.218, + "examples": 5000, + "loss": 2.13543974609375 + }, + "validation_seconds": 0.2992556095123291 + }, + { + "epoch": 20, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1602279196421303, + "train_seconds": 15.671279191970825, + "validation": { + "accuracy": 0.2136, + "examples": 5000, + "loss": 2.14320869140625 + }, + "validation_seconds": 0.29967164993286133 + }, + { + "epoch": 21, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.159357769478692, + "train_seconds": 15.638594388961792, + "validation": { + "accuracy": 0.2064, + "examples": 5000, + "loss": 2.159037548828125 + }, + "validation_seconds": 0.2990906238555908 + }, + { + "epoch": 22, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1632483467525905, + "train_seconds": 15.674551486968994, + "validation": { + "accuracy": 0.2232, + "examples": 5000, + "loss": 2.1262857421875 + }, + "validation_seconds": 0.2990391254425049 + }, + { + "epoch": 23, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.161040240605672, + "train_seconds": 15.671793460845947, + "validation": { + "accuracy": 0.219, + "examples": 5000, + "loss": 2.143005712890625 + }, + "validation_seconds": 0.29958534240722656 + }, + { + "epoch": 24, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.157951952955458, + "train_seconds": 15.660223960876465, + "validation": { + "accuracy": 0.2102, + "examples": 5000, + "loss": 2.15192724609375 + }, + "validation_seconds": 0.29978346824645996 + }, + { + "epoch": 25, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1627659381866455, + "train_seconds": 15.678407192230225, + "validation": { + "accuracy": 0.1964, + "examples": 5000, + "loss": 2.164074560546875 + }, + "validation_seconds": 0.29997992515563965 + }, + { + "epoch": 26, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.163635923576355, + "train_seconds": 15.671875715255737, + "validation": { + "accuracy": 0.2158, + "examples": 5000, + "loss": 2.151888037109375 + }, + "validation_seconds": 0.2996237277984619 + }, + { + "epoch": 27, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1635207350413004, + "train_seconds": 15.662642240524292, + "validation": { + "accuracy": 0.2136, + "examples": 5000, + "loss": 2.125229248046875 + }, + "validation_seconds": 0.2993924617767334 + }, + { + "epoch": 28, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.162565950393677, + "train_seconds": 15.636061668395996, + "validation": { + "accuracy": 0.1964, + "examples": 5000, + "loss": 2.16964326171875 + }, + "validation_seconds": 0.29975390434265137 + }, + { + "epoch": 29, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.163283192952474, + "train_seconds": 15.683935642242432, + "validation": { + "accuracy": 0.2076, + "examples": 5000, + "loss": 2.151294189453125 + }, + "validation_seconds": 0.29951977729797363 + }, + { + "epoch": 30, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.161205266825358, + "train_seconds": 15.69214129447937, + "validation": { + "accuracy": 0.2086, + "examples": 5000, + "loss": 2.14878486328125 + }, + "validation_seconds": 0.2994678020477295 + }, + { + "epoch": 31, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1700854666815865, + "train_seconds": 15.638405084609985, + "validation": { + "accuracy": 0.21, + "examples": 5000, + "loss": 2.168007275390625 + }, + "validation_seconds": 0.2993042469024658 + }, + { + "epoch": 32, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1673946551005048, + "train_seconds": 15.769407749176025, + "validation": { + "accuracy": 0.1872, + "examples": 5000, + "loss": 2.15320751953125 + }, + "validation_seconds": 0.2998044490814209 + }, + { + "epoch": 33, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.168034863026937, + "train_seconds": 15.698111772537231, + "validation": { + "accuracy": 0.2076, + "examples": 5000, + "loss": 2.149951318359375 + }, + "validation_seconds": 0.2999296188354492 + }, + { + "epoch": 34, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1725848429785835, + "train_seconds": 15.684096813201904, + "validation": { + "accuracy": 0.2216, + "examples": 5000, + "loss": 2.138531201171875 + }, + "validation_seconds": 0.2992548942565918 + }, + { + "epoch": 35, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.169652655537923, + "train_seconds": 15.643929243087769, + "validation": { + "accuracy": 0.2218, + "examples": 5000, + "loss": 2.138718603515625 + }, + "validation_seconds": 0.29943323135375977 + }, + { + "epoch": 36, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1699724278767905, + "train_seconds": 15.66665768623352, + "validation": { + "accuracy": 0.2084, + "examples": 5000, + "loss": 2.14305009765625 + }, + "validation_seconds": 0.29984354972839355 + }, + { + "epoch": 37, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1717837972429064, + "train_seconds": 15.693549633026123, + "validation": { + "accuracy": 0.2022, + "examples": 5000, + "loss": 2.137519580078125 + }, + "validation_seconds": 0.2995481491088867 + }, + { + "epoch": 38, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1724463059743244, + "train_seconds": 15.658748149871826, + "validation": { + "accuracy": 0.2186, + "examples": 5000, + "loss": 2.131788037109375 + }, + "validation_seconds": 0.29953789710998535 + }, + { + "epoch": 39, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.168097107272678, + "train_seconds": 15.651520252227783, + "validation": { + "accuracy": 0.2226, + "examples": 5000, + "loss": 2.13143837890625 + }, + "validation_seconds": 0.29965829849243164 + }, + { + "epoch": 40, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.170875356547038, + "train_seconds": 15.668555498123169, + "validation": { + "accuracy": 0.222, + "examples": 5000, + "loss": 2.124761669921875 + }, + "validation_seconds": 0.2994194030761719 + }, + { + "epoch": 41, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1716907165103487, + "train_seconds": 15.660097360610962, + "validation": { + "accuracy": 0.203, + "examples": 5000, + "loss": 2.168586181640625 + }, + "validation_seconds": 0.29941606521606445 + }, + { + "epoch": 42, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1702115387386747, + "train_seconds": 15.656601190567017, + "validation": { + "accuracy": 0.2056, + "examples": 5000, + "loss": 2.140987890625 + }, + "validation_seconds": 0.299454927444458 + }, + { + "epoch": 43, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1729301143222384, + "train_seconds": 15.676163911819458, + "validation": { + "accuracy": 0.1914, + "examples": 5000, + "loss": 2.18998251953125 + }, + "validation_seconds": 0.299652099609375 + }, + { + "epoch": 44, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1765141748216417, + "train_seconds": 15.67884373664856, + "validation": { + "accuracy": 0.209, + "examples": 5000, + "loss": 2.13410859375 + }, + "validation_seconds": 0.299485445022583 + }, + { + "epoch": 45, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.181572889751858, + "train_seconds": 15.63332986831665, + "validation": { + "accuracy": 0.2036, + "examples": 5000, + "loss": 2.15123642578125 + }, + "validation_seconds": 0.29934072494506836 + }, + { + "epoch": 46, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.175901057518853, + "train_seconds": 15.680678606033325, + "validation": { + "accuracy": 0.2054, + "examples": 5000, + "loss": 2.168797021484375 + }, + "validation_seconds": 0.29965710639953613 + }, + { + "epoch": 47, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1773429202185737, + "train_seconds": 15.656840324401855, + "validation": { + "accuracy": 0.2014, + "examples": 5000, + "loss": 2.13987265625 + }, + "validation_seconds": 0.2994050979614258 + }, + { + "epoch": 48, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1835591296725805, + "train_seconds": 15.684873819351196, + "validation": { + "accuracy": 0.2052, + "examples": 5000, + "loss": 2.15433330078125 + }, + "validation_seconds": 0.299971342086792 + }, + { + "epoch": 49, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.188736485120985, + "train_seconds": 15.647130250930786, + "validation": { + "accuracy": 0.2238, + "examples": 5000, + "loss": 2.130195849609375 + }, + "validation_seconds": 0.2995433807373047 + }, + { + "epoch": 50, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.180174428134494, + "train_seconds": 15.650701761245728, + "validation": { + "accuracy": 0.1848, + "examples": 5000, + "loss": 2.22315712890625 + }, + "validation_seconds": 0.2997779846191406 + }, + { + "epoch": 51, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.182067480383979, + "train_seconds": 15.674840927124023, + "validation": { + "accuracy": 0.1828, + "examples": 5000, + "loss": 2.165380908203125 + }, + "validation_seconds": 0.2998027801513672 + }, + { + "epoch": 52, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.179020531166924, + "train_seconds": 15.649744749069214, + "validation": { + "accuracy": 0.1946, + "examples": 5000, + "loss": 2.224890478515625 + }, + "validation_seconds": 0.29945802688598633 + }, + { + "epoch": 53, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.182496388244629, + "train_seconds": 15.68449592590332, + "validation": { + "accuracy": 0.2044, + "examples": 5000, + "loss": 2.159840478515625 + }, + "validation_seconds": 0.299649715423584 + }, + { + "epoch": 54, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.176760282389323, + "train_seconds": 15.676718711853027, + "validation": { + "accuracy": 0.2154, + "examples": 5000, + "loss": 2.154713623046875 + }, + "validation_seconds": 0.30002808570861816 + }, + { + "epoch": 55, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.181396118672689, + "train_seconds": 15.64739179611206, + "validation": { + "accuracy": 0.2118, + "examples": 5000, + "loss": 2.15625439453125 + }, + "validation_seconds": 0.299640417098999 + }, + { + "epoch": 56, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1928697071499297, + "train_seconds": 15.674021244049072, + "validation": { + "accuracy": 0.1838, + "examples": 5000, + "loss": 2.1972078125 + }, + "validation_seconds": 0.29950594902038574 + }, + { + "epoch": 57, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1952037180158825, + "train_seconds": 15.704137325286865, + "validation": { + "accuracy": 0.2096, + "examples": 5000, + "loss": 2.15042080078125 + }, + "validation_seconds": 0.29971742630004883 + }, + { + "epoch": 58, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.187411170917087, + "train_seconds": 15.676226615905762, + "validation": { + "accuracy": 0.2104, + "examples": 5000, + "loss": 2.142286865234375 + }, + "validation_seconds": 0.29947590827941895 + }, + { + "epoch": 59, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.1929678960164387, + "train_seconds": 15.638705253601074, + "validation": { + "accuracy": 0.1936, + "examples": 5000, + "loss": 2.195626708984375 + }, + "validation_seconds": 0.2998230457305908 + }, + { + "epoch": 60, + "lr": 0.003, + "output_lr": 0.003, + "train_examples": 45000, + "train_loss": 2.187026701990763, + "train_seconds": 15.711337566375732, + "validation": { + "accuracy": 0.2156, + "examples": 5000, + "loss": 2.13201728515625 + }, + "validation_seconds": 0.3002026081085205 + }, + { + "epoch": 61, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.153464202711317, + "train_seconds": 15.665104866027832, + "validation": { + "accuracy": 0.2306, + "examples": 5000, + "loss": 2.118577685546875 + }, + "validation_seconds": 0.29940152168273926 + }, + { + "epoch": 62, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.153363601769341, + "train_seconds": 15.632834196090698, + "validation": { + "accuracy": 0.2176, + "examples": 5000, + "loss": 2.125661083984375 + }, + "validation_seconds": 0.29975414276123047 + }, + { + "epoch": 63, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1547847679562038, + "train_seconds": 15.688811302185059, + "validation": { + "accuracy": 0.2246, + "examples": 5000, + "loss": 2.1178529296875 + }, + "validation_seconds": 0.2998015880584717 + }, + { + "epoch": 64, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1549198490990533, + "train_seconds": 15.655038595199585, + "validation": { + "accuracy": 0.2172, + "examples": 5000, + "loss": 2.1274083984375 + }, + "validation_seconds": 0.2996184825897217 + }, + { + "epoch": 65, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1497549418555364, + "train_seconds": 15.679463386535645, + "validation": { + "accuracy": 0.2244, + "examples": 5000, + "loss": 2.1182056640625 + }, + "validation_seconds": 0.29967188835144043 + }, + { + "epoch": 66, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1515708785163032, + "train_seconds": 15.667905807495117, + "validation": { + "accuracy": 0.2192, + "examples": 5000, + "loss": 2.13347197265625 + }, + "validation_seconds": 0.2998011112213135 + }, + { + "epoch": 67, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.150513512632582, + "train_seconds": 15.695557594299316, + "validation": { + "accuracy": 0.232, + "examples": 5000, + "loss": 2.12719873046875 + }, + "validation_seconds": 0.2998483180999756 + }, + { + "epoch": 68, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.156382405641344, + "train_seconds": 15.711182832717896, + "validation": { + "accuracy": 0.2216, + "examples": 5000, + "loss": 2.131969189453125 + }, + "validation_seconds": 0.29973673820495605 + }, + { + "epoch": 69, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1527923811594647, + "train_seconds": 15.65618896484375, + "validation": { + "accuracy": 0.2242, + "examples": 5000, + "loss": 2.128789599609375 + }, + "validation_seconds": 0.2994709014892578 + }, + { + "epoch": 70, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.153493192715115, + "train_seconds": 15.669008016586304, + "validation": { + "accuracy": 0.2196, + "examples": 5000, + "loss": 2.13042177734375 + }, + "validation_seconds": 0.29961538314819336 + }, + { + "epoch": 71, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.150450718943278, + "train_seconds": 15.71059536933899, + "validation": { + "accuracy": 0.2334, + "examples": 5000, + "loss": 2.118706591796875 + }, + "validation_seconds": 0.2995724678039551 + }, + { + "epoch": 72, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.152825714450412, + "train_seconds": 15.683857917785645, + "validation": { + "accuracy": 0.2224, + "examples": 5000, + "loss": 2.129543701171875 + }, + "validation_seconds": 0.2995285987854004 + }, + { + "epoch": 73, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1515955153571236, + "train_seconds": 15.6649649143219, + "validation": { + "accuracy": 0.2194, + "examples": 5000, + "loss": 2.12808408203125 + }, + "validation_seconds": 0.299257755279541 + }, + { + "epoch": 74, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1496700807359486, + "train_seconds": 15.722174882888794, + "validation": { + "accuracy": 0.216, + "examples": 5000, + "loss": 2.12822939453125 + }, + "validation_seconds": 0.29958152770996094 + }, + { + "epoch": 75, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1503506173027884, + "train_seconds": 15.714526891708374, + "validation": { + "accuracy": 0.224, + "examples": 5000, + "loss": 2.119987060546875 + }, + "validation_seconds": 0.2995455265045166 + }, + { + "epoch": 76, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1522617220136855, + "train_seconds": 15.697211027145386, + "validation": { + "accuracy": 0.2248, + "examples": 5000, + "loss": 2.12939140625 + }, + "validation_seconds": 0.2992062568664551 + }, + { + "epoch": 77, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.154192639456855, + "train_seconds": 15.73182463645935, + "validation": { + "accuracy": 0.2216, + "examples": 5000, + "loss": 2.131347509765625 + }, + "validation_seconds": 0.29962968826293945 + }, + { + "epoch": 78, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1487903494940865, + "train_seconds": 15.704017400741577, + "validation": { + "accuracy": 0.22, + "examples": 5000, + "loss": 2.124372314453125 + }, + "validation_seconds": 0.2995433807373047 + }, + { + "epoch": 79, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1514638863457574, + "train_seconds": 15.68510890007019, + "validation": { + "accuracy": 0.2244, + "examples": 5000, + "loss": 2.12896591796875 + }, + "validation_seconds": 0.2998325824737549 + }, + { + "epoch": 80, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1498017944335936, + "train_seconds": 15.664867162704468, + "validation": { + "accuracy": 0.2254, + "examples": 5000, + "loss": 2.124163525390625 + }, + "validation_seconds": 0.2989673614501953 + }, + { + "epoch": 81, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1490917604658337, + "train_seconds": 15.712203979492188, + "validation": { + "accuracy": 0.2296, + "examples": 5000, + "loss": 2.11083935546875 + }, + "validation_seconds": 0.29970264434814453 + }, + { + "epoch": 82, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1534319928487142, + "train_seconds": 15.689087390899658, + "validation": { + "accuracy": 0.2166, + "examples": 5000, + "loss": 2.13138134765625 + }, + "validation_seconds": 0.2988622188568115 + }, + { + "epoch": 83, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.150700324376424, + "train_seconds": 15.66288447380066, + "validation": { + "accuracy": 0.2264, + "examples": 5000, + "loss": 2.124049951171875 + }, + "validation_seconds": 0.2999234199523926 + }, + { + "epoch": 84, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1545015112559, + "train_seconds": 15.717640161514282, + "validation": { + "accuracy": 0.2256, + "examples": 5000, + "loss": 2.120006201171875 + }, + "validation_seconds": 0.2995924949645996 + }, + { + "epoch": 85, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1520926057603624, + "train_seconds": 15.6857271194458, + "validation": { + "accuracy": 0.217, + "examples": 5000, + "loss": 2.123595703125 + }, + "validation_seconds": 0.29950642585754395 + }, + { + "epoch": 86, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1539180565304226, + "train_seconds": 15.702425003051758, + "validation": { + "accuracy": 0.2266, + "examples": 5000, + "loss": 2.127067041015625 + }, + "validation_seconds": 0.299513578414917 + }, + { + "epoch": 87, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1541437976413302, + "train_seconds": 15.651525497436523, + "validation": { + "accuracy": 0.2174, + "examples": 5000, + "loss": 2.12785693359375 + }, + "validation_seconds": 0.2997150421142578 + }, + { + "epoch": 88, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.150302347056071, + "train_seconds": 15.696409463882446, + "validation": { + "accuracy": 0.221, + "examples": 5000, + "loss": 2.1273095703125 + }, + "validation_seconds": 0.2993946075439453 + }, + { + "epoch": 89, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.152333891635471, + "train_seconds": 15.691116094589233, + "validation": { + "accuracy": 0.2258, + "examples": 5000, + "loss": 2.12480576171875 + }, + "validation_seconds": 0.29930853843688965 + }, + { + "epoch": 90, + "lr": 0.00030000000000000003, + "output_lr": 0.00030000000000000003, + "train_examples": 45000, + "train_loss": 2.1507328274620905, + "train_seconds": 15.65622353553772, + "validation": { + "accuracy": 0.2304, + "examples": 5000, + "loss": 2.118491552734375 + }, + "validation_seconds": 0.2990148067474365 + }, + { + "epoch": 91, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.1486516242980955, + "train_seconds": 15.699949741363525, + "validation": { + "accuracy": 0.2204, + "examples": 5000, + "loss": 2.12403779296875 + }, + "validation_seconds": 0.2998237609863281 + }, + { + "epoch": 92, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.1497891513400607, + "train_seconds": 15.700400590896606, + "validation": { + "accuracy": 0.2206, + "examples": 5000, + "loss": 2.1224546875 + }, + "validation_seconds": 0.2998065948486328 + }, + { + "epoch": 93, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.1481778487735324, + "train_seconds": 15.68268871307373, + "validation": { + "accuracy": 0.2306, + "examples": 5000, + "loss": 2.119259814453125 + }, + "validation_seconds": 0.29954075813293457 + }, + { + "epoch": 94, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.1484259122212728, + "train_seconds": 15.650254726409912, + "validation": { + "accuracy": 0.2268, + "examples": 5000, + "loss": 2.122284521484375 + }, + "validation_seconds": 0.2995462417602539 + }, + { + "epoch": 95, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.1485114953782825, + "train_seconds": 15.716254949569702, + "validation": { + "accuracy": 0.2312, + "examples": 5000, + "loss": 2.1186466796875 + }, + "validation_seconds": 0.2994718551635742 + }, + { + "epoch": 96, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.14853645807902, + "train_seconds": 15.704525470733643, + "validation": { + "accuracy": 0.2256, + "examples": 5000, + "loss": 2.120378759765625 + }, + "validation_seconds": 0.2997438907623291 + }, + { + "epoch": 97, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.1462308433532713, + "train_seconds": 15.661826848983765, + "validation": { + "accuracy": 0.2294, + "examples": 5000, + "loss": 2.120892578125 + }, + "validation_seconds": 0.2998523712158203 + }, + { + "epoch": 98, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.149182392586602, + "train_seconds": 15.695777177810669, + "validation": { + "accuracy": 0.229, + "examples": 5000, + "loss": 2.1194109375 + }, + "validation_seconds": 0.29935646057128906 + }, + { + "epoch": 99, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.149133202785916, + "train_seconds": 15.691457509994507, + "validation": { + "accuracy": 0.2268, + "examples": 5000, + "loss": 2.116005859375 + }, + "validation_seconds": 0.29990196228027344 + }, + { + "epoch": 100, + "lr": 3.0000000000000004e-05, + "output_lr": 3.0000000000000004e-05, + "train_examples": 45000, + "train_loss": 2.1499680579715306, + "train_seconds": 15.672620058059692, + "validation": { + "accuracy": 0.228, + "examples": 5000, + "loss": 2.124036181640625 + }, + "validation_seconds": 0.29943299293518066 + } + ], + "evaluation_protocol": { + "split": "validation", + "test_evaluations": 0, + "test_used_for_selection": false + }, + "final": { + "accuracy": 0.228, + "evaluation_seconds": 0.2996077537536621, + "examples": 5000, + "finite": true, + "loss": 2.124036181640625 + }, + "first_nonfinite_step": null, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 3249364480, + "peak_memory_reserved_bytes": 4500488192, + "torch_version": "2.3.1+cu118" + }, + "layers": [], + "protocol_family": "resnet_local_learning_crossover", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "total_wall_seconds": 1598.730625152588, + "work": { + "candidate_label_evaluation_presentations": 0, + "completed_global_epochs": 100, + "feedforward_example_passes": 14005000, + "forward_macs_per_example": 125485696, + "forward_parameter_count": 853018, + "local_target_backward_example_evaluations": 0, + "local_vjp_example_evaluations": 0, + "logical_task_loss_queries": 0, + "num_trainable_layers": 56, + "ordinary_training_examples": 4500000, + "ordinary_validation_examples": 505000, + "relaxation_example_passes": 0, + "training_example_presentations": 9000000 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json.manifest.json new file mode 100644 index 0000000..f480d9e --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json.manifest.json @@ -0,0 +1,128 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/resnet_crossover_native.py", + "--method", + "pepita", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json", + "--feedback_seed", + "1729", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "100", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.003", + "--output_lr", + "0.003", + "--lr_schedule", + "pepita", + "--lr_milestones", + "60,90", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--pepita_projection_scale", + "0.05" + ], + "completed_unix_time": 1787004395.6670804, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 1602.2039358615875, + "epochs": 100, + "experiment_name": "resnet-r2-pepita-d56-lr0p003", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "pepita", + "method": "pepita", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-pepita-d56-lr0p003.json", + "output_exists": true, + "output_sha256": "dcebb1518d54fbefbe9bd0cb2a3b63a18591187b4dcad3cebeb3e93b1748c6e6", + "rate": 0.003, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json new file mode 100644 index 0000000..11c487e --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json @@ -0,0 +1,7256 @@ +{ + "architecture": { + "adaptive_apical_parameters": 376832, + "adaptive_feedback_parameters": 267904, + "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": 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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 1, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.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": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "ordinary_examples": 9000000, + "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": 0.9992486238479614, + "feedback_forward_cosine": [ + 0.9996339678764343, + 0.9996197819709778, + 0.999623715877533, + 0.9995846748352051, + 0.9997047185897827, + 0.9996135830879211, + 0.9995279908180237, + 0.9996839761734009, + 0.9996345043182373, + 0.999563455581665, + 0.9996508359909058, + 0.9994888305664062, + 0.999567985534668, + 0.9997429251670837, + 0.9997261762619019, + 0.9996747970581055, + 0.9997008442878723, + 0.9992879629135132, + 0.9999708533287048 + ], + "feedback_forward_norm_ratio": [ + 1.0003844499588013, + 1.0015509128570557, + 0.9996321797370911, + 1.001647710800171, + 0.9995453357696533, + 1.0023027658462524, + 0.9988731145858765, + 1.0006657838821411, + 0.999077558517456, + 1.0016008615493774, + 0.9999746084213257, + 1.0018646717071533, + 0.9991083145141602, + 1.000430703163147, + 1.0001863241195679, + 1.0009725093841553, + 1.0004057884216309, + 1.0017213821411133, + 1.0044785737991333 + ], + "feedback_forward_relative_error": [ + 0.027061661705374718, + 0.027639765292406082, + 0.02743203565478325, + 0.028893308714032173, + 0.024302596226334572, + 0.027923736721277237, + 0.03072751872241497, + 0.025158708915114403, + 0.027041001245379448, + 0.029615972191095352, + 0.026425801217556, + 0.03206216171383858, + 0.029396697878837585, + 0.02268228493630886, + 0.023404264822602272, + 0.025532469153404236, + 0.02446761168539524, + 0.037806157022714615, + 0.00886303000152111 + ], + "innovation_negative_gradient_cosine": [ + 0.9992955923080444, + 0.9991967678070068, + 0.9992392063140869, + 0.9991347193717957, + 0.9992673397064209, + 0.9993581175804138, + 0.9993082284927368, + 0.9993443489074707, + 0.9994051456451416, + 0.9993393421173096, + 0.9994563460350037, + 0.9994194507598877, + 0.9994933009147644, + 0.9996087551116943, + 0.9996581077575684, + 0.9997082948684692, + 0.9998230934143066, + 0.999924898147583, + 0.9078794717788696 + ], + "instruction_negative_gradient_cosine": [ + 0.9992955923080444, + 0.9991967678070068, + 0.9992392063140869, + 0.9991347789764404, + 0.9992673397064209, + 0.9993581175804138, + 0.9993082284927368, + 0.9993443489074707, + 0.9994051456451416, + 0.9993393421173096, + 0.9994563460350037, + 0.9994194507598877, + 0.9994933009147644, + 0.9996087551116943, + 0.9996581077575684, + 0.9997082948684692, + 0.9998230934143066, + 0.999924898147583, + 0.9349699020385742 + ], + "matched_negative_gradient_cosine": [ + 0.03966842591762543, + 0.012892607599496841, + 0.03667213395237923, + 0.04239062964916229, + 0.04645364731550217, + 0.06657349318265915, + 0.07031472027301788, + 0.0461806021630764, + 0.0483735054731369, + 0.11813706159591675, + 0.0720367431640625, + 0.23040176928043365, + 0.0929129347205162, + 0.08425292372703552, + 0.08489963412284851, + 0.2519374489784241, + 0.1920303851366043, + 0.2691277265548706, + 0.13027682900428772 + ], + "max_norm_match_direction_error": 1.7881393432617188e-07, + "max_norm_match_relative_error": 1.754590499558617e-07, + "mean_feedback_forward_cosine": 0.9996316621178075, + "mean_feedback_forward_relative_error": 0.026654567569494247, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.03110409900546074, + "max_absolute_post_projection_soma_slope": 5.2924469251536266e-09, + "max_absolute_pre_projection_soma_slope": 0.03110409900546074, + "max_positive_post_projection_soma_slope": 4.771845585338497e-09, + "min_post_projection_soma_slope": -5.2924469251536266e-09, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.4805662368076366e-08, + "pre_projection_traffic_rms_ratio": 0.007601542191251489 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.007601542191206739, + "raw_negative_gradient_cosine": [ + 0.03966842591762543, + 0.012892606668174267, + 0.03667213395237923, + 0.04239062964916229, + 0.04645364731550217, + 0.06657349318265915, + 0.07031472027301788, + 0.0461806021630764, + 0.048373498022556305, + 0.11813706904649734, + 0.0720367506146431, + 0.23040176928043365, + 0.0929129347205162, + 0.08425292372703552, + 0.0848996490240097, + 0.25193747878074646, + 0.1920303851366043, + 0.2691276967525482, + 0.13027682900428772 + ], + "teaching_negative_gradient_cosine": [ + 0.9992955923080444, + 0.9991967678070068, + 0.9992392063140869, + 0.9991347193717957, + 0.9992673397064209, + 0.9993581175804138, + 0.9993082284927368, + 0.9993443489074707, + 0.9994051456451416, + 0.9993393421173096, + 0.9994563460350037, + 0.9994194507598877, + 0.9994933009147644, + 0.9996087551116943, + 0.9996581077575684, + 0.9997082948684692, + 0.9998230934143066, + 0.999924898147583, + 0.9078794717788696 + ], + "traffic_instruction_rms_ratio": 9.3995664093872, + "used_negative_gradient_cosine": [ + 0.9992955923080444, + 0.9991967678070068, + 0.9992392063140869, + 0.9991347193717957, + 0.9992673397064209, + 0.9993581175804138, + 0.9993082284927368, + 0.9993443489074707, + 0.9994051456451416, + 0.9993393421173096, + 0.9994563460350037, + 0.9994194507598877, + 0.9994933009147644, + 0.9996087551116943, + 0.9996581077575684, + 0.9997082948684692, + 0.9998230934143066, + 0.999924898147583, + 0.9078794717788696 + ], + "wall_s": 0.10504412651062012 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.3116, + "eval_loss": 2.02555947265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3959459066390991, + "mean_feedback_forward_cosine": 0.12446963296909082, + "mean_feedback_forward_relative_error": 1.3425679206848145, + "min_feedback_forward_cosine": 0.03807289898395538 + }, + "lr": 0.02, + "mixed_apical": { + "innovation_rms": 0.004026297986185734, + "instruction_rms": 0.004026297985703689, + "raw_apical_rms": 0.018433059693736067, + "teaching_rms": 0.004026297986185734, + "traffic_rms": 0.01797717185451189 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104110181331635, + "maximum_absolute_post_projection_soma_slope": 1.8376869803660156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.029582466001542e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006025940798164677, + "minimum_observations": 72 + }, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 1.9139111452526516 + }, + { + "epoch": 2, + "eval_accuracy": 0.4306, + "eval_loss": 1.8821124755859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3497217893600464, + "mean_feedback_forward_cosine": 0.28874401983461884, + "mean_feedback_forward_relative_error": 1.2083154571683783, + "min_feedback_forward_cosine": 0.11273814737796783 + }, + "lr": 0.04, + "mixed_apical": { + "innovation_rms": 0.005270128443391427, + "instruction_rms": 0.005270128443363647, + "raw_apical_rms": 0.020287175500285828, + "teaching_rms": 0.005270128443391427, + "traffic_rms": 0.019583521396373933 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.4794059310929697e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.947214936240712e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006557262506205862, + "minimum_observations": 72 + }, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.5111413827260336 + }, + { + "epoch": 3, + "eval_accuracy": 0.5196, + "eval_loss": 1.8604373779296874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.298816442489624, + "mean_feedback_forward_cosine": 0.4102881244922939, + "mean_feedback_forward_relative_error": 1.0967007655846446, + "min_feedback_forward_cosine": 0.18582221865653992 + }, + "lr": 0.06000000000000001, + "mixed_apical": { + "innovation_rms": 0.005695637091846514, + "instruction_rms": 0.005695637093150145, + "raw_apical_rms": 0.02234436190563027, + "teaching_rms": 0.005695637091846514, + "traffic_rms": 0.021599725418751093 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3064925141748063e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8363789922639952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008079684570298312, + "minimum_observations": 72 + }, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.1980859601550633 + }, + { + "epoch": 4, + "eval_accuracy": 0.6292, + "eval_loss": 1.0531740112304688, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2463593482971191, + "mean_feedback_forward_cosine": 0.5076760571253928, + "mean_feedback_forward_relative_error": 0.9999459856434872, + "min_feedback_forward_cosine": 0.2567841410636902 + }, + "lr": 0.08, + "mixed_apical": { + "innovation_rms": 0.005602893273743301, + "instruction_rms": 0.005602893273696883, + "raw_apical_rms": 0.023992327474635113, + "teaching_rms": 0.005602893273743301, + "traffic_rms": 0.02332356349597622 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.384531866932548e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7928872981115507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008067533370801183, + "minimum_observations": 72 + }, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 1.002971761671702 + }, + { + "epoch": 5, + "eval_accuracy": 0.6908, + "eval_loss": 0.9096707763671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.186301827430725, + "mean_feedback_forward_cosine": 0.5885937731516989, + "mean_feedback_forward_relative_error": 0.9130983446773729, + "min_feedback_forward_cosine": 0.3299712836742401 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0052571458855342305, + "instruction_rms": 0.005257145887047622, + "raw_apical_rms": 0.025175455443176575, + "teaching_rms": 0.0052571458855342305, + "traffic_rms": 0.02461522301585032 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104108318686485, + "maximum_absolute_post_projection_soma_slope": 1.4709918616517825e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.762725017129495e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007898785010883138, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.8739679383595784 + }, + { + "epoch": 6, + "eval_accuracy": 0.6788, + "eval_loss": 0.9818740600585938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.137481927871704, + "mean_feedback_forward_cosine": 0.6417811418834486, + "mean_feedback_forward_relative_error": 0.8515592123332777, + "min_feedback_forward_cosine": 0.3858734965324402 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00511800449876425, + "instruction_rms": 0.005118004498437255, + "raw_apical_rms": 0.026285692882644283, + "teaching_rms": 0.00511800449876425, + "traffic_rms": 0.025778497751821908 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.2827462647635457e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7474045133965685e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008272124776851554, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7683374342070686 + }, + { + "epoch": 7, + "eval_accuracy": 0.697, + "eval_loss": 0.8805564208984376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0916686058044434, + "mean_feedback_forward_cosine": 0.6830935258614389, + "mean_feedback_forward_relative_error": 0.8006800193535654, + "min_feedback_forward_cosine": 0.4360256791114807 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005090637227480062, + "instruction_rms": 0.005090637226330095, + "raw_apical_rms": 0.02710135754344636, + "teaching_rms": 0.005090637227480062, + "traffic_rms": 0.026614705942795136 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.2188740683427568e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7337732104623406e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008402664048491541, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.700017037179735 + }, + { + "epoch": 8, + "eval_accuracy": 0.7582, + "eval_loss": 0.7145669799804687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0521509647369385, + "mean_feedback_forward_cosine": 0.7155027107188576, + "mean_feedback_forward_relative_error": 0.7585068752891139, + "min_feedback_forward_cosine": 0.4769858717918396 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005100889991883216, + "instruction_rms": 0.005100889989612128, + "raw_apical_rms": 0.02799491408125279, + "teaching_rms": 0.005100889991883216, + "traffic_rms": 0.02752234136034223 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2122367998301797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7188394376933303e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008349236372860024, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.6431639823065863 + }, + { + "epoch": 9, + "eval_accuracy": 0.714, + "eval_loss": 0.831437353515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.01520574092865, + "mean_feedback_forward_cosine": 0.7427803372081957, + "mean_feedback_forward_relative_error": 0.7210724385161149, + "min_feedback_forward_cosine": 0.5141153335571289 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005100730785606785, + "instruction_rms": 0.00510073078782863, + "raw_apical_rms": 0.028489280379951753, + "teaching_rms": 0.005100730785606785, + "traffic_rms": 0.028024566431863063 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1568369373549103e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7155556621068902e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008549445861177576, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5943075577841864 + }, + { + "epoch": 10, + "eval_accuracy": 0.7462, + "eval_loss": 0.8459345947265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9798910617828369, + "mean_feedback_forward_cosine": 0.7666835596686915, + "mean_feedback_forward_relative_error": 0.6867316757377825, + "min_feedback_forward_cosine": 0.5473746061325073 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005082316784895747, + "instruction_rms": 0.005082316785262091, + "raw_apical_rms": 0.028900599778495628, + "teaching_rms": 0.005082316784895747, + "traffic_rms": 0.028445411560596746 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104106456041336, + "maximum_absolute_post_projection_soma_slope": 1.2107207680855936e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7078874661949912e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008347870767561681, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5634444468127356 + }, + { + "epoch": 11, + "eval_accuracy": 0.7894, + "eval_loss": 0.6491141357421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9458162784576416, + "mean_feedback_forward_cosine": 0.787112085442794, + "mean_feedback_forward_relative_error": 0.6555717791381636, + "min_feedback_forward_cosine": 0.5784483551979065 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005123974302786634, + "instruction_rms": 0.005123974300594778, + "raw_apical_rms": 0.029270407704884427, + "teaching_rms": 0.005123974302786634, + "traffic_rms": 0.028813281735924453 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1776799091478551e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7015997192714634e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007876796654479682, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.5321417511728075 + }, + { + "epoch": 12, + "eval_accuracy": 0.7916, + "eval_loss": 0.6628656616210937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9140507578849792, + "mean_feedback_forward_cosine": 0.8051042274424904, + "mean_feedback_forward_relative_error": 0.6269298242895227, + "min_feedback_forward_cosine": 0.6064649820327759 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005089625208600715, + "instruction_rms": 0.005089625207450145, + "raw_apical_rms": 0.02950235115080955, + "teaching_rms": 0.005089625208600715, + "traffic_rms": 0.02905536991273402 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1587401260726438e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.697508297887163e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008881822530006531, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.5126900021764967 + }, + { + "epoch": 13, + "eval_accuracy": 0.7438, + "eval_loss": 0.8314508178710938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8837882280349731, + "mean_feedback_forward_cosine": 0.8213234888879877, + "mean_feedback_forward_relative_error": 0.5999040070332979, + "min_feedback_forward_cosine": 0.6320071220397949 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005123668963105766, + "instruction_rms": 0.005123668963803649, + "raw_apical_rms": 0.030016462374151044, + "teaching_rms": 0.005123668963105766, + "traffic_rms": 0.029570846120130275 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.243258118677204e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.69923774723968e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008417570457143489, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.4830100783560011 + }, + { + "epoch": 14, + "eval_accuracy": 0.8082, + "eval_loss": 0.6025417114257813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.854644775390625, + "mean_feedback_forward_cosine": 0.8355686758693895, + "mean_feedback_forward_relative_error": 0.5752676631274977, + "min_feedback_forward_cosine": 0.6560328006744385 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005139756653881987, + "instruction_rms": 0.005139756654406729, + "raw_apical_rms": 0.03028349071583645, + "teaching_rms": 0.005139756653881987, + "traffic_rms": 0.029838919366939717 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3025974077152114e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.696470114009614e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008806346988294534, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.463170660188463 + }, + { + "epoch": 15, + "eval_accuracy": 0.805, + "eval_loss": 0.6055401611328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8269039988517761, + "mean_feedback_forward_cosine": 0.8486043779473555, + "mean_feedback_forward_relative_error": 0.5517268729837317, + "min_feedback_forward_cosine": 0.6777786016464233 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005103417485489563, + "instruction_rms": 0.0051034174849611015, + "raw_apical_rms": 0.03061214404180109, + "teaching_rms": 0.005103417485489563, + "traffic_rms": 0.030178261546094633 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4411319249063581e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6896146117949813e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008763286921123518, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.4528525505039427 + }, + { + "epoch": 16, + "eval_accuracy": 0.7672, + "eval_loss": 0.7340782592773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7998066544532776, + "mean_feedback_forward_cosine": 0.8603590287660298, + "mean_feedback_forward_relative_error": 0.5296951375509563, + "min_feedback_forward_cosine": 0.6985958814620972 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005163668726531768, + "instruction_rms": 0.005163668725697001, + "raw_apical_rms": 0.030868834889039496, + "teaching_rms": 0.005163668726531768, + "traffic_rms": 0.030428273299187767 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2748145650220977e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6880693308318265e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009315669449123565, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.43222274622917173 + }, + { + "epoch": 17, + "eval_accuracy": 0.805, + "eval_loss": 0.6171972778320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7732645273208618, + "mean_feedback_forward_cosine": 0.8710650958512959, + "mean_feedback_forward_relative_error": 0.5087922160562716, + "min_feedback_forward_cosine": 0.7182590961456299 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005167402571857455, + "instruction_rms": 0.005167402572632576, + "raw_apical_rms": 0.03102051479275261, + "teaching_rms": 0.005167402571857455, + "traffic_rms": 0.03058101220474407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1738018557139185e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.682641138467952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0090117500951202, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.4198982719315423 + }, + { + "epoch": 18, + "eval_accuracy": 0.8326, + "eval_loss": 0.5000441101074219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7474690675735474, + "mean_feedback_forward_cosine": 0.8805968604589763, + "mean_feedback_forward_relative_error": 0.48924355130446584, + "min_feedback_forward_cosine": 0.7365822196006775 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005130215386217267, + "instruction_rms": 0.00513021538634713, + "raw_apical_rms": 0.03141726497064013, + "teaching_rms": 0.005130215386217267, + "traffic_rms": 0.0309897016667005 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104104593396187, + "maximum_absolute_post_projection_soma_slope": 1.1615486350535775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6820495615530207e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009317042634727535, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.407132633696662 + }, + { + "epoch": 19, + "eval_accuracy": 0.827, + "eval_loss": 0.5370485595703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7219812273979187, + "mean_feedback_forward_cosine": 0.8894854093852796, + "mean_feedback_forward_relative_error": 0.4705220602060619, + "min_feedback_forward_cosine": 0.7542740106582642 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005203601837145817, + "instruction_rms": 0.00520360183613507, + "raw_apical_rms": 0.031504752858528745, + "teaching_rms": 0.005203601837145817, + "traffic_rms": 0.031065647092155167 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.0613388390368073e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6718240273237736e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009022047554874449, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.39371106230417885 + }, + { + "epoch": 20, + "eval_accuracy": 0.8186, + "eval_loss": 0.5529235717773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.697158694267273, + "mean_feedback_forward_cosine": 0.8974085512914156, + "mean_feedback_forward_relative_error": 0.45305699972729935, + "min_feedback_forward_cosine": 0.7705336809158325 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005105721991124259, + "instruction_rms": 0.005105721989551596, + "raw_apical_rms": 0.03167865650987401, + "teaching_rms": 0.005105721991124259, + "traffic_rms": 0.03125853040121373 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.0572734687741558e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.685083141653272e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008546781791054408, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3791846862051222 + }, + { + "epoch": 21, + "eval_accuracy": 0.8418, + "eval_loss": 0.5103697387695313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6728779077529907, + "mean_feedback_forward_cosine": 0.9049001775289837, + "mean_feedback_forward_relative_error": 0.43597892554182754, + "min_feedback_forward_cosine": 0.7860504984855652 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005157035474318466, + "instruction_rms": 0.00515703547400123, + "raw_apical_rms": 0.032001719221733, + "teaching_rms": 0.005157035474318466, + "traffic_rms": 0.03157677551437884 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4813031690152911e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6796041782521214e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008514171408789961, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.37004463691711426 + }, + { + "epoch": 22, + "eval_accuracy": 0.832, + "eval_loss": 0.5088022521972656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6498216986656189, + "mean_feedback_forward_cosine": 0.9116312077170924, + "mean_feedback_forward_relative_error": 0.4200763890617772, + "min_feedback_forward_cosine": 0.8003801703453064 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005106973395853267, + "instruction_rms": 0.005106973397976795, + "raw_apical_rms": 0.03216938340004663, + "teaching_rms": 0.005106973395853267, + "traffic_rms": 0.0317553491245528 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.1584300629863264e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6823495916847997e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008189136306340775, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3618061388068729 + }, + { + "epoch": 23, + "eval_accuracy": 0.8436, + "eval_loss": 0.4814773193359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6265225410461426, + "mean_feedback_forward_cosine": 0.9179497612150092, + "mean_feedback_forward_relative_error": 0.4045995630716023, + "min_feedback_forward_cosine": 0.8143157958984375 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005055386002004759, + "instruction_rms": 0.005055386002356951, + "raw_apical_rms": 0.0322901928117579, + "teaching_rms": 0.005055386002004759, + "traffic_rms": 0.03188634064501978 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.0838371977683892e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6741981811622218e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00858211427408929, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.35613183804617987 + }, + { + "epoch": 24, + "eval_accuracy": 0.8344, + "eval_loss": 0.51261796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6042298078536987, + "mean_feedback_forward_cosine": 0.9237859217744124, + "mean_feedback_forward_relative_error": 0.3897753475528014, + "min_feedback_forward_cosine": 0.8270023465156555 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005026311796011831, + "instruction_rms": 0.005026311794119769, + "raw_apical_rms": 0.03239631130747826, + "teaching_rms": 0.005026311796011831, + "traffic_rms": 0.031997764570577646 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.115325076739282e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6719706724083858e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008635285323910395, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.3461827133178711 + }, + { + "epoch": 25, + "eval_accuracy": 0.8376, + "eval_loss": 0.4821809814453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5829207301139832, + "mean_feedback_forward_cosine": 0.9291425284586454, + "mean_feedback_forward_relative_error": 0.37564187300832647, + "min_feedback_forward_cosine": 0.8387241959571838 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005042792125698084, + "instruction_rms": 0.005042792126893753, + "raw_apical_rms": 0.03245913087995026, + "teaching_rms": 0.005042792125698084, + "traffic_rms": 0.03205849323489724 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.2731623755257715e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6674630115957838e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00855959767109689, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.3370492530769772 + }, + { + "epoch": 26, + "eval_accuracy": 0.8434, + "eval_loss": 0.495537744140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5627683997154236, + "mean_feedback_forward_cosine": 0.9340524171528063, + "mean_feedback_forward_relative_error": 0.3622163777288638, + "min_feedback_forward_cosine": 0.8494534492492676 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0049591596564131655, + "instruction_rms": 0.004959159657579335, + "raw_apical_rms": 0.03265736445835238, + "teaching_rms": 0.0049591596564131655, + "traffic_rms": 0.03227297794244435 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4674928827673739e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6686001064386828e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008467157723614333, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.3300551345931159 + }, + { + "epoch": 27, + "eval_accuracy": 0.8342, + "eval_loss": 0.5357415649414062, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5433613657951355, + "mean_feedback_forward_cosine": 0.9385507389118797, + "mean_feedback_forward_relative_error": 0.3495171783786071, + "min_feedback_forward_cosine": 0.8595328330993652 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0049489392695670105, + "instruction_rms": 0.004948939268289771, + "raw_apical_rms": 0.03277365859859984, + "teaching_rms": 0.0049489392695670105, + "traffic_rms": 0.03239129424735014 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.031104102730751038, + "maximum_absolute_post_projection_soma_slope": 1.1465401072996428e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6669398061146425e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008833973657661845, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.32522023562060465 + }, + { + "epoch": 28, + "eval_accuracy": 0.8584, + "eval_loss": 0.439326318359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5248848795890808, + "mean_feedback_forward_cosine": 0.9427224648626227, + "mean_feedback_forward_relative_error": 0.33729491735759537, + "min_feedback_forward_cosine": 0.8687142729759216 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005033935666566265, + "instruction_rms": 0.005033935664838083, + "raw_apical_rms": 0.03299414083397847, + "teaching_rms": 0.005033935666566265, + "traffic_rms": 0.032601410542060845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3903027173967075e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6688990737640255e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00887849304205244, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.3156820275094774 + }, + { + "epoch": 29, + "eval_accuracy": 0.8448, + "eval_loss": 0.4970781555175781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5062661170959473, + "mean_feedback_forward_cosine": 0.9466551416798642, + "mean_feedback_forward_relative_error": 0.3253671848460248, + "min_feedback_forward_cosine": 0.8777074813842773 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004954867700972754, + "instruction_rms": 0.004954867702669111, + "raw_apical_rms": 0.03305198524729274, + "teaching_rms": 0.004954867700972754, + "traffic_rms": 0.032672122765333926 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.325884468883487e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.657915232911312e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008695523989195557, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.31194691899087695 + }, + { + "epoch": 30, + "eval_accuracy": 0.858, + "eval_loss": 0.451676953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4891993999481201, + "mean_feedback_forward_cosine": 0.9501831343299464, + "mean_feedback_forward_relative_error": 0.31429174542427063, + "min_feedback_forward_cosine": 0.8856332898139954 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004896729235509657, + "instruction_rms": 0.00489672923643095, + "raw_apical_rms": 0.033082976683124876, + "teaching_rms": 0.004896729235509657, + "traffic_rms": 0.0327127175154366 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1145472988971505e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6612049666700516e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009022131846671251, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.30540858267678156 + }, + { + "epoch": 31, + "eval_accuracy": 0.8392, + "eval_loss": 0.5462397277832032, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4719200134277344, + "mean_feedback_forward_cosine": 0.9535760408953616, + "mean_feedback_forward_relative_error": 0.3032335952708596, + "min_feedback_forward_cosine": 0.8934358358383179 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004958124178690693, + "instruction_rms": 0.004958124178351577, + "raw_apical_rms": 0.033261925680794766, + "teaching_rms": 0.004958124178690693, + "traffic_rms": 0.03288311390801063 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1543311195794104e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6575105475981898e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008333529629313366, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.30266087454160057 + }, + { + "epoch": 32, + "eval_accuracy": 0.8036, + "eval_loss": 0.682837109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45478224754333496, + "mean_feedback_forward_cosine": 0.956767483761436, + "mean_feedback_forward_relative_error": 0.29247854414739105, + "min_feedback_forward_cosine": 0.9008991718292236 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004953896585750713, + "instruction_rms": 0.0049538965852565925, + "raw_apical_rms": 0.033257915393122, + "teaching_rms": 0.004953896585750713, + "traffic_rms": 0.03288020627538047 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1645643560598273e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6531492453302036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008216337933057904, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.29637968017260236 + }, + { + "epoch": 33, + "eval_accuracy": 0.8454, + "eval_loss": 0.49532133178710935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43830615282058716, + "mean_feedback_forward_cosine": 0.9597060868614599, + "mean_feedback_forward_relative_error": 0.2823241635372764, + "min_feedback_forward_cosine": 0.907834529876709 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004970880337995713, + "instruction_rms": 0.004970880337158342, + "raw_apical_rms": 0.033359007890836304, + "teaching_rms": 0.004970880337995713, + "traffic_rms": 0.032979909608127016 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2482241018574314e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6543463971984505e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008041048452023991, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.294988309123781 + }, + { + "epoch": 34, + "eval_accuracy": 0.8534, + "eval_loss": 0.4693566711425781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4227277338504791, + "mean_feedback_forward_cosine": 0.9624063247128537, + "mean_feedback_forward_relative_error": 0.2726022440351938, + "min_feedback_forward_cosine": 0.9141800999641418 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004892167475285149, + "instruction_rms": 0.004892167475637795, + "raw_apical_rms": 0.03358519006547927, + "teaching_rms": 0.004892167475285149, + "traffic_rms": 0.03321982884891977 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1676708489005705e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.653744592060938e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007941698955564375, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.29385936484071945 + }, + { + "epoch": 35, + "eval_accuracy": 0.8324, + "eval_loss": 0.5749091674804687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40767407417297363, + "mean_feedback_forward_cosine": 0.9649951959911146, + "mean_feedback_forward_relative_error": 0.262962150338449, + "min_feedback_forward_cosine": 0.9200702905654907 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004980105439659693, + "instruction_rms": 0.0049801054402052606, + "raw_apical_rms": 0.03347855324387941, + "teaching_rms": 0.004980105439659693, + "traffic_rms": 0.03309860209702309 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4705655360103265e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6504644458464934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007713812389435235, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.28374992145167455 + }, + { + "epoch": 36, + "eval_accuracy": 0.8426, + "eval_loss": 0.4943857971191406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.39344385266304016, + "mean_feedback_forward_cosine": 0.9673068209698326, + "mean_feedback_forward_relative_error": 0.25407097998418304, + "min_feedback_forward_cosine": 0.9254612922668457 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00480167981515975, + "instruction_rms": 0.004801679816594659, + "raw_apical_rms": 0.0337876464262403, + "teaching_rms": 0.00480167981515975, + "traffic_rms": 0.03343744331605359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.18527738734997e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6540505792612133e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008233388511928756, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.27896705027686225 + }, + { + "epoch": 37, + "eval_accuracy": 0.8496, + "eval_loss": 0.475590234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3787844777107239, + "mean_feedback_forward_cosine": 0.9695587503282648, + "mean_feedback_forward_relative_error": 0.24504558938114265, + "min_feedback_forward_cosine": 0.9308475255966187 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004897628023750723, + "instruction_rms": 0.004897628022957946, + "raw_apical_rms": 0.03391378027919636, + "teaching_rms": 0.004897628023750723, + "traffic_rms": 0.03355177070229845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4692020933182448e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.652624532061659e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0080768554028717, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.28317848139074114 + }, + { + "epoch": 38, + "eval_accuracy": 0.8388, + "eval_loss": 0.5251111755371094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.36561089754104614, + "mean_feedback_forward_cosine": 0.9715913094972309, + "mean_feedback_forward_relative_error": 0.23662489613420085, + "min_feedback_forward_cosine": 0.9354627132415771 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004867969835435969, + "instruction_rms": 0.004867969835411105, + "raw_apical_rms": 0.034011534417579974, + "teaching_rms": 0.004867969835435969, + "traffic_rms": 0.03365475892133878 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4360414191116888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.640010561246279e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007765667090728688, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.2726616321086884 + }, + { + "epoch": 39, + "eval_accuracy": 0.8516, + "eval_loss": 0.5048835876464843, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.35223522782325745, + "mean_feedback_forward_cosine": 0.9735078090115598, + "mean_feedback_forward_relative_error": 0.22845008381103216, + "min_feedback_forward_cosine": 0.9399992823600769 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004801664601956974, + "instruction_rms": 0.004801664602546571, + "raw_apical_rms": 0.03404760170530794, + "teaching_rms": 0.004801664601956974, + "traffic_rms": 0.033700256476151186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.088730616771727e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6503538103055278e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00793716446790211, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.27212047130266825 + }, + { + "epoch": 40, + "eval_accuracy": 0.8526, + "eval_loss": 0.4644252502441406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33981913328170776, + "mean_feedback_forward_cosine": 0.9752821106659738, + "mean_feedback_forward_relative_error": 0.22056759305690465, + "min_feedback_forward_cosine": 0.9440567493438721 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004814310388714407, + "instruction_rms": 0.004814310390229436, + "raw_apical_rms": 0.03398011142037965, + "teaching_rms": 0.004814310388714407, + "traffic_rms": 0.03363070354682049 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4484050403495985e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6436851105300998e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008074874997465701, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.26740257726775274 + }, + { + "epoch": 41, + "eval_accuracy": 0.8676, + "eval_loss": 0.4324529357910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32735109329223633, + "mean_feedback_forward_cosine": 0.976964185112401, + "mean_feedback_forward_relative_error": 0.21286306843945854, + "min_feedback_forward_cosine": 0.9480441212654114 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004837030249174624, + "instruction_rms": 0.004837030249051573, + "raw_apical_rms": 0.03398783325450356, + "teaching_rms": 0.004837030249174624, + "traffic_rms": 0.033634786648000445 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1495266960537265e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6450261607396043e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008407791255391525, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.26583909475538464 + }, + { + "epoch": 42, + "eval_accuracy": 0.8476, + "eval_loss": 0.5370885559082031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3157062828540802, + "mean_feedback_forward_cosine": 0.978500065050627, + "mean_feedback_forward_relative_error": 0.20558425548829531, + "min_feedback_forward_cosine": 0.9516211748123169 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004794390846475768, + "instruction_rms": 0.004794390848540248, + "raw_apical_rms": 0.03408740414116479, + "teaching_rms": 0.004794390846475768, + "traffic_rms": 0.033739975034218224 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.380817327145678e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641002205330793e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008418045090464961, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.2626739225387573 + }, + { + "epoch": 43, + "eval_accuracy": 0.8474, + "eval_loss": 0.5109385070800782, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.304364949464798, + "mean_feedback_forward_cosine": 0.9799450290830511, + "mean_feedback_forward_relative_error": 0.19850653723666542, + "min_feedback_forward_cosine": 0.9549909234046936 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004773650322553887, + "instruction_rms": 0.004773650322350811, + "raw_apical_rms": 0.03410092191690993, + "teaching_rms": 0.004773650322553887, + "traffic_rms": 0.03375879169793919 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2696138362855436e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.646613554126838e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008801516921465368, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.2619066998057895 + }, + { + "epoch": 44, + "eval_accuracy": 0.8382, + "eval_loss": 0.5117140441894531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29309210181236267, + "mean_feedback_forward_cosine": 0.9813267180794164, + "mean_feedback_forward_relative_error": 0.1915239977993463, + "min_feedback_forward_cosine": 0.9582096338272095 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0048636472342842105, + "instruction_rms": 0.0048636472349773505, + "raw_apical_rms": 0.03406931260972435, + "teaching_rms": 0.0048636472342842105, + "traffic_rms": 0.03371260529755212 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.663171822485765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641406773724961e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008529031631681387, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.258253004556232 + }, + { + "epoch": 45, + "eval_accuracy": 0.8524, + "eval_loss": 0.49885089111328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28267934918403625, + "mean_feedback_forward_cosine": 0.9825521017375746, + "mean_feedback_forward_relative_error": 0.18504575914458224, + "min_feedback_forward_cosine": 0.9610792398452759 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004776449581523731, + "instruction_rms": 0.004776449580832663, + "raw_apical_rms": 0.03405220258513439, + "teaching_rms": 0.004776449581523731, + "traffic_rms": 0.03370789796385943 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1919964570950015e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.646861215258964e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009121332944048951, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.25226737350887724 + }, + { + "epoch": 46, + "eval_accuracy": 0.8322, + "eval_loss": 0.5687988708496093, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27236035466194153, + "mean_feedback_forward_cosine": 0.9837380020241988, + "mean_feedback_forward_relative_error": 0.1786251789645145, + "min_feedback_forward_cosine": 0.9638214111328125 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004811620708506466, + "instruction_rms": 0.004811620707522746, + "raw_apical_rms": 0.03409326012300677, + "teaching_rms": 0.004811620708506466, + "traffic_rms": 0.033744104574039396 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3385352382044857e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6453304569253207e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008623667473082434, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.2542002528270086 + }, + { + "epoch": 47, + "eval_accuracy": 0.8648, + "eval_loss": 0.4691862121582031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2622673809528351, + "mean_feedback_forward_cosine": 0.9848273239637676, + "mean_feedback_forward_relative_error": 0.17249753718313418, + "min_feedback_forward_cosine": 0.966410756111145 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0047637483264667815, + "instruction_rms": 0.00476374832466967, + "raw_apical_rms": 0.03431797937572261, + "teaching_rms": 0.0047637483264667815, + "traffic_rms": 0.033978842904995504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2311806685261217e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.647349410140757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008973244018344448, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.2472731151368883 + }, + { + "epoch": 48, + "eval_accuracy": 0.8774, + "eval_loss": 0.390655712890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25277069211006165, + "mean_feedback_forward_cosine": 0.9858405025381791, + "mean_feedback_forward_relative_error": 0.16660764774209574, + "min_feedback_forward_cosine": 0.968781590461731 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004734057300343864, + "instruction_rms": 0.004734057300560648, + "raw_apical_rms": 0.034416352698988704, + "teaching_rms": 0.004734057300343864, + "traffic_rms": 0.034082861947264044 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.5418537557820855e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6381597280930896e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008873849990073814, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.24886706942982142 + }, + { + "epoch": 49, + "eval_accuracy": 0.8566, + "eval_loss": 0.4763729858398438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24390378594398499, + "mean_feedback_forward_cosine": 0.9867772993288542, + "mean_feedback_forward_relative_error": 0.1609509493175306, + "min_feedback_forward_cosine": 0.9709120392799377 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00472433885555469, + "instruction_rms": 0.0047243388555071745, + "raw_apical_rms": 0.034407930029486083, + "teaching_rms": 0.00472433885555469, + "traffic_rms": 0.03407515877199275 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.366304402949936e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6450807803613235e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008432729201866343, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.24381286273797353 + }, + { + "epoch": 50, + "eval_accuracy": 0.8546, + "eval_loss": 0.5009172668457031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23519784212112427, + "mean_feedback_forward_cosine": 0.9876453939237093, + "mean_feedback_forward_relative_error": 0.15553969811452062, + "min_feedback_forward_cosine": 0.9729282259941101 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004606681932301185, + "instruction_rms": 0.0046066819324219925, + "raw_apical_rms": 0.03445817750563184, + "teaching_rms": 0.004606681932301185, + "traffic_rms": 0.034141882132894624 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3151639777220225e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6364552935181166e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008785256886105718, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.24616385930114323 + }, + { + "epoch": 51, + "eval_accuracy": 0.84, + "eval_loss": 0.5515082946777343, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2269156277179718, + "mean_feedback_forward_cosine": 0.988458187956559, + "mean_feedback_forward_relative_error": 0.15030134508484289, + "min_feedback_forward_cosine": 0.9747790098190308 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0046441777956041635, + "instruction_rms": 0.004644177795562959, + "raw_apical_rms": 0.03444990691445392, + "teaching_rms": 0.0046441777956041635, + "traffic_rms": 0.034127614922136666 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.191775833575548e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.634738683197308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008759819974824733, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.23702717757754857 + }, + { + "epoch": 52, + "eval_accuracy": 0.8618, + "eval_loss": 0.4465389404296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21880142390727997, + "mean_feedback_forward_cosine": 0.9892266361336959, + "mean_feedback_forward_relative_error": 0.1451731159498817, + "min_feedback_forward_cosine": 0.9765316843986511 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004629176862530896, + "instruction_rms": 0.004629176863975987, + "raw_apical_rms": 0.034563132246923416, + "teaching_rms": 0.004629176862530896, + "traffic_rms": 0.034245299133160385 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4713018359202579e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6315995941463887e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008599320273287768, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.2346235758781433 + }, + { + "epoch": 53, + "eval_accuracy": 0.8478, + "eval_loss": 0.5581020751953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.21077044308185577, + "mean_feedback_forward_cosine": 0.9899541767019975, + "mean_feedback_forward_relative_error": 0.14016883134057648, + "min_feedback_forward_cosine": 0.9782070517539978 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004693637465156337, + "instruction_rms": 0.0046936374643177506, + "raw_apical_rms": 0.03475809665475123, + "teaching_rms": 0.004693637465156337, + "traffic_rms": 0.03443199783267906 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409714281559, + "maximum_absolute_post_projection_soma_slope": 1.0264071370613692e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6348604158506463e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007993278147103374, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.23640003298123677 + }, + { + "epoch": 54, + "eval_accuracy": 0.8128, + "eval_loss": 0.6650457153320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2032177746295929, + "mean_feedback_forward_cosine": 0.9906056209614402, + "mean_feedback_forward_relative_error": 0.13552678729358472, + "min_feedback_forward_cosine": 0.9797242879867554 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004569560428927146, + "instruction_rms": 0.004569560428360455, + "raw_apical_rms": 0.03459364877157499, + "teaching_rms": 0.004569560428927146, + "traffic_rms": 0.03428311090995729 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.284697415115943e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6305371842757283e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008608573884396474, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.2298361652188831 + }, + { + "epoch": 55, + "eval_accuracy": 0.8632, + "eval_loss": 0.47038958129882813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1957596242427826, + "mean_feedback_forward_cosine": 0.9912525101711875, + "mean_feedback_forward_relative_error": 0.13075828414998555, + "min_feedback_forward_cosine": 0.9811676740646362 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0046634500385850965, + "instruction_rms": 0.004663450037787709, + "raw_apical_rms": 0.03473260594777723, + "teaching_rms": 0.0046634500385850965, + "traffic_rms": 0.03440977488178481 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1937872912426428e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6346369070560357e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00887763029480918, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.23555502649413215 + }, + { + "epoch": 56, + "eval_accuracy": 0.8548, + "eval_loss": 0.47070689086914064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18890292942523956, + "mean_feedback_forward_cosine": 0.9918342389558491, + "mean_feedback_forward_relative_error": 0.12631950586249954, + "min_feedback_forward_cosine": 0.9824573397636414 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00462494777569615, + "instruction_rms": 0.0046249477761719626, + "raw_apical_rms": 0.03483555329347786, + "teaching_rms": 0.00462494777569615, + "traffic_rms": 0.034519437051945545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.5332638270137977e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6341594016900744e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008304118629929533, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.23004417071607378 + }, + { + "epoch": 57, + "eval_accuracy": 0.8586, + "eval_loss": 0.46194222412109376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18201766908168793, + "mean_feedback_forward_cosine": 0.9923800041801051, + "mean_feedback_forward_relative_error": 0.12201094215637759, + "min_feedback_forward_cosine": 0.9837061166763306 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004623286000372373, + "instruction_rms": 0.004623285999568594, + "raw_apical_rms": 0.0347715408729308, + "teaching_rms": 0.004623286000372373, + "traffic_rms": 0.03445529255125391 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4529461189738413e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6349651030211643e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008987323391636841, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.23447738461494447 + }, + { + "epoch": 58, + "eval_accuracy": 0.8776, + "eval_loss": 0.39425791625976564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17561373114585876, + "mean_feedback_forward_cosine": 0.9928765798869886, + "mean_feedback_forward_relative_error": 0.11794576519413998, + "min_feedback_forward_cosine": 0.9848186373710632 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004489613265887848, + "instruction_rms": 0.004489613266076248, + "raw_apical_rms": 0.03469464289829862, + "teaching_rms": 0.004489613265887848, + "traffic_rms": 0.03439626084777562 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.0672150274615433e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6355677912489722e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009090677963991408, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.22616133080588446 + }, + { + "epoch": 59, + "eval_accuracy": 0.8214, + "eval_loss": 0.6318106689453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16960246860980988, + "mean_feedback_forward_cosine": 0.9933565196238066, + "mean_feedback_forward_relative_error": 0.11387859382911732, + "min_feedback_forward_cosine": 0.9858263731002808 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00457532860681517, + "instruction_rms": 0.004575328606559777, + "raw_apical_rms": 0.03477156502703751, + "teaching_rms": 0.00457532860681517, + "traffic_rms": 0.03446172136504439 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3520507380349045e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6335615320188663e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009238969431445714, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.22260379991266463 + }, + { + "epoch": 60, + "eval_accuracy": 0.878, + "eval_loss": 0.4137625183105469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16372554004192352, + "mean_feedback_forward_cosine": 0.9937896226581774, + "mean_feedback_forward_relative_error": 0.11009067689117633, + "min_feedback_forward_cosine": 0.9867862462997437 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004529203460647302, + "instruction_rms": 0.004529203459621877, + "raw_apical_rms": 0.03463304539482143, + "teaching_rms": 0.004529203460647302, + "traffic_rms": 0.034327751484534805 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4452857577396117e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.633967674301304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009334968467973907, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.22208338684505888 + }, + { + "epoch": 61, + "eval_accuracy": 0.8534, + "eval_loss": 0.516475, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1577123999595642, + "mean_feedback_forward_cosine": 0.9942108737794977, + "mean_feedback_forward_relative_error": 0.1062790857333886, + "min_feedback_forward_cosine": 0.9877330660820007 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004550236786852152, + "instruction_rms": 0.004550236786631306, + "raw_apical_rms": 0.034765179154599105, + "teaching_rms": 0.004550236786852152, + "traffic_rms": 0.03445819330376294 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1467695237854514e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.631810116574943e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010080600596530591, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.22619280171129438 + }, + { + "epoch": 62, + "eval_accuracy": 0.8586, + "eval_loss": 0.442544677734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15221844613552094, + "mean_feedback_forward_cosine": 0.9945927325047945, + "mean_feedback_forward_relative_error": 0.10269431064003393, + "min_feedback_forward_cosine": 0.9885700941085815 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004521705374212668, + "instruction_rms": 0.004521705374268841, + "raw_apical_rms": 0.034761470487341545, + "teaching_rms": 0.004521705374212668, + "traffic_rms": 0.034458362299596604 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2999428200544116e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.632537082100106e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00985824219365451, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.21663891365263196 + }, + { + "epoch": 63, + "eval_accuracy": 0.855, + "eval_loss": 0.47691630249023437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14677031338214874, + "mean_feedback_forward_cosine": 0.9949565529823303, + "mean_feedback_forward_relative_error": 0.09916972682664268, + "min_feedback_forward_cosine": 0.9893686771392822 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00454800645576982, + "instruction_rms": 0.0045480064540945105, + "raw_apical_rms": 0.034746996973843966, + "teaching_rms": 0.00454800645576982, + "traffic_rms": 0.034440261084225494 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.231772195353642e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6402821822006248e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009285121608502507, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.2174586017343733 + }, + { + "epoch": 64, + "eval_accuracy": 0.8486, + "eval_loss": 0.5087167785644531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14159931242465973, + "mean_feedback_forward_cosine": 0.995289661382374, + "mean_feedback_forward_relative_error": 0.09583163555515439, + "min_feedback_forward_cosine": 0.9900988340377808 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004509878056473776, + "instruction_rms": 0.004509878057411221, + "raw_apical_rms": 0.03485475855180221, + "teaching_rms": 0.004509878056473776, + "traffic_rms": 0.03455458867414804 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4009420290506114e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6325254003801997e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009322428344413306, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.21440663772159152 + }, + { + "epoch": 65, + "eval_accuracy": 0.8446, + "eval_loss": 0.519659423828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1363605558872223, + "mean_feedback_forward_cosine": 0.9956220764862863, + "mean_feedback_forward_relative_error": 0.09237666682977426, + "min_feedback_forward_cosine": 0.9908114075660706 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0046595919890802945, + "instruction_rms": 0.00465959199033466, + "raw_apical_rms": 0.034852558301645464, + "teaching_rms": 0.0046595919890802945, + "traffic_rms": 0.03453176629947474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3025884371131724e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.636034339162178e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00948175308698507, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.22187283344003889 + }, + { + "epoch": 66, + "eval_accuracy": 0.8292, + "eval_loss": 0.5881011901855469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13144823908805847, + "mean_feedback_forward_cosine": 0.9959171947680021, + "mean_feedback_forward_relative_error": 0.08920571855024288, + "min_feedback_forward_cosine": 0.9914578199386597 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00458447076552368, + "instruction_rms": 0.004584470763683925, + "raw_apical_rms": 0.03497863145534947, + "teaching_rms": 0.00458447076552368, + "traffic_rms": 0.03467038425593533 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.143722805352354e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.635584598929081e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009966187537097727, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.216731480439504 + }, + { + "epoch": 67, + "eval_accuracy": 0.8674, + "eval_loss": 0.4599829772949219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12657660245895386, + "mean_feedback_forward_cosine": 0.9961990149397599, + "mean_feedback_forward_relative_error": 0.0860732936937558, + "min_feedback_forward_cosine": 0.9920773506164551 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004627145895933496, + "instruction_rms": 0.0046271458963005685, + "raw_apical_rms": 0.035152856415883046, + "teaching_rms": 0.004627145895933496, + "traffic_rms": 0.03483804983062403 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2837901408602193e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6286313526223627e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009312316831906867, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.2194969289912118 + }, + { + "epoch": 68, + "eval_accuracy": 0.8786, + "eval_loss": 0.39053208618164065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12211810797452927, + "mean_feedback_forward_cosine": 0.9964460071764494, + "mean_feedback_forward_relative_error": 0.08322563669399212, + "min_feedback_forward_cosine": 0.992622435092926 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004448222060488421, + "instruction_rms": 0.004448222060096612, + "raw_apical_rms": 0.03514675221017867, + "teaching_rms": 0.004448222060488421, + "traffic_rms": 0.03485672297521746 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.155842621614056e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.626279994344699e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00924205510891637, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.2129067963335249 + }, + { + "epoch": 69, + "eval_accuracy": 0.8682, + "eval_loss": 0.45594747314453127, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11784830689430237, + "mean_feedback_forward_cosine": 0.9966843002720883, + "mean_feedback_forward_relative_error": 0.08037736404098962, + "min_feedback_forward_cosine": 0.993126392364502 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004486322136916355, + "instruction_rms": 0.0044863221378644074, + "raw_apical_rms": 0.03523522815799778, + "teaching_rms": 0.004486322136916355, + "traffic_rms": 0.03494151811448261 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2787289449534e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6302305539835843e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009635511186971873, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.2087914967695872 + }, + { + "epoch": 70, + "eval_accuracy": 0.8676, + "eval_loss": 0.46101365966796876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11344674974679947, + "mean_feedback_forward_cosine": 0.9969156855031064, + "mean_feedback_forward_relative_error": 0.0775245220253342, + "min_feedback_forward_cosine": 0.9936275482177734 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0045926297245273235, + "instruction_rms": 0.004592629725361557, + "raw_apical_rms": 0.03517295844474793, + "teaching_rms": 0.0045926297245273235, + "traffic_rms": 0.034864906325467265 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3895605555092061e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6306149491126397e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010313904601028254, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.21919610788292354 + }, + { + "epoch": 71, + "eval_accuracy": 0.8624, + "eval_loss": 0.4905334533691406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10946271568536758, + "mean_feedback_forward_cosine": 0.9971191381153307, + "mean_feedback_forward_relative_error": 0.07491065189242363, + "min_feedback_forward_cosine": 0.9940650463104248 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004505364726091503, + "instruction_rms": 0.004505364726523862, + "raw_apical_rms": 0.035162062356154516, + "teaching_rms": 0.004505364726091503, + "traffic_rms": 0.03486381904572826 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.506950475516078e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6337605717311066e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009701712553423126, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.2061603936460283 + }, + { + "epoch": 72, + "eval_accuracy": 0.864, + "eval_loss": 0.45937369384765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10553739964962006, + "mean_feedback_forward_cosine": 0.9973140986342179, + "mean_feedback_forward_relative_error": 0.07233009212895443, + "min_feedback_forward_cosine": 0.9944818019866943 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004501120009738082, + "instruction_rms": 0.0045011200098905866, + "raw_apical_rms": 0.03523203479675473, + "teaching_rms": 0.004501120009738082, + "traffic_rms": 0.03493577899487651 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2663494253217777e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6291002096393187e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009789404177086205, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.20920034455723233 + }, + { + "epoch": 73, + "eval_accuracy": 0.8586, + "eval_loss": 0.49533743286132814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10173561424016953, + "mean_feedback_forward_cosine": 0.9974929408023232, + "mean_feedback_forward_relative_error": 0.0698717755118483, + "min_feedback_forward_cosine": 0.9948709607124329 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004452159829408188, + "instruction_rms": 0.004452159830043112, + "raw_apical_rms": 0.03521120323014893, + "teaching_rms": 0.004452159829408188, + "traffic_rms": 0.03492161503984042 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1709801128745312e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6294496252774047e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009982753524995102, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.20618399731847975 + }, + { + "epoch": 74, + "eval_accuracy": 0.8566, + "eval_loss": 0.4973161560058594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09814338386058807, + "mean_feedback_forward_cosine": 0.9976599059606853, + "mean_feedback_forward_relative_error": 0.06750352210120152, + "min_feedback_forward_cosine": 0.9952248334884644 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0044704650968832435, + "instruction_rms": 0.004470465096845052, + "raw_apical_rms": 0.03531770239480634, + "teaching_rms": 0.0044704650968832435, + "traffic_rms": 0.03502581041891627 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3497415629615261e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.628883202448988e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008991144034400933, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.2070677791595459 + }, + { + "epoch": 75, + "eval_accuracy": 0.8658, + "eval_loss": 0.44002984008789064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09478152543306351, + "mean_feedback_forward_cosine": 0.9978180370832744, + "mean_feedback_forward_relative_error": 0.06517410856720648, + "min_feedback_forward_cosine": 0.9955450296401978 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00454264692216653, + "instruction_rms": 0.004542646922724606, + "raw_apical_rms": 0.03538495497628492, + "teaching_rms": 0.00454264692216653, + "traffic_rms": 0.03508444157413726 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2500177781760158e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6245408513228918e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008954280103157295, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.20586039009624058 + }, + { + "epoch": 76, + "eval_accuracy": 0.8618, + "eval_loss": 0.5047085205078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09143461287021637, + "mean_feedback_forward_cosine": 0.9979652668300428, + "mean_feedback_forward_relative_error": 0.06293161380055703, + "min_feedback_forward_cosine": 0.9958520531654358 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004514038875925267, + "instruction_rms": 0.004514038875881944, + "raw_apical_rms": 0.03532667619049348, + "teaching_rms": 0.004514038875925267, + "traffic_rms": 0.035028996401786416 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2083551048647223e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.627984139976097e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008699853703649804, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.2045543732325236 + }, + { + "epoch": 77, + "eval_accuracy": 0.8472, + "eval_loss": 0.5431947448730469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08820625394582748, + "mean_feedback_forward_cosine": 0.9981047416988172, + "mean_feedback_forward_relative_error": 0.06073338058041899, + "min_feedback_forward_cosine": 0.9961393475532532 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004488939879508441, + "instruction_rms": 0.004488939879789814, + "raw_apical_rms": 0.03538401410533833, + "teaching_rms": 0.004488939879508441, + "traffic_rms": 0.035089421880606736 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.0987145415697341e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.626817068303352e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008779103060851395, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.20401948824723562 + }, + { + "epoch": 78, + "eval_accuracy": 0.883, + "eval_loss": 0.37970450439453124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08497180044651031, + "mean_feedback_forward_cosine": 0.9982345982601768, + "mean_feedback_forward_relative_error": 0.05861183295124456, + "min_feedback_forward_cosine": 0.9964160323143005 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004555921653594509, + "instruction_rms": 0.004555921654625884, + "raw_apical_rms": 0.03546975936725454, + "teaching_rms": 0.004555921653594509, + "traffic_rms": 0.035167905450692066 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.0623000257226067e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.621434190911347e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008460152146069481, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.20838114698727925 + }, + { + "epoch": 79, + "eval_accuracy": 0.8716, + "eval_loss": 0.4349828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08195152133703232, + "mean_feedback_forward_cosine": 0.9983531424873754, + "mean_feedback_forward_relative_error": 0.056606653293496685, + "min_feedback_forward_cosine": 0.996665358543396 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004485157069579081, + "instruction_rms": 0.004485157070267351, + "raw_apical_rms": 0.035540728599918, + "teaching_rms": 0.004485157069579081, + "traffic_rms": 0.035248704333594415 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2393795323362156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6170983902311632e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008678463501860812, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.20069760993321736 + }, + { + "epoch": 80, + "eval_accuracy": 0.8722, + "eval_loss": 0.4430991516113281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07911452651023865, + "mean_feedback_forward_cosine": 0.9984633263788725, + "mean_feedback_forward_relative_error": 0.05467654764652252, + "min_feedback_forward_cosine": 0.9968912601470947 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0045079332396336476, + "instruction_rms": 0.004507933238804002, + "raw_apical_rms": 0.035578414364218054, + "teaching_rms": 0.0045079332396336476, + "traffic_rms": 0.035283345328562615 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2718983199988543e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.626559093991407e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009245253864448234, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.19840980619854398 + }, + { + "epoch": 81, + "eval_accuracy": 0.8716, + "eval_loss": 0.4429517333984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07626857608556747, + "mean_feedback_forward_cosine": 0.9985685662219399, + "mean_feedback_forward_relative_error": 0.05277032405138016, + "min_feedback_forward_cosine": 0.9971103668212891 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0044664374004133875, + "instruction_rms": 0.004466437399506244, + "raw_apical_rms": 0.035601181321927945, + "teaching_rms": 0.0044664374004133875, + "traffic_rms": 0.03531270643760204 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.5420379639863313e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6280340575991765e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009392048812378368, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.20652776799466874 + }, + { + "epoch": 82, + "eval_accuracy": 0.8596, + "eval_loss": 0.4637076110839844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07356619089841843, + "mean_feedback_forward_cosine": 0.9986630929143805, + "mean_feedback_forward_relative_error": 0.051000150410752544, + "min_feedback_forward_cosine": 0.9973108172416687 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004467663117046105, + "instruction_rms": 0.004467663119095558, + "raw_apical_rms": 0.035543474657197496, + "teaching_rms": 0.004467663117046105, + "traffic_rms": 0.035254164064680245 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.5173352352348957e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.634132948920163e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009524069379349749, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.2017500485473209 + }, + { + "epoch": 83, + "eval_accuracy": 0.8726, + "eval_loss": 0.44147421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0709899514913559, + "mean_feedback_forward_cosine": 0.998748707143884, + "mean_feedback_forward_relative_error": 0.04933068960120803, + "min_feedback_forward_cosine": 0.9974950551986694 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004331244755192192, + "instruction_rms": 0.004331244754318333, + "raw_apical_rms": 0.0354581014473686, + "teaching_rms": 0.004331244755192192, + "traffic_rms": 0.03518428792538795 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1918544373656914e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.628112758512951e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009194250770854862, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.19447883835368687 + }, + { + "epoch": 84, + "eval_accuracy": 0.8432, + "eval_loss": 0.5976843139648438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06832236796617508, + "mean_feedback_forward_cosine": 0.9988360875531247, + "mean_feedback_forward_relative_error": 0.04757653666954292, + "min_feedback_forward_cosine": 0.9976797103881836 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00449569913828046, + "instruction_rms": 0.004495699139872506, + "raw_apical_rms": 0.03560047481926671, + "teaching_rms": 0.00449569913828046, + "traffic_rms": 0.03530655605531118 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3855610880852964e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6254458653860938e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009227602195459218, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.1982009590122435 + }, + { + "epoch": 85, + "eval_accuracy": 0.8714, + "eval_loss": 0.4689421752929688, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0658072829246521, + "mean_feedback_forward_cosine": 0.9989171059508073, + "mean_feedback_forward_relative_error": 0.04589343374889148, + "min_feedback_forward_cosine": 0.9978466033935547 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004473240100555046, + "instruction_rms": 0.004473240100649002, + "raw_apical_rms": 0.03558570370130643, + "teaching_rms": 0.004473240100555046, + "traffic_rms": 0.0352969631953337 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3908000084938976e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6322337465210142e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009480140194466701, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.1995692233191596 + }, + { + "epoch": 86, + "eval_accuracy": 0.8788, + "eval_loss": 0.4291405334472656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0634622797369957, + "mean_feedback_forward_cosine": 0.9989897796982213, + "mean_feedback_forward_relative_error": 0.04432482478257857, + "min_feedback_forward_cosine": 0.99799644947052 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004403505153147693, + "instruction_rms": 0.004403505154678212, + "raw_apical_rms": 0.0357053535971535, + "teaching_rms": 0.004403505153147693, + "traffic_rms": 0.03542529013894475 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2386266234898358e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6296237913063535e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009443209257356585, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.19470010391871134 + }, + { + "epoch": 87, + "eval_accuracy": 0.871, + "eval_loss": 0.45971768188476564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06109802424907684, + "mean_feedback_forward_cosine": 0.9990601978803936, + "mean_feedback_forward_relative_error": 0.042754401893992176, + "min_feedback_forward_cosine": 0.9981423616409302 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004557896740936863, + "instruction_rms": 0.004557896740819575, + "raw_apical_rms": 0.035865302055868825, + "teaching_rms": 0.004557896740936863, + "traffic_rms": 0.03556687724452723 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2751142364209045e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.623585945807858e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009776126863082948, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.20030388827853732 + }, + { + "epoch": 88, + "eval_accuracy": 0.8744, + "eval_loss": 0.4562751953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05893165245652199, + "mean_feedback_forward_cosine": 0.9991234195859808, + "mean_feedback_forward_relative_error": 0.041289182379841805, + "min_feedback_forward_cosine": 0.9982715845108032 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004438131279854983, + "instruction_rms": 0.004438131278594169, + "raw_apical_rms": 0.03587695328197024, + "teaching_rms": 0.004438131279854983, + "traffic_rms": 0.03559322356496659 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2551897299317716e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6242038664289318e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009567203178508883, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.19772963500552707 + }, + { + "epoch": 89, + "eval_accuracy": 0.8818, + "eval_loss": 0.393742041015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056795962154865265, + "mean_feedback_forward_cosine": 0.9991841222110548, + "mean_feedback_forward_relative_error": 0.039833312167933115, + "min_feedback_forward_cosine": 0.9983944296836853 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004542999277200115, + "instruction_rms": 0.004542999277433244, + "raw_apical_rms": 0.035812734370255755, + "teaching_rms": 0.004542999277200115, + "traffic_rms": 0.03551498619129561 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.105207303453426e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6283390844837256e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009153033826681829, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.19744758405950336 + }, + { + "epoch": 90, + "eval_accuracy": 0.827, + "eval_loss": 0.6168673828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05481269583106041, + "mean_feedback_forward_cosine": 0.9992394321843198, + "mean_feedback_forward_relative_error": 0.03845428224456938, + "min_feedback_forward_cosine": 0.9985045194625854 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00444154358005188, + "instruction_rms": 0.004441543578801351, + "raw_apical_rms": 0.036016278584755784, + "teaching_rms": 0.00444154358005188, + "traffic_rms": 0.03573315054412186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1648157105526025e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6269239789906622e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009388308772157275, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.19283736647499933 + }, + { + "epoch": 91, + "eval_accuracy": 0.8776, + "eval_loss": 0.4221093078613281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05281934514641762, + "mean_feedback_forward_cosine": 0.9992917525140863, + "mean_feedback_forward_relative_error": 0.03710895972816568, + "min_feedback_forward_cosine": 0.9986110925674438 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004461205840806688, + "instruction_rms": 0.004461205837663978, + "raw_apical_rms": 0.036108889058203204, + "teaching_rms": 0.004461205840806688, + "traffic_rms": 0.03582470494409129 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.5864106472918138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6244549495347382e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008758704244552483, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.20179404377937316 + }, + { + "epoch": 92, + "eval_accuracy": 0.8498, + "eval_loss": 0.5488065246582031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050882983952760696, + "mean_feedback_forward_cosine": 0.9993401922677693, + "mean_feedback_forward_relative_error": 0.0358158204900591, + "min_feedback_forward_cosine": 0.9987108707427979 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004385742457702917, + "instruction_rms": 0.0043857424557278875, + "raw_apical_rms": 0.0361424568017549, + "teaching_rms": 0.004385742457702917, + "traffic_rms": 0.03586691782618966 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4615670451689766e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618869183946591e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01006033014347003, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.19619109466340806 + }, + { + "epoch": 93, + "eval_accuracy": 0.874, + "eval_loss": 0.41386087646484376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04907582700252533, + "mean_feedback_forward_cosine": 0.9993855263057508, + "mean_feedback_forward_relative_error": 0.034564472636894175, + "min_feedback_forward_cosine": 0.9988006949424744 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004464036627903883, + "instruction_rms": 0.0044640366265155, + "raw_apical_rms": 0.03607449473952833, + "teaching_rms": 0.004464036627903883, + "traffic_rms": 0.03578905939075563 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4828236416519758e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6186613133143958e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01023274823824256, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.19600322678618962 + }, + { + "epoch": 94, + "eval_accuracy": 0.8606, + "eval_loss": 0.4576331787109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04737013205885887, + "mean_feedback_forward_cosine": 0.999426145302622, + "mean_feedback_forward_relative_error": 0.03339990821519965, + "min_feedback_forward_cosine": 0.9988824725151062 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004339615684676995, + "instruction_rms": 0.00433961568607273, + "raw_apical_rms": 0.036070355140655756, + "teaching_rms": 0.004339615684676995, + "traffic_rms": 0.03580020923633968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4304748496840602e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.620125759649221e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010584376748177247, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.19169836284187106 + }, + { + "epoch": 95, + "eval_accuracy": 0.8596, + "eval_loss": 0.4559836181640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04570562392473221, + "mean_feedback_forward_cosine": 0.9994649761601498, + "mean_feedback_forward_relative_error": 0.03224945597742733, + "min_feedback_forward_cosine": 0.9989596605300903 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004355237541317184, + "instruction_rms": 0.004355237540550336, + "raw_apical_rms": 0.03623222569830089, + "teaching_rms": 0.004355237541317184, + "traffic_rms": 0.03596110031917357 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1994017334870932e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6230141946270165e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010526418873724519, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.19197891221576266 + }, + { + "epoch": 96, + "eval_accuracy": 0.8804, + "eval_loss": 0.3978798095703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0440932996571064, + "mean_feedback_forward_cosine": 0.9995005256251285, + "mean_feedback_forward_relative_error": 0.031160078109487108, + "min_feedback_forward_cosine": 0.999031662940979 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004347173773633163, + "instruction_rms": 0.004347173773565615, + "raw_apical_rms": 0.036085388551255995, + "teaching_rms": 0.004347173773633163, + "traffic_rms": 0.03581637326586409 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.4107484069825205e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6188052547790338e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0105355196707146, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.18896560633977255 + }, + { + "epoch": 97, + "eval_accuracy": 0.8696, + "eval_loss": 0.45746707153320315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.04250432923436165, + "mean_feedback_forward_cosine": 0.9995339355970684, + "mean_feedback_forward_relative_error": 0.030100022442638874, + "min_feedback_forward_cosine": 0.9991000294685364 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004316854733873792, + "instruction_rms": 0.004316854732836267, + "raw_apical_rms": 0.03606182770129689, + "teaching_rms": 0.004316854733873792, + "traffic_rms": 0.0357937525580078 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.537128291317913e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619020163322017e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.010392886427435362, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.18978721633487278 + }, + { + "epoch": 98, + "eval_accuracy": 0.8804, + "eval_loss": 0.41344547729492187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.040988631546497345, + "mean_feedback_forward_cosine": 0.9995653001885665, + "mean_feedback_forward_relative_error": 0.029067374030618293, + "min_feedback_forward_cosine": 0.9991629123687744 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004360429312977475, + "instruction_rms": 0.00436042931312973, + "raw_apical_rms": 0.03617032916319902, + "teaching_rms": 0.004360429312977475, + "traffic_rms": 0.03589993955470201 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.5491643523546372e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.623243888849237e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009429446372638756, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.19110920924080743 + }, + { + "epoch": 99, + "eval_accuracy": 0.8758, + "eval_loss": 0.42149432373046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03946814313530922, + "mean_feedback_forward_cosine": 0.999595748750787, + "mean_feedback_forward_relative_error": 0.028032965467948662, + "min_feedback_forward_cosine": 0.9992236495018005 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0043877109872414895, + "instruction_rms": 0.004387710985843983, + "raw_apical_rms": 0.036154608328167366, + "teaching_rms": 0.0043877109872414895, + "traffic_rms": 0.03587890615478651 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3658580044761948e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6144190739499514e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009637188461033713, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.19423515079286363 + }, + { + "epoch": 100, + "eval_accuracy": 0.8842, + "eval_loss": 0.42995159912109376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03813067451119423, + "mean_feedback_forward_cosine": 0.9996227182840046, + "mean_feedback_forward_relative_error": 0.027078961463351, + "min_feedback_forward_cosine": 0.9992754459381104 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004330861420460429, + "instruction_rms": 0.004330861419590931, + "raw_apical_rms": 0.03606853535647668, + "teaching_rms": 0.004330861420460429, + "traffic_rms": 0.03579993205263799 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1662693033542837e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6230222065213373e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00962217048547801, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.1828259764790535 + }, + { + "epoch": 101, + "eval_accuracy": 0.9084, + "eval_loss": 0.3259307800292969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0381181500852108, + "mean_feedback_forward_cosine": 0.9996229786621896, + "mean_feedback_forward_relative_error": 0.027065618993028215, + "min_feedback_forward_cosine": 0.9992759227752686 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0033472765188487965, + "instruction_rms": 0.003347276517472718, + "raw_apical_rms": 0.03601295671222831, + "teaching_rms": 0.0033472765188487965, + "traffic_rms": 0.035850149572585595 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4773395839995374e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.614194736853349e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009414729635947462, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.11904574631386333 + }, + { + "epoch": 102, + "eval_accuracy": 0.9092, + "eval_loss": 0.3211487365722656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0381116084754467, + "mean_feedback_forward_cosine": 0.9996231825728166, + "mean_feedback_forward_relative_error": 0.027055564385495688, + "min_feedback_forward_cosine": 0.9992762207984924 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002990484514874979, + "instruction_rms": 0.0029904845151677907, + "raw_apical_rms": 0.035986723455086835, + "teaching_rms": 0.002990484514874979, + "traffic_rms": 0.03585656063547647 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409714281559, + "maximum_absolute_post_projection_soma_slope": 1.3441757928944753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6108888812942435e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009440852111375898, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.08980239703655243 + }, + { + "epoch": 103, + "eval_accuracy": 0.909, + "eval_loss": 0.3183172668457031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03810526058077812, + "mean_feedback_forward_cosine": 0.9996233362900583, + "mean_feedback_forward_relative_error": 0.027046052837058118, + "min_feedback_forward_cosine": 0.9992765188217163 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0029293037461553487, + "instruction_rms": 0.0029293037476591168, + "raw_apical_rms": 0.035936192474489675, + "teaching_rms": 0.0029293037461553487, + "traffic_rms": 0.035808943485872206 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4824728111761942e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6105168969895138e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009491064437173787, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.0815545727610588 + }, + { + "epoch": 104, + "eval_accuracy": 0.9118, + "eval_loss": 0.311673388671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0380992516875267, + "mean_feedback_forward_cosine": 0.9996235088298195, + "mean_feedback_forward_relative_error": 0.027036949305942182, + "min_feedback_forward_cosine": 0.9992767572402954 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002896674103148192, + "instruction_rms": 0.0028966741026558423, + "raw_apical_rms": 0.035942960685461434, + "teaching_rms": 0.002896674103148192, + "traffic_rms": 0.03581913271956188 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4247645729881242e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6097947301262527e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00951281183298253, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.07392787510024176 + }, + { + "epoch": 105, + "eval_accuracy": 0.9098, + "eval_loss": 0.32584077758789065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03809328004717827, + "mean_feedback_forward_cosine": 0.9996236782324942, + "mean_feedback_forward_relative_error": 0.027028057234067666, + "min_feedback_forward_cosine": 0.9992769360542297 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002982486869296268, + "instruction_rms": 0.0029824868702288607, + "raw_apical_rms": 0.03592376979909807, + "teaching_rms": 0.002982486869296268, + "traffic_rms": 0.035791721196680615 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409714281559, + "maximum_absolute_post_projection_soma_slope": 1.272930028051178e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6149943162567607e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009557600262048776, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.07141380354563395 + }, + { + "epoch": 106, + "eval_accuracy": 0.913, + "eval_loss": 0.32148916625976565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038087308406829834, + "mean_feedback_forward_cosine": 0.9996238162643031, + "mean_feedback_forward_relative_error": 0.027019406717858817, + "min_feedback_forward_cosine": 0.9992771148681641 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0029289830002966963, + "instruction_rms": 0.0029289830017090136, + "raw_apical_rms": 0.035914032439172634, + "teaching_rms": 0.0029289830002966963, + "traffic_rms": 0.035786792806693675 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409714281559, + "maximum_absolute_post_projection_soma_slope": 1.2585185338309657e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6155209415759927e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009643434889365563, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.06653599984910753 + }, + { + "epoch": 107, + "eval_accuracy": 0.9122, + "eval_loss": 0.32478743896484374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03808138519525528, + "mean_feedback_forward_cosine": 0.9996239856669777, + "mean_feedback_forward_relative_error": 0.02701078198458019, + "min_feedback_forward_cosine": 0.9992773532867432 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028665601539593255, + "instruction_rms": 0.0028665601528412975, + "raw_apical_rms": 0.03588042390045279, + "teaching_rms": 0.0028665601539593255, + "traffic_rms": 0.035756650487861315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3040343027626022e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.612231057199634e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009576412071197617, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.06255504687362247 + }, + { + "epoch": 108, + "eval_accuracy": 0.9122, + "eval_loss": 0.328871923828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038075514137744904, + "mean_feedback_forward_cosine": 0.9996241268358732, + "mean_feedback_forward_relative_error": 0.02700227861733813, + "min_feedback_forward_cosine": 0.9992775917053223 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028695633932737264, + "instruction_rms": 0.0028695633942765666, + "raw_apical_rms": 0.035890662211287484, + "teaching_rms": 0.0028695633932737264, + "traffic_rms": 0.035766018499236475 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1649767373000941e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6136342007962996e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00968816035848221, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.0597408724811342 + }, + { + "epoch": 109, + "eval_accuracy": 0.914, + "eval_loss": 0.33251861572265623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03806949779391289, + "mean_feedback_forward_cosine": 0.9996242868272882, + "mean_feedback_forward_relative_error": 0.026993888087178532, + "min_feedback_forward_cosine": 0.9992778301239014 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028985146835832764, + "instruction_rms": 0.002898514684873951, + "raw_apical_rms": 0.0358777356892019, + "teaching_rms": 0.0028985146835832764, + "traffic_rms": 0.0357524569798364 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2582765052115974e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.610909395309814e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009558368229309566, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.058068850870927176 + }, + { + "epoch": 110, + "eval_accuracy": 0.9128, + "eval_loss": 0.33395889892578123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038063690066337585, + "mean_feedback_forward_cosine": 0.999624468778309, + "mean_feedback_forward_relative_error": 0.02698576935615979, + "min_feedback_forward_cosine": 0.9992780685424805 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028696458178227983, + "instruction_rms": 0.002869645817842472, + "raw_apical_rms": 0.03585184658767721, + "teaching_rms": 0.0028696458178227983, + "traffic_rms": 0.035726987625552924 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4731561748249078e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6158843093443453e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009566121352255097, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.054469241478708054 + }, + { + "epoch": 111, + "eval_accuracy": 0.9138, + "eval_loss": 0.33631309814453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03805793449282646, + "mean_feedback_forward_cosine": 0.9996246475922433, + "mean_feedback_forward_relative_error": 0.026977706896631343, + "min_feedback_forward_cosine": 0.9992783069610596 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028511713941757337, + "instruction_rms": 0.0028511713934999392, + "raw_apical_rms": 0.03585176552770924, + "teaching_rms": 0.0028511713941757337, + "traffic_rms": 0.035727985689006425 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2175729757757381e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6156636693558254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009671452186644064, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.05229350429375967 + }, + { + "epoch": 112, + "eval_accuracy": 0.9154, + "eval_loss": 0.33809072265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038052406162023544, + "mean_feedback_forward_cosine": 0.9996248044465718, + "mean_feedback_forward_relative_error": 0.02696984668115252, + "min_feedback_forward_cosine": 0.9992785453796387 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028222139723085853, + "instruction_rms": 0.002822213971199403, + "raw_apical_rms": 0.035873038034578815, + "teaching_rms": 0.0028222139723085853, + "traffic_rms": 0.03575171930431188 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.392575832426246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6145431346167743e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009570548191714557, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.05088527640766568 + }, + { + "epoch": 113, + "eval_accuracy": 0.9102, + "eval_loss": 0.34676420288085935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038046613335609436, + "mean_feedback_forward_cosine": 0.999624933067121, + "mean_feedback_forward_relative_error": 0.026961923969027243, + "min_feedback_forward_cosine": 0.9992787837982178 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002842311355588269, + "instruction_rms": 0.0028423113554333825, + "raw_apical_rms": 0.03587723687063692, + "teaching_rms": 0.002842311355588269, + "traffic_rms": 0.03575509434187239 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4686383664752611e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6144555520498742e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009498977853440015, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.04946711292134391 + }, + { + "epoch": 114, + "eval_accuracy": 0.9122, + "eval_loss": 0.34561643676757814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03804095461964607, + "mean_feedback_forward_cosine": 0.9996250961956225, + "mean_feedback_forward_relative_error": 0.026953961700201035, + "min_feedback_forward_cosine": 0.9992790222167969 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002797731487760313, + "instruction_rms": 0.0027977314877417453, + "raw_apical_rms": 0.035867336452967774, + "teaching_rms": 0.002797731487760313, + "traffic_rms": 0.03574732062936145 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2948744299023929e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6190549016728118e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009511829493815006, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.04669660123321745 + }, + { + "epoch": 115, + "eval_accuracy": 0.9104, + "eval_loss": 0.3552204650878906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03803538531064987, + "mean_feedback_forward_cosine": 0.9996252342274314, + "mean_feedback_forward_relative_error": 0.026946195597319228, + "min_feedback_forward_cosine": 0.999279260635376 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002781539789875912, + "instruction_rms": 0.002781539789402498, + "raw_apical_rms": 0.03583110792841806, + "teaching_rms": 0.002781539789875912, + "traffic_rms": 0.03571137774542888 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2482407996117217e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6190075102145886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009584501072354873, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.043811548323763744 + }, + { + "epoch": 116, + "eval_accuracy": 0.913, + "eval_loss": 0.34888399658203123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03802938014268875, + "mean_feedback_forward_cosine": 0.9996254004930195, + "mean_feedback_forward_relative_error": 0.026938311069419508, + "min_feedback_forward_cosine": 0.9992794990539551 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0027589384972830176, + "instruction_rms": 0.002758938498364458, + "raw_apical_rms": 0.035822512187125864, + "teaching_rms": 0.0027589384972830176, + "traffic_rms": 0.0357032934364985 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4204090348357568e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6174936257567725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00952099290605394, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.04321008748345905 + }, + { + "epoch": 117, + "eval_accuracy": 0.9116, + "eval_loss": 0.351063623046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03802365064620972, + "mean_feedback_forward_cosine": 0.9996255730327807, + "mean_feedback_forward_relative_error": 0.026930446785531546, + "min_feedback_forward_cosine": 0.9992796778678894 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028121380910417835, + "instruction_rms": 0.0028121380914258374, + "raw_apical_rms": 0.03584902275998271, + "teaching_rms": 0.0028121380910417835, + "traffic_rms": 0.035727811590774095 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1390384635490136e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.615851491866325e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00953851163872493, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.041723162909348806 + }, + { + "epoch": 118, + "eval_accuracy": 0.9116, + "eval_loss": 0.3595480163574219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03801767900586128, + "mean_feedback_forward_cosine": 0.9996257110645896, + "mean_feedback_forward_relative_error": 0.026922483977518584, + "min_feedback_forward_cosine": 0.999279797077179 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0028066320925657522, + "instruction_rms": 0.0028066320926862657, + "raw_apical_rms": 0.03583823213845065, + "teaching_rms": 0.0028066320925657522, + "traffic_rms": 0.035717416974547646 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1584762482641509e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6138026794508964e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009567498567318766, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.04121919312477112 + }, + { + "epoch": 119, + "eval_accuracy": 0.9116, + "eval_loss": 0.36063224487304685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038011953234672546, + "mean_feedback_forward_cosine": 0.9996258741930911, + "mean_feedback_forward_relative_error": 0.026914845318778566, + "min_feedback_forward_cosine": 0.9992802143096924 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002766894998886884, + "instruction_rms": 0.0027668949982804733, + "raw_apical_rms": 0.035822942272094475, + "teaching_rms": 0.002766894998886884, + "traffic_rms": 0.035703425144166405 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1590560511365311e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6143377507427477e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009596634022104129, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.03897833952241474 + }, + { + "epoch": 120, + "eval_accuracy": 0.9122, + "eval_loss": 0.36014967651367186, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.038005948066711426, + "mean_feedback_forward_cosine": 0.9996260937891508, + "mean_feedback_forward_relative_error": 0.02690696235942213, + "min_feedback_forward_cosine": 0.9992803931236267 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002761162125425492, + "instruction_rms": 0.0027611621253415103, + "raw_apical_rms": 0.03583332192186645, + "teaching_rms": 0.002761162125425492, + "traffic_rms": 0.035714112015139084 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.258372517298767e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6169752667974602e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009646114910546006, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.038807947030994626 + }, + { + "epoch": 121, + "eval_accuracy": 0.912, + "eval_loss": 0.368635986328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03799998760223389, + "mean_feedback_forward_cosine": 0.9996262255467867, + "mean_feedback_forward_relative_error": 0.026899256302337898, + "min_feedback_forward_cosine": 0.9992806911468506 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00275476982346384, + "instruction_rms": 0.002754769822896326, + "raw_apical_rms": 0.03580108555403456, + "teaching_rms": 0.00275476982346384, + "traffic_rms": 0.03568257318023869 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.500537294418791e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6158118961452318e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009732036852036902, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.03859740868144565 + }, + { + "epoch": 122, + "eval_accuracy": 0.9112, + "eval_loss": 0.375572607421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03799373283982277, + "mean_feedback_forward_cosine": 0.9996264106348941, + "mean_feedback_forward_relative_error": 0.026891536863618774, + "min_feedback_forward_cosine": 0.9992808699607849 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0027601279683554577, + "instruction_rms": 0.002760127970405934, + "raw_apical_rms": 0.03581597123058649, + "teaching_rms": 0.0027601279683554577, + "traffic_rms": 0.03569536676631768 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3610550908538244e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.616823077868045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00965707582365918, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.03734560879071554 + }, + { + "epoch": 123, + "eval_accuracy": 0.9096, + "eval_loss": 0.3836206359863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03798767551779747, + "mean_feedback_forward_cosine": 0.9996265298441837, + "mean_feedback_forward_relative_error": 0.02688379639661626, + "min_feedback_forward_cosine": 0.9992810487747192 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0026655376026348985, + "instruction_rms": 0.0026655376010316545, + "raw_apical_rms": 0.03579517670364923, + "teaching_rms": 0.0026655376026348985, + "traffic_rms": 0.035683251075829196 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.6414423598121175e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6190680857384163e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00963298712955783, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.03550222207605839 + }, + { + "epoch": 124, + "eval_accuracy": 0.9124, + "eval_loss": 0.3698329162597656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03798159584403038, + "mean_feedback_forward_cosine": 0.9996267400289837, + "mean_feedback_forward_relative_error": 0.02687601612782792, + "min_feedback_forward_cosine": 0.9992812871932983 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0027511484741651825, + "instruction_rms": 0.0027511484752177208, + "raw_apical_rms": 0.03577062803804927, + "teaching_rms": 0.0027511484741651825, + "traffic_rms": 0.03565009219861281 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.120090598050183e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.616681270428917e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009707561803276307, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.03580282854802078 + }, + { + "epoch": 125, + "eval_accuracy": 0.9114, + "eval_loss": 0.3712406555175781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03797522932291031, + "mean_feedback_forward_cosine": 0.9996268937462255, + "mean_feedback_forward_relative_error": 0.026868038369636787, + "min_feedback_forward_cosine": 0.9992815852165222 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0027907677721793023, + "instruction_rms": 0.002790767771911957, + "raw_apical_rms": 0.03577578432471849, + "teaching_rms": 0.0027907677721793023, + "traffic_rms": 0.035652179953717646 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.379840330884008e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.614495218653508e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009818036049193164, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.036065083287821875 + }, + { + "epoch": 126, + "eval_accuracy": 0.913, + "eval_loss": 0.37410505981445313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037968847900629044, + "mean_feedback_forward_cosine": 0.9996270411892941, + "mean_feedback_forward_relative_error": 0.026860256336237256, + "min_feedback_forward_cosine": 0.9992817044258118 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0027855026442477598, + "instruction_rms": 0.0027855026440891987, + "raw_apical_rms": 0.03578097218964767, + "teaching_rms": 0.0027855026442477598, + "traffic_rms": 0.03565767145537793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.147241590615522e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6161130612054242e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009674539963637395, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.03503612222803964 + }, + { + "epoch": 127, + "eval_accuracy": 0.9104, + "eval_loss": 0.38507326049804685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03796249255537987, + "mean_feedback_forward_cosine": 0.9996272356886613, + "mean_feedback_forward_relative_error": 0.026852291273443324, + "min_feedback_forward_cosine": 0.9992821216583252 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0026975618837138983, + "instruction_rms": 0.002697561883383782, + "raw_apical_rms": 0.03579064949846312, + "teaching_rms": 0.0026975618837138983, + "traffic_rms": 0.03567367917199365 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3141908006275571e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618431664614786e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009527002395862164, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.03341853344672256 + }, + { + "epoch": 128, + "eval_accuracy": 0.911, + "eval_loss": 0.3920139404296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03795585036277771, + "mean_feedback_forward_cosine": 0.9996274207767687, + "mean_feedback_forward_relative_error": 0.026844228862931852, + "min_feedback_forward_cosine": 0.9992824196815491 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002750541218062686, + "instruction_rms": 0.0027505412191790746, + "raw_apical_rms": 0.03576918976281961, + "teaching_rms": 0.002750541218062686, + "traffic_rms": 0.0356481807025341 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2473168276017077e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6218493188187658e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009663760274781231, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.03287169239719709 + }, + { + "epoch": 129, + "eval_accuracy": 0.9134, + "eval_loss": 0.38938822021484376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037949901074171066, + "mean_feedback_forward_cosine": 0.9996275870423568, + "mean_feedback_forward_relative_error": 0.02683663461357355, + "min_feedback_forward_cosine": 0.9992826581001282 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0026238850680548745, + "instruction_rms": 0.0026238850678266677, + "raw_apical_rms": 0.035746906829503516, + "teaching_rms": 0.0026238850680548745, + "traffic_rms": 0.035636102017194186 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.0733844924004643e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6188728554294605e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009614808397064061, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.03094183478222953 + }, + { + "epoch": 130, + "eval_accuracy": 0.9134, + "eval_loss": 0.3928030151367187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03794398903846741, + "mean_feedback_forward_cosine": 0.9996277627192045, + "mean_feedback_forward_relative_error": 0.02682895414335163, + "min_feedback_forward_cosine": 0.9992827773094177 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002622710879992411, + "instruction_rms": 0.002622710880762901, + "raw_apical_rms": 0.035731069218762186, + "teaching_rms": 0.002622710879992411, + "traffic_rms": 0.03561991443288764 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2849611152887519e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6121007907933642e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009830500001566147, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.029354358004861407 + }, + { + "epoch": 131, + "eval_accuracy": 0.9114, + "eval_loss": 0.38648424072265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03793803229928017, + "mean_feedback_forward_cosine": 0.9996279195735329, + "mean_feedback_forward_relative_error": 0.026821194118575045, + "min_feedback_forward_cosine": 0.9992830157279968 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025914732781347487, + "instruction_rms": 0.0025914732781897134, + "raw_apical_rms": 0.0357203964741929, + "teaching_rms": 0.0025914732781347487, + "traffic_rms": 0.03561015827148135 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2277120653436668e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6177991820475174e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009724882261981085, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.028900067031052377 + }, + { + "epoch": 132, + "eval_accuracy": 0.9106, + "eval_loss": 0.3918890930175781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03793131932616234, + "mean_feedback_forward_cosine": 0.9996280921132941, + "mean_feedback_forward_relative_error": 0.02681309494533037, + "min_feedback_forward_cosine": 0.9992832541465759 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0026975961361970055, + "instruction_rms": 0.002697596135337974, + "raw_apical_rms": 0.035710380517198305, + "teaching_rms": 0.0026975961361970055, + "traffic_rms": 0.035592226690178135 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2983555564005655e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.622333585292056e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009668498919770402, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.030538018282088967 + }, + { + "epoch": 133, + "eval_accuracy": 0.9104, + "eval_loss": 0.39623680419921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037924885749816895, + "mean_feedback_forward_cosine": 0.9996282458305359, + "mean_feedback_forward_relative_error": 0.02680520125125584, + "min_feedback_forward_cosine": 0.999283492565155 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002635780184050218, + "instruction_rms": 0.0026357801832153664, + "raw_apical_rms": 0.03569409748753582, + "teaching_rms": 0.002635780184050218, + "traffic_rms": 0.03557964049918153 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3676385357541676e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.615643645286502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009754141415901071, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.029081654803454877 + }, + { + "epoch": 134, + "eval_accuracy": 0.9108, + "eval_loss": 0.3895315185546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037918198853731155, + "mean_feedback_forward_cosine": 0.999628462289509, + "mean_feedback_forward_relative_error": 0.02679695581134997, + "min_feedback_forward_cosine": 0.9992837905883789 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002742187509968168, + "instruction_rms": 0.0027421875111034097, + "raw_apical_rms": 0.03572439865144619, + "teaching_rms": 0.002742187509968168, + "traffic_rms": 0.03560007387385542 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3136595811147345e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6202907495852204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009760525172780596, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.028779723361465667 + }, + { + "epoch": 135, + "eval_accuracy": 0.9106, + "eval_loss": 0.3926762451171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037911418825387955, + "mean_feedback_forward_cosine": 0.9996286379663568, + "mean_feedback_forward_relative_error": 0.026788822914424696, + "min_feedback_forward_cosine": 0.9992839097976685 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0026868494036395157, + "instruction_rms": 0.002686849404910164, + "raw_apical_rms": 0.03569168232715309, + "teaching_rms": 0.0026868494036395157, + "traffic_rms": 0.03557283692593126 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3787125219266727e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6191191160133836e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00979281848443928, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.02846326970325576 + }, + { + "epoch": 136, + "eval_accuracy": 0.9084, + "eval_loss": 0.40878616943359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03790559247136116, + "mean_feedback_forward_cosine": 0.999628832465724, + "mean_feedback_forward_relative_error": 0.026781202195898482, + "min_feedback_forward_cosine": 0.9992842674255371 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025450656243850523, + "instruction_rms": 0.002545065624684874, + "raw_apical_rms": 0.03566917022717624, + "teaching_rms": 0.0025450656243850523, + "traffic_rms": 0.0355617712433612 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2010985095400883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6177253169118073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009744248196663233, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.025685858269201386 + }, + { + "epoch": 137, + "eval_accuracy": 0.9128, + "eval_loss": 0.3987355712890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03789890930056572, + "mean_feedback_forward_cosine": 0.9996290363763508, + "mean_feedback_forward_relative_error": 0.026773277278009214, + "min_feedback_forward_cosine": 0.999284565448761 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0026180515333599323, + "instruction_rms": 0.0026180515335020634, + "raw_apical_rms": 0.03570993695445157, + "teaching_rms": 0.0026180515333599323, + "traffic_rms": 0.03559666524284681 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.597332932590234e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6204578252730795e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009646679881658194, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.027844087640444437 + }, + { + "epoch": 138, + "eval_accuracy": 0.9102, + "eval_loss": 0.4109247741699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03789233788847923, + "mean_feedback_forward_cosine": 0.9996291806823329, + "mean_feedback_forward_relative_error": 0.026765429463825728, + "min_feedback_forward_cosine": 0.9992847442626953 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002638664010063119, + "instruction_rms": 0.0026386640096644647, + "raw_apical_rms": 0.03566768918823276, + "teaching_rms": 0.002638664010063119, + "traffic_rms": 0.03555359236882088 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2113433811578034e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.623062258955324e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00976945290872928, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.026772137907644114 + }, + { + "epoch": 139, + "eval_accuracy": 0.9144, + "eval_loss": 0.40518236694335935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03788561001420021, + "mean_feedback_forward_cosine": 0.9996293783187866, + "mean_feedback_forward_relative_error": 0.026757078539384038, + "min_feedback_forward_cosine": 0.999285101890564 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002612001682187245, + "instruction_rms": 0.0026120016827994065, + "raw_apical_rms": 0.03565216126697786, + "teaching_rms": 0.002612001682187245, + "traffic_rms": 0.035537428261166085 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1465832727708403e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6258578957992902e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009644668044545194, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.02580999892420239 + }, + { + "epoch": 140, + "eval_accuracy": 0.9104, + "eval_loss": 0.411278125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03787907212972641, + "mean_feedback_forward_cosine": 0.9996295696810672, + "mean_feedback_forward_relative_error": 0.026749126613140106, + "min_feedback_forward_cosine": 0.9992852210998535 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002611693291988267, + "instruction_rms": 0.002611693292765891, + "raw_apical_rms": 0.035591762399835865, + "teaching_rms": 0.002611693291988267, + "traffic_rms": 0.03547938614446786 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2840193797103439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6246608171250558e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009743369749401539, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.025524346191353266 + }, + { + "epoch": 141, + "eval_accuracy": 0.9106, + "eval_loss": 0.4062041748046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03787163272500038, + "mean_feedback_forward_cosine": 0.9996297767287806, + "mean_feedback_forward_relative_error": 0.026740355907302154, + "min_feedback_forward_cosine": 0.9992855787277222 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0027070270794916193, + "instruction_rms": 0.0027070270797813255, + "raw_apical_rms": 0.03563609802435912, + "teaching_rms": 0.0027070270794916193, + "traffic_rms": 0.035511611640737885 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4507527623663918e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6223096926328764e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009729349519671793, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.027296853695147567 + }, + { + "epoch": 142, + "eval_accuracy": 0.9118, + "eval_loss": 0.41186312255859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037864819169044495, + "mean_feedback_forward_cosine": 0.9996299618168881, + "mean_feedback_forward_relative_error": 0.026732132868155054, + "min_feedback_forward_cosine": 0.9992859363555908 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025778124945542435, + "instruction_rms": 0.0025778124941378392, + "raw_apical_rms": 0.03561647943645186, + "teaching_rms": 0.0025778124945542435, + "traffic_rms": 0.035505854937505556 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2427866735720272e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6168315613522945e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00971278371826236, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.02511273897687594 + }, + { + "epoch": 143, + "eval_accuracy": 0.9106, + "eval_loss": 0.4166827209472656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03785836696624756, + "mean_feedback_forward_cosine": 0.999630143767909, + "mean_feedback_forward_relative_error": 0.02672435906960776, + "min_feedback_forward_cosine": 0.9992861151695251 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0024070126294162093, + "instruction_rms": 0.0024070126293388137, + "raw_apical_rms": 0.03559019898416294, + "teaching_rms": 0.0024070126294162093, + "traffic_rms": 0.03549209008796763 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2012990602272566e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618936282268463e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009764600853767324, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.022995042044586606 + }, + { + "epoch": 144, + "eval_accuracy": 0.9126, + "eval_loss": 0.41340057373046873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03785092383623123, + "mean_feedback_forward_cosine": 0.9996303508156225, + "mean_feedback_forward_relative_error": 0.026715706151567008, + "min_feedback_forward_cosine": 0.999286413192749 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0026267032270786655, + "instruction_rms": 0.0026267032265151757, + "raw_apical_rms": 0.03561348328808878, + "teaching_rms": 0.0026267032270786655, + "traffic_rms": 0.03549544675313446 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4057949471180109e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6210168806989145e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009787132839748691, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.024796697247690624 + }, + { + "epoch": 145, + "eval_accuracy": 0.9126, + "eval_loss": 0.407359423828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03784457966685295, + "mean_feedback_forward_cosine": 0.9996305139441239, + "mean_feedback_forward_relative_error": 0.0267080179856796, + "min_feedback_forward_cosine": 0.9992866516113281 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002411227822155689, + "instruction_rms": 0.002411227823475597, + "raw_apical_rms": 0.03559807514429829, + "teaching_rms": 0.002411227822155689, + "traffic_rms": 0.0354983777965054 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.0809019457269642e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.620617980156489e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00980182348454157, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.022539040789835985 + }, + { + "epoch": 146, + "eval_accuracy": 0.91, + "eval_loss": 0.42620809326171877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03783755376935005, + "mean_feedback_forward_cosine": 0.9996307115805777, + "mean_feedback_forward_relative_error": 0.026699521137695564, + "min_feedback_forward_cosine": 0.9992868900299072 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002565811242517235, + "instruction_rms": 0.002565811242643303, + "raw_apical_rms": 0.035587567868722365, + "teaching_rms": 0.002565811242517235, + "traffic_rms": 0.03547522987971678 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4903037026670063e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6225682665495397e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009706511222796843, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.023578642381893264 + }, + { + "epoch": 147, + "eval_accuracy": 0.911, + "eval_loss": 0.42503335571289064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03783031553030014, + "mean_feedback_forward_cosine": 0.999630940587897, + "mean_feedback_forward_relative_error": 0.026690693670197538, + "min_feedback_forward_cosine": 0.9992870688438416 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025885059522246312, + "instruction_rms": 0.0025885059535557015, + "raw_apical_rms": 0.03555838610924396, + "teaching_rms": 0.0025885059522246312, + "traffic_rms": 0.035442078172238385 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2002680627176687e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6267581202852138e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00985412452162739, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.023740731816159353 + }, + { + "epoch": 148, + "eval_accuracy": 0.9122, + "eval_loss": 0.4174243408203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03782280161976814, + "mean_feedback_forward_cosine": 0.9996311382243508, + "mean_feedback_forward_relative_error": 0.026681908847470032, + "min_feedback_forward_cosine": 0.9992873668670654 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025628457988248884, + "instruction_rms": 0.0025628457987137013, + "raw_apical_rms": 0.03553239426766158, + "teaching_rms": 0.0025628457988248884, + "traffic_rms": 0.03542038082092827 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2999201715047093e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6248284634245207e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009733664121826479, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.023489599435859255 + }, + { + "epoch": 149, + "eval_accuracy": 0.915, + "eval_loss": 0.42041235961914064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037815794348716736, + "mean_feedback_forward_cosine": 0.9996313515462374, + "mean_feedback_forward_relative_error": 0.02667347900569439, + "min_feedback_forward_cosine": 0.9992876052856445 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00253459505010071, + "instruction_rms": 0.0025345950500882826, + "raw_apical_rms": 0.03551564592976059, + "teaching_rms": 0.00253459505010071, + "traffic_rms": 0.035407345219106684 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2784917124974982e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6207162199078193e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009688193952499912, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.022620651949776544 + }, + { + "epoch": 150, + "eval_accuracy": 0.9104, + "eval_loss": 0.42439256591796876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780863806605339, + "mean_feedback_forward_cosine": 0.9996315397714314, + "mean_feedback_forward_relative_error": 0.026664931229070613, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0025151916352342905, + "instruction_rms": 0.002515191635826509, + "raw_apical_rms": 0.03552600170105816, + "teaching_rms": 0.0025151916352342905, + "traffic_rms": 0.03541416019616737 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3124211939441466e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618080090131379e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009837820810131428, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.022285571916566955 + }, + { + "epoch": 151, + "eval_accuracy": 0.9112, + "eval_loss": 0.4160915588378906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780853748321533, + "mean_feedback_forward_cosine": 0.9996315460456046, + "mean_feedback_forward_relative_error": 0.0266646696254611, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0023690394798630107, + "instruction_rms": 0.002369039479136609, + "raw_apical_rms": 0.035500359686107825, + "teaching_rms": 0.0023690394798630107, + "traffic_rms": 0.03540311715202738 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3979216006987372e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6147353264949156e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009950262225916816, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.020033901221967406 + }, + { + "epoch": 152, + "eval_accuracy": 0.915, + "eval_loss": 0.412415478515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780847415328026, + "mean_feedback_forward_cosine": 0.999631542908518, + "mean_feedback_forward_relative_error": 0.026664435765460917, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002363181180109337, + "instruction_rms": 0.0023631811816877146, + "raw_apical_rms": 0.03550671226071755, + "teaching_rms": 0.002363181180109337, + "traffic_rms": 0.03540743308969674 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3078425453727505e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6143342377322934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009923390440856971, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.01982582550181283 + }, + { + "epoch": 153, + "eval_accuracy": 0.914, + "eval_loss": 0.4219018493652344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780841827392578, + "mean_feedback_forward_cosine": 0.9996315711422971, + "mean_feedback_forward_relative_error": 0.02666421739482566, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0022168579681471966, + "instruction_rms": 0.0022168579682489828, + "raw_apical_rms": 0.03549892093773543, + "teaching_rms": 0.0022168579681471966, + "traffic_rms": 0.03540893340906787 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1699555102495651e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6155479768674653e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00991793256075795, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.018583940839767457 + }, + { + "epoch": 154, + "eval_accuracy": 0.9154, + "eval_loss": 0.4142001647949219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037808362394571304, + "mean_feedback_forward_cosine": 0.9996315585939508, + "mean_feedback_forward_relative_error": 0.026664001033886484, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0022208326704436345, + "instruction_rms": 0.0022208326696203434, + "raw_apical_rms": 0.0354993800874489, + "teaching_rms": 0.0022208326704436345, + "traffic_rms": 0.03541088335135728 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.174215213950447e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6138077736505478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009778174847088088, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.018304667066037655 + }, + { + "epoch": 155, + "eval_accuracy": 0.9126, + "eval_loss": 0.41529046630859373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780830651521683, + "mean_feedback_forward_cosine": 0.9996315648681239, + "mean_feedback_forward_relative_error": 0.02666377888894395, + "min_feedback_forward_cosine": 0.9992879033088684 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020940083723578317, + "instruction_rms": 0.00209400837211589, + "raw_apical_rms": 0.03549419842454183, + "teaching_rms": 0.0020940083723578317, + "traffic_rms": 0.03541537284997737 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3867719417248736e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.615518187018473e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009840861291704164, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.017229355897671647 + }, + { + "epoch": 156, + "eval_accuracy": 0.9158, + "eval_loss": 0.4150919738769531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037808261811733246, + "mean_feedback_forward_cosine": 0.9996315617310373, + "mean_feedback_forward_relative_error": 0.02666356625329507, + "min_feedback_forward_cosine": 0.9992879033088684 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020718280070672877, + "instruction_rms": 0.0020718280072837747, + "raw_apical_rms": 0.03549177889857754, + "teaching_rms": 0.0020718280070672877, + "traffic_rms": 0.03541395171582714 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1908183772391112e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.616108327523135e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009694605252590052, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.016535244550307592 + }, + { + "epoch": 157, + "eval_accuracy": 0.9124, + "eval_loss": 0.4139591613769531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780820965766907, + "mean_feedback_forward_cosine": 0.9996315585939508, + "mean_feedback_forward_relative_error": 0.026663346510184437, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0021928518226127106, + "instruction_rms": 0.002192851823067784, + "raw_apical_rms": 0.03550177907923658, + "teaching_rms": 0.0021928518226127106, + "traffic_rms": 0.03541410326784334 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4171884110680821e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618542717085575e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009729103278611475, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.017914864792426426 + }, + { + "epoch": 158, + "eval_accuracy": 0.9136, + "eval_loss": 0.4135891052246094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780815750360489, + "mean_feedback_forward_cosine": 0.9996315366343448, + "mean_feedback_forward_relative_error": 0.026663127208226604, + "min_feedback_forward_cosine": 0.9992879033088684 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002134958372463622, + "instruction_rms": 0.0021349583719595688, + "raw_apical_rms": 0.03549795898305622, + "teaching_rms": 0.002134958372463622, + "traffic_rms": 0.03541718881392205 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2411835115244685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618284934513545e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009742366157895569, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.01753395065723194 + }, + { + "epoch": 159, + "eval_accuracy": 0.9156, + "eval_loss": 0.41060625610351564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780810534954071, + "mean_feedback_forward_cosine": 0.9996315680052105, + "mean_feedback_forward_relative_error": 0.026662918150817092, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0021345640691926306, + "instruction_rms": 0.0021345640694966726, + "raw_apical_rms": 0.03549960982866271, + "teaching_rms": 0.0021345640691926306, + "traffic_rms": 0.035419135821711384 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3708177704074842e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.621017741400353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009681370527874348, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.016990253758927185 + }, + { + "epoch": 160, + "eval_accuracy": 0.9152, + "eval_loss": 0.413508935546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037808045744895935, + "mean_feedback_forward_cosine": 0.9996315523197776, + "mean_feedback_forward_relative_error": 0.026662708848322694, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002075748873974312, + "instruction_rms": 0.002075748873711616, + "raw_apical_rms": 0.035495948403698525, + "teaching_rms": 0.002075748873974312, + "traffic_rms": 0.03541909665314382 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2328693621554976e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.623219162921405e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009798775898307792, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.016840938025050694 + }, + { + "epoch": 161, + "eval_accuracy": 0.915, + "eval_loss": 0.4121825561523437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780798986554146, + "mean_feedback_forward_cosine": 0.9996315774164701, + "mean_feedback_forward_relative_error": 0.02666250243782997, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002121357037740998, + "instruction_rms": 0.0021213570378834324, + "raw_apical_rms": 0.03550143453081703, + "teaching_rms": 0.002121357037740998, + "traffic_rms": 0.035419878280186855 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2869592502795513e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.617544037746847e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009704877673677476, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.016971301689743996 + }, + { + "epoch": 162, + "eval_accuracy": 0.9142, + "eval_loss": 0.41635877685546874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780794143676758, + "mean_feedback_forward_cosine": 0.9996315554568642, + "mean_feedback_forward_relative_error": 0.026662287743468033, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001991115441323907, + "instruction_rms": 0.001991115442340531, + "raw_apical_rms": 0.035491182938576275, + "teaching_rms": 0.001991115441323907, + "traffic_rms": 0.035420416209033016 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3590495839821415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619519543498664e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009655481431774027, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.015489534443534082 + }, + { + "epoch": 163, + "eval_accuracy": 0.9138, + "eval_loss": 0.4135448547363281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0378078892827034, + "mean_feedback_forward_cosine": 0.9996315774164701, + "mean_feedback_forward_relative_error": 0.02666207834293968, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020567535253763576, + "instruction_rms": 0.0020567535255817935, + "raw_apical_rms": 0.035494462825656777, + "teaching_rms": 0.0020567535253763576, + "traffic_rms": 0.03541904425544074 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2687332962002529e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6184658532040568e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009708826992103438, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.016186376640366183 + }, + { + "epoch": 164, + "eval_accuracy": 0.9144, + "eval_loss": 0.415368798828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780784085392952, + "mean_feedback_forward_cosine": 0.999631593101903, + "mean_feedback_forward_relative_error": 0.026661876294957965, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019375142163203856, + "instruction_rms": 0.0019375142165507645, + "raw_apical_rms": 0.035487892708861685, + "teaching_rms": 0.0019375142163203856, + "traffic_rms": 0.03541944844315403 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.291764029076603e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.61590983006741e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009717712448308678, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.015254196861386298 + }, + { + "epoch": 165, + "eval_accuracy": 0.9132, + "eval_loss": 0.4172639404296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780778497457504, + "mean_feedback_forward_cosine": 0.9996315742793837, + "mean_feedback_forward_relative_error": 0.026661668364938936, + "min_feedback_forward_cosine": 0.9992879033088684 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020433014823095573, + "instruction_rms": 0.0020433014818764506, + "raw_apical_rms": 0.03549243381270347, + "teaching_rms": 0.0020433014823095573, + "traffic_rms": 0.03541736025860466 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2993341513833911e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6160578069739604e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009716151742027465, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.01599680534899235 + }, + { + "epoch": 166, + "eval_accuracy": 0.9134, + "eval_loss": 0.41902383422851563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780774399638176, + "mean_feedback_forward_cosine": 0.9996315617310373, + "mean_feedback_forward_relative_error": 0.02666145984671618, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019939385671762506, + "instruction_rms": 0.001993938567715279, + "raw_apical_rms": 0.03548806314820587, + "teaching_rms": 0.0019939385671762506, + "traffic_rms": 0.03541724515608608 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.471319333035126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618911496952291e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00968140701438848, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.015350342567927307 + }, + { + "epoch": 167, + "eval_accuracy": 0.9138, + "eval_loss": 0.42023292236328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780769556760788, + "mean_feedback_forward_cosine": 0.9996315805535567, + "mean_feedback_forward_relative_error": 0.02666125147554435, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020467549993294968, + "instruction_rms": 0.00204675499844481, + "raw_apical_rms": 0.03549362319880884, + "teaching_rms": 0.0020467549993294968, + "traffic_rms": 0.03541744547909088 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3256748587764378e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6198538364641417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009709477833559205, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.016373495394074254 + }, + { + "epoch": 168, + "eval_accuracy": 0.9124, + "eval_loss": 0.41826709594726563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037807647138834, + "mean_feedback_forward_cosine": 0.9996316056502493, + "mean_feedback_forward_relative_error": 0.026661050898071966, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001984984467698352, + "instruction_rms": 0.0019849844670150416, + "raw_apical_rms": 0.03549006263098476, + "teaching_rms": 0.001984984467698352, + "traffic_rms": 0.03541851113385414 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2651434566635089e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6169507411084307e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009769807110601119, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.015453135259283914 + }, + { + "epoch": 169, + "eval_accuracy": 0.9138, + "eval_loss": 0.4206541809082031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780760243535042, + "mean_feedback_forward_cosine": 0.9996315899648165, + "mean_feedback_forward_relative_error": 0.026660842232798274, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00202351260849575, + "instruction_rms": 0.002023512608972183, + "raw_apical_rms": 0.03549055191927649, + "teaching_rms": 0.00202351260849575, + "traffic_rms": 0.03541667822482855 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.0796592952999617e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6186895543251947e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009821765196698794, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.015848362532920306 + }, + { + "epoch": 170, + "eval_accuracy": 0.914, + "eval_loss": 0.41541182861328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780755773186684, + "mean_feedback_forward_cosine": 0.9996315962389896, + "mean_feedback_forward_relative_error": 0.02666063675362813, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001959811377964408, + "instruction_rms": 0.0019598113775032918, + "raw_apical_rms": 0.035483915716881025, + "teaching_rms": 0.001959811377964408, + "traffic_rms": 0.035415321778480946 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3844236868010285e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.615777851086534e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0097210159161736, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.014944665809389617 + }, + { + "epoch": 171, + "eval_accuracy": 0.914, + "eval_loss": 0.4189151428222656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780750930309296, + "mean_feedback_forward_cosine": 0.999631615061509, + "mean_feedback_forward_relative_error": 0.026660423578792496, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020314639544911626, + "instruction_rms": 0.002031463954675679, + "raw_apical_rms": 0.03549389696505197, + "teaching_rms": 0.0020314639544911626, + "traffic_rms": 0.0354194518350967 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3361433737202333e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6189122268711468e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009821385755846523, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.01616706893179152 + }, + { + "epoch": 172, + "eval_accuracy": 0.9142, + "eval_loss": 0.41772709350585935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780745342373848, + "mean_feedback_forward_cosine": 0.999631615061509, + "mean_feedback_forward_relative_error": 0.02666021461941694, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020211037602693247, + "instruction_rms": 0.0020211037607435027, + "raw_apical_rms": 0.03549044453860664, + "teaching_rms": 0.0020211037602693247, + "traffic_rms": 0.03541693276093822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.447212927274677e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6164672638921847e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009711404564543315, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.015749647364185916 + }, + { + "epoch": 173, + "eval_accuracy": 0.9138, + "eval_loss": 0.42026092529296877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0378074124455452, + "mean_feedback_forward_cosine": 0.9996316181985956, + "mean_feedback_forward_relative_error": 0.026660006640380936, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020449684581992575, + "instruction_rms": 0.0020449684588546573, + "raw_apical_rms": 0.03549233281750145, + "teaching_rms": 0.0020449684581992575, + "traffic_rms": 0.0354157516906314 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2915743141661551e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6188593762277255e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009791714520651178, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.01570732250975238 + }, + { + "epoch": 174, + "eval_accuracy": 0.913, + "eval_loss": 0.41367199096679685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037807367742061615, + "mean_feedback_forward_cosine": 0.9996316213356821, + "mean_feedback_forward_relative_error": 0.026659800229888213, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001886148273150442, + "instruction_rms": 0.0018861482728113074, + "raw_apical_rms": 0.03548341077420893, + "teaching_rms": 0.001886148273150442, + "traffic_rms": 0.03541795860143163 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2818100358913398e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619612116302527e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009807925893813902, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.014504380411075221 + }, + { + "epoch": 175, + "eval_accuracy": 0.9138, + "eval_loss": 0.42116400146484373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037807315587997437, + "mean_feedback_forward_cosine": 0.999631615061509, + "mean_feedback_forward_relative_error": 0.02665959636827833, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001891071507064046, + "instruction_rms": 0.0018910715072027106, + "raw_apical_rms": 0.03548437812515092, + "teaching_rms": 0.001891071507064046, + "traffic_rms": 0.03541960000244503 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4794984792843024e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6202480277822806e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009742782585969964, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.014262648306621446 + }, + { + "epoch": 176, + "eval_accuracy": 0.9146, + "eval_loss": 0.41923724975585935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780727460980415, + "mean_feedback_forward_cosine": 0.9996316025131627, + "mean_feedback_forward_relative_error": 0.026659388095140457, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001960660959632219, + "instruction_rms": 0.0019606609598975095, + "raw_apical_rms": 0.03548896812625248, + "teaching_rms": 0.001960660959632219, + "traffic_rms": 0.035419966616758204 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.383726022652354e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6176810175820067e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009748636295267578, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.014850295876794392 + }, + { + "epoch": 177, + "eval_accuracy": 0.9146, + "eval_loss": 0.42256874389648436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037807222455739975, + "mean_feedback_forward_cosine": 0.999631615061509, + "mean_feedback_forward_relative_error": 0.026659192125263968, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018412803620673794, + "instruction_rms": 0.0018412803619506646, + "raw_apical_rms": 0.035484063996722395, + "teaching_rms": 0.0018412803620673794, + "traffic_rms": 0.03542176508889632 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4235107315130335e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6168921716787525e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009706403903176812, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.013851795967751079 + }, + { + "epoch": 178, + "eval_accuracy": 0.9146, + "eval_loss": 0.42203513793945313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780717775225639, + "mean_feedback_forward_cosine": 0.999631615061509, + "mean_feedback_forward_relative_error": 0.026658983509007254, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001962376031336217, + "instruction_rms": 0.001962376031402494, + "raw_apical_rms": 0.035488060418281044, + "teaching_rms": 0.001962376031336217, + "traffic_rms": 0.035417335802325374 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2479540068000006e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619737439999005e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00975698656244246, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.01524298331340154 + }, + { + "epoch": 179, + "eval_accuracy": 0.9148, + "eval_loss": 0.41612083740234374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037807125598192215, + "mean_feedback_forward_cosine": 0.999631615061509, + "mean_feedback_forward_relative_error": 0.026658773275190277, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002171803203326694, + "instruction_rms": 0.002171803203451999, + "raw_apical_rms": 0.03549754966817121, + "teaching_rms": 0.002171803203326694, + "traffic_rms": 0.03541387145940637 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3130986076248519e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6186041789509365e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009808782243198215, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.016495510240230295 + }, + { + "epoch": 180, + "eval_accuracy": 0.9144, + "eval_loss": 0.4196902770996094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780708089470863, + "mean_feedback_forward_cosine": 0.9996316087873358, + "mean_feedback_forward_relative_error": 0.026658564217780765, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018914864539903626, + "instruction_rms": 0.0018914864538349249, + "raw_apical_rms": 0.035480294072998284, + "teaching_rms": 0.0018914864539903626, + "traffic_rms": 0.035413231681480466 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3206939541987595e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6179269646261458e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009765187053182557, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.014542664614982076 + }, + { + "epoch": 181, + "eval_accuracy": 0.9144, + "eval_loss": 0.41798706665039065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780703619122505, + "mean_feedback_forward_cosine": 0.9996316307469418, + "mean_feedback_forward_relative_error": 0.026658367659700543, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019398865705368001, + "instruction_rms": 0.0019398865698302828, + "raw_apical_rms": 0.03548393447367662, + "teaching_rms": 0.0019398865705368001, + "traffic_rms": 0.035415769575345556 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.1239237096560828e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6171983030625008e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009820256248368584, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.014637808668613434 + }, + { + "epoch": 182, + "eval_accuracy": 0.9142, + "eval_loss": 0.42433768920898435, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037806980311870575, + "mean_feedback_forward_cosine": 0.9996316244727687, + "mean_feedback_forward_relative_error": 0.026658172621146628, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019189778150392788, + "instruction_rms": 0.0019189778147332775, + "raw_apical_rms": 0.03548243117479171, + "teaching_rms": 0.0019189778150392788, + "traffic_rms": 0.03541347583140527 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.5036496492371043e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6171335935078734e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009778777356824414, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.015194662693805166 + }, + { + "epoch": 183, + "eval_accuracy": 0.9132, + "eval_loss": 0.42002318115234377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0378069207072258, + "mean_feedback_forward_cosine": 0.9996316087873358, + "mean_feedback_forward_relative_error": 0.02665796532834831, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020331055821016237, + "instruction_rms": 0.0020331055826984887, + "raw_apical_rms": 0.03548177099625021, + "teaching_rms": 0.0020331055821016237, + "traffic_rms": 0.03540610143311877 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.3314930491503674e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6171520037725326e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009748853273521037, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.015491151970624924 + }, + { + "epoch": 184, + "eval_accuracy": 0.9136, + "eval_loss": 0.4230053771972656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780688717961311, + "mean_feedback_forward_cosine": 0.9996316213356821, + "mean_feedback_forward_relative_error": 0.02665776131968749, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018688744556348029, + "instruction_rms": 0.0018688744559231128, + "raw_apical_rms": 0.03547066975577514, + "teaching_rms": 0.0018688744556348029, + "traffic_rms": 0.03540849608681272 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2078251288016872e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6177227780445252e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00972546943895346, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.013875001421405209 + }, + { + "epoch": 185, + "eval_accuracy": 0.9142, + "eval_loss": 0.4274574951171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780684992671013, + "mean_feedback_forward_cosine": 0.9996316370211149, + "mean_feedback_forward_relative_error": 0.026657556477738053, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018596682569046118, + "instruction_rms": 0.0018596682571716025, + "raw_apical_rms": 0.03547805138678572, + "teaching_rms": 0.0018596682569046118, + "traffic_rms": 0.035413795228133545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2427649132007446e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619894433203317e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009703887064664369, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.014032311059700119 + }, + { + "epoch": 186, + "eval_accuracy": 0.9132, + "eval_loss": 0.42395435791015623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037806812673807144, + "mean_feedback_forward_cosine": 0.9996316307469418, + "mean_feedback_forward_relative_error": 0.02665735756684291, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018347543847815131, + "instruction_rms": 0.001834754385007048, + "raw_apical_rms": 0.03546985095177895, + "teaching_rms": 0.0018347543847815131, + "traffic_rms": 0.03540835359612842 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4097837564008842e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6180625176744536e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009759561785606016, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.013660452849997414 + }, + { + "epoch": 187, + "eval_accuracy": 0.914, + "eval_loss": 0.42424832153320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780675679445267, + "mean_feedback_forward_cosine": 0.9996316244727687, + "mean_feedback_forward_relative_error": 0.026657164488968096, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002006498395177536, + "instruction_rms": 0.0020064983952284275, + "raw_apical_rms": 0.03548441304733765, + "teaching_rms": 0.002006498395177536, + "traffic_rms": 0.03540871518407568 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3115205810265707e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6189264360511836e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009766192089174008, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.015087417976723777 + }, + { + "epoch": 188, + "eval_accuracy": 0.913, + "eval_loss": 0.4244384094238281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780670836567879, + "mean_feedback_forward_cosine": 0.9996316213356821, + "mean_feedback_forward_relative_error": 0.026656959696035636, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019518414181727644, + "instruction_rms": 0.00195184141939903, + "raw_apical_rms": 0.0354842783765005, + "teaching_rms": 0.0019518414181727644, + "traffic_rms": 0.03541406576694399 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2689654660391625e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6187451529159547e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00966205666092081, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.014488841858175066 + }, + { + "epoch": 189, + "eval_accuracy": 0.9128, + "eval_loss": 0.42518949584960936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780665621161461, + "mean_feedback_forward_cosine": 0.9996316401582015, + "mean_feedback_forward_relative_error": 0.026656765343719406, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019539568888954336, + "instruction_rms": 0.0019539568878578707, + "raw_apical_rms": 0.03548744358893964, + "teaching_rms": 0.0019539568888954336, + "traffic_rms": 0.03541403026304128 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3715903968147813e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6187100559918243e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009720175671811639, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.015103299892942111 + }, + { + "epoch": 190, + "eval_accuracy": 0.912, + "eval_loss": 0.4249855224609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037806615233421326, + "mean_feedback_forward_cosine": 0.9996316338840284, + "mean_feedback_forward_relative_error": 0.026656565746586574, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019031509788945931, + "instruction_rms": 0.001903150978429193, + "raw_apical_rms": 0.03548138778264967, + "teaching_rms": 0.0019031509788945931, + "traffic_rms": 0.03541441407733898 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.3956911182333442e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6212568066008135e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009812432448323301, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.014304727699855963 + }, + { + "epoch": 191, + "eval_accuracy": 0.913, + "eval_loss": 0.4252706420898438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780658170580864, + "mean_feedback_forward_cosine": 0.9996316558436343, + "mean_feedback_forward_relative_error": 0.026656368796370532, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018229452660776897, + "instruction_rms": 0.0018229452649682286, + "raw_apical_rms": 0.035472715103697164, + "teaching_rms": 0.0018229452660776897, + "traffic_rms": 0.03541313239646838 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4061011910371235e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6228162384022934e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009659836963272264, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.013309210489193599 + }, + { + "epoch": 192, + "eval_accuracy": 0.9136, + "eval_loss": 0.4258235900878906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780653327703476, + "mean_feedback_forward_cosine": 0.9996316558436343, + "mean_feedback_forward_relative_error": 0.026656166650354862, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018731226617284761, + "instruction_rms": 0.0018731226609877986, + "raw_apical_rms": 0.03547900176202172, + "teaching_rms": 0.0018731226617284761, + "traffic_rms": 0.03541178669681627 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4582523633066558e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619988173584462e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009646573340541543, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.013877916736486886 + }, + { + "epoch": 193, + "eval_accuracy": 0.9122, + "eval_loss": 0.4229419250488281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780648484826088, + "mean_feedback_forward_cosine": 0.9996316370211149, + "mean_feedback_forward_relative_error": 0.026655971023597215, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019435254211202456, + "instruction_rms": 0.0019435254209826719, + "raw_apical_rms": 0.035481766034522494, + "teaching_rms": 0.0019435254211202456, + "traffic_rms": 0.03541070988285963 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4557861582886744e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6205606463700388e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00970974169740098, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.01430190967188941 + }, + { + "epoch": 194, + "eval_accuracy": 0.9126, + "eval_loss": 0.4232492980957031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0378064326941967, + "mean_feedback_forward_cosine": 0.9996316464323747, + "mean_feedback_forward_relative_error": 0.026655769269717366, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020056041251614136, + "instruction_rms": 0.002005604125471447, + "raw_apical_rms": 0.03548394231694213, + "teaching_rms": 0.0020056041251614136, + "traffic_rms": 0.03540891461726112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2684434835819047e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.618433978649905e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009615365573339791, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.014883263099855847 + }, + { + "epoch": 195, + "eval_accuracy": 0.9136, + "eval_loss": 0.42721522216796876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037806395441293716, + "mean_feedback_forward_cosine": 0.9996316527065477, + "mean_feedback_forward_relative_error": 0.0266555695255336, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00193867040984838, + "instruction_rms": 0.0019386704093032282, + "raw_apical_rms": 0.03547711030343143, + "teaching_rms": 0.00193867040984838, + "traffic_rms": 0.03540596497952997 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.2688057715593004e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.619342135782023e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009858859734747524, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.014291609461274411 + }, + { + "epoch": 196, + "eval_accuracy": 0.9138, + "eval_loss": 0.42337653198242187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780634328722954, + "mean_feedback_forward_cosine": 0.9996316589807209, + "mean_feedback_forward_relative_error": 0.026655367134433044, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020019146099371065, + "instruction_rms": 0.002001914610000151, + "raw_apical_rms": 0.03547915490680592, + "teaching_rms": 0.0020019146099371065, + "traffic_rms": 0.035404544313311116 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.4446251306310387e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.622310184832768e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00974169248560291, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.015002086046669218 + }, + { + "epoch": 197, + "eval_accuracy": 0.9132, + "eval_loss": 0.42314592895507813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780629485845566, + "mean_feedback_forward_cosine": 0.9996316558436343, + "mean_feedback_forward_relative_error": 0.02665516979208118, + "min_feedback_forward_cosine": 0.999288022518158 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019091024603237609, + "instruction_rms": 0.0019091024594505403, + "raw_apical_rms": 0.03546936178603817, + "teaching_rms": 0.0019091024603237609, + "traffic_rms": 0.03540039946082456 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.174481578658515e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.61961026386494e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009675740384794889, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.014087256733576456 + }, + { + "epoch": 198, + "eval_accuracy": 0.9138, + "eval_loss": 0.4227126831054687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.03780624642968178, + "mean_feedback_forward_cosine": 0.9996316621178075, + "mean_feedback_forward_relative_error": 0.026654975586815885, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019059753502149801, + "instruction_rms": 0.0019059753496980464, + "raw_apical_rms": 0.035465963658777015, + "teaching_rms": 0.0019059753502149801, + "traffic_rms": 0.03539591078722619 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.2380260372424345e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6286669231052677e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009671483722060802, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.01432547069489956 + }, + { + "epoch": 199, + "eval_accuracy": 0.9114, + "eval_loss": 0.4212016235351563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037806201726198196, + "mean_feedback_forward_cosine": 0.9996316621178075, + "mean_feedback_forward_relative_error": 0.02665476863713641, + "min_feedback_forward_cosine": 0.9992880821228027 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019068836071506673, + "instruction_rms": 0.0019068836076270475, + "raw_apical_rms": 0.03546137798021448, + "teaching_rms": 0.0019068836071506673, + "traffic_rms": 0.035395210269209734 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110409900546074, + "maximum_absolute_post_projection_soma_slope": 1.6277230230343775e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6185724249417297e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009620592725691523, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.013825085100448794 + }, + { + "epoch": 200, + "eval_accuracy": 0.9132, + "eval_loss": 0.4233872314453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.037806157022714615, + "mean_feedback_forward_cosine": 0.9996316621178075, + "mean_feedback_forward_relative_error": 0.026654567569494247, + "min_feedback_forward_cosine": 0.9992879629135132 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001971850176861854, + "instruction_rms": 0.0019718501772984765, + "raw_apical_rms": 0.035469325266448755, + "teaching_rms": 0.001971850176861854, + "traffic_rms": 0.03539758458336501 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.03110410086810589, + "maximum_absolute_post_projection_soma_slope": 1.1857881787591396e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6222237096517437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009666236036098026, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.014849238505380022 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9132, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.4233872314453125, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2013018112, + "peak_memory_reserved_bytes": 2774532096, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 1.6958033199898637e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 1.6958033199898637e-08, + "min_residual_soma_slope": -1.5074077097665395e-08, + "mse": 1.4716528582992302e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.197853883807226e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.197853883807226e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 21.13534927368164, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.024861574172973633, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3003.0224483013153, + "train_wall_s": 2980.52508187294 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.008468643762171268, + 0.006615260150283575, + 0.003986957482993603, + 0.004385664593428373, + 0.0027129307854920626, + 0.0031575150787830353, + 0.002003862289711833, + 0.0037573478184640408, + 0.0027080816216766834, + 0.0030197801534086466, + 0.0020560885313898325, + 0.0023898854851722717, + 0.0016841775504872203, + 0.0034015050623565912, + 0.0025495023000985384, + 0.003054187633097172, + 0.0020917924121022224, + 0.002605562563985586, + 0.001891607535071671 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.000000476837158, + 4.0, + 4.0, + 4.000000476837158 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.04418104141950607, + 0.03573717549443245, + 0.013813067227602005, + 0.022754332050681114, + 0.00723816454410553, + 0.01724369451403618, + 0.004617006052285433, + 0.020357992500066757, + 0.007312532514333725, + 0.01600862480700016, + 0.004682047292590141, + 0.012491599656641483, + 0.00330126890912652, + 0.017684802412986755, + 0.006067128852009773, + 0.016301598399877548, + 0.004425332881510258, + 0.013932524248957634, + 0.0035721727181226015 + ], + "unscaled_traffic_rms": [ + 0.7667219638824463, + 0.7404345870018005, + 1.1545466184616089, + 0.7709590792655945, + 1.499236822128296, + 0.7324451804161072, + 1.7360707521438599, + 0.7382550835609436, + 1.4813371896743774, + 0.7545382976531982, + 1.7565721273422241, + 0.765277624130249, + 2.040642738342285, + 0.7693622708320618, + 1.6808624267578125, + 0.7494204044342041, + 1.890743613243103, + 0.7480518221855164, + 2.118159055709839 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 360980614955008, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 360978048000000, + "local_weight_correlation_macs": 364959360000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 364959360000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 2595266560 + }, + "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": 40698241875968, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 9000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 9000064, + "total_forward_equivalent_examples": 9000064, + "total_macs_estimate": 1451879978221568 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json.manifest.json new file mode 100644 index 0000000..a751944 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json.manifest.json @@ -0,0 +1,150 @@ +{ + "architecture": "resnet20", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "kp_traffic", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "20", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--traffic_rule", + "innovation", + "--predictor_mode", + "closed_form", + "--neutral_projection", + "1", + "--traffic_seed", + "5000", + "--traffic_ratio", + "4", + "--traffic_calibration_examples", + "64", + "--learn_P", + "1", + "--eta_P", + "0.1", + "--predictor_warmup_steps", + "1", + "--predictor_every", + "0" + ], + "completed_unix_time": 1787034193.1208334, + "cuda_visible_devices": "0", + "depth": 20, + "driver_wall_seconds": 3006.6630001068115, + "epochs": 200, + "experiment_name": "resnet-r2-sdil-d20-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "sdil", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d20-lr0p1.json", + "output_exists": true, + "output_sha256": "b0be7a2e395429c4bb29209bdce5d250dccd484048551fde04102553ed2ad272", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json new file mode 100644 index 0000000..83cf766 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json @@ -0,0 +1,7472 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 1, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.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": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "ordinary_examples": 9000000, + "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": 0.9990641891956329, + "feedback_forward_cosine": [ + 0.9995248913764954, + 0.999525785446167, + 0.9993621110916138, + 0.9992882013320923, + 0.9992794990539551, + 0.9991533756256104, + 0.9995105266571045, + 0.9992536306381226, + 0.9993960857391357, + 0.9991365671157837, + 0.9993268847465515, + 0.9995407462120056, + 0.9993659257888794, + 0.9992746114730835, + 0.9994126558303833, + 0.9992644786834717, + 0.9995049834251404, + 0.9993177652359009, + 0.9994486570358276, + 0.9992416501045227, + 0.999461829662323, + 0.9996631145477295, + 0.9995964169502258, + 0.9995673298835754, + 0.9996181726455688, + 0.9994949102401733, + 0.9995560646057129, + 0.9993234872817993, + 0.9994652271270752, + 0.9987451434135437, + 0.9999646544456482 + ], + "feedback_forward_norm_ratio": [ + 0.9988867044448853, + 1.0008950233459473, + 0.9997606873512268, + 1.0032583475112915, + 0.9985292553901672, + 1.0024914741516113, + 0.998984158039093, + 1.0015926361083984, + 0.9991874098777771, + 1.0027505159378052, + 0.9982722401618958, + 1.0011152029037476, + 0.9988278746604919, + 1.0015062093734741, + 0.9987968802452087, + 1.0027403831481934, + 0.9992345571517944, + 1.0019012689590454, + 0.9989014267921448, + 1.0029679536819458, + 0.9991740584373474, + 1.0008102655410767, + 0.9995593428611755, + 1.001561164855957, + 1.0000790357589722, + 1.0014222860336304, + 0.9998015761375427, + 1.0020161867141724, + 0.999867856502533, + 1.0024110078811646, + 1.0053294897079468 + ], + "feedback_forward_relative_error": [ + 0.030829928815364838, + 0.03082478605210781, + 0.03571289777755737, + 0.037930626422166824, + 0.037959691137075424, + 0.041274625808000565, + 0.03128883242607117, + 0.03869929164648056, + 0.03474787622690201, + 0.04170121252536774, + 0.03669923543930054, + 0.030344000086188316, + 0.0356106236577034, + 0.038146961480379105, + 0.034275274723768234, + 0.03850400820374489, + 0.031461749225854874, + 0.0370202399790287, + 0.03320717811584473, + 0.039115216583013535, + 0.03280482068657875, + 0.02598089911043644, + 0.028407160192728043, + 0.029481027275323868, + 0.027640528976917267, + 0.03183675557374954, + 0.02979353256523609, + 0.036875028163194656, + 0.032700978219509125, + 0.05021461844444275, + 0.009983262047171593 + ], + "innovation_negative_gradient_cosine": [ + 0.9991377592086792, + 0.9990687966346741, + 0.9991105794906616, + 0.9988575577735901, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993926882743835, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.9999379515647888, + 0.7521367073059082 + ], + "instruction_negative_gradient_cosine": [ + 0.999137818813324, + 0.9990687966346741, + 0.9991105794906616, + 0.9988574981689453, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993927478790283, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.999937891960144, + 0.755734920501709 + ], + "matched_negative_gradient_cosine": [ + 0.004778770729899406, + 0.005657309666275978, + 0.0010874532163143158, + 0.018822133541107178, + 0.00964898057281971, + 0.016325540840625763, + 0.017202919349074364, + 0.017850112169981003, + 0.023066185414791107, + 0.041207410395145416, + 0.020595937967300415, + -0.04097568243741989, + -0.003883502446115017, + 0.03719305247068405, + 0.01840835250914097, + 0.05979178845882416, + 0.025737924501299858, + 0.10348624736070633, + 0.022145891562104225, + 0.05590894818305969, + 0.056562818586826324, + 0.0805809497833252, + 0.044539716094732285, + 0.16223689913749695, + 0.06719361245632172, + 0.2673830986022949, + 0.09676877409219742, + 0.24156257510185242, + 0.09677475690841675, + 0.1181703507900238, + 0.12388422340154648 + ], + "max_norm_match_direction_error": 1.7881393432617188e-07, + "max_norm_match_relative_error": 1.6217698828313587e-07, + "mean_feedback_forward_cosine": 0.9994059801101685, + "mean_feedback_forward_relative_error": 0.03390557637378093, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.06770927459001541, + "max_absolute_post_projection_soma_slope": 1.9101433323953643e-08, + "max_absolute_pre_projection_soma_slope": 0.06770927459001541, + "max_positive_post_projection_soma_slope": 8.284928298962768e-09, + "min_post_projection_soma_slope": -1.9101433323953643e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.5151598565298195e-08, + "pre_projection_traffic_rms_ratio": 0.011312205665592186 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.011312205665559983, + "raw_negative_gradient_cosine": [ + 0.004778770264238119, + 0.005657309666275978, + 0.0010874529834836721, + 0.018822133541107178, + 0.00964897871017456, + 0.016325540840625763, + 0.017202919349074364, + 0.0178501158952713, + 0.023066185414791107, + 0.041207410395145416, + 0.020595937967300415, + -0.04097568243741989, + -0.0038835033774375916, + 0.03719305619597435, + 0.01840835250914097, + 0.05979178845882416, + 0.025737926363945007, + 0.10348625481128693, + 0.022145891562104225, + 0.05590894818305969, + 0.056562818586826324, + 0.0805809497833252, + 0.044539712369441986, + 0.16223689913749695, + 0.06719361245632172, + 0.2673830986022949, + 0.09676877409219742, + 0.24156257510185242, + 0.09677475690841675, + 0.1181703582406044, + 0.12388423085212708 + ], + "teaching_negative_gradient_cosine": [ + 0.9991377592086792, + 0.9990687966346741, + 0.9991105794906616, + 0.9988575577735901, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993926882743835, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.9999379515647888, + 0.7521367073059082 + ], + "traffic_instruction_rms_ratio": 23.54690779022645, + "used_negative_gradient_cosine": [ + 0.9991377592086792, + 0.9990687966346741, + 0.9991105794906616, + 0.9988575577735901, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993926882743835, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.9999379515647888, + 0.7521367073059082 + ], + "wall_s": 0.1316971778869629 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.388, + "eval_loss": 1.6842045654296876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3977231979370117, + "mean_feedback_forward_cosine": 0.10176758456134027, + "mean_feedback_forward_relative_error": 1.3543651142428, + "min_feedback_forward_cosine": 0.03091779351234436 + }, + "lr": 0.02, + "mixed_apical": { + "innovation_rms": 0.004567544741317272, + "instruction_rms": 0.004567544740593721, + "raw_apical_rms": 0.03513459748022013, + "teaching_rms": 0.004567544741317272, + "traffic_rms": 0.03482145024585586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677093043923378, + "maximum_absolute_post_projection_soma_slope": 3.303899731577076e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.048501115545045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007256994196668291, + "minimum_observations": 72 + }, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.0257865279939438 + }, + { + "epoch": 2, + "eval_accuracy": 0.474, + "eval_loss": 1.59408642578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3580535650253296, + "mean_feedback_forward_cosine": 0.24024918771559192, + "mean_feedback_forward_relative_error": 1.2470388316339063, + "min_feedback_forward_cosine": 0.09387169778347015 + }, + "lr": 0.04, + "mixed_apical": { + "innovation_rms": 0.005307031705626174, + "instruction_rms": 0.005307031705694782, + "raw_apical_rms": 0.040016573325931976, + "teaching_rms": 0.005307031705626174, + "traffic_rms": 0.039657659576885225 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.864828895279061e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.9509758007218805e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009549755634508534, + "minimum_observations": 72 + }, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.5833565533955891 + }, + { + "epoch": 3, + "eval_accuracy": 0.5938, + "eval_loss": 1.21504912109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3161413669586182, + "mean_feedback_forward_cosine": 0.3675699810827932, + "mean_feedback_forward_relative_error": 1.1383580238588396, + "min_feedback_forward_cosine": 0.15631060302257538 + }, + "lr": 0.06000000000000001, + "mixed_apical": { + "innovation_rms": 0.005956704153025323, + "instruction_rms": 0.005956704150186274, + "raw_apical_rms": 0.04542083796175327, + "teaching_rms": 0.005956704153025323, + "traffic_rms": 0.04502470954108027 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4744238658058748e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8605195850272107e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00897240334648494, + "minimum_observations": 72 + }, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.1766535493850707 + }, + { + "epoch": 4, + "eval_accuracy": 0.6, + "eval_loss": 1.2461006103515626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2697638273239136, + "mean_feedback_forward_cosine": 0.46007657051086426, + "mean_feedback_forward_relative_error": 1.0509521999666769, + "min_feedback_forward_cosine": 0.2210283875465393 + }, + "lr": 0.08, + "mixed_apical": { + "innovation_rms": 0.005509644266008199, + "instruction_rms": 0.005509644266709996, + "raw_apical_rms": 0.04825561232693312, + "teaching_rms": 0.005509644266008199, + "traffic_rms": 0.047937556891173266 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.671688020290276e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8255515906846483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007464922171899973, + "minimum_observations": 72 + }, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 0.9523900893847148 + }, + { + "epoch": 5, + "eval_accuracy": 0.7276, + "eval_loss": 0.7826447875976562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2237781286239624, + "mean_feedback_forward_cosine": 0.5366203938761065, + "mean_feedback_forward_relative_error": 0.973043107217358, + "min_feedback_forward_cosine": 0.27918457984924316 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005091983841681295, + "instruction_rms": 0.00509198384066812, + "raw_apical_rms": 0.05179744654933709, + "teaching_rms": 0.005091983841681295, + "traffic_rms": 0.05154422627249688 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.839830059997439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7884721745800226e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007986453419867427, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.815070916832818 + }, + { + "epoch": 6, + "eval_accuracy": 0.7534, + "eval_loss": 0.7137961181640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1856058835983276, + "mean_feedback_forward_cosine": 0.5884564268973566, + "mean_feedback_forward_relative_error": 0.9165485866608158, + "min_feedback_forward_cosine": 0.32551515102386475 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005027530190887641, + "instruction_rms": 0.005027530189482462, + "raw_apical_rms": 0.05350177493955319, + "teaching_rms": 0.005027530190887641, + "traffic_rms": 0.05326255006739937 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9742937002197323e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7749626564124105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008138300633405591, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7121988859388564 + }, + { + "epoch": 7, + "eval_accuracy": 0.719, + "eval_loss": 0.8742424682617187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1527107954025269, + "mean_feedback_forward_cosine": 0.6276193309214807, + "mean_feedback_forward_relative_error": 0.8711237195999392, + "min_feedback_forward_cosine": 0.3639024794101715 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004874586884066469, + "instruction_rms": 0.004874586884370733, + "raw_apical_rms": 0.05472647430139628, + "teaching_rms": 0.004874586884066469, + "traffic_rms": 0.054506565328151145 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6285386311997172e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.760684749344245e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008401987143937221, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6420152118682861 + }, + { + "epoch": 8, + "eval_accuracy": 0.7544, + "eval_loss": 0.7437487548828124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1242905855178833, + "mean_feedback_forward_cosine": 0.6590101055560573, + "mean_feedback_forward_relative_error": 0.8331726616428744, + "min_feedback_forward_cosine": 0.3955609202384949 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0046604658948121765, + "instruction_rms": 0.00466046589273467, + "raw_apical_rms": 0.05602660090017935, + "teaching_rms": 0.0046604658948121765, + "traffic_rms": 0.05583040206733472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9735030437905152e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7574253798858607e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00839715642009344, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.596587984000312 + }, + { + "epoch": 9, + "eval_accuracy": 0.7946, + "eval_loss": 0.60154697265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0955640077590942, + "mean_feedback_forward_cosine": 0.6862208708640067, + "mean_feedback_forward_relative_error": 0.7990370306276506, + "min_feedback_forward_cosine": 0.42728692293167114 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004562242773027457, + "instruction_rms": 0.004562242772240408, + "raw_apical_rms": 0.05656990566652518, + "teaching_rms": 0.004562242773027457, + "traffic_rms": 0.05638386271176832 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.863355952342772e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7548746334258725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008591035332460245, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5534716961648729 + }, + { + "epoch": 10, + "eval_accuracy": 0.7944, + "eval_loss": 0.6164186157226562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0699032545089722, + "mean_feedback_forward_cosine": 0.7099020461882314, + "mean_feedback_forward_relative_error": 0.7679428881214511, + "min_feedback_forward_cosine": 0.4542800188064575 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004563739370769258, + "instruction_rms": 0.004563739371583582, + "raw_apical_rms": 0.05763044514406948, + "teaching_rms": 0.004563739370769258, + "traffic_rms": 0.05744746317299652 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6442590339570415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.747594984543883e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008666427106516958, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5181713239722782 + }, + { + "epoch": 11, + "eval_accuracy": 0.814, + "eval_loss": 0.5557191284179688, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0450198650360107, + "mean_feedback_forward_cosine": 0.7315100171873646, + "mean_feedback_forward_relative_error": 0.7385791319031869, + "min_feedback_forward_cosine": 0.4804466664791107 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004469426111132943, + "instruction_rms": 0.004469426108247147, + "raw_apical_rms": 0.05807760817231134, + "teaching_rms": 0.004469426111132943, + "traffic_rms": 0.057902990804586524 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.267847598602657e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7500813606099063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00868155962782433, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.49058791041904026 + }, + { + "epoch": 12, + "eval_accuracy": 0.8144, + "eval_loss": 0.5658951782226562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0195896625518799, + "mean_feedback_forward_cosine": 0.7507525624767426, + "mean_feedback_forward_relative_error": 0.7111556318498427, + "min_feedback_forward_cosine": 0.5057989954948425 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004474259591132489, + "instruction_rms": 0.004474259590376193, + "raw_apical_rms": 0.058553984622773635, + "teaching_rms": 0.004474259591132489, + "traffic_rms": 0.05838041382212191 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4798591624630717e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.736381041438012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008935329311463529, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.4756378570795059 + }, + { + "epoch": 13, + "eval_accuracy": 0.783, + "eval_loss": 0.6674646728515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9945835471153259, + "mean_feedback_forward_cosine": 0.7681244919376988, + "mean_feedback_forward_relative_error": 0.6856772361263153, + "min_feedback_forward_cosine": 0.5301342010498047 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004440338589064446, + "instruction_rms": 0.004440338590856765, + "raw_apical_rms": 0.05932498650991804, + "teaching_rms": 0.004440338589064446, + "traffic_rms": 0.05915643318048489 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.289915184723213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7339047457168407e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009083964661700545, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.4465404761367374 + }, + { + "epoch": 14, + "eval_accuracy": 0.8244, + "eval_loss": 0.5664008117675782, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9696456789970398, + "mean_feedback_forward_cosine": 0.7843829989433289, + "mean_feedback_forward_relative_error": 0.6608943179730447, + "min_feedback_forward_cosine": 0.5536136031150818 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004406709406016659, + "instruction_rms": 0.0044067094081843795, + "raw_apical_rms": 0.06019784191137481, + "teaching_rms": 0.004406709406016659, + "traffic_rms": 0.0600341441766403 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.981530400347765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.726925638732207e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009200338237390608, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.43072997092141047 + }, + { + "epoch": 15, + "eval_accuracy": 0.819, + "eval_loss": 0.5506152587890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9452754855155945, + "mean_feedback_forward_cosine": 0.7992827103983972, + "mean_feedback_forward_relative_error": 0.6372344061251609, + "min_feedback_forward_cosine": 0.5759081840515137 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004326748108247472, + "instruction_rms": 0.004326748109206437, + "raw_apical_rms": 0.06030161495920084, + "teaching_rms": 0.004326748108247472, + "traffic_rms": 0.060144057022412205 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.354581774000053e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7276153912306198e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009178980940521502, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.4152876753939523 + }, + { + "epoch": 16, + "eval_accuracy": 0.8252, + "eval_loss": 0.5355547790527344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9220921993255615, + "mean_feedback_forward_cosine": 0.8126598596572876, + "mean_feedback_forward_relative_error": 0.6151499200251794, + "min_feedback_forward_cosine": 0.5968871116638184 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004373927017103245, + "instruction_rms": 0.00437392701692843, + "raw_apical_rms": 0.060641845687035054, + "teaching_rms": 0.004373927017103245, + "traffic_rms": 0.06048179286830912 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.218938360305401e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7335190174540215e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009351303294551121, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.396997609117296 + }, + { + "epoch": 17, + "eval_accuracy": 0.8344, + "eval_loss": 0.49461121215820314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8985036015510559, + "mean_feedback_forward_cosine": 0.8253427820820962, + "mean_feedback_forward_relative_error": 0.5936438220162545, + "min_feedback_forward_cosine": 0.6173129081726074 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004339730517574678, + "instruction_rms": 0.0043397305169046016, + "raw_apical_rms": 0.06086485465183153, + "teaching_rms": 0.004339730517574678, + "traffic_rms": 0.06070770022252158 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7485743459010337e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.727211790915874e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009320647707649931, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.381810572581821 + }, + { + "epoch": 18, + "eval_accuracy": 0.8494, + "eval_loss": 0.44682206420898435, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8745174407958984, + "mean_feedback_forward_cosine": 0.8368679458095182, + "mean_feedback_forward_relative_error": 0.5733353779200585, + "min_feedback_forward_cosine": 0.6376420259475708 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004295681396640564, + "instruction_rms": 0.004295681397735242, + "raw_apical_rms": 0.061066769983406814, + "teaching_rms": 0.004295681396640564, + "traffic_rms": 0.06091296846652354 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.872836901701703e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7253816849063617e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009109225358336104, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.372761104800966 + }, + { + "epoch": 19, + "eval_accuracy": 0.8438, + "eval_loss": 0.5001562072753907, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8507093787193298, + "mean_feedback_forward_cosine": 0.8478189783711587, + "mean_feedback_forward_relative_error": 0.5536001593835892, + "min_feedback_forward_cosine": 0.6569257974624634 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00430706482340492, + "instruction_rms": 0.004307064824701717, + "raw_apical_rms": 0.06156562532931305, + "teaching_rms": 0.00430706482340492, + "traffic_rms": 0.06141233044377472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.177505820100123e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7231557651690612e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009345401061877862, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.36028081623183356 + }, + { + "epoch": 20, + "eval_accuracy": 0.8364, + "eval_loss": 0.5102594970703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8273930549621582, + "mean_feedback_forward_cosine": 0.8576913264489943, + "mean_feedback_forward_relative_error": 0.5349951569111117, + "min_feedback_forward_cosine": 0.6755627393722534 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004224719036629221, + "instruction_rms": 0.004224719038728415, + "raw_apical_rms": 0.061968769954962054, + "teaching_rms": 0.004224719036629221, + "traffic_rms": 0.061822003201752664 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.2150183909275256e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7203478299842856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009124211101281124, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3493091468864017 + }, + { + "epoch": 21, + "eval_accuracy": 0.8274, + "eval_loss": 0.5551596557617188, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8041688203811646, + "mean_feedback_forward_cosine": 0.8668903343139156, + "mean_feedback_forward_relative_error": 0.517221026843594, + "min_feedback_forward_cosine": 0.6933361291885376 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004196544092540644, + "instruction_rms": 0.0041965440922780905, + "raw_apical_rms": 0.0622468100375289, + "teaching_rms": 0.004196544092540644, + "traffic_rms": 0.06210289419973713 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 1.9702342868299638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7178518780542035e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00913035517495945, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.33633084367116295 + }, + { + "epoch": 22, + "eval_accuracy": 0.8174, + "eval_loss": 0.5925237060546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7810643315315247, + "mean_feedback_forward_cosine": 0.8756693716972105, + "mean_feedback_forward_relative_error": 0.49964048016455864, + "min_feedback_forward_cosine": 0.7106687426567078 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004198794541783616, + "instruction_rms": 0.0041987945443736205, + "raw_apical_rms": 0.062473956208242275, + "teaching_rms": 0.004198794541783616, + "traffic_rms": 0.06233022832452929 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4426169531466257e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.715997397441632e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009176835462509384, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3273309647401174 + }, + { + "epoch": 23, + "eval_accuracy": 0.8468, + "eval_loss": 0.4822109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7584139704704285, + "mean_feedback_forward_cosine": 0.8838707900816395, + "mean_feedback_forward_relative_error": 0.48255138868285763, + "min_feedback_forward_cosine": 0.7267588376998901 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004134608060019288, + "instruction_rms": 0.004134608059179785, + "raw_apical_rms": 0.06299561503763626, + "teaching_rms": 0.004134608060019288, + "traffic_rms": 0.06285719573841968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.3491308454026694e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7131197218623048e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009187891024563598, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.3197494336181217 + }, + { + "epoch": 24, + "eval_accuracy": 0.8548, + "eval_loss": 0.47645836791992185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7370368242263794, + "mean_feedback_forward_cosine": 0.8913365890902858, + "mean_feedback_forward_relative_error": 0.4664817803329037, + "min_feedback_forward_cosine": 0.7420064806938171 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004068535321030107, + "instruction_rms": 0.004068535321475998, + "raw_apical_rms": 0.06336379104187902, + "teaching_rms": 0.004068535321030107, + "traffic_rms": 0.06323072750014327 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.353047889869231e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7124145812856903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009404880109961884, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.30924332915941877 + }, + { + "epoch": 25, + "eval_accuracy": 0.8522, + "eval_loss": 0.4688670715332031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.716600239276886, + "mean_feedback_forward_cosine": 0.8982402047803325, + "mean_feedback_forward_relative_error": 0.4512202475340136, + "min_feedback_forward_cosine": 0.7559291124343872 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004049261991271629, + "instruction_rms": 0.0040492619898989425, + "raw_apical_rms": 0.06351800839271533, + "teaching_rms": 0.004049261991271629, + "traffic_rms": 0.06338627554501733 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 2.3370080981521824e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7090493041667627e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00951750746944136, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.3014814792844984 + }, + { + "epoch": 26, + "eval_accuracy": 0.8586, + "eval_loss": 0.4448318176269531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6953040957450867, + "mean_feedback_forward_cosine": 0.9047312967238887, + "mean_feedback_forward_relative_error": 0.4362514446819982, + "min_feedback_forward_cosine": 0.7700546383857727 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004038551876452253, + "instruction_rms": 0.004038551875714417, + "raw_apical_rms": 0.06293602933749305, + "teaching_rms": 0.004038551876452253, + "traffic_rms": 0.06280328592312978 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.783110808479705e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7167502013725602e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009235098366898327, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.2990670070648193 + }, + { + "epoch": 27, + "eval_accuracy": 0.8444, + "eval_loss": 0.5124790283203124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6745215058326721, + "mean_feedback_forward_cosine": 0.9107907472118255, + "mean_feedback_forward_relative_error": 0.42197790549647424, + "min_feedback_forward_cosine": 0.7833300828933716 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004002224396894647, + "instruction_rms": 0.004002224397603151, + "raw_apical_rms": 0.06332909380438531, + "teaching_rms": 0.004002224396894647, + "traffic_rms": 0.06319963985975724 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7412404790538858e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7122489179406515e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009382553349983636, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.2910355213324229 + }, + { + "epoch": 28, + "eval_accuracy": 0.855, + "eval_loss": 0.45888658447265623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6546169519424438, + "mean_feedback_forward_cosine": 0.9165496057079684, + "mean_feedback_forward_relative_error": 0.40787488510531766, + "min_feedback_forward_cosine": 0.7957733869552612 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00405124829155215, + "instruction_rms": 0.004051248292788438, + "raw_apical_rms": 0.06371395587046495, + "teaching_rms": 0.00405124829155215, + "traffic_rms": 0.06358227999593934 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4874308834910153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7035052282601972e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009621776229063744, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.27963798101743065 + }, + { + "epoch": 29, + "eval_accuracy": 0.8446, + "eval_loss": 0.5147614685058594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6332876682281494, + "mean_feedback_forward_cosine": 0.9220803841467826, + "mean_feedback_forward_relative_error": 0.3939231518776186, + "min_feedback_forward_cosine": 0.8088099956512451 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0040142515849250875, + "instruction_rms": 0.0040142515845335396, + "raw_apical_rms": 0.06378851239153105, + "teaching_rms": 0.0040142515849250875, + "traffic_rms": 0.0636594051496206 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.671583487767748e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7058148057442953e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00963041799819651, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.28177171817620594 + }, + { + "epoch": 30, + "eval_accuracy": 0.8634, + "eval_loss": 0.42816656494140626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6144068837165833, + "mean_feedback_forward_cosine": 0.9271933878621748, + "mean_feedback_forward_relative_error": 0.38059873734751054, + "min_feedback_forward_cosine": 0.8199130296707153 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004047472878011346, + "instruction_rms": 0.00404747287724283, + "raw_apical_rms": 0.0639902815971012, + "teaching_rms": 0.004047472878011346, + "traffic_rms": 0.06385950306625889 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.661687759086817e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7010070054018895e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009610093831761419, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.27227373503314123 + }, + { + "epoch": 31, + "eval_accuracy": 0.8554, + "eval_loss": 0.47649244384765627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5947118401527405, + "mean_feedback_forward_cosine": 0.9319607653925496, + "mean_feedback_forward_relative_error": 0.3677683886020414, + "min_feedback_forward_cosine": 0.8311227560043335 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003970332986572767, + "instruction_rms": 0.0039703329867750966, + "raw_apical_rms": 0.06403122964700471, + "teaching_rms": 0.003970332986572767, + "traffic_rms": 0.06390497861215359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6646426337938465e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7014060207564076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009607793921135612, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.26816212072902257 + }, + { + "epoch": 32, + "eval_accuracy": 0.821, + "eval_loss": 0.5862628173828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.575197160243988, + "mean_feedback_forward_cosine": 0.9365121702994069, + "mean_feedback_forward_relative_error": 0.35506819621209174, + "min_feedback_forward_cosine": 0.841862678527832 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004041092568796793, + "instruction_rms": 0.004041092568857933, + "raw_apical_rms": 0.06434880632694744, + "teaching_rms": 0.004041092568796793, + "traffic_rms": 0.06421914223623654 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.928237918808918e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.700544054890139e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00970343761641436, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.2629036833021376 + }, + { + "epoch": 33, + "eval_accuracy": 0.8614, + "eval_loss": 0.4444266784667969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5564904808998108, + "mean_feedback_forward_cosine": 0.9406849799617645, + "mean_feedback_forward_relative_error": 0.3430360435478149, + "min_feedback_forward_cosine": 0.8517687320709229 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00403412739904424, + "instruction_rms": 0.0040341274000684395, + "raw_apical_rms": 0.06480477899317917, + "teaching_rms": 0.00403412739904424, + "traffic_rms": 0.06467634344777613 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.132125314664336e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6950566314144252e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009332616035410659, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.2601968579000897 + }, + { + "epoch": 34, + "eval_accuracy": 0.8706, + "eval_loss": 0.4194605712890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.538070559501648, + "mean_feedback_forward_cosine": 0.9445796762743304, + "mean_feedback_forward_relative_error": 0.33144712976871, + "min_feedback_forward_cosine": 0.8612483143806458 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003983203975616594, + "instruction_rms": 0.00398320397674507, + "raw_apical_rms": 0.06473500908896433, + "teaching_rms": 0.003983203975616594, + "traffic_rms": 0.0646094019501837 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.1507609915261128e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6970872730951366e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00974665736367693, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.2567433127456241 + }, + { + "epoch": 35, + "eval_accuracy": 0.859, + "eval_loss": 0.4610693786621094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5206394195556641, + "mean_feedback_forward_cosine": 0.9482109585115986, + "mean_feedback_forward_relative_error": 0.3202533092229597, + "min_feedback_forward_cosine": 0.8700209856033325 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003933592381475546, + "instruction_rms": 0.003933592382589914, + "raw_apical_rms": 0.06499513738354835, + "teaching_rms": 0.003933592381475546, + "traffic_rms": 0.06487294588293592 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.1977518471771873e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.700391664730098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009886096025631225, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.2506347870879703 + }, + { + "epoch": 36, + "eval_accuracy": 0.8618, + "eval_loss": 0.4532576416015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5033196806907654, + "mean_feedback_forward_cosine": 0.9514969010506907, + "mean_feedback_forward_relative_error": 0.30979928566563514, + "min_feedback_forward_cosine": 0.878451943397522 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003896671065599783, + "instruction_rms": 0.003896671065423597, + "raw_apical_rms": 0.06514360563307033, + "teaching_rms": 0.003896671065599783, + "traffic_rms": 0.0650237827795661 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4190093483866804e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.690560100655861e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009983557577443883, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.24602477144135368 + }, + { + "epoch": 37, + "eval_accuracy": 0.8672, + "eval_loss": 0.41887584838867187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.486396461725235, + "mean_feedback_forward_cosine": 0.9546678143162881, + "mean_feedback_forward_relative_error": 0.29938822239637375, + "min_feedback_forward_cosine": 0.8864339590072632 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003893000874761815, + "instruction_rms": 0.0038930008759012007, + "raw_apical_rms": 0.06537661699246185, + "teaching_rms": 0.003893000874761815, + "traffic_rms": 0.06525766422221822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 1.950840022857392e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.69036997062304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009698828726604075, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.2446405156109068 + }, + { + "epoch": 38, + "eval_accuracy": 0.8566, + "eval_loss": 0.4433000061035156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.471229612827301, + "mean_feedback_forward_cosine": 0.9574937397433866, + "mean_feedback_forward_relative_error": 0.289746877406874, + "min_feedback_forward_cosine": 0.8933587074279785 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003888784879666175, + "instruction_rms": 0.0038887848787239744, + "raw_apical_rms": 0.06561847299448788, + "teaching_rms": 0.003888784879666175, + "traffic_rms": 0.06550056080262137 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.945887445093831e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6862998767583053e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009867522792933625, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.23308202556769053 + }, + { + "epoch": 39, + "eval_accuracy": 0.852, + "eval_loss": 0.5213807067871093, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4555465579032898, + "mean_feedback_forward_cosine": 0.9602498469814178, + "mean_feedback_forward_relative_error": 0.28014876885760215, + "min_feedback_forward_cosine": 0.9002827405929565 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003943040376982913, + "instruction_rms": 0.00394304037560111, + "raw_apical_rms": 0.06544258303738437, + "teaching_rms": 0.003943040376982913, + "traffic_rms": 0.06532078750332516 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.1349510603840827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.691445734257437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009442867302361439, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.22730018768575455 + }, + { + "epoch": 40, + "eval_accuracy": 0.8622, + "eval_loss": 0.4563710205078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4407300651073456, + "mean_feedback_forward_cosine": 0.9627599639277304, + "mean_feedback_forward_relative_error": 0.27106243035485666, + "min_feedback_forward_cosine": 0.9065840840339661 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038601983349361778, + "instruction_rms": 0.0038601983341153136, + "raw_apical_rms": 0.06575174846595079, + "teaching_rms": 0.0038601983349361778, + "traffic_rms": 0.06563548604744925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.8701164112021615e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6795436523467688e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009744524816008222, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.22868775055673388 + }, + { + "epoch": 41, + "eval_accuracy": 0.8586, + "eval_loss": 0.4705298645019531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4253489375114441, + "mean_feedback_forward_cosine": 0.9651982476634364, + "mean_feedback_forward_relative_error": 0.26196353954653584, + "min_feedback_forward_cosine": 0.912874162197113 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038398989149705247, + "instruction_rms": 0.0038398989148078623, + "raw_apical_rms": 0.06570998625869473, + "teaching_rms": 0.0038398989149705247, + "traffic_rms": 0.06559499804204008 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.507201912749224e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6877465057365064e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00982549091408656, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.2281675344573127 + }, + { + "epoch": 42, + "eval_accuracy": 0.8692, + "eval_loss": 0.4207826354980469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41090819239616394, + "mean_feedback_forward_cosine": 0.9674562023532006, + "mean_feedback_forward_relative_error": 0.25319438043140596, + "min_feedback_forward_cosine": 0.9185685515403748 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003900329559278322, + "instruction_rms": 0.003900329559352517, + "raw_apical_rms": 0.06577654280179057, + "teaching_rms": 0.003900329559278322, + "traffic_rms": 0.06565717687693617 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7220789178272753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.684044156398938e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009841091216831013, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.22557188702424366 + }, + { + "epoch": 43, + "eval_accuracy": 0.854, + "eval_loss": 0.4854722900390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3975459933280945, + "mean_feedback_forward_cosine": 0.9695534186978494, + "mean_feedback_forward_relative_error": 0.24481760037522163, + "min_feedback_forward_cosine": 0.9236574172973633 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038587651591665, + "instruction_rms": 0.003858765160739692, + "raw_apical_rms": 0.06620083047706758, + "teaching_rms": 0.0038587651591665, + "traffic_rms": 0.06608502767678665 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9172804616450776e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6827225905703166e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009815370464705665, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.22222819827132756 + }, + { + "epoch": 44, + "eval_accuracy": 0.8568, + "eval_loss": 0.477352685546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3839803636074066, + "mean_feedback_forward_cosine": 0.9715552522290137, + "mean_feedback_forward_relative_error": 0.23654738549263246, + "min_feedback_forward_cosine": 0.9287176132202148 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038450055363336328, + "instruction_rms": 0.003845005535372914, + "raw_apical_rms": 0.06648666215531655, + "teaching_rms": 0.0038450055363336328, + "traffic_rms": 0.06637190449648765 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.916029551158772e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6823816055350486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009656137922083871, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.22109255310694376 + }, + { + "epoch": 45, + "eval_accuracy": 0.8624, + "eval_loss": 0.4692766296386719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3707510530948639, + "mean_feedback_forward_cosine": 0.9734137385122238, + "mean_feedback_forward_relative_error": 0.2286106754214533, + "min_feedback_forward_cosine": 0.9334725141525269 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038689133874628477, + "instruction_rms": 0.0038689133876962404, + "raw_apical_rms": 0.06663124662534974, + "teaching_rms": 0.0038689133874628477, + "traffic_rms": 0.06651578034124844 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6223659688184853e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6800576791592795e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009691724331843043, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.21285866431660122 + }, + { + "epoch": 46, + "eval_accuracy": 0.858, + "eval_loss": 0.4939498779296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3577093183994293, + "mean_feedback_forward_cosine": 0.975148985462804, + "mean_feedback_forward_relative_error": 0.22097072293681483, + "min_feedback_forward_cosine": 0.9380070567131042 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003831199979431922, + "instruction_rms": 0.003831199978927395, + "raw_apical_rms": 0.06692136029826536, + "teaching_rms": 0.003831199979431922, + "traffic_rms": 0.06680845661552556 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.1969630498451806e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.681096194116952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009524549023077611, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.21487696191999647 + }, + { + "epoch": 47, + "eval_accuracy": 0.872, + "eval_loss": 0.472917919921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3452940285205841, + "mean_feedback_forward_cosine": 0.9767827987670898, + "mean_feedback_forward_relative_error": 0.21350583530241443, + "min_feedback_forward_cosine": 0.9421908855438232 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038345523695841512, + "instruction_rms": 0.0038345523689545485, + "raw_apical_rms": 0.0670646653799901, + "teaching_rms": 0.0038345523695841512, + "traffic_rms": 0.06695165768518628 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.331855324972821e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6728564451423404e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009493565128231124, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.21163174434767829 + }, + { + "epoch": 48, + "eval_accuracy": 0.8688, + "eval_loss": 0.44108818359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33314257860183716, + "mean_feedback_forward_cosine": 0.9782731494595928, + "mean_feedback_forward_relative_error": 0.20647436476522876, + "min_feedback_forward_cosine": 0.9461215734481812 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003764190321564789, + "instruction_rms": 0.0037641903210092367, + "raw_apical_rms": 0.0670691289351605, + "teaching_rms": 0.003764190321564789, + "traffic_rms": 0.06696056626715648 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.51406370668883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.674099971128485e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009744211307131654, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.20839293567339578 + }, + { + "epoch": 49, + "eval_accuracy": 0.8862, + "eval_loss": 0.37858214111328126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32079362869262695, + "mean_feedback_forward_cosine": 0.9797187486002522, + "mean_feedback_forward_relative_error": 0.19946648492928473, + "min_feedback_forward_cosine": 0.9499779939651489 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003796721602114192, + "instruction_rms": 0.0037967216032537697, + "raw_apical_rms": 0.06727567276809461, + "teaching_rms": 0.003796721602114192, + "traffic_rms": 0.06716515229479343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.709113454737235e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6801163922097045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009502915382779762, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.20797124909559886 + }, + { + "epoch": 50, + "eval_accuracy": 0.8714, + "eval_loss": 0.4533976684570313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3097163140773773, + "mean_feedback_forward_cosine": 0.9810209947247659, + "mean_feedback_forward_relative_error": 0.19289126487508898, + "min_feedback_forward_cosine": 0.9533424377441406 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003790985533641146, + "instruction_rms": 0.0037909855329868116, + "raw_apical_rms": 0.06751110619547468, + "teaching_rms": 0.003790985533641146, + "traffic_rms": 0.06740139136510716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.339059600511064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6697007933493875e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009328096615369125, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.20432197641796537 + }, + { + "epoch": 51, + "eval_accuracy": 0.8674, + "eval_loss": 0.442272802734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29837894439697266, + "mean_feedback_forward_cosine": 0.9822761243389498, + "mean_feedback_forward_relative_error": 0.18638599760109378, + "min_feedback_forward_cosine": 0.9566454887390137 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003752230948103122, + "instruction_rms": 0.003752230948219423, + "raw_apical_rms": 0.06714921863417124, + "teaching_rms": 0.003752230948103122, + "traffic_rms": 0.06704126659320363 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1463550815024064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6738598084129747e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009195574078491352, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.20099864941967857 + }, + { + "epoch": 52, + "eval_accuracy": 0.8576, + "eval_loss": 0.49777291259765627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28850314021110535, + "mean_feedback_forward_cosine": 0.9834090625086138, + "mean_feedback_forward_relative_error": 0.18029032479370793, + "min_feedback_forward_cosine": 0.9594382047653198 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037215057599723268, + "instruction_rms": 0.003721505758804505, + "raw_apical_rms": 0.06721710055288083, + "teaching_rms": 0.0037215057599723268, + "traffic_rms": 0.06711074629889201 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9702910353535117e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.673041938791294e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009731579987993283, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.19583829987843832 + }, + { + "epoch": 53, + "eval_accuracy": 0.8728, + "eval_loss": 0.4199692565917969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27802181243896484, + "mean_feedback_forward_cosine": 0.9845092200463818, + "mean_feedback_forward_relative_error": 0.17417119635689643, + "min_feedback_forward_cosine": 0.962307333946228 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037648703348918803, + "instruction_rms": 0.0037648703360029915, + "raw_apical_rms": 0.06741783690888394, + "teaching_rms": 0.0037648703348918803, + "traffic_rms": 0.06730939466965924 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1614129031586344e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.674570193260333e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009629303913869847, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.19910230238967472 + }, + { + "epoch": 54, + "eval_accuracy": 0.8592, + "eval_loss": 0.48668612670898437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26857489347457886, + "mean_feedback_forward_cosine": 0.9855129680325908, + "mean_feedback_forward_relative_error": 0.16836787543950543, + "min_feedback_forward_cosine": 0.9647953510284424 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037347801412208263, + "instruction_rms": 0.0037347801417342602, + "raw_apical_rms": 0.06763126728019535, + "teaching_rms": 0.0037347801412208263, + "traffic_rms": 0.06752490857710151 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.12550163528158e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6735748017196808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009732354467041557, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.19133018584781222 + }, + { + "epoch": 55, + "eval_accuracy": 0.879, + "eval_loss": 0.4195354675292969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2584693133831024, + "mean_feedback_forward_cosine": 0.9865040856022989, + "mean_feedback_forward_relative_error": 0.1624837938335634, + "min_feedback_forward_cosine": 0.967373251914978 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003791996760680481, + "instruction_rms": 0.003791996760556632, + "raw_apical_rms": 0.06805450134213765, + "teaching_rms": 0.003791996760680481, + "traffic_rms": 0.06794524829129649 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.24055697831227e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.678159205079645e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009456198256669738, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.2014359640068478 + }, + { + "epoch": 56, + "eval_accuracy": 0.8786, + "eval_loss": 0.3782403259277344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24904784560203552, + "mean_feedback_forward_cosine": 0.9874036119830224, + "mean_feedback_forward_relative_error": 0.1569397098595096, + "min_feedback_forward_cosine": 0.9696934819221497 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037843500449065964, + "instruction_rms": 0.0037843500451697504, + "raw_apical_rms": 0.06776119506531587, + "teaching_rms": 0.0037843500449065964, + "traffic_rms": 0.06765212949389462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.656533920093352e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.676397019751039e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009796610734344686, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.19607376774284574 + }, + { + "epoch": 57, + "eval_accuracy": 0.8538, + "eval_loss": 0.5347755859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24004074931144714, + "mean_feedback_forward_cosine": 0.9882267367455267, + "mean_feedback_forward_relative_error": 0.1517001005911058, + "min_feedback_forward_cosine": 0.9718272089958191 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037077486941967386, + "instruction_rms": 0.0037077486959924476, + "raw_apical_rms": 0.06816540951085935, + "teaching_rms": 0.0037077486941967386, + "traffic_rms": 0.0680611902470229 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 2.8713960986692655e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6686866317146235e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009484137712073895, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.1932436572737164 + }, + { + "epoch": 58, + "eval_accuracy": 0.867, + "eval_loss": 0.4802153991699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.231326624751091, + "mean_feedback_forward_cosine": 0.9890082401614035, + "mean_feedback_forward_relative_error": 0.1465402961738648, + "min_feedback_forward_cosine": 0.9738050699234009 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003742871989211473, + "instruction_rms": 0.0037428719918443055, + "raw_apical_rms": 0.06845554943573484, + "teaching_rms": 0.003742871989211473, + "traffic_rms": 0.06835025965135379 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.002559623155321e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6692705455518372e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009609183502983362, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.1894873697307375 + }, + { + "epoch": 59, + "eval_accuracy": 0.869, + "eval_loss": 0.46880623779296876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22357633709907532, + "mean_feedback_forward_cosine": 0.9897216020091888, + "mean_feedback_forward_relative_error": 0.1416760240831683, + "min_feedback_forward_cosine": 0.9755195379257202 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037117681191604895, + "instruction_rms": 0.0037117681183885614, + "raw_apical_rms": 0.06885039561989607, + "teaching_rms": 0.0037117681191604895, + "traffic_rms": 0.0687473068281067 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.011009042097612e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6794532005972392e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009946817394249772, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.18652594333754646 + }, + { + "epoch": 60, + "eval_accuracy": 0.8708, + "eval_loss": 0.45631016845703126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2157677412033081, + "mean_feedback_forward_cosine": 0.9903988684377363, + "mean_feedback_forward_relative_error": 0.13691440188596327, + "min_feedback_forward_cosine": 0.9771890044212341 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037127934702630514, + "instruction_rms": 0.00371279346821834, + "raw_apical_rms": 0.0686732189890834, + "teaching_rms": 0.0037127934702630514, + "traffic_rms": 0.06856912437735939 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 2.8024947695826086e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6693847376850912e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009393395066284187, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.18327303275797102 + }, + { + "epoch": 61, + "eval_accuracy": 0.8718, + "eval_loss": 0.4464225769042969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20803159475326538, + "mean_feedback_forward_cosine": 0.991038772367662, + "mean_feedback_forward_relative_error": 0.13224729607182165, + "min_feedback_forward_cosine": 0.9787874221801758 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036716750526718453, + "instruction_rms": 0.0036716750520768868, + "raw_apical_rms": 0.0687325340227867, + "teaching_rms": 0.0036716750526718453, + "traffic_rms": 0.068631049144822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.627700451147575e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.679051637414625e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009369368561217686, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.18633612112204234 + }, + { + "epoch": 62, + "eval_accuracy": 0.876, + "eval_loss": 0.4284907897949219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20074361562728882, + "mean_feedback_forward_cosine": 0.9916233131962437, + "mean_feedback_forward_relative_error": 0.1278414705828313, + "min_feedback_forward_cosine": 0.9802231192588806 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036913750583233446, + "instruction_rms": 0.0036913750564064947, + "raw_apical_rms": 0.06897525226004436, + "teaching_rms": 0.0036913750583233446, + "traffic_rms": 0.0688732098451882 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.788874991528246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6674829261820502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009119546716608301, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.1868861578941345 + }, + { + "epoch": 63, + "eval_accuracy": 0.8764, + "eval_loss": 0.4421176025390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19378747045993805, + "mean_feedback_forward_cosine": 0.9921654770451207, + "mean_feedback_forward_relative_error": 0.12360262750617919, + "min_feedback_forward_cosine": 0.9815561771392822 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003675538245747924, + "instruction_rms": 0.0036755382454474795, + "raw_apical_rms": 0.06919303102746925, + "teaching_rms": 0.003675538245747924, + "traffic_rms": 0.06909157947821462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5002013731855186e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6631604351137743e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009591347883657127, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.18030287986596424 + }, + { + "epoch": 64, + "eval_accuracy": 0.8606, + "eval_loss": 0.4947023864746094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18710586428642273, + "mean_feedback_forward_cosine": 0.9926713493562513, + "mean_feedback_forward_relative_error": 0.11952662311734692, + "min_feedback_forward_cosine": 0.9827964305877686 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036685665645455005, + "instruction_rms": 0.003668566564958399, + "raw_apical_rms": 0.06899592224688407, + "teaching_rms": 0.0036685665645455005, + "traffic_rms": 0.06889483733996477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.01061895413568e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6715958232016822e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009469844400082687, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.179588240453932 + }, + { + "epoch": 65, + "eval_accuracy": 0.8856, + "eval_loss": 0.38915611572265624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18032169342041016, + "mean_feedback_forward_cosine": 0.9931720533678609, + "mean_feedback_forward_relative_error": 0.11536907224405196, + "min_feedback_forward_cosine": 0.9840102791786194 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003749223339813597, + "instruction_rms": 0.003749223340732459, + "raw_apical_rms": 0.06937663825925827, + "teaching_rms": 0.003749223339813597, + "traffic_rms": 0.069272070622568 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.875665655073135e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.670059678456853e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009408578479519461, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.1879544606897566 + }, + { + "epoch": 66, + "eval_accuracy": 0.877, + "eval_loss": 0.414552197265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17396798729896545, + "mean_feedback_forward_cosine": 0.9936253255413424, + "mean_feedback_forward_relative_error": 0.11146722529684344, + "min_feedback_forward_cosine": 0.9851115345954895 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003655129286387944, + "instruction_rms": 0.0036551292857377946, + "raw_apical_rms": 0.06990854412409622, + "teaching_rms": 0.003655129286387944, + "traffic_rms": 0.0698102541672528 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.541091331271673e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.674265402304113e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00986880363510469, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.17746702675554488 + }, + { + "epoch": 67, + "eval_accuracy": 0.8752, + "eval_loss": 0.45304346923828126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1677214503288269, + "mean_feedback_forward_cosine": 0.9940506239091197, + "mean_feedback_forward_relative_error": 0.10766576723225656, + "min_feedback_forward_cosine": 0.9861553311347961 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00364378350989667, + "instruction_rms": 0.003643783510117977, + "raw_apical_rms": 0.06982380075625995, + "teaching_rms": 0.00364378350989667, + "traffic_rms": 0.0697253093055836 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6275363157756146e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6698265352807232e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009722678805803367, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.17719422046078576 + }, + { + "epoch": 68, + "eval_accuracy": 0.8842, + "eval_loss": 0.3809119873046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16186578571796417, + "mean_feedback_forward_cosine": 0.9944415861560453, + "mean_feedback_forward_relative_error": 0.10406146186494059, + "min_feedback_forward_cosine": 0.9870941042900085 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036261218173337774, + "instruction_rms": 0.0036261218168315516, + "raw_apical_rms": 0.06978934825449166, + "teaching_rms": 0.0036261218173337774, + "traffic_rms": 0.069691834882275 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.079424715224377e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6676291423852776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01003843050606504, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.171932643699646 + }, + { + "epoch": 69, + "eval_accuracy": 0.8854, + "eval_loss": 0.3779376342773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1558031588792801, + "mean_feedback_forward_cosine": 0.9948173030730216, + "mean_feedback_forward_relative_error": 0.10046967719831774, + "min_feedback_forward_cosine": 0.9880361557006836 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003679628838903734, + "instruction_rms": 0.0036796288372781265, + "raw_apical_rms": 0.0700862147160062, + "teaching_rms": 0.003679628838903734, + "traffic_rms": 0.06998683022391407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7146118242171724e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6697609063299328e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009998567842615923, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.17809613135523267 + }, + { + "epoch": 70, + "eval_accuracy": 0.8754, + "eval_loss": 0.42756370849609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15017694234848022, + "mean_feedback_forward_cosine": 0.9951669727602312, + "mean_feedback_forward_relative_error": 0.09701778066735114, + "min_feedback_forward_cosine": 0.9888768792152405 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036259993143688937, + "instruction_rms": 0.003625999314513519, + "raw_apical_rms": 0.07033731329615805, + "teaching_rms": 0.0036259993143688937, + "traffic_rms": 0.07024063978982102 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0599758222015225e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6606407015584944e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009573684923473932, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.17603131600353453 + }, + { + "epoch": 71, + "eval_accuracy": 0.8722, + "eval_loss": 0.46291195068359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14495930075645447, + "mean_feedback_forward_cosine": 0.9954846212940831, + "mean_feedback_forward_relative_error": 0.09377915147812135, + "min_feedback_forward_cosine": 0.9896330833435059 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003627911470636652, + "instruction_rms": 0.003627911471955973, + "raw_apical_rms": 0.07046372460091366, + "teaching_rms": 0.003627911470636652, + "traffic_rms": 0.07036733456818349 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.187840391660757e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.655666551295883e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009569554454389147, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.17087456733385722 + }, + { + "epoch": 72, + "eval_accuracy": 0.8742, + "eval_loss": 0.41198401489257813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13980896770954132, + "mean_feedback_forward_cosine": 0.9957855786046674, + "mean_feedback_forward_relative_error": 0.09058924752377695, + "min_feedback_forward_cosine": 0.9903514385223389 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036313193832344533, + "instruction_rms": 0.003631319381984086, + "raw_apical_rms": 0.07061298827791872, + "teaching_rms": 0.0036313193832344533, + "traffic_rms": 0.0705164096689381 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.685270399955698e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6516195090030695e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0093422294375728, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.16949140020608902 + }, + { + "epoch": 73, + "eval_accuracy": 0.8708, + "eval_loss": 0.4712065979003906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13476146757602692, + "mean_feedback_forward_cosine": 0.9960707118434291, + "mean_feedback_forward_relative_error": 0.08746329466662099, + "min_feedback_forward_cosine": 0.9910346269607544 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00362695184190896, + "instruction_rms": 0.0036269518412897135, + "raw_apical_rms": 0.07047801939763346, + "teaching_rms": 0.00362695184190896, + "traffic_rms": 0.07038163708152971 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.089782817435662e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.660622801091213e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00924721079102012, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.1752525559955173 + }, + { + "epoch": 74, + "eval_accuracy": 0.8716, + "eval_loss": 0.4177692199707031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12982839345932007, + "mean_feedback_forward_cosine": 0.9963331895489846, + "mean_feedback_forward_relative_error": 0.08447716481262638, + "min_feedback_forward_cosine": 0.9916701316833496 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003608410703062105, + "instruction_rms": 0.003608410702433176, + "raw_apical_rms": 0.07073821852932612, + "teaching_rms": 0.003608410703062105, + "traffic_rms": 0.07064285059448229 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.683637572748921e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.666976742797306e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009209060033436514, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.17139004131423102 + }, + { + "epoch": 75, + "eval_accuracy": 0.8912, + "eval_loss": 0.37933065185546877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12515747547149658, + "mean_feedback_forward_cosine": 0.9965746345058564, + "mean_feedback_forward_relative_error": 0.08164186985983003, + "min_feedback_forward_cosine": 0.9922549724578857 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036249178952964113, + "instruction_rms": 0.0036249178952551097, + "raw_apical_rms": 0.07086841209431215, + "teaching_rms": 0.0036249178952964113, + "traffic_rms": 0.07077238220220579 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.203561587701188e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.666084813064403e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009161638659102488, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.16958862608273825 + }, + { + "epoch": 76, + "eval_accuracy": 0.8812, + "eval_loss": 0.42446873779296873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12056756764650345, + "mean_feedback_forward_cosine": 0.9968070810840975, + "mean_feedback_forward_relative_error": 0.07882179145611101, + "min_feedback_forward_cosine": 0.9928070306777954 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036848084598481933, + "instruction_rms": 0.0036848084591509693, + "raw_apical_rms": 0.07092062240412486, + "teaching_rms": 0.0036848084598481933, + "traffic_rms": 0.07082153183190511 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5348755034192436e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6594730304805588e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009131354313175306, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.1727997622595893 + }, + { + "epoch": 77, + "eval_accuracy": 0.8844, + "eval_loss": 0.3910231018066406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11632562428712845, + "mean_feedback_forward_cosine": 0.997019596638218, + "mean_feedback_forward_relative_error": 0.07614792000141836, + "min_feedback_forward_cosine": 0.9933022260665894 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003596840719711829, + "instruction_rms": 0.0035968407196002488, + "raw_apical_rms": 0.07113584166134894, + "teaching_rms": 0.003596840719711829, + "traffic_rms": 0.0710414819832535 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6839654882214745e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.662309649777286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00984470253604775, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.16724212563302782 + }, + { + "epoch": 78, + "eval_accuracy": 0.882, + "eval_loss": 0.3831909851074219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11226841062307358, + "mean_feedback_forward_cosine": 0.9972206527186979, + "mean_feedback_forward_relative_error": 0.07352818993310775, + "min_feedback_forward_cosine": 0.9937601685523987 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036420830789038183, + "instruction_rms": 0.0036420830789345047, + "raw_apical_rms": 0.07121224261868106, + "teaching_rms": 0.0036420830789038183, + "traffic_rms": 0.07111619993692685 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.624504429922126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.66994563362321e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009543623209031843, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.17047343085077074 + }, + { + "epoch": 79, + "eval_accuracy": 0.8678, + "eval_loss": 0.44401912841796876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10821803659200668, + "mean_feedback_forward_cosine": 0.9974058635773198, + "mean_feedback_forward_relative_error": 0.07102661531779074, + "min_feedback_forward_cosine": 0.9942016005516052 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036054255838827444, + "instruction_rms": 0.0036054255846796443, + "raw_apical_rms": 0.07135991644014633, + "teaching_rms": 0.0036054255838827444, + "traffic_rms": 0.07126583954021509 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.244164403781724e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6555204898444926e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009561302727504852, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.1651164001862208 + }, + { + "epoch": 80, + "eval_accuracy": 0.8776, + "eval_loss": 0.4436854675292969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10441537946462631, + "mean_feedback_forward_cosine": 0.9975783498056473, + "mean_feedback_forward_relative_error": 0.06861883546075513, + "min_feedback_forward_cosine": 0.9946001172065735 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035964310372020012, + "instruction_rms": 0.003596431037086029, + "raw_apical_rms": 0.07132739490731534, + "teaching_rms": 0.0035964310372020012, + "traffic_rms": 0.07123316787910322 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.702899320501274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6588788726412646e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009810967358417407, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.16397848246362473 + }, + { + "epoch": 81, + "eval_accuracy": 0.8776, + "eval_loss": 0.4233134704589844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10063564032316208, + "mean_feedback_forward_cosine": 0.9977411646996776, + "mean_feedback_forward_relative_error": 0.06626483101037241, + "min_feedback_forward_cosine": 0.9949819445610046 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003571335612018009, + "instruction_rms": 0.0035713356125407895, + "raw_apical_rms": 0.07163790382306386, + "teaching_rms": 0.003571335612018009, + "traffic_rms": 0.07154558551446401 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6408196518445948e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.656869259830858e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009733629286904164, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.1652719701078203 + }, + { + "epoch": 82, + "eval_accuracy": 0.8726, + "eval_loss": 0.44266607055664065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09689801186323166, + "mean_feedback_forward_cosine": 0.9979009993614689, + "mean_feedback_forward_relative_error": 0.06387322474150889, + "min_feedback_forward_cosine": 0.9953445196151733 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036833393936125253, + "instruction_rms": 0.003683339394709965, + "raw_apical_rms": 0.07160597339924625, + "teaching_rms": 0.0036833393936125253, + "traffic_rms": 0.07150795155961165 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.135566473133622e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6618241543435608e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009807831831993425, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.1679715323501163 + }, + { + "epoch": 83, + "eval_accuracy": 0.8666, + "eval_loss": 0.5030519104003907, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0934315100312233, + "mean_feedback_forward_cosine": 0.9980408107080767, + "mean_feedback_forward_relative_error": 0.06170915283503071, + "min_feedback_forward_cosine": 0.995671272277832 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003554035555607399, + "instruction_rms": 0.0035540355554000607, + "raw_apical_rms": 0.07172303359046345, + "teaching_rms": 0.003554035555607399, + "traffic_rms": 0.07163143160456198 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.2331570309906965e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6646891743525186e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009595793136706894, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.16389296588367885 + }, + { + "epoch": 84, + "eval_accuracy": 0.8506, + "eval_loss": 0.5674269653320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0899653360247612, + "mean_feedback_forward_cosine": 0.998176436270437, + "mean_feedback_forward_relative_error": 0.05953777473299734, + "min_feedback_forward_cosine": 0.9959843754768372 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036174064463937107, + "instruction_rms": 0.0036174064458526912, + "raw_apical_rms": 0.07178923803194671, + "teaching_rms": 0.0036174064463937107, + "traffic_rms": 0.0716947802706199 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.38670602673119e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6597809357246895e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009311851541260572, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.1703937453561359 + }, + { + "epoch": 85, + "eval_accuracy": 0.8916, + "eval_loss": 0.3849796752929687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08688816428184509, + "mean_feedback_forward_cosine": 0.9982977009588673, + "mean_feedback_forward_relative_error": 0.05752050305806821, + "min_feedback_forward_cosine": 0.9962527751922607 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003597530378386697, + "instruction_rms": 0.003597530377097671, + "raw_apical_rms": 0.0720722297302695, + "teaching_rms": 0.003597530378386697, + "traffic_rms": 0.07197911854139254 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.321809754515016e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6607927820744984e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00932571139363934, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.16465434232287937 + }, + { + "epoch": 86, + "eval_accuracy": 0.893, + "eval_loss": 0.37870723266601564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08391399681568146, + "mean_feedback_forward_cosine": 0.9984068870544434, + "mean_feedback_forward_relative_error": 0.05564795470526142, + "min_feedback_forward_cosine": 0.9965047240257263 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035002124601188843, + "instruction_rms": 0.0035002124596508924, + "raw_apical_rms": 0.07226275588302505, + "teaching_rms": 0.0035002124601188843, + "traffic_rms": 0.07217460705782518 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.464167264155549e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.656678765377475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00945063351076668, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.1550357510195838 + }, + { + "epoch": 87, + "eval_accuracy": 0.878, + "eval_loss": 0.42328782958984373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08091253787279129, + "mean_feedback_forward_cosine": 0.998512170007152, + "mean_feedback_forward_relative_error": 0.0537769113697352, + "min_feedback_forward_cosine": 0.9967488050460815 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035730514073078613, + "instruction_rms": 0.0035730514078760292, + "raw_apical_rms": 0.07224889820710166, + "teaching_rms": 0.0035730514073078613, + "traffic_rms": 0.0721564351530741 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.410326954167431e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.668245969136442e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009184684011630855, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.16146544399261475 + }, + { + "epoch": 88, + "eval_accuracy": 0.8832, + "eval_loss": 0.42677001953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07806982100009918, + "mean_feedback_forward_cosine": 0.9986101600431627, + "mean_feedback_forward_relative_error": 0.0519687618699766, + "min_feedback_forward_cosine": 0.9969735741615295 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003487752935243331, + "instruction_rms": 0.003487752934749015, + "raw_apical_rms": 0.07215588748529213, + "teaching_rms": 0.003487752935243331, + "traffic_rms": 0.07206799095668322 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.830646377878111e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.660071052193981e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009321216980913925, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.16241874133480919 + }, + { + "epoch": 89, + "eval_accuracy": 0.872, + "eval_loss": 0.46930485229492186, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07518605142831802, + "mean_feedback_forward_cosine": 0.9987055313202643, + "mean_feedback_forward_relative_error": 0.0501535901980054, + "min_feedback_forward_cosine": 0.9971925616264343 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035720601461635072, + "instruction_rms": 0.0035720601469758596, + "raw_apical_rms": 0.07244703868875312, + "teaching_rms": 0.0035720601461635072, + "traffic_rms": 0.07235575863016837 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.132954600777339e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6593580799944038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009185582520955396, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.163161230580012 + }, + { + "epoch": 90, + "eval_accuracy": 0.8746, + "eval_loss": 0.4608991455078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07252176105976105, + "mean_feedback_forward_cosine": 0.9987933501120536, + "mean_feedback_forward_relative_error": 0.04842322115455904, + "min_feedback_forward_cosine": 0.9973869919776917 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035652092843790066, + "instruction_rms": 0.003565209284478349, + "raw_apical_rms": 0.07258737420948207, + "teaching_rms": 0.0035652092843790066, + "traffic_rms": 0.07249694187569879 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.627100042535858e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.66188932376684e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009473667757679357, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.1598022187034289 + }, + { + "epoch": 91, + "eval_accuracy": 0.8812, + "eval_loss": 0.41463336181640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06982004642486572, + "mean_feedback_forward_cosine": 0.9988757660312038, + "mean_feedback_forward_relative_error": 0.04674123125451227, + "min_feedback_forward_cosine": 0.9975777268409729 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035763155716032986, + "instruction_rms": 0.0035763155708816653, + "raw_apical_rms": 0.07281488449730049, + "teaching_rms": 0.0035763155716032986, + "traffic_rms": 0.072723704848616 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.574161056008052e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6492086060836144e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009355969896967818, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.16621972343126934 + }, + { + "epoch": 92, + "eval_accuracy": 0.88, + "eval_loss": 0.45427987060546876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06721863150596619, + "mean_feedback_forward_cosine": 0.9989541172981262, + "mean_feedback_forward_relative_error": 0.04508476139556977, + "min_feedback_forward_cosine": 0.9977545142173767 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036187418742760565, + "instruction_rms": 0.003618741871613057, + "raw_apical_rms": 0.07319300477937919, + "teaching_rms": 0.0036187418742760565, + "traffic_rms": 0.07309999209704568 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.3080969902575816e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6510856252466656e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009249153718346741, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.16049772697289785 + }, + { + "epoch": 93, + "eval_accuracy": 0.8842, + "eval_loss": 0.40804675903320314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06487451493740082, + "mean_feedback_forward_cosine": 0.9990244584698831, + "mean_feedback_forward_relative_error": 0.04354102500984745, + "min_feedback_forward_cosine": 0.9979076981544495 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003525129912335197, + "instruction_rms": 0.0035251299124563706, + "raw_apical_rms": 0.07322521898968991, + "teaching_rms": 0.003525129912335197, + "traffic_rms": 0.07313673749452419 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6598152064470924e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.650618967878567e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008946322908574451, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.15889199346436395 + }, + { + "epoch": 94, + "eval_accuracy": 0.8774, + "eval_loss": 0.4351710021972656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0624665692448616, + "mean_feedback_forward_cosine": 0.9990899043698465, + "mean_feedback_forward_relative_error": 0.04205331461684358, + "min_feedback_forward_cosine": 0.9980596899986267 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035437455909674825, + "instruction_rms": 0.003543745589771452, + "raw_apical_rms": 0.07332060943510744, + "teaching_rms": 0.0035437455909674825, + "traffic_rms": 0.07323156261518245 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.717973839911792e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6439794761812034e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008958942905872078, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.16275668821732203 + }, + { + "epoch": 95, + "eval_accuracy": 0.8834, + "eval_loss": 0.40972940673828123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.060325946658849716, + "mean_feedback_forward_cosine": 0.9991485514948445, + "mean_feedback_forward_relative_error": 0.04066988294042887, + "min_feedback_forward_cosine": 0.9981906414031982 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034620313225154366, + "instruction_rms": 0.003462031322280486, + "raw_apical_rms": 0.07341350454202532, + "teaching_rms": 0.0034620313225154366, + "traffic_rms": 0.07332870689643071 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5315704138838555e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.665924058018557e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009248571732534813, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.1549742729028066 + }, + { + "epoch": 96, + "eval_accuracy": 0.8874, + "eval_loss": 0.3855793701171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.058349937200546265, + "mean_feedback_forward_cosine": 0.9992040837964704, + "mean_feedback_forward_relative_error": 0.03932008099171423, + "min_feedback_forward_cosine": 0.9983067512512207 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034875497038143145, + "instruction_rms": 0.003487549704974714, + "raw_apical_rms": 0.07365003879253579, + "teaching_rms": 0.0034875497038143145, + "traffic_rms": 0.07356420552182218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 4.434439659917189e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6458171200670373e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00937028715119181, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.14878102054595949 + }, + { + "epoch": 97, + "eval_accuracy": 0.8886, + "eval_loss": 0.3953185546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056168243288993835, + "mean_feedback_forward_cosine": 0.9992594199795877, + "mean_feedback_forward_relative_error": 0.0379281306879655, + "min_feedback_forward_cosine": 0.9984309077262878 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035957013318513083, + "instruction_rms": 0.0035957013326367734, + "raw_apical_rms": 0.0733056118526088, + "teaching_rms": 0.0035957013318513083, + "traffic_rms": 0.07321380774767293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9192386225295195e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.650805863676102e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009429207017063875, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.15785606362554763 + }, + { + "epoch": 98, + "eval_accuracy": 0.8842, + "eval_loss": 0.41677572631835935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05430087074637413, + "mean_feedback_forward_cosine": 0.9993068614313679, + "mean_feedback_forward_relative_error": 0.03669139430407555, + "min_feedback_forward_cosine": 0.9985330700874329 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003495759529560324, + "instruction_rms": 0.003495759529936718, + "raw_apical_rms": 0.07341240825660733, + "teaching_rms": 0.003495759529560324, + "traffic_rms": 0.07332585321366156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.8842856042720086e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.650165960022111e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009514835634155968, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.1526362185160319 + }, + { + "epoch": 99, + "eval_accuracy": 0.8944, + "eval_loss": 0.3704293640136719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05233167111873627, + "mean_feedback_forward_cosine": 0.9993537126048919, + "mean_feedback_forward_relative_error": 0.03543054300450509, + "min_feedback_forward_cosine": 0.9986371397972107 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003504387026545859, + "instruction_rms": 0.0035043870275930696, + "raw_apical_rms": 0.07377413682739672, + "teaching_rms": 0.003504387026545859, + "traffic_rms": 0.07368727359965827 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.901835654573915e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6477510326228728e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009302213669263211, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.15529265949726104 + }, + { + "epoch": 100, + "eval_accuracy": 0.882, + "eval_loss": 0.4139551330566406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050475358963012695, + "mean_feedback_forward_cosine": 0.9993980373105695, + "mean_feedback_forward_relative_error": 0.03419552521119194, + "min_feedback_forward_cosine": 0.998731791973114 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003581658465411083, + "instruction_rms": 0.0035816584649890552, + "raw_apical_rms": 0.07379287180973258, + "teaching_rms": 0.003581658465411083, + "traffic_rms": 0.0737022879116428 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6121676316724916e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6539305751499018e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009126019775016257, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.15373161282274458 + }, + { + "epoch": 101, + "eval_accuracy": 0.9166, + "eval_loss": 0.296131787109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05046016722917557, + "mean_feedback_forward_cosine": 0.999398410320282, + "mean_feedback_forward_relative_error": 0.034182172779354356, + "min_feedback_forward_cosine": 0.9987326264381409 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002503702058118298, + "instruction_rms": 0.0025037020573505588, + "raw_apical_rms": 0.07391774718212694, + "teaching_rms": 0.002503702058118298, + "traffic_rms": 0.07387247257723574 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 4.1269547779165805e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.636368132771414e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008899219379347055, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.08968734825385942 + }, + { + "epoch": 102, + "eval_accuracy": 0.9174, + "eval_loss": 0.29645615234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05045269802212715, + "mean_feedback_forward_cosine": 0.9993986218206344, + "mean_feedback_forward_relative_error": 0.034173633602838364, + "min_feedback_forward_cosine": 0.9987329244613647 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0021944999453406113, + "instruction_rms": 0.002194499946388587, + "raw_apical_rms": 0.07379360535339856, + "teaching_rms": 0.0021944999453406113, + "traffic_rms": 0.07375858013327992 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.687942040642156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.640653593188287e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008763187863656765, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.06357012153069179 + }, + { + "epoch": 103, + "eval_accuracy": 0.9176, + "eval_loss": 0.29464952392578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05044633895158768, + "mean_feedback_forward_cosine": 0.9993987890981859, + "mean_feedback_forward_relative_error": 0.034165723489657525, + "min_feedback_forward_cosine": 0.9987332820892334 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002103950848009906, + "instruction_rms": 0.0021039508485826415, + "raw_apical_rms": 0.07372151276783091, + "teaching_rms": 0.002103950848009906, + "traffic_rms": 0.07368904657404154 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5674055709478125e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6338468449839717e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008848141077239939, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.053366583776805136 + }, + { + "epoch": 104, + "eval_accuracy": 0.9194, + "eval_loss": 0.29925628662109377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05044025182723999, + "mean_feedback_forward_cosine": 0.9993989756030421, + "mean_feedback_forward_relative_error": 0.03415835798988419, + "min_feedback_forward_cosine": 0.9987336993217468 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002059360329649582, + "instruction_rms": 0.002059360328937544, + "raw_apical_rms": 0.07365805133806275, + "teaching_rms": 0.002059360329649582, + "traffic_rms": 0.07362615429563846 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.2976984698507295e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6291663740232836e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008897021931968223, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.047539353975984786 + }, + { + "epoch": 105, + "eval_accuracy": 0.9178, + "eval_loss": 0.3023534362792969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05043430253863335, + "mean_feedback_forward_cosine": 0.9993991640306288, + "mean_feedback_forward_relative_error": 0.03415105467842471, + "min_feedback_forward_cosine": 0.9987341165542603 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020590809279698, + "instruction_rms": 0.002059080928048657, + "raw_apical_rms": 0.07356907459457078, + "teaching_rms": 0.0020590809279698, + "traffic_rms": 0.07353715436783345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.57055824906638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6422143636760872e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009101876718829222, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.043767406027846865 + }, + { + "epoch": 106, + "eval_accuracy": 0.9214, + "eval_loss": 0.3021748046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05042882636189461, + "mean_feedback_forward_cosine": 0.9993993466900241, + "mean_feedback_forward_relative_error": 0.03414431998446103, + "min_feedback_forward_cosine": 0.9987342953681946 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001980464299154211, + "instruction_rms": 0.0019804642994078595, + "raw_apical_rms": 0.07350965809847608, + "teaching_rms": 0.001980464299154211, + "traffic_rms": 0.07348000631026358 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.873495302286756e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.637047972949355e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008936199094665764, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.039899225935671065 + }, + { + "epoch": 107, + "eval_accuracy": 0.9212, + "eval_loss": 0.3047294006347656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050423093140125275, + "mean_feedback_forward_cosine": 0.9993994985857317, + "mean_feedback_forward_relative_error": 0.03413757090006144, + "min_feedback_forward_cosine": 0.9987345933914185 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019584899976528005, + "instruction_rms": 0.00195848999787617, + "raw_apical_rms": 0.07348220034259192, + "teaching_rms": 0.0019584899976528005, + "traffic_rms": 0.07345267258990162 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.621288513500076e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6403314786546443e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008975976658212542, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.0374519247604741 + }, + { + "epoch": 108, + "eval_accuracy": 0.9196, + "eval_loss": 0.3181728271484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05041760951280594, + "mean_feedback_forward_cosine": 0.9993996524041698, + "mean_feedback_forward_relative_error": 0.034131177580885345, + "min_feedback_forward_cosine": 0.9987348914146423 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019020930719547322, + "instruction_rms": 0.0019020930718985536, + "raw_apical_rms": 0.07347984602626077, + "teaching_rms": 0.0019020930719547322, + "traffic_rms": 0.0734517919918475 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.617163812918989e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6394616077975364e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009199146910600758, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.0350735355509652 + }, + { + "epoch": 109, + "eval_accuracy": 0.9182, + "eval_loss": 0.3142703857421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05041196942329407, + "mean_feedback_forward_cosine": 0.9993998177589909, + "mean_feedback_forward_relative_error": 0.034124843535884734, + "min_feedback_forward_cosine": 0.9987351298332214 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018793636401583705, + "instruction_rms": 0.0018793636398705615, + "raw_apical_rms": 0.07344885593576457, + "teaching_rms": 0.0018793636401583705, + "traffic_rms": 0.07342106875997677 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9828502334794393e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6351668861964216e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009067827007928238, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.031972534075048235 + }, + { + "epoch": 110, + "eval_accuracy": 0.9198, + "eval_loss": 0.3206904541015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05040634423494339, + "mean_feedback_forward_cosine": 0.9993999773456205, + "mean_feedback_forward_relative_error": 0.034118656760021564, + "min_feedback_forward_cosine": 0.9987354874610901 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018419724681878135, + "instruction_rms": 0.0018419724681582822, + "raw_apical_rms": 0.07341183097499239, + "teaching_rms": 0.0018419724681878135, + "traffic_rms": 0.07338468595792112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.3905721724968316e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.632848519724072e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00905213726479151, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.03032042328119278 + }, + { + "epoch": 111, + "eval_accuracy": 0.9198, + "eval_loss": 0.32301548461914065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050401121377944946, + "mean_feedback_forward_cosine": 0.9994001542368243, + "mean_feedback_forward_relative_error": 0.03411253298362417, + "min_feedback_forward_cosine": 0.9987357258796692 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018661703224029228, + "instruction_rms": 0.0018661703226740712, + "raw_apical_rms": 0.07339671282876449, + "teaching_rms": 0.0018661703224029228, + "traffic_rms": 0.07336889522340968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.526874081671849e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6362773445108326e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009158179741872831, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.029052887268861134 + }, + { + "epoch": 112, + "eval_accuracy": 0.9198, + "eval_loss": 0.3287255310058594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05039596930146217, + "mean_feedback_forward_cosine": 0.9994002945961491, + "mean_feedback_forward_relative_error": 0.034106574922559724, + "min_feedback_forward_cosine": 0.9987360239028931 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018375925809945362, + "instruction_rms": 0.001837592581293839, + "raw_apical_rms": 0.07334994477907034, + "teaching_rms": 0.0018375925809945362, + "traffic_rms": 0.07332291949214792 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5672488962745774e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6330009403873294e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009117030862201607, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.028106970258553823 + }, + { + "epoch": 113, + "eval_accuracy": 0.9196, + "eval_loss": 0.3258778381347656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050390880554914474, + "mean_feedback_forward_cosine": 0.9994004368782043, + "mean_feedback_forward_relative_error": 0.03410087453742181, + "min_feedback_forward_cosine": 0.9987362027168274 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001753882422299409, + "instruction_rms": 0.001753882421137723, + "raw_apical_rms": 0.0733300083731065, + "teaching_rms": 0.001753882422299409, + "traffic_rms": 0.07330515633365575 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.279366467268119e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383620473785716e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009157197080994243, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.025644417638580004 + }, + { + "epoch": 114, + "eval_accuracy": 0.9194, + "eval_loss": 0.3260003479003906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05038650706410408, + "mean_feedback_forward_cosine": 0.9994005695466073, + "mean_feedback_forward_relative_error": 0.034095589251768206, + "min_feedback_forward_cosine": 0.9987364411354065 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016507215790715448, + "instruction_rms": 0.0016507215784339788, + "raw_apical_rms": 0.07326140117464916, + "teaching_rms": 0.0016507215790715448, + "traffic_rms": 0.0732391549592118 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.903891254708469e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.640263418771098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009264603588974272, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.022590324265095923 + }, + { + "epoch": 115, + "eval_accuracy": 0.9228, + "eval_loss": 0.3252786376953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0503816194832325, + "mean_feedback_forward_cosine": 0.9994007272105063, + "mean_feedback_forward_relative_error": 0.03409001198146613, + "min_feedback_forward_cosine": 0.9987367391586304 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001714554079436268, + "instruction_rms": 0.0017145540793167373, + "raw_apical_rms": 0.07326508180322594, + "teaching_rms": 0.001714554079436268, + "traffic_rms": 0.07324081471506289 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0334526829610695e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6416379274780682e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009108295278790897, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.023377065229415895 + }, + { + "epoch": 116, + "eval_accuracy": 0.9218, + "eval_loss": 0.3257878662109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05037691816687584, + "mean_feedback_forward_cosine": 0.9994008925653273, + "mean_feedback_forward_relative_error": 0.03408442137222136, + "min_feedback_forward_cosine": 0.9987369179725647 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001714413491093942, + "instruction_rms": 0.0017144134921156249, + "raw_apical_rms": 0.07323759252924394, + "teaching_rms": 0.001714413491093942, + "traffic_rms": 0.07321339172730634 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.17054195622768e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6411490192345613e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009176613455319031, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.022235628300905228 + }, + { + "epoch": 117, + "eval_accuracy": 0.9212, + "eval_loss": 0.3269914794921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050372641533613205, + "mean_feedback_forward_cosine": 0.9994010060064255, + "mean_feedback_forward_relative_error": 0.034079157176517674, + "min_feedback_forward_cosine": 0.998737096786499 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016238884290508307, + "instruction_rms": 0.0016238884287559203, + "raw_apical_rms": 0.07328271932961998, + "teaching_rms": 0.0016238884290508307, + "traffic_rms": 0.07326048242890883 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.549012106418559e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6376774560338555e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0092378843029572, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.020025251053439247 + }, + { + "epoch": 118, + "eval_accuracy": 0.9224, + "eval_loss": 0.33299217529296876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05036759003996849, + "mean_feedback_forward_cosine": 0.999401169438516, + "mean_feedback_forward_relative_error": 0.03407376680162645, + "min_feedback_forward_cosine": 0.9987374544143677 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016813589612660286, + "instruction_rms": 0.0016813589623253867, + "raw_apical_rms": 0.07328028315887815, + "teaching_rms": 0.0016813589612660286, + "traffic_rms": 0.07325650206488526 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.0073060486301983e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6392372447890742e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009127665566403929, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.021251448687952427 + }, + { + "epoch": 119, + "eval_accuracy": 0.9218, + "eval_loss": 0.33599387817382814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05036266893148422, + "mean_feedback_forward_cosine": 0.9994013251796845, + "mean_feedback_forward_relative_error": 0.034068433087198965, + "min_feedback_forward_cosine": 0.9987376928329468 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016793044916914072, + "instruction_rms": 0.0016793044920357545, + "raw_apical_rms": 0.07320856823210507, + "teaching_rms": 0.0016793044916914072, + "traffic_rms": 0.0731843035091355 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.590515262885674e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6402947751876797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009107391532663028, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.0202204686784082 + }, + { + "epoch": 120, + "eval_accuracy": 0.922, + "eval_loss": 0.3391268798828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050358157604932785, + "mean_feedback_forward_cosine": 0.999401459770818, + "mean_feedback_forward_relative_error": 0.03406337510433889, + "min_feedback_forward_cosine": 0.9987379312515259 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015648042851772904, + "instruction_rms": 0.0015648042850207287, + "raw_apical_rms": 0.07318844722301304, + "teaching_rms": 0.0015648042851772904, + "traffic_rms": 0.07316722756940557 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.5761722472216206e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.649314614754579e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00920271237294771, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.018327527635296187 + }, + { + "epoch": 121, + "eval_accuracy": 0.9204, + "eval_loss": 0.34658875732421873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05035347491502762, + "mean_feedback_forward_cosine": 0.9994015885937598, + "mean_feedback_forward_relative_error": 0.03405820316965542, + "min_feedback_forward_cosine": 0.998738169670105 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001574837191435334, + "instruction_rms": 0.0015748371918382507, + "raw_apical_rms": 0.07316225684028811, + "teaching_rms": 0.001574837191435334, + "traffic_rms": 0.07314047159639461 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.4941205484528837e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.645228259004105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009332900625260264, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.01816672069641451 + }, + { + "epoch": 122, + "eval_accuracy": 0.9202, + "eval_loss": 0.34789984130859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05034828931093216, + "mean_feedback_forward_cosine": 0.9994017520258504, + "mean_feedback_forward_relative_error": 0.03405260583085398, + "min_feedback_forward_cosine": 0.9987384676933289 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001662376227696435, + "instruction_rms": 0.0016623762277789668, + "raw_apical_rms": 0.07318676003390975, + "teaching_rms": 0.001662376227696435, + "traffic_rms": 0.0731627205265084 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.041743650873286e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.635776606278276e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009076971193582126, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.019140352391192897 + }, + { + "epoch": 123, + "eval_accuracy": 0.922, + "eval_loss": 0.35231456298828123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05034376308321953, + "mean_feedback_forward_cosine": 0.9994018885397142, + "mean_feedback_forward_relative_error": 0.03404748508886945, + "min_feedback_forward_cosine": 0.998738706111908 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015854376293903563, + "instruction_rms": 0.001585437629724215, + "raw_apical_rms": 0.0731196388448265, + "teaching_rms": 0.0015854376293903563, + "traffic_rms": 0.0730979823387364 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.055430918332604e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6426968928782518e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009258208004321764, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.017558129587107233 + }, + { + "epoch": 124, + "eval_accuracy": 0.9232, + "eval_loss": 0.34616173706054687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05033883452415466, + "mean_feedback_forward_cosine": 0.9994020442808828, + "mean_feedback_forward_relative_error": 0.03404202627679033, + "min_feedback_forward_cosine": 0.9987388253211975 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015925764422237579, + "instruction_rms": 0.0015925764416803519, + "raw_apical_rms": 0.07313878644961624, + "teaching_rms": 0.0015925764422237579, + "traffic_rms": 0.073115404526152 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.379950269139954e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6441229389360017e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009274810200918145, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.016589551832692492 + }, + { + "epoch": 125, + "eval_accuracy": 0.9216, + "eval_loss": 0.3495522033691406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05033367872238159, + "mean_feedback_forward_cosine": 0.9994022038675123, + "mean_feedback_forward_relative_error": 0.034036547636553165, + "min_feedback_forward_cosine": 0.9987391829490662 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015926960648308965, + "instruction_rms": 0.001592696064703634, + "raw_apical_rms": 0.07310383161859343, + "teaching_rms": 0.0015926960648308965, + "traffic_rms": 0.07308115086094133 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.832180439644617e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642661790056256e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00918368779003918, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.017056103925241366 + }, + { + "epoch": 126, + "eval_accuracy": 0.9216, + "eval_loss": 0.35535875244140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05032917112112045, + "mean_feedback_forward_cosine": 0.99940235576322, + "mean_feedback_forward_relative_error": 0.03403143389450927, + "min_feedback_forward_cosine": 0.9987393021583557 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001532232881593061, + "instruction_rms": 0.0015322328819102713, + "raw_apical_rms": 0.07310194505776765, + "teaching_rms": 0.001532232881593061, + "traffic_rms": 0.07308083408976394 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.162943494951833e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641713835608175e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009221676830231031, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.015211372663577397 + }, + { + "epoch": 127, + "eval_accuracy": 0.9216, + "eval_loss": 0.35379696044921877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05032459646463394, + "mean_feedback_forward_cosine": 0.9994024980452753, + "mean_feedback_forward_relative_error": 0.034026270762326255, + "min_feedback_forward_cosine": 0.9987396001815796 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015258038387716882, + "instruction_rms": 0.0015258038388503378, + "raw_apical_rms": 0.07307827300679864, + "teaching_rms": 0.0015258038387716882, + "traffic_rms": 0.07305702099121773 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.6272851389185234e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6435527168634897e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009130628426376854, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.015641128700640466 + }, + { + "epoch": 128, + "eval_accuracy": 0.9222, + "eval_loss": 0.35718463745117185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05031966418027878, + "mean_feedback_forward_cosine": 0.9994026287909477, + "mean_feedback_forward_relative_error": 0.03402112616646674, + "min_feedback_forward_cosine": 0.9987397193908691 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015096897160306836, + "instruction_rms": 0.0015096897163843694, + "raw_apical_rms": 0.07305015673281999, + "teaching_rms": 0.0015096897160306836, + "traffic_rms": 0.07302892372884263 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.438694745090288e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6452464620442568e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00925408248877178, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.015048105148474376 + }, + { + "epoch": 129, + "eval_accuracy": 0.9222, + "eval_loss": 0.35528975830078124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05031528323888779, + "mean_feedback_forward_cosine": 0.9994027787639249, + "mean_feedback_forward_relative_error": 0.03401619577479939, + "min_feedback_forward_cosine": 0.9987400770187378 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001467657122464643, + "instruction_rms": 0.0014676571223915448, + "raw_apical_rms": 0.07306057313662985, + "teaching_rms": 0.001467657122464643, + "traffic_rms": 0.07304073399270669 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.058058339888703e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6418627325526776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009300725272727057, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.01413145841260751 + }, + { + "epoch": 130, + "eval_accuracy": 0.9212, + "eval_loss": 0.35853994750976564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05031029134988785, + "mean_feedback_forward_cosine": 0.9994029498869373, + "mean_feedback_forward_relative_error": 0.03401098742840752, + "min_feedback_forward_cosine": 0.9987403154373169 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014523157466972074, + "instruction_rms": 0.0014523157468996032, + "raw_apical_rms": 0.0730532038867201, + "teaching_rms": 0.0014523157466972074, + "traffic_rms": 0.07303325276072449 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5567165213024055e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6394125717665666e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009217598344823634, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.014199002430588007 + }, + { + "epoch": 131, + "eval_accuracy": 0.9226, + "eval_loss": 0.35520433349609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05030573904514313, + "mean_feedback_forward_cosine": 0.9994030998599145, + "mean_feedback_forward_relative_error": 0.03400599851363128, + "min_feedback_forward_cosine": 0.9987406134605408 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014316561964262543, + "instruction_rms": 0.0014316561966733975, + "raw_apical_rms": 0.07306319828164792, + "teaching_rms": 0.0014316561964262543, + "traffic_rms": 0.07304411608409099 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9560333675581205e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6430615926998932e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009323446125707572, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.013526438065618276 + }, + { + "epoch": 132, + "eval_accuracy": 0.9236, + "eval_loss": 0.36005784912109373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05030115693807602, + "mean_feedback_forward_cosine": 0.9994032306055869, + "mean_feedback_forward_relative_error": 0.03400100389074895, + "min_feedback_forward_cosine": 0.9987407922744751 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014382695782894992, + "instruction_rms": 0.0014382695778631363, + "raw_apical_rms": 0.07307802445809876, + "teaching_rms": 0.0014382695782894992, + "traffic_rms": 0.07305819674230177 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.118337315617282e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642000280238486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009196017944382502, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.013548072084453371 + }, + { + "epoch": 133, + "eval_accuracy": 0.9214, + "eval_loss": 0.36589906005859374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05029694363474846, + "mean_feedback_forward_cosine": 0.9994033517376069, + "mean_feedback_forward_relative_error": 0.03399634926069167, + "min_feedback_forward_cosine": 0.998741090297699 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013199388175307158, + "instruction_rms": 0.0013199388177408222, + "raw_apical_rms": 0.0730877369466105, + "teaching_rms": 0.0013199388175307158, + "traffic_rms": 0.07307120245075292 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.3117224518264265e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641820543158784e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009270692331866448, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.011600939117082291 + }, + { + "epoch": 134, + "eval_accuracy": 0.922, + "eval_loss": 0.3665016418457031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05029251053929329, + "mean_feedback_forward_cosine": 0.9994034997878536, + "mean_feedback_forward_relative_error": 0.033991458885852366, + "min_feedback_forward_cosine": 0.998741090297699 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001424050863529644, + "instruction_rms": 0.0014240508637996918, + "raw_apical_rms": 0.07304906485194272, + "teaching_rms": 0.001424050863529644, + "traffic_rms": 0.07302919407436295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.679631532804706e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6435836091159347e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009450031870753627, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.01283611283360256 + }, + { + "epoch": 135, + "eval_accuracy": 0.9224, + "eval_loss": 0.3726066833496094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05028746649622917, + "mean_feedback_forward_cosine": 0.9994036516835613, + "mean_feedback_forward_relative_error": 0.033986137368746346, + "min_feedback_forward_cosine": 0.9987415075302124 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001479838315273072, + "instruction_rms": 0.0014798383162909754, + "raw_apical_rms": 0.07298577671049526, + "teaching_rms": 0.001479838315273072, + "traffic_rms": 0.07296484025762337 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.291577499453524e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6416886489302135e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009370776115057027, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.013333194206986162 + }, + { + "epoch": 136, + "eval_accuracy": 0.923, + "eval_loss": 0.3642227478027344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05028301104903221, + "mean_feedback_forward_cosine": 0.9994038228065737, + "mean_feedback_forward_relative_error": 0.033981250418770696, + "min_feedback_forward_cosine": 0.9987417459487915 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013616196736991172, + "instruction_rms": 0.0013616196737312946, + "raw_apical_rms": 0.07297882030325077, + "teaching_rms": 0.0013616196736991172, + "traffic_rms": 0.07296049545588662 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9056946121718283e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642310615525088e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009484388583996937, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.011724352257781559 + }, + { + "epoch": 137, + "eval_accuracy": 0.9236, + "eval_loss": 0.36472017822265623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05027836188673973, + "mean_feedback_forward_cosine": 0.999403957397707, + "mean_feedback_forward_relative_error": 0.03397634877793251, + "min_feedback_forward_cosine": 0.9987421035766602 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001342822403989234, + "instruction_rms": 0.0013428224040910272, + "raw_apical_rms": 0.0729822857742375, + "teaching_rms": 0.001342822403989234, + "traffic_rms": 0.07296335682268686 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.8794461804391176e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6351492112505394e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009328691442835048, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.011808820550143718 + }, + { + "epoch": 138, + "eval_accuracy": 0.921, + "eval_loss": 0.3714885986328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05027312412858009, + "mean_feedback_forward_cosine": 0.9994041208297976, + "mean_feedback_forward_relative_error": 0.03397086391886396, + "min_feedback_forward_cosine": 0.9987423419952393 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014630978232266752, + "instruction_rms": 0.0014630978235427672, + "raw_apical_rms": 0.07301775381295596, + "teaching_rms": 0.0014630978232266752, + "traffic_rms": 0.0729966487923406 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.932416430847297e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.637262266643452e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009256529178537804, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.012508721241851649 + }, + { + "epoch": 139, + "eval_accuracy": 0.9248, + "eval_loss": 0.36586863403320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05026783421635628, + "mean_feedback_forward_cosine": 0.9994043227164976, + "mean_feedback_forward_relative_error": 0.03396528625800725, + "min_feedback_forward_cosine": 0.9987425804138184 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014520432637409404, + "instruction_rms": 0.0014520432634355302, + "raw_apical_rms": 0.0730057830481586, + "teaching_rms": 0.0014520432637409404, + "traffic_rms": 0.07298413660330243 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1338825695902415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.635901133150022e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009179780753142939, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.012142394143508541 + }, + { + "epoch": 140, + "eval_accuracy": 0.9224, + "eval_loss": 0.376943359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05026276409626007, + "mean_feedback_forward_cosine": 0.9994044380803262, + "mean_feedback_forward_relative_error": 0.03396030199984389, + "min_feedback_forward_cosine": 0.9987427592277527 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013495141283864642, + "instruction_rms": 0.001349514128180668, + "raw_apical_rms": 0.07296538552645916, + "teaching_rms": 0.0013495141283864642, + "traffic_rms": 0.07294706516908853 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.5952443628038964e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6394641085446702e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009383143381762667, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.011706718593173557 + }, + { + "epoch": 141, + "eval_accuracy": 0.9246, + "eval_loss": 0.36509554443359377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05025757476687431, + "mean_feedback_forward_cosine": 0.9994046015124167, + "mean_feedback_forward_relative_error": 0.03395504402297159, + "min_feedback_forward_cosine": 0.9987429976463318 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013513653636865344, + "instruction_rms": 0.0013513653627506327, + "raw_apical_rms": 0.07296948617265846, + "teaching_rms": 0.0013513653636865344, + "traffic_rms": 0.07295083631529936 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.399305370521688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6372433844617886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009366212179531961, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.011151374283226 + }, + { + "epoch": 142, + "eval_accuracy": 0.9242, + "eval_loss": 0.37224436645507814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050253115594387054, + "mean_feedback_forward_cosine": 0.999404739949011, + "mean_feedback_forward_relative_error": 0.03395028944097219, + "min_feedback_forward_cosine": 0.9987432360649109 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012526717766469747, + "instruction_rms": 0.001252671775948499, + "raw_apical_rms": 0.07294409291308425, + "teaching_rms": 0.0012526717766469747, + "traffic_rms": 0.07292772282963204 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1604881317880427e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642280184927957e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009348331796996548, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.009966925987145967 + }, + { + "epoch": 143, + "eval_accuracy": 0.9234, + "eval_loss": 0.38368392333984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050248317420482635, + "mean_feedback_forward_cosine": 0.9994048918447187, + "mean_feedback_forward_relative_error": 0.033945409641150504, + "min_feedback_forward_cosine": 0.99874347448349 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001286189656071483, + "instruction_rms": 0.0012861896555978786, + "raw_apical_rms": 0.07294821075933687, + "teaching_rms": 0.001286189656071483, + "traffic_rms": 0.07293130182317659 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.6473304255177936e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6375652749029974e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009365291032438687, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.010365842316134109 + }, + { + "epoch": 144, + "eval_accuracy": 0.9232, + "eval_loss": 0.383241064453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050243910402059555, + "mean_feedback_forward_cosine": 0.9994050168221996, + "mean_feedback_forward_relative_error": 0.03394059752745013, + "min_feedback_forward_cosine": 0.9987436532974243 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001276116556537042, + "instruction_rms": 0.0012761165568696384, + "raw_apical_rms": 0.07292939398541631, + "teaching_rms": 0.001276116556537042, + "traffic_rms": 0.07291261704185605 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.370652379430794e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642175233349516e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0091762770400409, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.00985723092638784 + }, + { + "epoch": 145, + "eval_accuracy": 0.9242, + "eval_loss": 0.37986845703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05023950710892677, + "mean_feedback_forward_cosine": 0.9994051514133331, + "mean_feedback_forward_relative_error": 0.03393592354991744, + "min_feedback_forward_cosine": 0.9987438321113586 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012351150970882408, + "instruction_rms": 0.0012351150978570596, + "raw_apical_rms": 0.07293325650420523, + "teaching_rms": 0.0012351150970882408, + "traffic_rms": 0.07291708632823415 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.577432477503862e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6395470741251502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009204737415403922, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.009665863710972998 + }, + { + "epoch": 146, + "eval_accuracy": 0.923, + "eval_loss": 0.38360340576171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05023485794663429, + "mean_feedback_forward_cosine": 0.9994053302272674, + "mean_feedback_forward_relative_error": 0.03393060140191547, + "min_feedback_forward_cosine": 0.9987441301345825 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012814315542508852, + "instruction_rms": 0.001281431554315269, + "raw_apical_rms": 0.07287779382947894, + "teaching_rms": 0.0012814315542508852, + "traffic_rms": 0.07286031956870602 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.3912446184558576e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6364850692780833e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00932224725261283, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.009358042332612806 + }, + { + "epoch": 147, + "eval_accuracy": 0.9232, + "eval_loss": 0.3824809326171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05023019015789032, + "mean_feedback_forward_cosine": 0.9994054609729398, + "mean_feedback_forward_relative_error": 0.033925831738498904, + "min_feedback_forward_cosine": 0.9987443089485168 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012637335666448634, + "instruction_rms": 0.0012637335668080345, + "raw_apical_rms": 0.07287240860193166, + "teaching_rms": 0.0012637335666448634, + "traffic_rms": 0.07285584999262126 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.68501895056761e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641440568636575e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009306209725214905, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.009624289382663038 + }, + { + "epoch": 148, + "eval_accuracy": 0.9222, + "eval_loss": 0.38150341796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05022507533431053, + "mean_feedback_forward_cosine": 0.9994056224822998, + "mean_feedback_forward_relative_error": 0.03392021559299, + "min_feedback_forward_cosine": 0.998744547367096 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013185545309442087, + "instruction_rms": 0.0013185545306520783, + "raw_apical_rms": 0.07286871263122031, + "teaching_rms": 0.0013185545309442087, + "traffic_rms": 0.07285093943959148 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.156742505756483e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6395440969528884e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009326790193415034, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.01013564128163788 + }, + { + "epoch": 149, + "eval_accuracy": 0.9204, + "eval_loss": 0.38396596069335936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050220828503370285, + "mean_feedback_forward_cosine": 0.9994057378461284, + "mean_feedback_forward_relative_error": 0.03391571517192548, + "min_feedback_forward_cosine": 0.9987448453903198 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011543139337207263, + "instruction_rms": 0.0011543139338205891, + "raw_apical_rms": 0.07282128451272232, + "teaching_rms": 0.0011543139337207263, + "traffic_rms": 0.0728072899533545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0628336250847497e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63922328067474e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00925633488855441, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.008327673875954416 + }, + { + "epoch": 150, + "eval_accuracy": 0.9214, + "eval_loss": 0.38366978759765624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502161830663681, + "mean_feedback_forward_cosine": 0.9994059032009494, + "mean_feedback_forward_relative_error": 0.03391084285272706, + "min_feedback_forward_cosine": 0.9987449645996094 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001262952555487919, + "instruction_rms": 0.001262952555335136, + "raw_apical_rms": 0.07285522659217394, + "teaching_rms": 0.001262952555487919, + "traffic_rms": 0.07283838657013365 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.671799930771158e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6390587803751088e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009380069419239363, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.009336604690717325 + }, + { + "epoch": 151, + "eval_accuracy": 0.922, + "eval_loss": 0.3788788818359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021611601114273, + "mean_feedback_forward_cosine": 0.9994059051236799, + "mean_feedback_forward_relative_error": 0.03391070210284764, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011623118793956482, + "instruction_rms": 0.0011623118799001153, + "raw_apical_rms": 0.07284443897297617, + "teaching_rms": 0.0011623118793956482, + "traffic_rms": 0.0728301011721793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.582945851121622e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6423221738816907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009589571673323487, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.008326395595901542 + }, + { + "epoch": 152, + "eval_accuracy": 0.9234, + "eval_loss": 0.3746349792480469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021607503294945, + "mean_feedback_forward_cosine": 0.9994059262737152, + "mean_feedback_forward_relative_error": 0.03391058439569127, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011402023765820348, + "instruction_rms": 0.0011402023772544708, + "raw_apical_rms": 0.07284256320758146, + "teaching_rms": 0.0011402023765820348, + "traffic_rms": 0.07282858728433825 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9096363479984575e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6379234733280538e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009477752188391394, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.00844092088109917 + }, + { + "epoch": 153, + "eval_accuracy": 0.9216, + "eval_loss": 0.37851375732421877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021604523062706, + "mean_feedback_forward_cosine": 0.9994059224282542, + "mean_feedback_forward_relative_error": 0.03391046849109473, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010433452263418365, + "instruction_rms": 0.0010433452267703878, + "raw_apical_rms": 0.07283875120750984, + "teaching_rms": 0.0010433452263418365, + "traffic_rms": 0.07282653566874077 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.2417408252504174e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6392866256194406e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009271696744672719, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.007334845918748114 + }, + { + "epoch": 154, + "eval_accuracy": 0.9218, + "eval_loss": 0.3769091857910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021600425243378, + "mean_feedback_forward_cosine": 0.999405941655559, + "mean_feedback_forward_relative_error": 0.03391035447918599, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010592078552617693, + "instruction_rms": 0.0010592078555815578, + "raw_apical_rms": 0.0728372357261761, + "teaching_rms": 0.0010592078552617693, + "traffic_rms": 0.0728243414193174 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9540491769685104e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6361725490104595e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009313667526166636, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.007543874210533169 + }, + { + "epoch": 155, + "eval_accuracy": 0.9214, + "eval_loss": 0.37496671142578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021597072482109, + "mean_feedback_forward_cosine": 0.9994059051236799, + "mean_feedback_forward_relative_error": 0.033910241699026476, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010345856088621523, + "instruction_rms": 0.0010345856090650898, + "raw_apical_rms": 0.0728365529366768, + "teaching_rms": 0.0010345856088621523, + "traffic_rms": 0.07282455131023481 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.8732434976273e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638016824703517e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009354873480270223, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.007022323927697209 + }, + { + "epoch": 156, + "eval_accuracy": 0.9212, + "eval_loss": 0.375509814453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021592974662781, + "mean_feedback_forward_cosine": 0.9994059262737152, + "mean_feedback_forward_relative_error": 0.03391013267419992, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001016232065481463, + "instruction_rms": 0.0010162320656357457, + "raw_apical_rms": 0.07282690160517119, + "teaching_rms": 0.001016232065481463, + "traffic_rms": 0.07281513292762037 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1301920106207035e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6382168537115793e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009296229783639766, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.006862791797270377 + }, + { + "epoch": 157, + "eval_accuracy": 0.922, + "eval_loss": 0.3778203308105469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021589994430542, + "mean_feedback_forward_cosine": 0.9994059301191761, + "mean_feedback_forward_relative_error": 0.03391001493700089, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010073372393457304, + "instruction_rms": 0.0010073372392150456, + "raw_apical_rms": 0.07282213680545249, + "teaching_rms": 0.0010073372393457304, + "traffic_rms": 0.07281038702447838 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.664934311586876e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6348023378875587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009261372796807976, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.007121293383919531 + }, + { + "epoch": 158, + "eval_accuracy": 0.9224, + "eval_loss": 0.3786832702636719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215862691402435, + "mean_feedback_forward_cosine": 0.9994059281964456, + "mean_feedback_forward_relative_error": 0.03390990684349691, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009815531392525657, + "instruction_rms": 0.0009815531391464748, + "raw_apical_rms": 0.07282119378546283, + "teaching_rms": 0.0009815531392525657, + "traffic_rms": 0.072810389042785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.7329989766021754e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6373333040449887e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009549073501831705, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0069797852007879154 + }, + { + "epoch": 159, + "eval_accuracy": 0.9238, + "eval_loss": 0.374742138671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215814262628555, + "mean_feedback_forward_cosine": 0.9994059281964456, + "mean_feedback_forward_relative_error": 0.033909794604105335, + "min_feedback_forward_cosine": 0.9987450242042542 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010119271228425945, + "instruction_rms": 0.0010119271229596044, + "raw_apical_rms": 0.07282141540928444, + "teaching_rms": 0.0010119271228425945, + "traffic_rms": 0.07280968992056051 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1533684935757265e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63807837813095e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009365390021928427, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.007130757482639617 + }, + { + "epoch": 160, + "eval_accuracy": 0.9234, + "eval_loss": 0.37453284912109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215788185596466, + "mean_feedback_forward_cosine": 0.999405941655559, + "mean_feedback_forward_relative_error": 0.033909681403348525, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010204216858368256, + "instruction_rms": 0.0010204216861331291, + "raw_apical_rms": 0.07281743260799332, + "teaching_rms": 0.0010204216858368256, + "traffic_rms": 0.0728055896562762 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.4760297751290636e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.636717054432205e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009435143997698714, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.007176053299734162 + }, + { + "epoch": 161, + "eval_accuracy": 0.9244, + "eval_loss": 0.3757555358886719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021575093269348, + "mean_feedback_forward_cosine": 0.9994059397328284, + "mean_feedback_forward_relative_error": 0.03390957120685808, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009790675754742858, + "instruction_rms": 0.0009790675752193514, + "raw_apical_rms": 0.07281026428724338, + "teaching_rms": 0.0009790675754742858, + "traffic_rms": 0.07279962064713329 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.136117499025204e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639591172495471e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009261054781229678, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.0067126722757187155 + }, + { + "epoch": 162, + "eval_accuracy": 0.9214, + "eval_loss": 0.37511260375976563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021572858095169, + "mean_feedback_forward_cosine": 0.9994059320419065, + "mean_feedback_forward_relative_error": 0.03390947230640919, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008729864156213054, + "instruction_rms": 0.0008729864154180971, + "raw_apical_rms": 0.07280216728271632, + "teaching_rms": 0.0008729864156213054, + "traffic_rms": 0.07279293527425085 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.601536846531417e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6374553336649922e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009376385483806822, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.006105931394402351 + }, + { + "epoch": 163, + "eval_accuracy": 0.9228, + "eval_loss": 0.3784532287597656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502156987786293, + "mean_feedback_forward_cosine": 0.9994059147373322, + "mean_feedback_forward_relative_error": 0.03390936520431311, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001044296922700967, + "instruction_rms": 0.0010442969225778944, + "raw_apical_rms": 0.07280095812172736, + "teaching_rms": 0.001044296922700967, + "traffic_rms": 0.07278821216121345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 3.5597199854464634e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6371352304211653e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009319832459995793, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.007222668740732802 + }, + { + "epoch": 164, + "eval_accuracy": 0.9208, + "eval_loss": 0.3755808349609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021568015217781, + "mean_feedback_forward_cosine": 0.9994059339646371, + "mean_feedback_forward_relative_error": 0.03390925957430755, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009426020207152024, + "instruction_rms": 0.0009426020209865341, + "raw_apical_rms": 0.07279991349257399, + "teaching_rms": 0.0009426020207152024, + "traffic_rms": 0.07278962840524515 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.144392834997234e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639210997898199e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009353989720007325, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.006089698071612252 + }, + { + "epoch": 165, + "eval_accuracy": 0.923, + "eval_loss": 0.37624190063476565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215646624565125, + "mean_feedback_forward_cosine": 0.9994059397328284, + "mean_feedback_forward_relative_error": 0.03390915436489928, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009311302803166897, + "instruction_rms": 0.000931130280125532, + "raw_apical_rms": 0.07280005213689716, + "teaching_rms": 0.0009311302803166897, + "traffic_rms": 0.07278956904321929 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.6460868929898425e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638568239501903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009272727348378633, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.0064649609093864755 + }, + { + "epoch": 166, + "eval_accuracy": 0.922, + "eval_loss": 0.3788002136230469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021560937166214, + "mean_feedback_forward_cosine": 0.9994059339646371, + "mean_feedback_forward_relative_error": 0.03390904711258988, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000951799096096911, + "instruction_rms": 0.0009517990966686603, + "raw_apical_rms": 0.07279990962474571, + "teaching_rms": 0.000951799096096911, + "traffic_rms": 0.07278931803572794 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.383095759090793e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6378296091705208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009335598022536896, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.006698690681656201 + }, + { + "epoch": 167, + "eval_accuracy": 0.9224, + "eval_loss": 0.3780606689453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021558329463005, + "mean_feedback_forward_cosine": 0.9994059608828637, + "mean_feedback_forward_relative_error": 0.033908947491117065, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008896655708649884, + "instruction_rms": 0.0008896655708132983, + "raw_apical_rms": 0.07279797216228476, + "teaching_rms": 0.0008896655708649884, + "traffic_rms": 0.07278825528259195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.671396342497246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6407614709913652e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009282221743622334, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.006275268090681897 + }, + { + "epoch": 168, + "eval_accuracy": 0.9224, + "eval_loss": 0.378640185546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021555349230766, + "mean_feedback_forward_cosine": 0.9994059320419065, + "mean_feedback_forward_relative_error": 0.033908843483415343, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009209827422351977, + "instruction_rms": 0.0009209827423907628, + "raw_apical_rms": 0.0727964476001508, + "teaching_rms": 0.0009209827422351977, + "traffic_rms": 0.07278594311247243 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.299055961747399e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638953556226787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00936715458701123, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.006200297907160388 + }, + { + "epoch": 169, + "eval_accuracy": 0.923, + "eval_loss": 0.37887355346679685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021551996469498, + "mean_feedback_forward_cosine": 0.9994059551146722, + "mean_feedback_forward_relative_error": 0.033908742149510694, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009711995502875797, + "instruction_rms": 0.0009711995498512051, + "raw_apical_rms": 0.07279644517307739, + "teaching_rms": 0.0009711995502875797, + "traffic_rms": 0.07278561888103524 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1259067273813343e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6411877288308163e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009288771153875197, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.0066708320299784345 + }, + { + "epoch": 170, + "eval_accuracy": 0.9224, + "eval_loss": 0.37823710327148435, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021549388766289, + "mean_feedback_forward_cosine": 0.9994059281964456, + "mean_feedback_forward_relative_error": 0.033908638021638314, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009068830681940569, + "instruction_rms": 0.0009068830681429819, + "raw_apical_rms": 0.07279731809941654, + "teaching_rms": 0.0009068830681940569, + "traffic_rms": 0.07278748630276233 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1985181553627626e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6417179504213202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00934180178498991, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.005992806352302432 + }, + { + "epoch": 171, + "eval_accuracy": 0.9222, + "eval_loss": 0.3777138671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502154640853405, + "mean_feedback_forward_cosine": 0.9994059685737856, + "mean_feedback_forward_relative_error": 0.033908533683467294, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009218364303518459, + "instruction_rms": 0.0009218364301352782, + "raw_apical_rms": 0.07279907283003745, + "teaching_rms": 0.0009218364303518459, + "traffic_rms": 0.07278936574241918 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.051080537692542e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63948518168127e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009399914198352297, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.006182804728216595 + }, + { + "epoch": 172, + "eval_accuracy": 0.9234, + "eval_loss": 0.37877182006835936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021543800830841, + "mean_feedback_forward_cosine": 0.9994059781874379, + "mean_feedback_forward_relative_error": 0.033908430216533524, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008757880830134921, + "instruction_rms": 0.0008757880829026351, + "raw_apical_rms": 0.0727909295614263, + "teaching_rms": 0.0008757880830134921, + "traffic_rms": 0.07278188463544438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.394451735834082e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639762290664436e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00939934921304198, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.005911937013144295 + }, + { + "epoch": 173, + "eval_accuracy": 0.923, + "eval_loss": 0.37629825439453124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215404480695724, + "mean_feedback_forward_cosine": 0.9994059551146722, + "mean_feedback_forward_relative_error": 0.033908325397679885, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000973114015998277, + "instruction_rms": 0.0009731140163191162, + "raw_apical_rms": 0.07279270316450596, + "teaching_rms": 0.000973114015998277, + "traffic_rms": 0.0727814818875947 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.15684695553864e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6398265330088304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009303746652720676, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.006478916634453668 + }, + { + "epoch": 174, + "eval_accuracy": 0.9222, + "eval_loss": 0.3773844543457031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215382128953934, + "mean_feedback_forward_cosine": 0.9994059493464809, + "mean_feedback_forward_relative_error": 0.033908218565967777, + "min_feedback_forward_cosine": 0.9987452626228333 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009448674316323791, + "instruction_rms": 0.0009448674318010515, + "raw_apical_rms": 0.07279174859924621, + "teaching_rms": 0.0009448674316323791, + "traffic_rms": 0.07278141079728034 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.272727949228283e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6418756013933296e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009380181118971624, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.006255613459543221 + }, + { + "epoch": 175, + "eval_accuracy": 0.923, + "eval_loss": 0.3780188659667969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021534487605095, + "mean_feedback_forward_cosine": 0.9994059551146722, + "mean_feedback_forward_relative_error": 0.033908120416585474, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008671276321794992, + "instruction_rms": 0.0008671276314825126, + "raw_apical_rms": 0.07279193205743949, + "teaching_rms": 0.0008671276321794992, + "traffic_rms": 0.07278263841080512 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.534815462558072e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6410912552650877e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009327561037759096, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.006052574612283044 + }, + { + "epoch": 176, + "eval_accuracy": 0.9226, + "eval_loss": 0.3774455078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021531879901886, + "mean_feedback_forward_cosine": 0.999405974341977, + "mean_feedback_forward_relative_error": 0.03390802196677654, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009107835894467998, + "instruction_rms": 0.000910783589541975, + "raw_apical_rms": 0.07279263183190296, + "teaching_rms": 0.0009107835894467998, + "traffic_rms": 0.07278268993508986 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.600662523695064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6396375303343103e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009284183196265975, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.006236939753923151 + }, + { + "epoch": 177, + "eval_accuracy": 0.9246, + "eval_loss": 0.3832741027832031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021528899669647, + "mean_feedback_forward_cosine": 0.9994059531919418, + "mean_feedback_forward_relative_error": 0.03390791567583238, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009633222918402966, + "instruction_rms": 0.0009633222919751328, + "raw_apical_rms": 0.07280173134169926, + "teaching_rms": 0.0009633222918402966, + "traffic_rms": 0.07279096102970496 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.57055824906638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638033545376644e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009280104216846495, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.006763021509597699 + }, + { + "epoch": 178, + "eval_accuracy": 0.9244, + "eval_loss": 0.3776351867675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215255469083786, + "mean_feedback_forward_cosine": 0.9994059647283247, + "mean_feedback_forward_relative_error": 0.033907806140280536, + "min_feedback_forward_cosine": 0.998745322227478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009586253475755312, + "instruction_rms": 0.000958625347693359, + "raw_apical_rms": 0.07280250700127605, + "teaching_rms": 0.0009586253475755312, + "traffic_rms": 0.07279176941753737 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.719749841479825e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6398562880900587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00944796904913204, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.006566029266516367 + }, + { + "epoch": 179, + "eval_accuracy": 0.9234, + "eval_loss": 0.37907015380859377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502152293920517, + "mean_feedback_forward_cosine": 0.9994059666510551, + "mean_feedback_forward_relative_error": 0.03390770447590659, + "min_feedback_forward_cosine": 0.9987449645996094 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008790105723924426, + "instruction_rms": 0.0008790105720771522, + "raw_apical_rms": 0.07279674862877619, + "teaching_rms": 0.0008790105723924426, + "traffic_rms": 0.07278790064682795 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.993412806517881e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6405543416716802e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009364554261458803, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.005639267408682241 + }, + { + "epoch": 180, + "eval_accuracy": 0.9234, + "eval_loss": 0.3765585388183594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021520331501961, + "mean_feedback_forward_cosine": 0.9994059724192466, + "mean_feedback_forward_relative_error": 0.03390759593176265, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008834507133467337, + "instruction_rms": 0.0008834507134877638, + "raw_apical_rms": 0.07279706868424211, + "teaching_rms": 0.0008834507133467337, + "traffic_rms": 0.07278753699624642 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1553420260243e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6402966158823036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009301459478879287, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.005865725637806786 + }, + { + "epoch": 181, + "eval_accuracy": 0.9238, + "eval_loss": 0.3767367126464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021516606211662, + "mean_feedback_forward_cosine": 0.9994059397328284, + "mean_feedback_forward_relative_error": 0.03390748928030653, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009059746277008423, + "instruction_rms": 0.0009059746276981573, + "raw_apical_rms": 0.07280177074087833, + "teaching_rms": 0.0009059746277008423, + "traffic_rms": 0.07279164452071894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.174826173652946e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639320196533143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00936177934788029, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.006068159006536007 + }, + { + "epoch": 182, + "eval_accuracy": 0.9246, + "eval_loss": 0.3834036926269531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215139985084534, + "mean_feedback_forward_cosine": 0.9994059647283247, + "mean_feedback_forward_relative_error": 0.03390739034981497, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009031406778541334, + "instruction_rms": 0.0009031406777414056, + "raw_apical_rms": 0.07279577581150613, + "teaching_rms": 0.0009031406778541334, + "traffic_rms": 0.07278685137317638 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.092711591852094e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6351176539766795e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009371290056919238, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.006130151518848207 + }, + { + "epoch": 183, + "eval_accuracy": 0.9226, + "eval_loss": 0.38016424560546874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021510645747185, + "mean_feedback_forward_cosine": 0.9994059724192466, + "mean_feedback_forward_relative_error": 0.033907286882881194, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009198295107836894, + "instruction_rms": 0.0009198295107451514, + "raw_apical_rms": 0.07279343644432036, + "teaching_rms": 0.0009198295107836894, + "traffic_rms": 0.07278314696048392 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.4009541849400193e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383791361941714e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009477513198940484, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.006037287366555797 + }, + { + "epoch": 184, + "eval_accuracy": 0.9222, + "eval_loss": 0.37987239379882815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021507665514946, + "mean_feedback_forward_cosine": 0.9994059724192466, + "mean_feedback_forward_relative_error": 0.03390718251466751, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008858086400521034, + "instruction_rms": 0.0008858086399444548, + "raw_apical_rms": 0.07279044923353264, + "teaching_rms": 0.0008858086400521034, + "traffic_rms": 0.07278121669807068 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.3412252520292895e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6363775237892308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009410989802460348, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.0060372411655883 + }, + { + "epoch": 185, + "eval_accuracy": 0.9238, + "eval_loss": 0.3833714599609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215043127536774, + "mean_feedback_forward_cosine": 0.9994059685737856, + "mean_feedback_forward_relative_error": 0.03390708415498657, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000824911630066183, + "instruction_rms": 0.0008249116295215703, + "raw_apical_rms": 0.07279042653889208, + "teaching_rms": 0.000824911630066183, + "traffic_rms": 0.07278195475173309 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.077775355744052e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6367446567474114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009301936742821348, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.005888968814992243 + }, + { + "epoch": 186, + "eval_accuracy": 0.924, + "eval_loss": 0.3802086730957031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021502077579498, + "mean_feedback_forward_cosine": 0.9994059666510551, + "mean_feedback_forward_relative_error": 0.03390698176958869, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00089744708681856, + "instruction_rms": 0.0008974470863714073, + "raw_apical_rms": 0.07278875310093366, + "teaching_rms": 0.00089744708681856, + "traffic_rms": 0.0727795669934635 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.340492765209092e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6354787432597972e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009585534170728247, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.005864890639566713 + }, + { + "epoch": 187, + "eval_accuracy": 0.9234, + "eval_loss": 0.3811006103515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502149872481823, + "mean_feedback_forward_cosine": 0.999405982032899, + "mean_feedback_forward_relative_error": 0.03390688521246756, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000916919782043793, + "instruction_rms": 0.0009169197822677663, + "raw_apical_rms": 0.0727924594210436, + "teaching_rms": 0.000916919782043793, + "traffic_rms": 0.07278210274929854 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 4.196382263899068e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383436759923482e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009338556265023695, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.006068551631271839 + }, + { + "epoch": 188, + "eval_accuracy": 0.9244, + "eval_loss": 0.38284039916992185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021495372056961, + "mean_feedback_forward_cosine": 0.9994059839556294, + "mean_feedback_forward_relative_error": 0.03390678682274396, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008812204341248592, + "instruction_rms": 0.0008812204343298186, + "raw_apical_rms": 0.07278655818563996, + "teaching_rms": 0.0008812204341248592, + "traffic_rms": 0.07277747691378347 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 4.115873153409666e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383831814105143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009336102800834519, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.00590759240090847 + }, + { + "epoch": 189, + "eval_accuracy": 0.924, + "eval_loss": 0.38746590576171874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021492764353752, + "mean_feedback_forward_cosine": 0.9994059608828637, + "mean_feedback_forward_relative_error": 0.03390668215410363, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000944013221245689, + "instruction_rms": 0.000944013221230113, + "raw_apical_rms": 0.07279046141483732, + "teaching_rms": 0.000944013221245689, + "traffic_rms": 0.07277950835120359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.1209863016765667e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6368435021535036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009340577283093121, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.00627474368541605 + }, + { + "epoch": 190, + "eval_accuracy": 0.9242, + "eval_loss": 0.3782501953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021490156650543, + "mean_feedback_forward_cosine": 0.9994059781874379, + "mean_feedback_forward_relative_error": 0.033906579858833746, + "min_feedback_forward_cosine": 0.998745322227478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009228944355058448, + "instruction_rms": 0.0009228944359649598, + "raw_apical_rms": 0.07278233118548318, + "teaching_rms": 0.0009228944355058448, + "traffic_rms": 0.07277222320074045 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.167023703871564e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6409180082698417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00929254526828421, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.006109715344508489 + }, + { + "epoch": 191, + "eval_accuracy": 0.9236, + "eval_loss": 0.381222705078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214868038892746, + "mean_feedback_forward_cosine": 0.999405974341977, + "mean_feedback_forward_relative_error": 0.033906479366123676, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009125403572763838, + "instruction_rms": 0.0009125403574922392, + "raw_apical_rms": 0.07277732131233654, + "teaching_rms": 0.0009125403572763838, + "traffic_rms": 0.07276739159796018 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.160996169844111e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6368375345992205e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009306256766792505, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.005790863518623842 + }, + { + "epoch": 192, + "eval_accuracy": 0.924, + "eval_loss": 0.37977269897460936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021483451128006, + "mean_feedback_forward_cosine": 0.9994059897238209, + "mean_feedback_forward_relative_error": 0.03390637695068313, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008634372559339853, + "instruction_rms": 0.0008634372557363993, + "raw_apical_rms": 0.07277595362353002, + "teaching_rms": 0.0008634372559339853, + "traffic_rms": 0.07276696431622127 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.573541818013837e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6418940704509305e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009294865945736475, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.0055582207631423245 + }, + { + "epoch": 193, + "eval_accuracy": 0.9232, + "eval_loss": 0.38124073486328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021481215953827, + "mean_feedback_forward_cosine": 0.9994060051056647, + "mean_feedback_forward_relative_error": 0.03390627768972228, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000852372129311672, + "instruction_rms": 0.0008523721291413863, + "raw_apical_rms": 0.07277602368993115, + "teaching_rms": 0.000852372129311672, + "traffic_rms": 0.07276717651199026 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0884806651320105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63812014918529e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009340981492915228, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.005360517854657438 + }, + { + "epoch": 194, + "eval_accuracy": 0.9228, + "eval_loss": 0.3823780700683594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214774906635284, + "mean_feedback_forward_cosine": 0.9994059878010904, + "mean_feedback_forward_relative_error": 0.033906179780681285, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008369135264463215, + "instruction_rms": 0.0008369135264636173, + "raw_apical_rms": 0.07277888262188072, + "teaching_rms": 0.0008369135264463215, + "traffic_rms": 0.07277062523838189 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.558799477332286e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6423136984203254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009384849108658035, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.005631521541501085 + }, + { + "epoch": 195, + "eval_accuracy": 0.9236, + "eval_loss": 0.3792059997558594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214748829603195, + "mean_feedback_forward_cosine": 0.9994059858783599, + "mean_feedback_forward_relative_error": 0.033906079918867155, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008770085033274025, + "instruction_rms": 0.0008770085031596293, + "raw_apical_rms": 0.07277805580198711, + "teaching_rms": 0.0008770085033274025, + "traffic_rms": 0.0727688635778701 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.804241982014901e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6397546781794007e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009314376633056905, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.00563206836965142 + }, + { + "epoch": 196, + "eval_accuracy": 0.9254, + "eval_loss": 0.37827137451171877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214722752571106, + "mean_feedback_forward_cosine": 0.9994059916465513, + "mean_feedback_forward_relative_error": 0.033905980087095694, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000863613228595432, + "instruction_rms": 0.0008636132283742729, + "raw_apical_rms": 0.07277717260645288, + "teaching_rms": 0.000863613228595432, + "traffic_rms": 0.07276802659885094 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.043533863296034e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6419506779648067e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00930628926678014, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.005555464403993553 + }, + { + "epoch": 197, + "eval_accuracy": 0.9222, + "eval_loss": 0.37736356811523436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021468549966812, + "mean_feedback_forward_cosine": 0.9994059878010904, + "mean_feedback_forward_relative_error": 0.033905881607244094, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000900053276765676, + "instruction_rms": 0.0009000532764372408, + "raw_apical_rms": 0.07277405393904711, + "teaching_rms": 0.000900053276765676, + "traffic_rms": 0.07276462906558215 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.7921108858872685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6407543926550647e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009467302561984978, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.0058654394092659155 + }, + { + "epoch": 198, + "eval_accuracy": 0.9244, + "eval_loss": 0.37802588500976564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021467059850693, + "mean_feedback_forward_cosine": 0.9994059839556294, + "mean_feedback_forward_relative_error": 0.033905782015813936, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000811120721353702, + "instruction_rms": 0.0008111207213972554, + "raw_apical_rms": 0.07277230863245111, + "teaching_rms": 0.000811120721353702, + "traffic_rms": 0.07276464107952894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.067615577696415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6391286366009206e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009343913485228653, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.005283751911007696 + }, + { + "epoch": 199, + "eval_accuracy": 0.9224, + "eval_loss": 0.3791741271972656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021464079618454, + "mean_feedback_forward_cosine": 0.9994059839556294, + "mean_feedback_forward_relative_error": 0.033905680531695966, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008688715170549379, + "instruction_rms": 0.0008688715163338572, + "raw_apical_rms": 0.07277670697438565, + "teaching_rms": 0.0008688715170549379, + "traffic_rms": 0.07276745220179037 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.409411419852404e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638858979983995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009275071927212605, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.005599399071849055 + }, + { + "epoch": 200, + "eval_accuracy": 0.9236, + "eval_loss": 0.3822314697265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021461844444275, + "mean_feedback_forward_cosine": 0.9994059801101685, + "mean_feedback_forward_relative_error": 0.03390557637378093, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008638414231774739, + "instruction_rms": 0.0008638414230827563, + "raw_apical_rms": 0.0727772008884368, + "teaching_rms": 0.0008638414231774739, + "traffic_rms": 0.07276843471150797 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 2.943882115857832e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6374803630092046e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00941933233090907, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.005543160841696792 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9236, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.3822314697265625, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2329526272, + "peak_memory_reserved_bytes": 3575644160, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 4.4873448956650464e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 4.4873448956650464e-08, + "min_residual_soma_slope": -4.4004142552012127e-08, + "mse": 5.220318620602023e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.213750884517225e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.213750884517225e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 34.19122052192688, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.6957113742828369, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 4736.380233287811, + "train_wall_s": 4699.746609687805 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.020279373973608017, + 0.01588572934269905, + 0.009585018269717693, + 0.01053142361342907, + 0.006489923689514399, + 0.007580762729048729, + 0.004748271778225899, + 0.0056568942964077, + 0.0037147034890949726, + 0.004711497575044632, + 0.0028121231589466333, + 0.005797920748591423, + 0.004072000738233328, + 0.004937782883644104, + 0.0032058130018413067, + 0.004144369624555111, + 0.002652793889865279, + 0.0034981879871338606, + 0.0022634966298937798, + 0.003057616064324975, + 0.0020107743330299854, + 0.004512992687523365, + 0.003177475184202194, + 0.004072428680956364, + 0.002793447347357869, + 0.0033600262831896544, + 0.0025724933948367834, + 0.0031278016977012157, + 0.0024165462236851454, + 0.0028540485072880983, + 0.0024009225890040398 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.10579779744148254, + 0.08581840991973877, + 0.03320790454745293, + 0.05464063584804535, + 0.017315272241830826, + 0.0413997545838356, + 0.010940272361040115, + 0.029754366725683212, + 0.007736234460026026, + 0.025829754769802094, + 0.005339059978723526, + 0.03059656172990799, + 0.009305471554398537, + 0.026019593700766563, + 0.006473024841398001, + 0.02154967002570629, + 0.004868118092417717, + 0.01855073869228363, + 0.003812844166532159, + 0.01592211239039898, + 0.0030986128840595484, + 0.024262895807623863, + 0.006371741183102131, + 0.021971462294459343, + 0.005072675179690123, + 0.01743730716407299, + 0.004265559371560812, + 0.01702703908085823, + 0.0037584376987069845, + 0.01519029587507248, + 0.0034847366623580456 + ], + "unscaled_traffic_rms": [ + 0.7667219638824463, + 0.7404345870018005, + 1.1545466184616089, + 0.7709590792655945, + 1.499236822128296, + 0.7324451804161072, + 1.7360707521438599, + 0.760479211807251, + 1.9206777811050415, + 0.7296232581138611, + 2.10683012008667, + 0.7579833269119263, + 1.750368356704712, + 0.7590868473052979, + 1.9810292720794678, + 0.7692683339118958, + 2.1797285079956055, + 0.7542961835861206, + 2.3746018409729004, + 0.768143355846405, + 2.5957090854644775, + 0.7440155148506165, + 1.9947296380996704, + 0.7414032816886902, + 2.2027409076690674, + 0.7707672119140625, + 2.4123384952545166, + 0.7347846031188965, + 2.5718624591827393, + 0.7515451908111572, + 2.755929946899414 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 615786394894336, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 615782016000000, + "local_weight_correlation_macs": 619763328000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 619763328000000, + "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": 65471084756992, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 9000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 9000064, + "total_forward_equivalent_examples": 9000064, + "total_macs_estimate": 2471099474100224 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json.manifest.json new file mode 100644 index 0000000..1f19819 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json.manifest.json @@ -0,0 +1,150 @@ +{ + "architecture": "resnet32", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "kp_traffic", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "32", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--traffic_rule", + "innovation", + "--predictor_mode", + "closed_form", + "--neutral_projection", + "1", + "--traffic_seed", + "5000", + "--traffic_ratio", + "4", + "--traffic_calibration_examples", + "64", + "--learn_P", + "1", + "--eta_P", + "0.1", + "--predictor_warmup_steps", + "1", + "--predictor_every", + "0" + ], + "completed_unix_time": 1787028606.225115, + "cuda_visible_devices": "0", + "depth": 32, + "driver_wall_seconds": 4740.1479415893555, + "epochs": 200, + "experiment_name": "resnet-r2-sdil-d32-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-8c47bc34-8284-ba08-eb44-8f381122ed6d", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "sdil", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json", + "output_exists": true, + "output_sha256": "516a62577e7b024b7686e816ebb55bb95f9de07204117af1905ade3b67f14f09", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json new file mode 100644 index 0000000..f742aaa --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json @@ -0,0 +1,7904 @@ +{ + "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": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 1, + "loader_seed": 0, + "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": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.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": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "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": 9000000, + "ordinary_examples": 9000000, + "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": 0.9988019201490614, + "feedback_forward_cosine": [ + 0.9994082450866699, + 0.9993633031845093, + 0.9988594055175781, + 0.9988021850585938, + 0.9989034533500671, + 0.9987430572509766, + 0.9988512992858887, + 0.9986526966094971, + 0.9989784359931946, + 0.9988847374916077, + 0.9990218281745911, + 0.998627781867981, + 0.9988900423049927, + 0.9984332323074341, + 0.9992363452911377, + 0.9988893270492554, + 0.9993957281112671, + 0.9991062879562378, + 0.9990890622138977, + 0.9993386268615723, + 0.9988206624984741, + 0.9987505674362183, + 0.9989955425262451, + 0.9988095164299011, + 0.9991083741188049, + 0.9988147616386414, + 0.9991130828857422, + 0.9988299012184143, + 0.9990316033363342, + 0.9987291693687439, + 0.9991141557693481, + 0.998677134513855, + 0.9991635084152222, + 0.9987846612930298, + 0.9991999864578247, + 0.9988572001457214, + 0.999275267124176, + 0.9995086193084717, + 0.9993034601211548, + 0.9992854595184326, + 0.9993381500244141, + 0.9992458820343018, + 0.999390721321106, + 0.9992026090621948, + 0.999376654624939, + 0.9990814924240112, + 0.9993203282356262, + 0.998900294303894, + 0.9990503191947937, + 0.9984703063964844, + 0.998579740524292, + 0.9977589845657349, + 0.9986475706100464, + 0.9975122213363647, + 0.9999625086784363 + ], + "feedback_forward_norm_ratio": [ + 0.9990415573120117, + 1.000943660736084, + 0.9984080791473389, + 1.0028012990951538, + 0.9976716041564941, + 1.0041753053665161, + 0.9988983869552612, + 1.0034799575805664, + 0.9985147714614868, + 1.0039784908294678, + 0.9979335069656372, + 1.0033996105194092, + 0.9982885718345642, + 1.0039857625961304, + 0.9969147443771362, + 1.0046236515045166, + 0.9986658096313477, + 1.0027612447738647, + 0.9974470138549805, + 1.001051664352417, + 0.9977295398712158, + 1.0032198429107666, + 0.9985910058021545, + 1.0023927688598633, + 0.9979374408721924, + 1.00351881980896, + 0.9980901479721069, + 1.0027828216552734, + 0.9979441165924072, + 1.0024582147598267, + 0.9984496235847473, + 1.002616047859192, + 0.9983158707618713, + 1.003697156906128, + 0.998077929019928, + 1.0036767721176147, + 0.9981355667114258, + 1.0010948181152344, + 0.9994342923164368, + 1.0016738176345825, + 0.9992212057113647, + 1.0019851922988892, + 0.9994403719902039, + 1.0023409128189087, + 0.9992360472679138, + 1.00275719165802, + 0.9991130232810974, + 1.0027894973754883, + 0.9987571239471436, + 1.0030573606491089, + 0.998371958732605, + 1.0036388635635376, + 0.9987108111381531, + 1.0035512447357178, + 1.0061359405517578 + ], + "feedback_forward_relative_error": [ + 0.0343991219997406, + 0.035711634904146194, + 0.04774899035692215, + 0.04909480735659599, + 0.046834658831357956, + 0.05041529983282089, + 0.047919176518917084, + 0.05211520567536354, + 0.04519075155258179, + 0.04748785123229027, + 0.044232290238142014, + 0.05258563905954361, + 0.047105684876441956, + 0.056232575327157974, + 0.039142630994319916, + 0.047465842217206955, + 0.03476492688059807, + 0.04242744669318199, + 0.04270471632480621, + 0.0364033468067646, + 0.048563480377197266, + 0.05017266422510147, + 0.04481198638677597, + 0.0489116869866848, + 0.04223593324422836, + 0.04889996349811554, + 0.042119480669498444, + 0.04852398484945297, + 0.04401214048266411, + 0.05053481459617615, + 0.04208802059292793, + 0.051570046693086624, + 0.04090111702680588, + 0.049531225115060806, + 0.04000690579414368, + 0.04803706333041191, + 0.03808150812983513, + 0.03138519451022148, + 0.037317704409360886, + 0.03787294030189514, + 0.03637689724564552, + 0.038922250270843506, + 0.034905411303043365, + 0.04005312919616699, + 0.035302672535181046, + 0.04300840571522713, + 0.03686448559165001, + 0.04704458639025688, + 0.04357275739312172, + 0.055480051785707474, + 0.05327729135751724, + 0.06716600060462952, + 0.05199092626571655, + 0.07075086236000061, + 0.01063375361263752 + ], + "innovation_negative_gradient_cosine": [ + 0.9989402890205383, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248114585876, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985315203666687, + 0.9990233182907104, + 0.9983853697776794, + 0.9990407824516296, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990947842597961, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991479516029358, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.999323308467865, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992249011993408, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.999476432800293, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8107086420059204 + ], + "instruction_negative_gradient_cosine": [ + 0.9989403486251831, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248710632324, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985314607620239, + 0.9990233182907104, + 0.9983853101730347, + 0.9990407228469849, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990948438644409, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991480112075806, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.9993233680725098, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992247819900513, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.9994763731956482, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8192939758300781 + ], + "matched_negative_gradient_cosine": [ + 0.008659751154482365, + 0.004593271762132645, + 0.006168784573674202, + 0.019933126866817474, + -0.006501247175037861, + -0.002974468283355236, + -0.019822947680950165, + 0.006686114706099033, + -0.026584379374980927, + 0.013955975882709026, + -0.031510911881923676, + -0.02690306305885315, + -0.035326167941093445, + -0.009890604764223099, + -0.031980231404304504, + 0.007654354441910982, + -0.03439664840698242, + -0.022561632096767426, + -0.02662527561187744, + -0.017831280827522278, + -0.057944610714912415, + -0.009949561208486557, + -0.057219497859478, + 0.010901489295065403, + -0.05287568271160126, + 0.041184812784194946, + -0.0532805360853672, + 0.03105335868895054, + -0.06220395117998123, + 0.038169000297784805, + -0.06922180950641632, + 0.04855819046497345, + -0.07716953754425049, + 0.03764723241329193, + -0.05986480787396431, + 0.061009179800748825, + -0.04746749624609947, + 0.04781028628349304, + -0.06197467818856239, + 0.07869354635477066, + -0.06947663426399231, + 0.08068685978651047, + -0.06836137920618057, + 0.07235347479581833, + -0.04931625351309776, + 0.10434074699878693, + -0.05912315845489502, + 0.0692649632692337, + -0.04305483400821686, + 0.05201111361384392, + -0.027310216799378395, + -0.010442962870001793, + 0.01267608255147934, + 0.03919082134962082, + 0.11070923507213593 + ], + "max_norm_match_direction_error": 2.384185791015625e-07, + "max_norm_match_relative_error": 2.003661876415208e-07, + "mean_feedback_forward_cosine": 0.9989724636077881, + "mean_feedback_forward_relative_error": 0.044561998918652534, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.15488560497760773, + "max_absolute_post_projection_soma_slope": 2.4402833531667056e-08, + "max_absolute_pre_projection_soma_slope": 0.15488560497760773, + "max_positive_post_projection_soma_slope": 2.331008275291424e-08, + "min_post_projection_soma_slope": -2.4402833531667056e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.4101872854046475e-08, + "pre_projection_traffic_rms_ratio": 0.00682154858224028 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.006821548582191776, + "raw_negative_gradient_cosine": [ + 0.008659751154482365, + 0.004593271762132645, + 0.006168785504996777, + 0.019933126866817474, + -0.006501247175037861, + -0.0029744685161858797, + -0.019822947680950165, + 0.006686114706099033, + -0.026584379374980927, + 0.013955974951386452, + -0.031510911881923676, + -0.02690306305885315, + -0.035326167941093445, + -0.009890605695545673, + -0.031980231404304504, + 0.00765435304492712, + -0.03439664840698242, + -0.022561632096767426, + -0.02662527561187744, + -0.017831280827522278, + -0.057944610714912415, + -0.009949559345841408, + -0.057219497859478, + 0.010901490226387978, + -0.05287568271160126, + 0.041184812784194946, + -0.053280532360076904, + 0.03105335868895054, + -0.06220395117998123, + 0.038169000297784805, + -0.06922180950641632, + 0.04855819791555405, + -0.07716953754425049, + 0.03764723241329193, + -0.05986480787396431, + 0.061009179800748825, + -0.047467492520809174, + 0.04781028628349304, + -0.06197467818856239, + 0.07869355380535126, + -0.06947663426399231, + 0.08068685978651047, + -0.06836137920618057, + 0.07235348224639893, + -0.04931625351309776, + 0.10434074699878693, + -0.05912315100431442, + 0.0692649632692337, + -0.04305483028292656, + 0.05201110988855362, + -0.027310214936733246, + -0.010442961938679218, + 0.012676079757511616, + 0.03919082134962082, + 0.11070923507213593 + ], + "teaching_negative_gradient_cosine": [ + 0.9989402890205383, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248114585876, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985315203666687, + 0.9990233182907104, + 0.9983853697776794, + 0.9990407824516296, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990947842597961, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991479516029358, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.999323308467865, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992249011993408, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.999476432800293, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8107086420059204 + ], + "traffic_instruction_rms_ratio": 215.5495647029216, + "used_negative_gradient_cosine": [ + 0.9989402890205383, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248114585876, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985315203666687, + 0.9990233182907104, + 0.9983853697776794, + 0.9990407824516296, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990947842597961, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991479516029358, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.999323308467865, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992249011993408, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.999476432800293, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8107086420059204 + ], + "wall_s": 0.4923238754272461 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.3038, + "eval_loss": 1.9626150634765624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.4722906351089478, + "mean_feedback_forward_cosine": 0.08352781883017583, + "mean_feedback_forward_relative_error": 1.3621288386258212, + "min_feedback_forward_cosine": 0.01647607609629631 + }, + "lr": 0.02, + "mixed_apical": { + "innovation_rms": 0.004356151012226876, + "instruction_rms": 0.004356151012217689, + "raw_apical_rms": 0.06061163826323626, + "teaching_rms": 0.004356151012226876, + "traffic_rms": 0.060407860278224694 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.1548856645822525, + "maximum_absolute_post_projection_soma_slope": 9.257554722807981e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.039406479428281e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006108313938754286, + "minimum_observations": 72 + }, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.133220109324985 + }, + { + "epoch": 2, + "eval_accuracy": 0.4618, + "eval_loss": 1.5612975830078124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3664660453796387, + "mean_feedback_forward_cosine": 0.16610823449763384, + "mean_feedback_forward_relative_error": 1.2985886638814752, + "min_feedback_forward_cosine": 0.07316796481609344 + }, + "lr": 0.04, + "mixed_apical": { + "innovation_rms": 0.0034584715151784293, + "instruction_rms": 0.0034584715164519913, + "raw_apical_rms": 0.06749097762423845, + "teaching_rms": 0.0034584715151784293, + "traffic_rms": 0.0674012550651796 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.701681425942297e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.9294788966921167e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007481774138670017, + "minimum_observations": 72 + }, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.689423344315423 + }, + { + "epoch": 3, + "eval_accuracy": 0.4822, + "eval_loss": 1.8836912109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.335602045059204, + "mean_feedback_forward_cosine": 0.26583654568953946, + "mean_feedback_forward_relative_error": 1.218086308782751, + "min_feedback_forward_cosine": 0.11645860224962234 + }, + "lr": 0.06000000000000001, + "mixed_apical": { + "innovation_rms": 0.004173249144323395, + "instruction_rms": 0.004173249145743122, + "raw_apical_rms": 0.07839940296300797, + "teaching_rms": 0.004173249144323395, + "traffic_rms": 0.07828697291035139 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.905030940946745e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8459908782882565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008971926468541658, + "minimum_observations": 72 + }, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.3186583264244927 + }, + { + "epoch": 4, + "eval_accuracy": 0.6074, + "eval_loss": 1.1148765869140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3030064105987549, + "mean_feedback_forward_cosine": 0.3486238040707328, + "mean_feedback_forward_relative_error": 1.1469927928664467, + "min_feedback_forward_cosine": 0.16100770235061646 + }, + "lr": 0.08, + "mixed_apical": { + "innovation_rms": 0.0040261518638995695, + "instruction_rms": 0.004026151863866751, + "raw_apical_rms": 0.08739736235297962, + "teaching_rms": 0.0040261518638995695, + "traffic_rms": 0.08730287845580852 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.533971313478105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7700298915347967e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008616507541436013, + "minimum_observations": 72 + }, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 1.0482054900699191 + }, + { + "epoch": 5, + "eval_accuracy": 0.634, + "eval_loss": 1.18164111328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.274701714515686, + "mean_feedback_forward_cosine": 0.4180870847268538, + "mean_feedback_forward_relative_error": 1.0841255122965032, + "min_feedback_forward_cosine": 0.20132866501808167 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00344417015497597, + "instruction_rms": 0.00344417015405666, + "raw_apical_rms": 0.0936506411050976, + "teaching_rms": 0.00344417015497597, + "traffic_rms": 0.09358633788170477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.826502385592903e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.725172805987594e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007545408892908994, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.8692952327728272 + }, + { + "epoch": 6, + "eval_accuracy": 0.7196, + "eval_loss": 0.8544650756835938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2515367269515991, + "mean_feedback_forward_cosine": 0.46626698022538965, + "mean_feedback_forward_relative_error": 1.0380515185269443, + "min_feedback_forward_cosine": 0.23207521438598633 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003258748170606174, + "instruction_rms": 0.003258748168094511, + "raw_apical_rms": 0.09746526204424157, + "teaching_rms": 0.003258748170606174, + "traffic_rms": 0.09741016939437787 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.529219286401712e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7000357580493475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007613073673005167, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7370866859753926 + }, + { + "epoch": 7, + "eval_accuracy": 0.7446, + "eval_loss": 0.7813277587890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2309913635253906, + "mean_feedback_forward_cosine": 0.5051967512477528, + "mean_feedback_forward_relative_error": 0.9995049790902572, + "min_feedback_forward_cosine": 0.2589109241962433 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031316636807330396, + "instruction_rms": 0.0031316636808242466, + "raw_apical_rms": 0.10071810745913976, + "teaching_rms": 0.0031316636807330396, + "traffic_rms": 0.10066887091447983 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.269455520348856e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6890729415319725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00749110066189653, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6620292295561896 + }, + { + "epoch": 8, + "eval_accuracy": 0.7572, + "eval_loss": 0.7254913330078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2107770442962646, + "mean_feedback_forward_cosine": 0.5387616542252627, + "mean_feedback_forward_relative_error": 0.965096912058917, + "min_feedback_forward_cosine": 0.2846664488315582 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030568417743375748, + "instruction_rms": 0.003056841775456467, + "raw_apical_rms": 0.10333989084849575, + "teaching_rms": 0.0030568417743375748, + "traffic_rms": 0.10329416930041671 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.03358714101887e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6730053206974266e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007363044872899511, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.6020555562443204 + }, + { + "epoch": 9, + "eval_accuracy": 0.8042, + "eval_loss": 0.6033475341796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1921597719192505, + "mean_feedback_forward_cosine": 0.5684559215198863, + "mean_feedback_forward_relative_error": 0.9336268262429671, + "min_feedback_forward_cosine": 0.30742526054382324 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003028564959114619, + "instruction_rms": 0.003028564959126992, + "raw_apical_rms": 0.10490871755052607, + "teaching_rms": 0.003028564959114619, + "traffic_rms": 0.10486444962000695 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.068348312737726e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6725913714204886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007720522522345244, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.551194996770223 + }, + { + "epoch": 10, + "eval_accuracy": 0.7772, + "eval_loss": 0.6904501342773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1731219291687012, + "mean_feedback_forward_cosine": 0.5953169167041779, + "mean_feedback_forward_relative_error": 0.9040230079130693, + "min_feedback_forward_cosine": 0.3304404616355896 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00298673036466182, + "instruction_rms": 0.002986730365661981, + "raw_apical_rms": 0.10737435769335861, + "teaching_rms": 0.00298673036466182, + "traffic_rms": 0.10733233793075171 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.138749597539572e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6607941445358757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006803165530875269, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5172687877972921 + }, + { + "epoch": 11, + "eval_accuracy": 0.8256, + "eval_loss": 0.5396714233398437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.154249906539917, + "mean_feedback_forward_cosine": 0.61987811868841, + "mean_feedback_forward_relative_error": 0.8761465267701583, + "min_feedback_forward_cosine": 0.3531190752983093 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029536718346400105, + "instruction_rms": 0.002953671834104379, + "raw_apical_rms": 0.11032681417903577, + "teaching_rms": 0.0029536718346400105, + "traffic_rms": 0.11028671176853683 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.04118576777546e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.648240257180984e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006627346222083808, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.4821384444978502 + }, + { + "epoch": 12, + "eval_accuracy": 0.7906, + "eval_loss": 0.629708203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1347694396972656, + "mean_feedback_forward_cosine": 0.6426574105566198, + "mean_feedback_forward_relative_error": 0.8493387200615623, + "min_feedback_forward_cosine": 0.375612735748291 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029327823604470378, + "instruction_rms": 0.00293278236035286, + "raw_apical_rms": 0.11178448445325447, + "teaching_rms": 0.0029327823604470378, + "traffic_rms": 0.11174537583442526 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.192748003286397e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.645749663223053e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006768195241669802, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.46439163663652205 + }, + { + "epoch": 13, + "eval_accuracy": 0.7176, + "eval_loss": 1.2164780517578124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.116695761680603, + "mean_feedback_forward_cosine": 0.6637622697786851, + "mean_feedback_forward_relative_error": 0.8237342596054077, + "min_feedback_forward_cosine": 0.3959151804447174 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002959575737588152, + "instruction_rms": 0.0029595757375393207, + "raw_apical_rms": 0.11238785921812094, + "teaching_rms": 0.002959575737588152, + "traffic_rms": 0.11234834424661272 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.434101061993715e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639734164071579e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007556765464320654, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.431485446272956 + }, + { + "epoch": 14, + "eval_accuracy": 0.8306, + "eval_loss": 0.527476123046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0985679626464844, + "mean_feedback_forward_cosine": 0.6832012095234611, + "mean_feedback_forward_relative_error": 0.7993083840066736, + "min_feedback_forward_cosine": 0.41602540016174316 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029475493765237, + "instruction_rms": 0.002947549376620449, + "raw_apical_rms": 0.11334492394345563, + "teaching_rms": 0.0029475493765237, + "traffic_rms": 0.11330598076956266 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.384948392224032e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.635391588293572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007641804239876833, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.4132346541987525 + }, + { + "epoch": 15, + "eval_accuracy": 0.8368, + "eval_loss": 0.5055118041992187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0807214975357056, + "mean_feedback_forward_cosine": 0.7016383577476848, + "mean_feedback_forward_relative_error": 0.7754036990079013, + "min_feedback_forward_cosine": 0.43487748503685 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029272191084330516, + "instruction_rms": 0.002927219109138898, + "raw_apical_rms": 0.11455419807492699, + "teaching_rms": 0.0029272191084330516, + "traffic_rms": 0.11451626827829946 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.27533154329285e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6350265657651034e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0076884928417303685, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.39594793183803556 + }, + { + "epoch": 16, + "eval_accuracy": 0.8376, + "eval_loss": 0.49911519165039064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.062043309211731, + "mean_feedback_forward_cosine": 0.7190157613971017, + "mean_feedback_forward_relative_error": 0.7521753657947887, + "min_feedback_forward_cosine": 0.45529162883758545 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029609705366583886, + "instruction_rms": 0.0029609705357800196, + "raw_apical_rms": 0.11580640097877634, + "teaching_rms": 0.0029609705366583886, + "traffic_rms": 0.11576787682414112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.146004238871683e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6201921493682125e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0073920124447960185, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.37886198110050623 + }, + { + "epoch": 17, + "eval_accuracy": 0.8362, + "eval_loss": 0.5104703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0426470041275024, + "mean_feedback_forward_cosine": 0.7352717670527371, + "mean_feedback_forward_relative_error": 0.7296695384112272, + "min_feedback_forward_cosine": 0.47560685873031616 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029236920534005634, + "instruction_rms": 0.0029236920541952276, + "raw_apical_rms": 0.11692784126086204, + "teaching_rms": 0.0029236920534005634, + "traffic_rms": 0.11689065791646244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.37444310061619e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6217122107274594e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007021743952473259, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.36305745289060803 + }, + { + "epoch": 18, + "eval_accuracy": 0.8484, + "eval_loss": 0.4793226806640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0218993425369263, + "mean_feedback_forward_cosine": 0.7508972482247787, + "mean_feedback_forward_relative_error": 0.7074883742765947, + "min_feedback_forward_cosine": 0.4965720772743225 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029090364970267677, + "instruction_rms": 0.0029090364967422036, + "raw_apical_rms": 0.11792658249965413, + "teaching_rms": 0.0029090364970267677, + "traffic_rms": 0.11788996664729536 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.450660888252969e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6166336038137776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006906754591517689, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.3581241923438178 + }, + { + "epoch": 19, + "eval_accuracy": 0.8336, + "eval_loss": 0.5050219970703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0014246702194214, + "mean_feedback_forward_cosine": 0.7655846194787459, + "mean_feedback_forward_relative_error": 0.686012976007028, + "min_feedback_forward_cosine": 0.5174472332000732 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029182402393648416, + "instruction_rms": 0.0029182402380410533, + "raw_apical_rms": 0.11888263616236855, + "teaching_rms": 0.0029182402393648416, + "traffic_rms": 0.11884613260719862 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.689170106710662e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6154687636359724e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006872142968705691, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.33987536052068074 + }, + { + "epoch": 20, + "eval_accuracy": 0.843, + "eval_loss": 0.4968329345703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9816786646842957, + "mean_feedback_forward_cosine": 0.7795224471525712, + "mean_feedback_forward_relative_error": 0.6649386021223935, + "min_feedback_forward_cosine": 0.5363730192184448 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002928576557955807, + "instruction_rms": 0.0029285765574235023, + "raw_apical_rms": 0.11968668469044214, + "teaching_rms": 0.002928576557955807, + "traffic_rms": 0.1196502353407163 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.13406427873997e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.605554074264612e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0073758731777270985, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3248981048266093 + }, + { + "epoch": 21, + "eval_accuracy": 0.8438, + "eval_loss": 0.4893141540527344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9613377451896667, + "mean_feedback_forward_cosine": 0.7925911849195307, + "mean_feedback_forward_relative_error": 0.6446035880934108, + "min_feedback_forward_cosine": 0.5557418465614319 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002895096260092414, + "instruction_rms": 0.00289509625882239, + "raw_apical_rms": 0.12005545551222437, + "teaching_rms": 0.002895096260092414, + "traffic_rms": 0.12001988512162164 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.405740376498215e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.604683605215343e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0070417378300881295, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.3175357684188419 + }, + { + "epoch": 22, + "eval_accuracy": 0.8296, + "eval_loss": 0.54222138671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9419049620628357, + "mean_feedback_forward_cosine": 0.8045841130343351, + "mean_feedback_forward_relative_error": 0.6253614441914992, + "min_feedback_forward_cosine": 0.5741333961486816 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002856555226427493, + "instruction_rms": 0.002856555226869535, + "raw_apical_rms": 0.12087110037492259, + "teaching_rms": 0.002856555226427493, + "traffic_rms": 0.12083667601829956 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.764066557389924e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6102840511883144e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007370012425247404, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.30489753291871813 + }, + { + "epoch": 23, + "eval_accuracy": 0.8574, + "eval_loss": 0.4504137390136719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9202107787132263, + "mean_feedback_forward_cosine": 0.8164499846371738, + "mean_feedback_forward_relative_error": 0.6057515130801634, + "min_feedback_forward_cosine": 0.593589186668396 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028725876334469535, + "instruction_rms": 0.00287258763288266, + "raw_apical_rms": 0.12143231882816564, + "teaching_rms": 0.0028725876334469535, + "traffic_rms": 0.12139765691039796 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.967339689050277e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5969387802185478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0071395836607705205, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.30050436871846514 + }, + { + "epoch": 24, + "eval_accuracy": 0.8556, + "eval_loss": 0.4592766662597656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8989444971084595, + "mean_feedback_forward_cosine": 0.8273687156763944, + "mean_feedback_forward_relative_error": 0.5870982256802646, + "min_feedback_forward_cosine": 0.6122702360153198 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002864436829320474, + "instruction_rms": 0.002864436829603412, + "raw_apical_rms": 0.12229681651152781, + "teaching_rms": 0.002864436829320474, + "traffic_rms": 0.12226248775156293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.430216442116944e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5990416529969212e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007536012273628919, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.2934852758381102 + }, + { + "epoch": 25, + "eval_accuracy": 0.868, + "eval_loss": 0.44546570434570315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8786854147911072, + "mean_feedback_forward_cosine": 0.8374691377986562, + "mean_feedback_forward_relative_error": 0.5693387931043451, + "min_feedback_forward_cosine": 0.6295708417892456 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002829651196747579, + "instruction_rms": 0.0028296511963980223, + "raw_apical_rms": 0.12340358394249921, + "teaching_rms": 0.002829651196747579, + "traffic_rms": 0.12337042871511392 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.942688912658923e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5987336739498122e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007212634374511028, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.28005119054052563 + }, + { + "epoch": 26, + "eval_accuracy": 0.8544, + "eval_loss": 0.47900587768554687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8583404421806335, + "mean_feedback_forward_cosine": 0.8471909403800965, + "mean_feedback_forward_relative_error": 0.5516635892066089, + "min_feedback_forward_cosine": 0.6465649604797363 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002848069834085287, + "instruction_rms": 0.0028480698337839737, + "raw_apical_rms": 0.1239141001535537, + "teaching_rms": 0.002848069834085287, + "traffic_rms": 0.12388064220186878 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.506996896860073e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6022283644044384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007315140923491566, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.27400048888524375 + }, + { + "epoch": 27, + "eval_accuracy": 0.8618, + "eval_loss": 0.4623237609863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8366420269012451, + "mean_feedback_forward_cosine": 0.8568273132497614, + "mean_feedback_forward_relative_error": 0.5336897641420364, + "min_feedback_forward_cosine": 0.6642788648605347 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029013843485941938, + "instruction_rms": 0.0029013843489158237, + "raw_apical_rms": 0.12409414597047937, + "teaching_rms": 0.0029013843485941938, + "traffic_rms": 0.12405940981755702 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.39095185053884e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5923831749284995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007635893560403707, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.2738130042764876 + }, + { + "epoch": 28, + "eval_accuracy": 0.8622, + "eval_loss": 0.4267572326660156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8168511390686035, + "mean_feedback_forward_cosine": 0.8653341174125672, + "mean_feedback_forward_relative_error": 0.5172499830072577, + "min_feedback_forward_cosine": 0.6802014112472534 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00283451752625823, + "instruction_rms": 0.002834517526025941, + "raw_apical_rms": 0.1251018315361786, + "teaching_rms": 0.00283451752625823, + "traffic_rms": 0.12506885912587118 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.042048366656672e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.584903010045293e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0072322718528165875, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.255333635351393 + }, + { + "epoch": 29, + "eval_accuracy": 0.8388, + "eval_loss": 0.5879559814453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.795055091381073, + "mean_feedback_forward_cosine": 0.8737094499848106, + "mean_feedback_forward_relative_error": 0.5007094104181636, + "min_feedback_forward_cosine": 0.6969889402389526 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028299987552350097, + "instruction_rms": 0.002829998755785549, + "raw_apical_rms": 0.12658011802634273, + "teaching_rms": 0.0028299987552350097, + "traffic_rms": 0.12654780600309287 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.592021861706598e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5873349930203038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0077517706763474605, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.25626799415747326 + }, + { + "epoch": 30, + "eval_accuracy": 0.8718, + "eval_loss": 0.39867880859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7741381525993347, + "mean_feedback_forward_cosine": 0.8812859730287032, + "mean_feedback_forward_relative_error": 0.4851856117898768, + "min_feedback_forward_cosine": 0.7126460075378418 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027692479169138754, + "instruction_rms": 0.002769247917051474, + "raw_apical_rms": 0.12651338429739048, + "teaching_rms": 0.0027692479169138754, + "traffic_rms": 0.1264823177734938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.423312098353563e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.57817413257678e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007284180781127167, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.2529096601009369 + }, + { + "epoch": 31, + "eval_accuracy": 0.8496, + "eval_loss": 0.5267428649902344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7539872527122498, + "mean_feedback_forward_cosine": 0.888480608029799, + "mean_feedback_forward_relative_error": 0.47003339474851435, + "min_feedback_forward_cosine": 0.7276049852371216 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002777148944164583, + "instruction_rms": 0.0027771489435388587, + "raw_apical_rms": 0.1268157266566673, + "teaching_rms": 0.002777148944164583, + "traffic_rms": 0.12678454579159426 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.542652724623622e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5803557519019572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007124422477869925, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.24340678247875638 + }, + { + "epoch": 32, + "eval_accuracy": 0.8584, + "eval_loss": 0.4592197570800781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7332122325897217, + "mean_feedback_forward_cosine": 0.8954121329567649, + "mean_feedback_forward_relative_error": 0.4549131179397756, + "min_feedback_forward_cosine": 0.7420244216918945 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028007994985260643, + "instruction_rms": 0.0028007994988742528, + "raw_apical_rms": 0.12702853135736303, + "teaching_rms": 0.0028007994985260643, + "traffic_rms": 0.12699686934374244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.398566026495246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.580052024833378e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007594909668813672, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.23412086204952665 + }, + { + "epoch": 33, + "eval_accuracy": 0.8736, + "eval_loss": 0.4201716491699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7124688625335693, + "mean_feedback_forward_cosine": 0.9018361796032298, + "mean_feedback_forward_relative_error": 0.4405053669756109, + "min_feedback_forward_cosine": 0.7564120292663574 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027640386068428814, + "instruction_rms": 0.0027640386061699306, + "raw_apical_rms": 0.12794259905336494, + "teaching_rms": 0.0027640386068428814, + "traffic_rms": 0.1279120062439201 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.957049774631741e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5846975676409703e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007508806525524007, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.2334535021940867 + }, + { + "epoch": 34, + "eval_accuracy": 0.8588, + "eval_loss": 0.4472315795898438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6925833821296692, + "mean_feedback_forward_cosine": 0.9078588420694524, + "mean_feedback_forward_relative_error": 0.42656055174090646, + "min_feedback_forward_cosine": 0.7698280811309814 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027950519476158046, + "instruction_rms": 0.002795051948121846, + "raw_apical_rms": 0.12850303163694832, + "teaching_rms": 0.0027950519476158046, + "traffic_rms": 0.12847181246060527 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.994501854431292e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.58093701741168e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007139756940525537, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.23220217074288263 + }, + { + "epoch": 35, + "eval_accuracy": 0.8782, + "eval_loss": 0.4101513122558594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6739795804023743, + "mean_feedback_forward_cosine": 0.9134018529545177, + "mean_feedback_forward_relative_error": 0.41330894218249753, + "min_feedback_forward_cosine": 0.7817182540893555 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002732587766544747, + "instruction_rms": 0.0027325877671195275, + "raw_apical_rms": 0.12855414815159502, + "teaching_rms": 0.002732587766544747, + "traffic_rms": 0.128524331035323 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.077385160731865e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5769118836785654e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007295633148785567, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.218170282819536 + }, + { + "epoch": 36, + "eval_accuracy": 0.8558, + "eval_loss": 0.4931694885253906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6537808775901794, + "mean_feedback_forward_cosine": 0.9188592899929393, + "mean_feedback_forward_relative_error": 0.3998900077559731, + "min_feedback_forward_cosine": 0.7946034669876099 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002769010866746977, + "instruction_rms": 0.0027690108679081483, + "raw_apical_rms": 0.12890115929377205, + "teaching_rms": 0.002769010866746977, + "traffic_rms": 0.12887049951636173 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.622905767197153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5788079860832783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007345657039752453, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.22459270853466456 + }, + { + "epoch": 37, + "eval_accuracy": 0.8776, + "eval_loss": 0.40465287475585937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.634764552116394, + "mean_feedback_forward_cosine": 0.9238546707413413, + "mean_feedback_forward_relative_error": 0.38725803738290615, + "min_feedback_forward_cosine": 0.8060342073440552 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027250066728778913, + "instruction_rms": 0.002725006674491068, + "raw_apical_rms": 0.1296990459836724, + "teaching_rms": 0.0027250066728778913, + "traffic_rms": 0.12966971913174946 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.402741341091314e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5756175122683314e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007059058562824513, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.21901262276437547 + }, + { + "epoch": 38, + "eval_accuracy": 0.8746, + "eval_loss": 0.4201908935546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6174896359443665, + "mean_feedback_forward_cosine": 0.9285104047168385, + "mean_feedback_forward_relative_error": 0.37506215965205975, + "min_feedback_forward_cosine": 0.816362202167511 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002758986976064875, + "instruction_rms": 0.0027589869757443514, + "raw_apical_rms": 0.13022396191701172, + "teaching_rms": 0.002758986976064875, + "traffic_rms": 0.13019391394380037 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.348520369987455e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.581195620728288e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006967136351535924, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.20924033000204298 + }, + { + "epoch": 39, + "eval_accuracy": 0.8458, + "eval_loss": 0.5435376342773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5990682244300842, + "mean_feedback_forward_cosine": 0.9329551046544855, + "mean_feedback_forward_relative_error": 0.36302519400011407, + "min_feedback_forward_cosine": 0.8269726037979126 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027273322335747492, + "instruction_rms": 0.0027273322343506355, + "raw_apical_rms": 0.13095048467476908, + "teaching_rms": 0.0027273322335747492, + "traffic_rms": 0.13092118149951398 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.58836825095932e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5812156990171665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007481262871491552, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.20766454303529527 + }, + { + "epoch": 40, + "eval_accuracy": 0.874, + "eval_loss": 0.43536727294921873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5813534259796143, + "mean_feedback_forward_cosine": 0.9371668880636042, + "mean_feedback_forward_relative_error": 0.3513018066232855, + "min_feedback_forward_cosine": 0.8368985056877136 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027336582577564207, + "instruction_rms": 0.002733658257960806, + "raw_apical_rms": 0.13118291598587228, + "teaching_rms": 0.0027336582577564207, + "traffic_rms": 0.13115348568521756 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.182926964333092e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.579920515301132e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007629555184683061, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.20535460352367824 + }, + { + "epoch": 41, + "eval_accuracy": 0.8628, + "eval_loss": 0.4404171691894531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5636888742446899, + "mean_feedback_forward_cosine": 0.9411124727942727, + "mean_feedback_forward_relative_error": 0.3399224884130738, + "min_feedback_forward_cosine": 0.8466179966926575 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002744078526469126, + "instruction_rms": 0.002744078525429673, + "raw_apical_rms": 0.1316687178999931, + "teaching_rms": 0.002744078526469126, + "traffic_rms": 0.1316392812613315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.20359870140419e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5820787800352437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008055984411047367, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.2067696696360906 + }, + { + "epoch": 42, + "eval_accuracy": 0.8614, + "eval_loss": 0.4805859619140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5467633605003357, + "mean_feedback_forward_cosine": 0.9447931029579856, + "mean_feedback_forward_relative_error": 0.32900360971689224, + "min_feedback_forward_cosine": 0.8555374145507812 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002707756478686592, + "instruction_rms": 0.002707756479250177, + "raw_apical_rms": 0.13231939173287324, + "teaching_rms": 0.002707756478686592, + "traffic_rms": 0.13229083411503598 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.411892883306791e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5674920046896586e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007553069431776005, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.19649066227674483 + }, + { + "epoch": 43, + "eval_accuracy": 0.883, + "eval_loss": 0.38886524658203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5288952589035034, + "mean_feedback_forward_cosine": 0.9484201853925531, + "mean_feedback_forward_relative_error": 0.31789819408546793, + "min_feedback_forward_cosine": 0.864727795124054 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002765419179195218, + "instruction_rms": 0.002765419179768991, + "raw_apical_rms": 0.1326336693797406, + "teaching_rms": 0.002765419179195218, + "traffic_rms": 0.13260393202195417 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.192547712975284e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5708784935248106e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007264604335689377, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.20149312126371596 + }, + { + "epoch": 44, + "eval_accuracy": 0.8612, + "eval_loss": 0.50747626953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5118293762207031, + "mean_feedback_forward_cosine": 0.9517511595379222, + "mean_feedback_forward_relative_error": 0.3073600098490715, + "min_feedback_forward_cosine": 0.8732614517211914 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027203200955642414, + "instruction_rms": 0.0027203200957793164, + "raw_apical_rms": 0.13295848846007663, + "teaching_rms": 0.0027203200955642414, + "traffic_rms": 0.13292969522240394 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.60637206553838e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5711413034686096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007564231186813065, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.19517155153486465 + }, + { + "epoch": 45, + "eval_accuracy": 0.8878, + "eval_loss": 0.36498590087890626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4960921108722687, + "mean_feedback_forward_cosine": 0.9547883597287264, + "mean_feedback_forward_relative_error": 0.2974040485241196, + "min_feedback_forward_cosine": 0.8808695077896118 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026858554466063487, + "instruction_rms": 0.002685855445555374, + "raw_apical_rms": 0.13319069534527686, + "teaching_rms": 0.0026858554466063487, + "traffic_rms": 0.13316264565196495 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.15449451402128e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5824865636116507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0071554614644181004, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.18736503223843046 + }, + { + "epoch": 46, + "eval_accuracy": 0.8796, + "eval_loss": 0.4100537841796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.479801744222641, + "mean_feedback_forward_cosine": 0.9577374501661821, + "mean_feedback_forward_relative_error": 0.28743237392468884, + "min_feedback_forward_cosine": 0.8884043097496033 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002689226776063514, + "instruction_rms": 0.0026892267768847474, + "raw_apical_rms": 0.13383122359073962, + "teaching_rms": 0.002689226776063514, + "traffic_rms": 0.1338033331991483 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.937294554949403e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5711693143053502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007326037402498839, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.1950026453812917 + }, + { + "epoch": 47, + "eval_accuracy": 0.8474, + "eval_loss": 0.5450638427734374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4652605950832367, + "mean_feedback_forward_cosine": 0.9603951844302091, + "mean_feedback_forward_relative_error": 0.2781396528536623, + "min_feedback_forward_cosine": 0.895006537437439 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027059804394138154, + "instruction_rms": 0.0027059804383819715, + "raw_apical_rms": 0.13411446537974409, + "teaching_rms": 0.0027059804394138154, + "traffic_rms": 0.13408619038165967 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.11744689851912e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.575908293843505e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007480411684179155, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.18229786556031968 + }, + { + "epoch": 48, + "eval_accuracy": 0.881, + "eval_loss": 0.40165243530273437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45054587721824646, + "mean_feedback_forward_cosine": 0.9629461017521945, + "mean_feedback_forward_relative_error": 0.26894587508656764, + "min_feedback_forward_cosine": 0.9014789462089539 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002693908686589515, + "instruction_rms": 0.0026939086853265126, + "raw_apical_rms": 0.13502132911330866, + "teaching_rms": 0.002693908686589515, + "traffic_rms": 0.1349935960286189 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.1770195129656713e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.570783537931346e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006955519458433687, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.1800456293159061 + }, + { + "epoch": 49, + "eval_accuracy": 0.8728, + "eval_loss": 0.4312097717285156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43543577194213867, + "mean_feedback_forward_cosine": 0.9654507127675143, + "mean_feedback_forward_relative_error": 0.25960782170295715, + "min_feedback_forward_cosine": 0.9078564643859863 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027744577659100825, + "instruction_rms": 0.0027744577652871103, + "raw_apical_rms": 0.13506830415447602, + "teaching_rms": 0.0027744577659100825, + "traffic_rms": 0.1350388678008716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.2062105270160828e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5670809307587052e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00705050694589141, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.187842915391922 + }, + { + "epoch": 50, + "eval_accuracy": 0.882, + "eval_loss": 0.4146776062011719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42082518339157104, + "mean_feedback_forward_cosine": 0.9676603544842113, + "mean_feedback_forward_relative_error": 0.25113575214689426, + "min_feedback_forward_cosine": 0.9138859510421753 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026742389896612447, + "instruction_rms": 0.0026742389904405466, + "raw_apical_rms": 0.13622161122472387, + "teaching_rms": 0.0026742389896612447, + "traffic_rms": 0.13619444996311894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.755938469244938e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5681843587762813e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006928600617987082, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.17939616974724665 + }, + { + "epoch": 51, + "eval_accuracy": 0.8604, + "eval_loss": 0.5345984741210937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40693342685699463, + "mean_feedback_forward_cosine": 0.9697972904552113, + "mean_feedback_forward_relative_error": 0.24261356687003915, + "min_feedback_forward_cosine": 0.9194050431251526 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027662623067010294, + "instruction_rms": 0.002766262307855727, + "raw_apical_rms": 0.1360560502837404, + "teaching_rms": 0.0027662623067010294, + "traffic_rms": 0.13602696343853785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.263002999025048e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5683109799004856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007084063566015009, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.1764916455878152 + }, + { + "epoch": 52, + "eval_accuracy": 0.8872, + "eval_loss": 0.4156570983886719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3940022587776184, + "mean_feedback_forward_cosine": 0.9716921968893572, + "mean_feedback_forward_relative_error": 0.23481411662968724, + "min_feedback_forward_cosine": 0.9243811964988708 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002697988098239205, + "instruction_rms": 0.0026979880970540156, + "raw_apical_rms": 0.13647367571385105, + "teaching_rms": 0.002697988098239205, + "traffic_rms": 0.13644610343628633 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.969654802764126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5684261488329718e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007448526446113501, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.17460551758077408 + }, + { + "epoch": 53, + "eval_accuracy": 0.8586, + "eval_loss": 0.523772998046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38133642077445984, + "mean_feedback_forward_cosine": 0.9734930721196261, + "mean_feedback_forward_relative_error": 0.2271831222555854, + "min_feedback_forward_cosine": 0.929129958152771 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026592711490205072, + "instruction_rms": 0.0026592711492970256, + "raw_apical_rms": 0.1371275600132018, + "teaching_rms": 0.0026592711490205072, + "traffic_rms": 0.13710088693649733 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.430827426129326e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5650635465185947e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007512666166711919, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.16919431093004014 + }, + { + "epoch": 54, + "eval_accuracy": 0.8722, + "eval_loss": 0.4497177734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3682323396205902, + "mean_feedback_forward_cosine": 0.9752173001115972, + "mean_feedback_forward_relative_error": 0.21963099959221752, + "min_feedback_forward_cosine": 0.9339085817337036 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026435040710614346, + "instruction_rms": 0.002643504071121708, + "raw_apical_rms": 0.13807241382641705, + "teaching_rms": 0.0026435040710614346, + "traffic_rms": 0.138046214003334 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.264454010031841e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5631190660242787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00658986963550856, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.17247383966445923 + }, + { + "epoch": 55, + "eval_accuracy": 0.8752, + "eval_loss": 0.4191963256835938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3554491400718689, + "mean_feedback_forward_cosine": 0.9768868413838473, + "mean_feedback_forward_relative_error": 0.21204523864117536, + "min_feedback_forward_cosine": 0.9383438229560852 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027582438894246095, + "instruction_rms": 0.0027582438892282783, + "raw_apical_rms": 0.13812332969784827, + "teaching_rms": 0.0027582438894246095, + "traffic_rms": 0.13809479552881931 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.316208521819135e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5671207016380378e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0070906782215834656, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.17254678353203667 + }, + { + "epoch": 56, + "eval_accuracy": 0.8906, + "eval_loss": 0.37485194091796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.34314805269241333, + "mean_feedback_forward_cosine": 0.9784383351152593, + "mean_feedback_forward_relative_error": 0.20477277054028079, + "min_feedback_forward_cosine": 0.9425045251846313 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027295839618689363, + "instruction_rms": 0.0027295839617653022, + "raw_apical_rms": 0.13859104716160184, + "teaching_rms": 0.0027295839618689363, + "traffic_rms": 0.13856313390866049 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.156334109164163e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.566586840540628e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0073859677514521915, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.17252226725286907 + }, + { + "epoch": 57, + "eval_accuracy": 0.8628, + "eval_loss": 0.516157080078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33128440380096436, + "mean_feedback_forward_cosine": 0.9798409548672763, + "mean_feedback_forward_relative_error": 0.19796063107523051, + "min_feedback_forward_cosine": 0.9463651180267334 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026615503012947686, + "instruction_rms": 0.0026615503021120116, + "raw_apical_rms": 0.13962380587554718, + "teaching_rms": 0.0026615503012947686, + "traffic_rms": 0.1395975968215392 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.387510774809925e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568659816810336e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0068918217166773765, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.16815749957296583 + }, + { + "epoch": 58, + "eval_accuracy": 0.8842, + "eval_loss": 0.39102216796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3196627199649811, + "mean_feedback_forward_cosine": 0.9811797521331094, + "mean_feedback_forward_relative_error": 0.19124092947353016, + "min_feedback_forward_cosine": 0.950006902217865 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026910145468295055, + "instruction_rms": 0.002691014546299131, + "raw_apical_rms": 0.14000806357468948, + "teaching_rms": 0.0026910145468295055, + "traffic_rms": 0.13998117499928586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.236707233460038e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5617678302335342e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007050061817657493, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.16205002730157642 + }, + { + "epoch": 59, + "eval_accuracy": 0.8536, + "eval_loss": 0.5629090026855469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30899161100387573, + "mean_feedback_forward_cosine": 0.9823998147791082, + "mean_feedback_forward_relative_error": 0.18491454686630857, + "min_feedback_forward_cosine": 0.9532514810562134 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026542196189168255, + "instruction_rms": 0.0026542196190562158, + "raw_apical_rms": 0.1408215516346445, + "teaching_rms": 0.0026542196189168255, + "traffic_rms": 0.14079564388851343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.042845050544202e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5614803441729725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006744619242398327, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.16429483944310083 + }, + { + "epoch": 60, + "eval_accuracy": 0.8662, + "eval_loss": 0.4899179992675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.298313170671463, + "mean_feedback_forward_cosine": 0.9836080356077714, + "mean_feedback_forward_relative_error": 0.17843473235314541, + "min_feedback_forward_cosine": 0.9563993811607361 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027398928836518495, + "instruction_rms": 0.0027398928831306704, + "raw_apical_rms": 0.1412928126177813, + "teaching_rms": 0.0027398928836518495, + "traffic_rms": 0.14126510981437895 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.259701900215077e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5647205410476997e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006542335349927483, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.16467524908648598 + }, + { + "epoch": 61, + "eval_accuracy": 0.8658, + "eval_loss": 0.4761692321777344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2877059280872345, + "mean_feedback_forward_cosine": 0.9847212650559165, + "mean_feedback_forward_relative_error": 0.17225712991573594, + "min_feedback_forward_cosine": 0.9594355821609497 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026823350079127613, + "instruction_rms": 0.002682335009338564, + "raw_apical_rms": 0.14184140255649313, + "teaching_rms": 0.0026823350079127613, + "traffic_rms": 0.14181507741524693 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.723926387370739e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5598152529847118e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006406189595088024, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.1637429113732444 + }, + { + "epoch": 62, + "eval_accuracy": 0.874, + "eval_loss": 0.4491427978515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27832525968551636, + "mean_feedback_forward_cosine": 0.985716106674888, + "mean_feedback_forward_relative_error": 0.16652044640346006, + "min_feedback_forward_cosine": 0.9620136022567749 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026883491568320092, + "instruction_rms": 0.0026883491571807086, + "raw_apical_rms": 0.14177491808751969, + "teaching_rms": 0.0026883491568320092, + "traffic_rms": 0.14174836359817858 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.302907789608071e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5672505030602204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0063795182501372415, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.15938696756495369 + }, + { + "epoch": 63, + "eval_accuracy": 0.8836, + "eval_loss": 0.416558642578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26862141489982605, + "mean_feedback_forward_cosine": 0.9866406364874406, + "mean_feedback_forward_relative_error": 0.16102071370590818, + "min_feedback_forward_cosine": 0.9645927548408508 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026748116780336787, + "instruction_rms": 0.0026748116780000463, + "raw_apical_rms": 0.14231261999482483, + "teaching_rms": 0.0026748116780336787, + "traffic_rms": 0.1422866051708621 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.027869771519363e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5610342012518305e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006332642947062814, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.1564018759330114 + }, + { + "epoch": 64, + "eval_accuracy": 0.8804, + "eval_loss": 0.4190369567871094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2594032287597656, + "mean_feedback_forward_cosine": 0.9875170274214311, + "mean_feedback_forward_relative_error": 0.1556335460055958, + "min_feedback_forward_cosine": 0.9669644832611084 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00265454344144199, + "instruction_rms": 0.002654543442255734, + "raw_apical_rms": 0.14241026046077243, + "teaching_rms": 0.00265454344144199, + "traffic_rms": 0.14238456670964295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.0995398724844563e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558226096933054e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006739131120864999, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.15596432549556097 + }, + { + "epoch": 65, + "eval_accuracy": 0.876, + "eval_loss": 0.43032020874023436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25037944316864014, + "mean_feedback_forward_cosine": 0.9883387251333757, + "mean_feedback_forward_relative_error": 0.15041881521994419, + "min_feedback_forward_cosine": 0.9692103862762451 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002660427249889889, + "instruction_rms": 0.0026604272507897045, + "raw_apical_rms": 0.14271273509772173, + "teaching_rms": 0.002660427249889889, + "traffic_rms": 0.14268685941881587 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.739331036622389e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5603208316482917e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00676777397236648, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.15521818123393588 + }, + { + "epoch": 66, + "eval_accuracy": 0.8644, + "eval_loss": 0.4574156127929688, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.241669699549675, + "mean_feedback_forward_cosine": 0.9891077453439886, + "mean_feedback_forward_relative_error": 0.14536555781960486, + "min_feedback_forward_cosine": 0.9712958931922913 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026611669122147048, + "instruction_rms": 0.0026611669117068064, + "raw_apical_rms": 0.14344285344871494, + "teaching_rms": 0.0026611669122147048, + "traffic_rms": 0.14341728438724502 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.2062049847827438e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5584886618607892e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006897932759860844, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.15296029583613077 + }, + { + "epoch": 67, + "eval_accuracy": 0.871, + "eval_loss": 0.470939794921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23326486349105835, + "mean_feedback_forward_cosine": 0.9898286559365013, + "mean_feedback_forward_relative_error": 0.14045636931603606, + "min_feedback_forward_cosine": 0.9732484221458435 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002687069164880757, + "instruction_rms": 0.0026870691648951677, + "raw_apical_rms": 0.14413292472521066, + "teaching_rms": 0.002687069164880757, + "traffic_rms": 0.14410686004612444 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.190936367531322e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5566121694916376e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006414959937737721, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.1511115888701545 + }, + { + "epoch": 68, + "eval_accuracy": 0.8824, + "eval_loss": 0.4170122924804687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22500966489315033, + "mean_feedback_forward_cosine": 0.9904924251816489, + "mean_feedback_forward_relative_error": 0.13578086393800648, + "min_feedback_forward_cosine": 0.975089430809021 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002664870463922071, + "instruction_rms": 0.0026648704640862457, + "raw_apical_rms": 0.14456652877016374, + "teaching_rms": 0.002664870463922071, + "traffic_rms": 0.14454092368457447 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.0229710767362121e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5550262469623638e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006032912033059801, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.14787748815748428 + }, + { + "epoch": 69, + "eval_accuracy": 0.8874, + "eval_loss": 0.408225830078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2171240746974945, + "mean_feedback_forward_cosine": 0.9911184993657198, + "mean_feedback_forward_relative_error": 0.13122503445907072, + "min_feedback_forward_cosine": 0.9767956137657166 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002645522793587049, + "instruction_rms": 0.002645522793048388, + "raw_apical_rms": 0.14505173493671394, + "teaching_rms": 0.002645522793587049, + "traffic_rms": 0.14502663517708173 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.356143237937431e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5635902284659227e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006305431332244356, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.15232352664205764 + }, + { + "epoch": 70, + "eval_accuracy": 0.882, + "eval_loss": 0.4383751708984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2094569355249405, + "mean_feedback_forward_cosine": 0.9917126774787903, + "mean_feedback_forward_relative_error": 0.12675295553424143, + "min_feedback_forward_cosine": 0.9783931374549866 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026780029889819277, + "instruction_rms": 0.002678002988411314, + "raw_apical_rms": 0.14585905041703728, + "teaching_rms": 0.0026780029889819277, + "traffic_rms": 0.14583342579872963 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.433038007320647e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558632909576437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005996440704448436, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.1509448115931617 + }, + { + "epoch": 71, + "eval_accuracy": 0.8922, + "eval_loss": 0.3718404113769531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20150041580200195, + "mean_feedback_forward_cosine": 0.9922957691279325, + "mean_feedback_forward_relative_error": 0.12221148996190591, + "min_feedback_forward_cosine": 0.9799900054931641 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027130463732828367, + "instruction_rms": 0.0027130463740779698, + "raw_apical_rms": 0.14632378821722586, + "teaching_rms": 0.0027130463732828367, + "traffic_rms": 0.14629759574932538 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.710679573667221e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5560421661491046e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005826156288919894, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.1526694540924496 + }, + { + "epoch": 72, + "eval_accuracy": 0.8636, + "eval_loss": 0.4955664001464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19431686401367188, + "mean_feedback_forward_cosine": 0.9928037383339622, + "mean_feedback_forward_relative_error": 0.1181078931147402, + "min_feedback_forward_cosine": 0.9813801646232605 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002664071129739347, + "instruction_rms": 0.0026640711295466517, + "raw_apical_rms": 0.14673525991584466, + "teaching_rms": 0.002664071129739347, + "traffic_rms": 0.1467101764499398 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.126009731768136e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5553409037084792e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006670596422769175, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.1505700844446818 + }, + { + "epoch": 73, + "eval_accuracy": 0.8908, + "eval_loss": 0.39359422607421873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1880732625722885, + "mean_feedback_forward_cosine": 0.9932585954666138, + "mean_feedback_forward_relative_error": 0.11430438337000934, + "min_feedback_forward_cosine": 0.9825577735900879 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025850056044761024, + "instruction_rms": 0.0025850056042038927, + "raw_apical_rms": 0.14750154417055503, + "teaching_rms": 0.0025850056044761024, + "traffic_rms": 0.14747790681118195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.290473291732269e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.547891947157436e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005991140922600184, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.13924342036247253 + }, + { + "epoch": 74, + "eval_accuracy": 0.8754, + "eval_loss": 0.4568603271484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1814851313829422, + "mean_feedback_forward_cosine": 0.9937111085111444, + "mean_feedback_forward_relative_error": 0.1103922130370682, + "min_feedback_forward_cosine": 0.983745813369751 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002688761067319657, + "instruction_rms": 0.0026887610679484295, + "raw_apical_rms": 0.1482457856465738, + "teaching_rms": 0.002688761067319657, + "traffic_rms": 0.14822035964027308 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.460766937583685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5501059713435917e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006072493714815955, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.14499519442982145 + }, + { + "epoch": 75, + "eval_accuracy": 0.8836, + "eval_loss": 0.44175254516601564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17547009885311127, + "mean_feedback_forward_cosine": 0.9941201881928877, + "mean_feedback_forward_relative_error": 0.10673196173527023, + "min_feedback_forward_cosine": 0.9847981929779053 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026194071485565852, + "instruction_rms": 0.0026194071484548567, + "raw_apical_rms": 0.14859383211916885, + "teaching_rms": 0.0026194071485565852, + "traffic_rms": 0.14856960693123666 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.422455183814236e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5488001708791423e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006249918872554391, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.14262372442086538 + }, + { + "epoch": 76, + "eval_accuracy": 0.8904, + "eval_loss": 0.4010156982421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16894270479679108, + "mean_feedback_forward_cosine": 0.9945245125076988, + "mean_feedback_forward_relative_error": 0.10299218760972673, + "min_feedback_forward_cosine": 0.9859023690223694 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026902458711936606, + "instruction_rms": 0.002690245872698505, + "raw_apical_rms": 0.14890981101763492, + "teaching_rms": 0.0026902458711936606, + "traffic_rms": 0.14888451675335065 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.025010345136252e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5514415971110283e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006046222514356379, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.14856343885262807 + }, + { + "epoch": 77, + "eval_accuracy": 0.888, + "eval_loss": 0.3846547119140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16314925253391266, + "mean_feedback_forward_cosine": 0.9948757789351723, + "mean_feedback_forward_relative_error": 0.09962514144453136, + "min_feedback_forward_cosine": 0.9868456125259399 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026074787589298602, + "instruction_rms": 0.0026074787585614817, + "raw_apical_rms": 0.1487536503573009, + "teaching_rms": 0.0026074787589298602, + "traffic_rms": 0.14872976845351318 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.563444105405324e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550309463471458e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006040926408161601, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.14225712541474236 + }, + { + "epoch": 78, + "eval_accuracy": 0.8872, + "eval_loss": 0.42608485717773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15741144120693207, + "mean_feedback_forward_cosine": 0.99521865086122, + "mean_feedback_forward_relative_error": 0.09623410363088955, + "min_feedback_forward_cosine": 0.9877543449401855 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026644436601364232, + "instruction_rms": 0.002664443659893001, + "raw_apical_rms": 0.1491940605432092, + "teaching_rms": 0.0026644436601364232, + "traffic_rms": 0.14916935816296806 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.823434534590888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545706453746411e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006045228585124519, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.14104981551302803 + }, + { + "epoch": 79, + "eval_accuracy": 0.8934, + "eval_loss": 0.3808540100097656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15193703770637512, + "mean_feedback_forward_cosine": 0.9955377231944691, + "mean_feedback_forward_relative_error": 0.09295748018405654, + "min_feedback_forward_cosine": 0.9885869026184082 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002669457632122175, + "instruction_rms": 0.0026694576316966968, + "raw_apical_rms": 0.14943006841902376, + "teaching_rms": 0.002669457632122175, + "traffic_rms": 0.149405098505395 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.857086987110961e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5490430314861486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006168012162918069, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.1423822466108534 + }, + { + "epoch": 80, + "eval_accuracy": 0.8898, + "eval_loss": 0.3982021240234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14650283753871918, + "mean_feedback_forward_cosine": 0.9958310755816373, + "mean_feedback_forward_relative_error": 0.08984928002411668, + "min_feedback_forward_cosine": 0.9893860220909119 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026474612464264867, + "instruction_rms": 0.0026474612460560456, + "raw_apical_rms": 0.14947968514396548, + "teaching_rms": 0.0026474612464264867, + "traffic_rms": 0.1494552083984354 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.022184433935763e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.543926067696769e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006218834414476274, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.14141934985054863 + }, + { + "epoch": 81, + "eval_accuracy": 0.8836, + "eval_loss": 0.4241302490234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1411639302968979, + "mean_feedback_forward_cosine": 0.9961040648547086, + "mean_feedback_forward_relative_error": 0.08685493939979509, + "min_feedback_forward_cosine": 0.9901441335678101 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025987194917833226, + "instruction_rms": 0.0025987194914536575, + "raw_apical_rms": 0.14968001705164785, + "teaching_rms": 0.0025987194917833226, + "traffic_rms": 0.14965637338464843 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.208247865264639e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545970864669793e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0060768839228992275, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.13652204445468055 + }, + { + "epoch": 82, + "eval_accuracy": 0.8654, + "eval_loss": 0.4978619323730469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13596899807453156, + "mean_feedback_forward_cosine": 0.9963705561377786, + "mean_feedback_forward_relative_error": 0.08382239555093375, + "min_feedback_forward_cosine": 0.9908521771430969 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027191995463910993, + "instruction_rms": 0.0027191995464462345, + "raw_apical_rms": 0.15058060300650536, + "teaching_rms": 0.0027191995463910993, + "traffic_rms": 0.1505549825375809 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.142924118852534e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5464224443863883e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0064172513863293805, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.1410555468744702 + }, + { + "epoch": 83, + "eval_accuracy": 0.8796, + "eval_loss": 0.45452017211914064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13104906678199768, + "mean_feedback_forward_cosine": 0.9966188224879178, + "mean_feedback_forward_relative_error": 0.08089822991327807, + "min_feedback_forward_cosine": 0.9914987087249756 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002680861010178563, + "instruction_rms": 0.0026808610101687782, + "raw_apical_rms": 0.15046814068250927, + "teaching_rms": 0.002680861010178563, + "traffic_rms": 0.15044312746630847 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.218634496235609e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541083891368073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006708994528578518, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.14174490228758918 + }, + { + "epoch": 84, + "eval_accuracy": 0.8776, + "eval_loss": 0.43775721435546877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12613120675086975, + "mean_feedback_forward_cosine": 0.9968532724813981, + "mean_feedback_forward_relative_error": 0.0780410936948928, + "min_feedback_forward_cosine": 0.9921197891235352 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002686031433754745, + "instruction_rms": 0.0026860314337306635, + "raw_apical_rms": 0.150985361330108, + "teaching_rms": 0.002686031433754745, + "traffic_rms": 0.15096039520575474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.15800885839235e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5473791648249723e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006400952485962795, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.14036392624113295 + }, + { + "epoch": 85, + "eval_accuracy": 0.8952, + "eval_loss": 0.39643720703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12179127335548401, + "mean_feedback_forward_cosine": 0.9970596129244024, + "mean_feedback_forward_relative_error": 0.07543538016351786, + "min_feedback_forward_cosine": 0.9926533699035645 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002612883120352127, + "instruction_rms": 0.0026128831216062705, + "raw_apical_rms": 0.15135610358838467, + "teaching_rms": 0.002612883120352127, + "traffic_rms": 0.15133266316972901 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.039478250841057e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541004753195001e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006462480140216263, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.13974663161171808 + }, + { + "epoch": 86, + "eval_accuracy": 0.8948, + "eval_loss": 0.4052803955078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1174011081457138, + "mean_feedback_forward_cosine": 0.997258413921703, + "mean_feedback_forward_relative_error": 0.0728431259366599, + "min_feedback_forward_cosine": 0.993171751499176 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026489839610354844, + "instruction_rms": 0.0026489839610946936, + "raw_apical_rms": 0.15199564036918278, + "teaching_rms": 0.0026489839610354844, + "traffic_rms": 0.15197157663331626 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.204347778928422e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541520011626562e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006599853107754889, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.1359469562371572 + }, + { + "epoch": 87, + "eval_accuracy": 0.899, + "eval_loss": 0.35590679931640623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11299395561218262, + "mean_feedback_forward_cosine": 0.9974545836448669, + "mean_feedback_forward_relative_error": 0.07018705796111714, + "min_feedback_forward_cosine": 0.9936717748641968 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027426366580564406, + "instruction_rms": 0.002742636658924247, + "raw_apical_rms": 0.15216527045753545, + "teaching_rms": 0.0027426366580564406, + "traffic_rms": 0.15213959259253293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.208154784166254e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5491023232544303e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006589823912851985, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.1432716718620724 + }, + { + "epoch": 88, + "eval_accuracy": 0.8828, + "eval_loss": 0.44636400146484373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10919094830751419, + "mean_feedback_forward_cosine": 0.9976166118275035, + "mean_feedback_forward_relative_error": 0.06791295161978765, + "min_feedback_forward_cosine": 0.9940903782844543 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002568962072145562, + "instruction_rms": 0.002568962072070146, + "raw_apical_rms": 0.15291764332156205, + "teaching_rms": 0.002568962072145562, + "traffic_rms": 0.15289512250203224 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.191164286268759e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539541243301297e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006509948402681217, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.13177943440013462 + }, + { + "epoch": 89, + "eval_accuracy": 0.8844, + "eval_loss": 0.4189477661132813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10539348423480988, + "mean_feedback_forward_cosine": 0.9977782498706471, + "mean_feedback_forward_relative_error": 0.06556682921946048, + "min_feedback_forward_cosine": 0.9944922924041748 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026700113278570257, + "instruction_rms": 0.0026700113274014236, + "raw_apical_rms": 0.15335840010050059, + "teaching_rms": 0.0026700113278570257, + "traffic_rms": 0.15333401514771836 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.212346986307239e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5447497966286252e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006545180418373123, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.13200987011061774 + }, + { + "epoch": 90, + "eval_accuracy": 0.8722, + "eval_loss": 0.47785639038085936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10174806416034698, + "mean_feedback_forward_cosine": 0.9979229591109536, + "mean_feedback_forward_relative_error": 0.06339765506034548, + "min_feedback_forward_cosine": 0.9948654770851135 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025989833391281162, + "instruction_rms": 0.0025989833395016213, + "raw_apical_rms": 0.15412977703443345, + "teaching_rms": 0.0025989833391281162, + "traffic_rms": 0.15410691832897805 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.776132721597605e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5438698962622395e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006579018614455897, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.13670137802229987 + }, + { + "epoch": 91, + "eval_accuracy": 0.8918, + "eval_loss": 0.4050358825683594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09814438968896866, + "mean_feedback_forward_cosine": 0.998060323975303, + "mean_feedback_forward_relative_error": 0.061264423179355536, + "min_feedback_forward_cosine": 0.9952216148376465 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002605491879054624, + "instruction_rms": 0.0026054918786173427, + "raw_apical_rms": 0.15449997791814127, + "teaching_rms": 0.002605491879054624, + "traffic_rms": 0.15447700282163823 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.898832083379602e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5388043103855907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006240335922778175, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.1353029336505466 + }, + { + "epoch": 92, + "eval_accuracy": 0.8886, + "eval_loss": 0.41280927734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0946253091096878, + "mean_feedback_forward_cosine": 0.9981919646263122, + "mean_feedback_forward_relative_error": 0.059149937975135715, + "min_feedback_forward_cosine": 0.9955571889877319 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026225778220454157, + "instruction_rms": 0.0026225778222808077, + "raw_apical_rms": 0.1546422124371648, + "teaching_rms": 0.0026225778220454157, + "traffic_rms": 0.15461872216771375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.40321084435891e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5417959049635595e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006154486073687765, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.13231551391018762 + }, + { + "epoch": 93, + "eval_accuracy": 0.8864, + "eval_loss": 0.440756201171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09125298261642456, + "mean_feedback_forward_cosine": 0.998314782706174, + "mean_feedback_forward_relative_error": 0.05710795244032686, + "min_feedback_forward_cosine": 0.9958686828613281 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026295626089894945, + "instruction_rms": 0.0026295626091631798, + "raw_apical_rms": 0.15497635629214854, + "teaching_rms": 0.0026295626089894945, + "traffic_rms": 0.15495296632301486 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.132918173307189e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545857133853126e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0060287912510769, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.13290769317944845 + }, + { + "epoch": 94, + "eval_accuracy": 0.8788, + "eval_loss": 0.4635826843261719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08786870539188385, + "mean_feedback_forward_cosine": 0.9984281691637906, + "mean_feedback_forward_relative_error": 0.0551531354134733, + "min_feedback_forward_cosine": 0.9961684346199036 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026328994008423196, + "instruction_rms": 0.0026328994005010184, + "raw_apical_rms": 0.1553290348678801, + "teaching_rms": 0.0026328994008423196, + "traffic_rms": 0.15530565884898515 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.298514421132495e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5375114678351604e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006197718110068988, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.13446046793195937 + }, + { + "epoch": 95, + "eval_accuracy": 0.8726, + "eval_loss": 0.4548491027832031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08478537946939468, + "mean_feedback_forward_cosine": 0.998533051664179, + "mean_feedback_forward_relative_error": 0.053281935693865475, + "min_feedback_forward_cosine": 0.9964317083358765 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002580716866748122, + "instruction_rms": 0.0025807168673490683, + "raw_apical_rms": 0.15564059803859528, + "teaching_rms": 0.002580716866748122, + "traffic_rms": 0.15561814308990682 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.592698381131413e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.530378163367741e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006073440122871595, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.1281175126393636 + }, + { + "epoch": 96, + "eval_accuracy": 0.8854, + "eval_loss": 0.4196091125488281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0817914828658104, + "mean_feedback_forward_cosine": 0.9986320441419428, + "mean_feedback_forward_relative_error": 0.05145530422980135, + "min_feedback_forward_cosine": 0.9966788291931152 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002601581738691615, + "instruction_rms": 0.0026015817386345575, + "raw_apical_rms": 0.1561557088080724, + "teaching_rms": 0.002601581738691615, + "traffic_rms": 0.1561330561640619 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.298388737808637e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534340281688561e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0061203401449506445, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.1325615198334058 + }, + { + "epoch": 97, + "eval_accuracy": 0.8898, + "eval_loss": 0.38385213623046877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07893867790699005, + "mean_feedback_forward_cosine": 0.9987242817878723, + "mean_feedback_forward_relative_error": 0.04968901854008436, + "min_feedback_forward_cosine": 0.9969061017036438 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026407954787508905, + "instruction_rms": 0.00264079547831246, + "raw_apical_rms": 0.156541361440323, + "teaching_rms": 0.0026407954787508905, + "traffic_rms": 0.15651807622354477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.303297960310374e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534870101288664e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0059349221765024164, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.13027635041210386 + }, + { + "epoch": 98, + "eval_accuracy": 0.8778, + "eval_loss": 0.4473033386230469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07644806802272797, + "mean_feedback_forward_cosine": 0.998805943402377, + "mean_feedback_forward_relative_error": 0.048069786178794774, + "min_feedback_forward_cosine": 0.997097373008728 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002568411945691171, + "instruction_rms": 0.0025684119463123905, + "raw_apical_rms": 0.15654332971914975, + "teaching_rms": 0.002568411945691171, + "traffic_rms": 0.15652118693536912 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.238367689093138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360332500682992e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005915503647992677, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.12253387283749051 + }, + { + "epoch": 99, + "eval_accuracy": 0.8902, + "eval_loss": 0.39856865234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07376903295516968, + "mean_feedback_forward_cosine": 0.9988860639658841, + "mean_feedback_forward_relative_error": 0.046428847753188826, + "min_feedback_forward_cosine": 0.9972964525222778 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026522896668071372, + "instruction_rms": 0.0026522896669382012, + "raw_apical_rms": 0.15687143960009156, + "teaching_rms": 0.0026522896668071372, + "traffic_rms": 0.1568479856923433 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.49559711468828e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535734913191686e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005805982409407321, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.12809759814474317 + }, + { + "epoch": 100, + "eval_accuracy": 0.8754, + "eval_loss": 0.47369862060546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07100122421979904, + "mean_feedback_forward_cosine": 0.9989642349156467, + "mean_feedback_forward_relative_error": 0.044769683361730796, + "min_feedback_forward_cosine": 0.9974948167800903 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002654077659886959, + "instruction_rms": 0.0026540776599525647, + "raw_apical_rms": 0.15676781284296182, + "teaching_rms": 0.002654077659886959, + "traffic_rms": 0.15674415038878697 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.196233925948036e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53599456264671e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005805873848419814, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.13096666544874508 + }, + { + "epoch": 101, + "eval_accuracy": 0.9234, + "eval_loss": 0.27488249816894533, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0709770992398262, + "mean_feedback_forward_cosine": 0.9989649100737138, + "mean_feedback_forward_relative_error": 0.04475421941077167, + "min_feedback_forward_cosine": 0.9974966049194336 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018681300327095065, + "instruction_rms": 0.0018681300327021025, + "raw_apical_rms": 0.15710599554554727, + "teaching_rms": 0.0018681300327095065, + "traffic_rms": 0.15709385198047787 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.328763729219645e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541821079967624e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005917019782837112, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.07250560165014532 + }, + { + "epoch": 102, + "eval_accuracy": 0.9236, + "eval_loss": 0.27401832275390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07096787542104721, + "mean_feedback_forward_cosine": 0.9989651679992676, + "mean_feedback_forward_relative_error": 0.044747015613723885, + "min_feedback_forward_cosine": 0.9974972009658813 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015019077679723524, + "instruction_rms": 0.0015019077677623576, + "raw_apical_rms": 0.1570744741692994, + "teaching_rms": 0.0015019077679723524, + "traffic_rms": 0.1570666584012238 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.297803333334741e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534955971026108e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00589227916419364, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.04508370224568579 + }, + { + "epoch": 103, + "eval_accuracy": 0.9272, + "eval_loss": 0.2832565734863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07095963507890701, + "mean_feedback_forward_cosine": 0.9989653944969177, + "mean_feedback_forward_relative_error": 0.04474038678136739, + "min_feedback_forward_cosine": 0.9974977970123291 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001402311703568493, + "instruction_rms": 0.001402311703717671, + "raw_apical_rms": 0.1571234752168091, + "teaching_rms": 0.001402311703568493, + "traffic_rms": 0.15711630915522729 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 6.596710733219879e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360707500546335e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005898395072908181, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.03774840233557754 + }, + { + "epoch": 104, + "eval_accuracy": 0.925, + "eval_loss": 0.2829547271728516, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07095282524824142, + "mean_feedback_forward_cosine": 0.9989656166596846, + "mean_feedback_forward_relative_error": 0.04473473756489429, + "min_feedback_forward_cosine": 0.9974982738494873 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012750670215814861, + "instruction_rms": 0.0012750670214078238, + "raw_apical_rms": 0.1570559243132205, + "teaching_rms": 0.0012750670215814861, + "traffic_rms": 0.15704995341997907 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.518710193859079e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535662505794863e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00590715919836829, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.030839167262448206 + }, + { + "epoch": 105, + "eval_accuracy": 0.9252, + "eval_loss": 0.2876938934326172, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07094645500183105, + "mean_feedback_forward_cosine": 0.9989657987247814, + "mean_feedback_forward_relative_error": 0.04472943408922716, + "min_feedback_forward_cosine": 0.9974987506866455 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012315131502262538, + "instruction_rms": 0.0012315131500998835, + "raw_apical_rms": 0.15701118973065295, + "teaching_rms": 0.0012315131502262538, + "traffic_rms": 0.1570055661357571 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.324130990582489e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538040215241406e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005888449746465611, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.027694817138049336 + }, + { + "epoch": 106, + "eval_accuracy": 0.9274, + "eval_loss": 0.28754229736328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07094030827283859, + "mean_feedback_forward_cosine": 0.9989660057154569, + "mean_feedback_forward_relative_error": 0.044724249264056035, + "min_feedback_forward_cosine": 0.9974992275238037 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012348020577495457, + "instruction_rms": 0.0012348020577501633, + "raw_apical_rms": 0.1569274317159094, + "teaching_rms": 0.0012348020577495457, + "traffic_rms": 0.15692170682291784 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.317851213883841e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538755021751766e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005776124519394298, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.025074613791704177 + }, + { + "epoch": 107, + "eval_accuracy": 0.9274, + "eval_loss": 0.28696969299316405, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.070933997631073, + "mean_feedback_forward_cosine": 0.9989662062038075, + "mean_feedback_forward_relative_error": 0.04471908270974051, + "min_feedback_forward_cosine": 0.9974997043609619 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012253141640255732, + "instruction_rms": 0.0012253141635396715, + "raw_apical_rms": 0.1568674549808664, + "teaching_rms": 0.0012253141640255732, + "traffic_rms": 0.15686162558067718 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.282807246156153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536815027355969e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058119740785779524, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.023852526133590273 + }, + { + "epoch": 108, + "eval_accuracy": 0.9278, + "eval_loss": 0.2932923095703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07092788070440292, + "mean_feedback_forward_cosine": 0.9989663926037875, + "mean_feedback_forward_relative_error": 0.044714076359840954, + "min_feedback_forward_cosine": 0.9975001215934753 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001164947536507392, + "instruction_rms": 0.0011649475362220188, + "raw_apical_rms": 0.15685390294079102, + "teaching_rms": 0.001164947536507392, + "traffic_rms": 0.15684856122585833 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.801549000736486e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537907331784123e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005743717111929405, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.02110563539125853 + }, + { + "epoch": 109, + "eval_accuracy": 0.9278, + "eval_loss": 0.3009987487792969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07092272490262985, + "mean_feedback_forward_cosine": 0.9989665714177218, + "mean_feedback_forward_relative_error": 0.04470948527821086, + "min_feedback_forward_cosine": 0.9975005388259888 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010989609109914053, + "instruction_rms": 0.0010989609108569018, + "raw_apical_rms": 0.15679209054773383, + "teaching_rms": 0.0010989609109914053, + "traffic_rms": 0.15678733926444163 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.114888944670383e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355600671949557e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005833617276694973, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.018469695284962655 + }, + { + "epoch": 110, + "eval_accuracy": 0.9322, + "eval_loss": 0.2989709991455078, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07091721147298813, + "mean_feedback_forward_cosine": 0.9989667556502603, + "mean_feedback_forward_relative_error": 0.0447049082680182, + "min_feedback_forward_cosine": 0.9975008964538574 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011106760717449825, + "instruction_rms": 0.001110676071854277, + "raw_apical_rms": 0.15674526207655107, + "teaching_rms": 0.0011106760717449825, + "traffic_rms": 0.15674021156413928 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.926139661800335e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5407929228254788e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005785295681242412, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.017986075497666995 + }, + { + "epoch": 111, + "eval_accuracy": 0.9312, + "eval_loss": 0.2975237548828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07091204077005386, + "mean_feedback_forward_cosine": 0.9989669182083823, + "mean_feedback_forward_relative_error": 0.04470046184618365, + "min_feedback_forward_cosine": 0.9975012540817261 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010458979697199353, + "instruction_rms": 0.0010458979695164385, + "raw_apical_rms": 0.15675384503128525, + "teaching_rms": 0.0010458979697199353, + "traffic_rms": 0.15674930876130302 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.771234239977275e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5387045144860748e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005776015332794581, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.01620114773346318 + }, + { + "epoch": 112, + "eval_accuracy": 0.9324, + "eval_loss": 0.3009730224609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07090695202350616, + "mean_feedback_forward_cosine": 0.9989670731804587, + "mean_feedback_forward_relative_error": 0.04469619986008514, + "min_feedback_forward_cosine": 0.9975014925003052 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010423431884891887, + "instruction_rms": 0.001042343188609069, + "raw_apical_rms": 0.1567646603551989, + "teaching_rms": 0.0010423431884891887, + "traffic_rms": 0.15676002837858788 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 6.636144433969093e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5359185499159353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00577949730688871, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.01542122282071246 + }, + { + "epoch": 113, + "eval_accuracy": 0.9304, + "eval_loss": 0.3087217346191406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07090183347463608, + "mean_feedback_forward_cosine": 0.9989672541618347, + "mean_feedback_forward_relative_error": 0.04469187412072312, + "min_feedback_forward_cosine": 0.9975018501281738 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010069105057670844, + "instruction_rms": 0.0010069105051730399, + "raw_apical_rms": 0.15674903436380308, + "teaching_rms": 0.0010069105057670844, + "traffic_rms": 0.15674476255955913 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.054693185182259e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5373920009531483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005810685867104819, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.014235611286262672 + }, + { + "epoch": 114, + "eval_accuracy": 0.931, + "eval_loss": 0.314695849609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07089726626873016, + "mean_feedback_forward_cosine": 0.9989674113013528, + "mean_feedback_forward_relative_error": 0.044688034074550326, + "min_feedback_forward_cosine": 0.9975022077560425 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009729803436573065, + "instruction_rms": 0.000972980343533662, + "raw_apical_rms": 0.15672653995108682, + "teaching_rms": 0.0009729803436573065, + "traffic_rms": 0.1567224886940586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.398914275176139e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538041464532142e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005762569430221804, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.012990544721484185 + }, + { + "epoch": 115, + "eval_accuracy": 0.9284, + "eval_loss": 0.3111776794433594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07089243829250336, + "mean_feedback_forward_cosine": 0.9989675565199418, + "mean_feedback_forward_relative_error": 0.044684005274691366, + "min_feedback_forward_cosine": 0.9975023865699768 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009499882285774866, + "instruction_rms": 0.0009499882290781333, + "raw_apical_rms": 0.15675961693258456, + "teaching_rms": 0.0009499882285774866, + "traffic_rms": 0.15675556098779417 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.990103346675824e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5365512890623903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005769194975041045, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.012409936531053649 + }, + { + "epoch": 116, + "eval_accuracy": 0.9318, + "eval_loss": 0.3148053466796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07088766992092133, + "mean_feedback_forward_cosine": 0.9989677049896933, + "mean_feedback_forward_relative_error": 0.04467999024147337, + "min_feedback_forward_cosine": 0.997502863407135 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009559801532536852, + "instruction_rms": 0.0009559801530943921, + "raw_apical_rms": 0.15673376741355344, + "teaching_rms": 0.0009559801532536852, + "traffic_rms": 0.1567297328532247 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.778653809964453e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5335779749294922e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005744965806277495, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.012501899681488673 + }, + { + "epoch": 117, + "eval_accuracy": 0.9322, + "eval_loss": 0.3190792907714844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07088316231966019, + "mean_feedback_forward_cosine": 0.9989678827199069, + "mean_feedback_forward_relative_error": 0.04467604162002152, + "min_feedback_forward_cosine": 0.9975031018257141 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009494193280651492, + "instruction_rms": 0.0009494193279696085, + "raw_apical_rms": 0.15675959832923328, + "teaching_rms": 0.0009494193280651492, + "traffic_rms": 0.15675568312690336 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.358942528095213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.533411126948754e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005877179956535044, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.011608204885323842 + }, + { + "epoch": 118, + "eval_accuracy": 0.9298, + "eval_loss": 0.32249583129882814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087850570678711, + "mean_feedback_forward_cosine": 0.9989680203524502, + "mean_feedback_forward_relative_error": 0.044672052206640896, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009212186690548354, + "instruction_rms": 0.0009212186691952662, + "raw_apical_rms": 0.1567560234107983, + "teaching_rms": 0.0009212186690548354, + "traffic_rms": 0.15675203733659918 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 8.317160649085054e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5406398805912062e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005816009318486093, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.011224714832069974 + }, + { + "epoch": 119, + "eval_accuracy": 0.932, + "eval_loss": 0.32551557006835935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087486982345581, + "mean_feedback_forward_cosine": 0.9989681460640647, + "mean_feedback_forward_relative_error": 0.0446686068888415, + "min_feedback_forward_cosine": 0.9975037574768066 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008470678397713774, + "instruction_rms": 0.0008470678396755314, + "raw_apical_rms": 0.15666118073060328, + "teaching_rms": 0.0008470678397713774, + "traffic_rms": 0.15665776704803078 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.599363354733214e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541024069794903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005867086167817335, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.009133812613785267 + }, + { + "epoch": 120, + "eval_accuracy": 0.931, + "eval_loss": 0.3299600830078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087042182683945, + "mean_feedback_forward_cosine": 0.9989683086221869, + "mean_feedback_forward_relative_error": 0.044664719616147604, + "min_feedback_forward_cosine": 0.9975039958953857 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008949044177462138, + "instruction_rms": 0.0008949044175458051, + "raw_apical_rms": 0.15663499626042995, + "teaching_rms": 0.0008949044177462138, + "traffic_rms": 0.15663112010010294 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.459373480538488e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5368161104063114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005843225887937226, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.009784037462208007 + }, + { + "epoch": 121, + "eval_accuracy": 0.9292, + "eval_loss": 0.32630562744140623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07086643576622009, + "mean_feedback_forward_cosine": 0.9989684527570551, + "mean_feedback_forward_relative_error": 0.044661184095523575, + "min_feedback_forward_cosine": 0.9975043535232544 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008145382890291357, + "instruction_rms": 0.0008145382889972386, + "raw_apical_rms": 0.1566612766499781, + "teaching_rms": 0.0008145382890291357, + "traffic_rms": 0.15665788889761117 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 6.135365993031883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5382620985560022e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005812800528565876, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.009073666857431332 + }, + { + "epoch": 122, + "eval_accuracy": 0.9302, + "eval_loss": 0.32952420654296877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07086223363876343, + "mean_feedback_forward_cosine": 0.9989685979756442, + "mean_feedback_forward_relative_error": 0.044657568464224986, + "min_feedback_forward_cosine": 0.9975045919418335 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000828141355639425, + "instruction_rms": 0.0008281413559018742, + "raw_apical_rms": 0.15664252699002104, + "teaching_rms": 0.000828141355639425, + "traffic_rms": 0.15663918014748687 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 5.9446421118991566e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5356018937259457e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005891065101916085, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.009004947647555835 + }, + { + "epoch": 123, + "eval_accuracy": 0.9304, + "eval_loss": 0.33173782958984377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07085803151130676, + "mean_feedback_forward_cosine": 0.9989687345244668, + "mean_feedback_forward_relative_error": 0.044654008221219885, + "min_feedback_forward_cosine": 0.9975050091743469 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008190901937503409, + "instruction_rms": 0.0008190901936549919, + "raw_apical_rms": 0.15661127665436178, + "teaching_rms": 0.0008190901937503409, + "traffic_rms": 0.15660783013726773 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 7.179834682347064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539510267280735e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005910763090338856, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.0088103628065851 + }, + { + "epoch": 124, + "eval_accuracy": 0.9312, + "eval_loss": 0.33396995239257815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07085384428501129, + "mean_feedback_forward_cosine": 0.9989688656546852, + "mean_feedback_forward_relative_error": 0.044650416922840205, + "min_feedback_forward_cosine": 0.9975050687789917 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008038917385117785, + "instruction_rms": 0.0008038917382657697, + "raw_apical_rms": 0.15659737569674384, + "teaching_rms": 0.0008038917385117785, + "traffic_rms": 0.1565940664957959 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.250633788340565e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5383427378892843e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005881761370236408, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.00854498468455341 + }, + { + "epoch": 125, + "eval_accuracy": 0.9296, + "eval_loss": 0.33593333740234377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07084941118955612, + "mean_feedback_forward_cosine": 0.9989690412174571, + "mean_feedback_forward_relative_error": 0.044646709530868314, + "min_feedback_forward_cosine": 0.9975054264068604 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000820239429463051, + "instruction_rms": 0.0008202394295456283, + "raw_apical_rms": 0.1565658226530157, + "teaching_rms": 0.000820239429463051, + "traffic_rms": 0.15656242031847997 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.694543313618851e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5364542856645387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005836558479366115, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.008565741331958109 + }, + { + "epoch": 126, + "eval_accuracy": 0.93, + "eval_loss": 0.3363638122558594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708455815911293, + "mean_feedback_forward_cosine": 0.9989691755988381, + "mean_feedback_forward_relative_error": 0.044643311642787675, + "min_feedback_forward_cosine": 0.9975056648254395 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007721182285074628, + "instruction_rms": 0.0007721182286492539, + "raw_apical_rms": 0.15655390561368063, + "teaching_rms": 0.0007721182285074628, + "traffic_rms": 0.1565506457901433 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.784327410969127e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5363067740951922e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005898069685332526, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.007548671312464608 + }, + { + "epoch": 127, + "eval_accuracy": 0.9322, + "eval_loss": 0.33970142822265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07084105908870697, + "mean_feedback_forward_cosine": 0.998969309980219, + "mean_feedback_forward_relative_error": 0.04463963938707655, + "min_feedback_forward_cosine": 0.9975060224533081 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007952847310662166, + "instruction_rms": 0.000795284731119907, + "raw_apical_rms": 0.15649639667063278, + "teaching_rms": 0.0007952847310662166, + "traffic_rms": 0.1564932482741599 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.252860712014808e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5363136251838993e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005954777612223128, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.007928385191245212 + }, + { + "epoch": 128, + "eval_accuracy": 0.93, + "eval_loss": 0.34044767456054686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07083731144666672, + "mean_feedback_forward_cosine": 0.998969432440671, + "mean_feedback_forward_relative_error": 0.044636607339436356, + "min_feedback_forward_cosine": 0.9975062608718872 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006701769134514322, + "instruction_rms": 0.0006701769134532176, + "raw_apical_rms": 0.1564775455340859, + "teaching_rms": 0.0006701769134514322, + "traffic_rms": 0.15647496268213462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.703119481699105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5411468270352428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005862443182673322, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.006844878549252947 + }, + { + "epoch": 129, + "eval_accuracy": 0.9308, + "eval_loss": 0.3454027954101562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07083291560411453, + "mean_feedback_forward_cosine": 0.9989695819941434, + "mean_feedback_forward_relative_error": 0.04463285764848644, + "min_feedback_forward_cosine": 0.9975066781044006 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008105760127383864, + "instruction_rms": 0.0008105760127973803, + "raw_apical_rms": 0.15649128965338405, + "teaching_rms": 0.0008105760127383864, + "traffic_rms": 0.15648792415711557 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.919364636814862e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5357132180343736e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005847970109920921, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.00751388033653299 + }, + { + "epoch": 130, + "eval_accuracy": 0.9312, + "eval_loss": 0.3445435241699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07082948088645935, + "mean_feedback_forward_cosine": 0.9989696947011081, + "mean_feedback_forward_relative_error": 0.044630019857802176, + "min_feedback_forward_cosine": 0.997506856918335 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006418838499680199, + "instruction_rms": 0.0006418838497189099, + "raw_apical_rms": 0.15642447055607922, + "teaching_rms": 0.0006418838499680199, + "traffic_rms": 0.15642219504731347 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.848455586487944e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5415882295120003e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00581332910731474, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.005915861031826999 + }, + { + "epoch": 131, + "eval_accuracy": 0.9312, + "eval_loss": 0.3421359436035156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07082522660493851, + "mean_feedback_forward_cosine": 0.9989698323336514, + "mean_feedback_forward_relative_error": 0.04462645375593142, + "min_feedback_forward_cosine": 0.9975071549415588 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000737686763895328, + "instruction_rms": 0.000737686764102491, + "raw_apical_rms": 0.15641385282290396, + "teaching_rms": 0.000737686763895328, + "traffic_rms": 0.15641073588207244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 8.277032037540266e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5418303191121762e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005896190408225254, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.006755238181166351 + }, + { + "epoch": 132, + "eval_accuracy": 0.9308, + "eval_loss": 0.3458807006835937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07082144916057587, + "mean_feedback_forward_cosine": 0.9989699743010781, + "mean_feedback_forward_relative_error": 0.0446231294084679, + "min_feedback_forward_cosine": 0.9975073933601379 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007122230452864486, + "instruction_rms": 0.000712223045463341, + "raw_apical_rms": 0.15639152371594156, + "teaching_rms": 0.0007122230452864486, + "traffic_rms": 0.1563886586829158 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.187219352999819e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5373814648263543e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005826939346422026, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.006470164734808107 + }, + { + "epoch": 133, + "eval_accuracy": 0.9326, + "eval_loss": 0.34408472900390624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07081776857376099, + "mean_feedback_forward_cosine": 0.9989700924266468, + "mean_feedback_forward_relative_error": 0.04462002458220178, + "min_feedback_forward_cosine": 0.997507631778717 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006537943374323771, + "instruction_rms": 0.0006537943376771356, + "raw_apical_rms": 0.1563871952729514, + "teaching_rms": 0.0006537943374323771, + "traffic_rms": 0.15638462368018344 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.090667336091428e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536691766808993e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005813838483222319, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.005801812709122896 + }, + { + "epoch": 134, + "eval_accuracy": 0.9306, + "eval_loss": 0.3515958251953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07081308215856552, + "mean_feedback_forward_cosine": 0.9989702679894187, + "mean_feedback_forward_relative_error": 0.044616280546919865, + "min_feedback_forward_cosine": 0.9975079298019409 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007491625438862237, + "instruction_rms": 0.0007491625436995141, + "raw_apical_rms": 0.15636806297910094, + "teaching_rms": 0.0007491625438862237, + "traffic_rms": 0.15636475892267998 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.767201992692208e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5358320999069396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005863827401303301, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.006816023340572914 + }, + { + "epoch": 135, + "eval_accuracy": 0.9322, + "eval_loss": 0.3474887512207031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708087608218193, + "mean_feedback_forward_cosine": 0.9989703969521956, + "mean_feedback_forward_relative_error": 0.04461274790492925, + "min_feedback_forward_cosine": 0.9975084066390991 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007182916366688837, + "instruction_rms": 0.000718291636966421, + "raw_apical_rms": 0.15639258387918176, + "teaching_rms": 0.0007182916366688837, + "traffic_rms": 0.15638971540673072 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.585887746268781e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5351030459360198e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005775043328720412, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.00620322556015518 + }, + { + "epoch": 136, + "eval_accuracy": 0.93, + "eval_loss": 0.3605854431152344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708053857088089, + "mean_feedback_forward_cosine": 0.9989705400033431, + "mean_feedback_forward_relative_error": 0.044609861262142655, + "min_feedback_forward_cosine": 0.9975085258483887 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006322296166297042, + "instruction_rms": 0.0006322296167166218, + "raw_apical_rms": 0.1564355575673448, + "teaching_rms": 0.0006322296166297042, + "traffic_rms": 0.15643309830365237 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.498969267771827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537036167272147e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005937657389980397, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.00543699383230673 + }, + { + "epoch": 137, + "eval_accuracy": 0.9328, + "eval_loss": 0.35638165893554685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07080122083425522, + "mean_feedback_forward_cosine": 0.9989706635475158, + "mean_feedback_forward_relative_error": 0.0446064278991385, + "min_feedback_forward_cosine": 0.9975087642669678 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006742951027451404, + "instruction_rms": 0.0006742951026956253, + "raw_apical_rms": 0.156422869061531, + "teaching_rms": 0.0006742951027451404, + "traffic_rms": 0.15642013032309593 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.080841868400967e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5443088115768734e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0059150798739307595, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.005787847943024503 + }, + { + "epoch": 138, + "eval_accuracy": 0.93, + "eval_loss": 0.36132265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707966685295105, + "mean_feedback_forward_cosine": 0.9989708206870339, + "mean_feedback_forward_relative_error": 0.04460295835001902, + "min_feedback_forward_cosine": 0.9975091218948364 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006623774581377063, + "instruction_rms": 0.0006623774580451413, + "raw_apical_rms": 0.15636598707233249, + "teaching_rms": 0.0006623774581377063, + "traffic_rms": 0.1563632726246208 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.556190612627688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536713059570718e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005861925669562594, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.005622351417938868 + }, + { + "epoch": 139, + "eval_accuracy": 0.932, + "eval_loss": 0.35857147216796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07079344242811203, + "mean_feedback_forward_cosine": 0.9989709236405112, + "mean_feedback_forward_relative_error": 0.04460024108940905, + "min_feedback_forward_cosine": 0.9975094199180603 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005859143428559092, + "instruction_rms": 0.0005859143429779436, + "raw_apical_rms": 0.15636178605641754, + "teaching_rms": 0.0005859143428559092, + "traffic_rms": 0.15635984807257855 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.712134302868435e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534403999828429e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005829575657965549, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.004863424326479435 + }, + { + "epoch": 140, + "eval_accuracy": 0.931, + "eval_loss": 0.36219781494140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07078968733549118, + "mean_feedback_forward_cosine": 0.9989710688591004, + "mean_feedback_forward_relative_error": 0.044596916471015324, + "min_feedback_forward_cosine": 0.9975095987319946 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006475315319334628, + "instruction_rms": 0.0006475315319258127, + "raw_apical_rms": 0.15632147733190216, + "teaching_rms": 0.0006475315319334628, + "traffic_rms": 0.15631878209154426 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.692059173474263e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5363117901857483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005813491649850502, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.005354508005910449 + }, + { + "epoch": 141, + "eval_accuracy": 0.9308, + "eval_loss": 0.35549102783203124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07078536599874496, + "mean_feedback_forward_cosine": 0.998971240086989, + "mean_feedback_forward_relative_error": 0.04459290558641607, + "min_feedback_forward_cosine": 0.9975098967552185 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007365572518808323, + "instruction_rms": 0.0007365572518513475, + "raw_apical_rms": 0.15628183672241677, + "teaching_rms": 0.0007365572518808323, + "traffic_rms": 0.15627873155211572 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.233802534756251e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5344348896844735e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005786564292966944, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.005712501531182271 + }, + { + "epoch": 142, + "eval_accuracy": 0.9328, + "eval_loss": 0.3628958251953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07078168541193008, + "mean_feedback_forward_cosine": 0.9989713582125577, + "mean_feedback_forward_relative_error": 0.044589934091676366, + "min_feedback_forward_cosine": 0.9975100755691528 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005875167950028665, + "instruction_rms": 0.0005875167950690892, + "raw_apical_rms": 0.15631119822140535, + "teaching_rms": 0.0005875167950028665, + "traffic_rms": 0.1563089432548003 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.872849584420692e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5394041854233672e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058716752505594254, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.004544817179813981 + }, + { + "epoch": 143, + "eval_accuracy": 0.9324, + "eval_loss": 0.35895855712890623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07077816128730774, + "mean_feedback_forward_cosine": 0.9989714720032432, + "mean_feedback_forward_relative_error": 0.04458676227791743, + "min_feedback_forward_cosine": 0.9975103735923767 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006086106565886836, + "instruction_rms": 0.0006086106570433067, + "raw_apical_rms": 0.15632151175090112, + "teaching_rms": 0.0006086106565886836, + "traffic_rms": 0.15631922770924686 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.099928550109325e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5343036354917024e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005783009342741666, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.004699092442211178 + }, + { + "epoch": 144, + "eval_accuracy": 0.9296, + "eval_loss": 0.36100722045898437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07077439874410629, + "mean_feedback_forward_cosine": 0.9989716204729947, + "mean_feedback_forward_relative_error": 0.044583823006938805, + "min_feedback_forward_cosine": 0.9975106716156006 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005788136725466557, + "instruction_rms": 0.0005788136728108593, + "raw_apical_rms": 0.15631840573988948, + "teaching_rms": 0.0005788136725466557, + "traffic_rms": 0.15631622774328052 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.939645740862034e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5343369875683058e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058867044920564516, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.004655390022281144 + }, + { + "epoch": 145, + "eval_accuracy": 0.9302, + "eval_loss": 0.3650362976074219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07077116519212723, + "mean_feedback_forward_cosine": 0.9989717201753097, + "mean_feedback_forward_relative_error": 0.04458092269911008, + "min_feedback_forward_cosine": 0.9975109100341797 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005656258353966133, + "instruction_rms": 0.0005656258354606748, + "raw_apical_rms": 0.15630051021225427, + "teaching_rms": 0.0005656258353966133, + "traffic_rms": 0.15629844015168792 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.635451654801727e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5341038056365583e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0057506056176664455, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.004451047736489111 + }, + { + "epoch": 146, + "eval_accuracy": 0.9322, + "eval_loss": 0.3686236083984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07076733559370041, + "mean_feedback_forward_cosine": 0.9989718556404114, + "mean_feedback_forward_relative_error": 0.04457757064903324, + "min_feedback_forward_cosine": 0.9975112676620483 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006118507959450226, + "instruction_rms": 0.0006118507956350694, + "raw_apical_rms": 0.15629460654171135, + "teaching_rms": 0.0006118507959450226, + "traffic_rms": 0.15629206355538705 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.179304617466187e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355107701321243e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005792362550473921, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.004486601050115294 + }, + { + "epoch": 147, + "eval_accuracy": 0.9312, + "eval_loss": 0.3708784484863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07076365500688553, + "mean_feedback_forward_cosine": 0.9989719791845841, + "mean_feedback_forward_relative_error": 0.044574448466300966, + "min_feedback_forward_cosine": 0.9975113868713379 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005917439001439649, + "instruction_rms": 0.0005917439003842913, + "raw_apical_rms": 0.1562591996962405, + "teaching_rms": 0.0005917439001439649, + "traffic_rms": 0.15625692916408343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.0164442443283406e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5339368077170327e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005743765676402482, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.004553148381991519 + }, + { + "epoch": 148, + "eval_accuracy": 0.933, + "eval_loss": 0.35687979736328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075996696949005, + "mean_feedback_forward_cosine": 0.9989721233194525, + "mean_feedback_forward_relative_error": 0.04457118883728981, + "min_feedback_forward_cosine": 0.9975116848945618 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000569190896028838, + "instruction_rms": 0.0005691908960499179, + "raw_apical_rms": 0.15621313399809966, + "teaching_rms": 0.000569190896028838, + "traffic_rms": 0.15621085011549987 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.872203073271521e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5366421053637076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811858511124198, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.004216510014779245 + }, + { + "epoch": 149, + "eval_accuracy": 0.9332, + "eval_loss": 0.36613702392578124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707562193274498, + "mean_feedback_forward_cosine": 0.9989722728729248, + "mean_feedback_forward_relative_error": 0.04456792431460185, + "min_feedback_forward_cosine": 0.9975119233131409 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005833205318357344, + "instruction_rms": 0.0005833205316495131, + "raw_apical_rms": 0.15622415389041572, + "teaching_rms": 0.0005833205318357344, + "traffic_rms": 0.15622189222459812 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.04262541009848e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5417710321320986e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811621761220833, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.004337725640171104 + }, + { + "epoch": 150, + "eval_accuracy": 0.9306, + "eval_loss": 0.36401038818359377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075214385986328, + "mean_feedback_forward_cosine": 0.998972417007793, + "mean_feedback_forward_relative_error": 0.04456454075195573, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005859148486319701, + "instruction_rms": 0.0005859148484117649, + "raw_apical_rms": 0.15617846671077237, + "teaching_rms": 0.0005859148486319701, + "traffic_rms": 0.15617611209861504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.601874247280648e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355904813225256e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005827398066955642, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.004388301985752251 + }, + { + "epoch": 151, + "eval_accuracy": 0.9318, + "eval_loss": 0.36422869873046876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707520842552185, + "mean_feedback_forward_cosine": 0.9989724159240723, + "mean_feedback_forward_relative_error": 0.044564471613954416, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005565280756125213, + "instruction_rms": 0.0005565280757680321, + "raw_apical_rms": 0.15618927699685, + "teaching_rms": 0.0005565280756125213, + "traffic_rms": 0.156187117889629 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.69225741489754e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5346876037915427e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005892246531822571, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.003987497380210293 + }, + { + "epoch": 152, + "eval_accuracy": 0.9332, + "eval_loss": 0.36020493774414064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075206190347672, + "mean_feedback_forward_cosine": 0.9989724224263972, + "mean_feedback_forward_relative_error": 0.044564413160763004, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005095751335896143, + "instruction_rms": 0.0005095751334227084, + "raw_apical_rms": 0.15618482505259446, + "teaching_rms": 0.0005095751335896143, + "traffic_rms": 0.15618299150455436 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.206531111274671e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539423829837073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005756206280104596, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.0036039010546687574 + }, + { + "epoch": 153, + "eval_accuracy": 0.9316, + "eval_loss": 0.3657336242675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075202465057373, + "mean_feedback_forward_cosine": 0.9989724180915139, + "mean_feedback_forward_relative_error": 0.044564356282353404, + "min_feedback_forward_cosine": 0.99751216173172 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005144573056545982, + "instruction_rms": 0.0005144573057573081, + "raw_apical_rms": 0.15618785228516027, + "teaching_rms": 0.0005144573056545982, + "traffic_rms": 0.15618585031669577 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.958780846526679e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5397136800678003e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005782406628060175, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.003619852273642189 + }, + { + "epoch": 154, + "eval_accuracy": 0.9312, + "eval_loss": 0.361766259765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075200229883194, + "mean_feedback_forward_cosine": 0.9989724180915139, + "mean_feedback_forward_relative_error": 0.04456430270590565, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047462126330647184, + "instruction_rms": 0.00047462126314686654, + "raw_apical_rms": 0.1561874935552836, + "teaching_rms": 0.00047462126330647184, + "traffic_rms": 0.15618592888199648 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.252059930351606e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5359469216651334e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005833945475461095, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.003418941561298238 + }, + { + "epoch": 155, + "eval_accuracy": 0.9322, + "eval_loss": 0.3601679626464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075197249650955, + "mean_feedback_forward_cosine": 0.9989724332636053, + "mean_feedback_forward_relative_error": 0.044564251889559356, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004348900655850833, + "instruction_rms": 0.00043489006562027013, + "raw_apical_rms": 0.15618763321450302, + "teaching_rms": 0.0004348900655850833, + "traffic_rms": 0.15618636070868136 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.102023857896711e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360289173569167e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005871366295406924, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.0030920101912381747 + }, + { + "epoch": 156, + "eval_accuracy": 0.9334, + "eval_loss": 0.36134964599609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075195014476776, + "mean_feedback_forward_cosine": 0.9989724256775596, + "mean_feedback_forward_relative_error": 0.04456419878723947, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004496863362942985, + "instruction_rms": 0.000449686335963409, + "raw_apical_rms": 0.15618638196317103, + "teaching_rms": 0.0004496863362942985, + "traffic_rms": 0.15618490617362066 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.387681139585766e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5328539620243115e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005807898817690802, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.0033026031192392113 + }, + { + "epoch": 157, + "eval_accuracy": 0.933, + "eval_loss": 0.36484820556640624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075192034244537, + "mean_feedback_forward_cosine": 0.9989724191752347, + "mean_feedback_forward_relative_error": 0.04456414290788499, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048817587151015504, + "instruction_rms": 0.0004881758713967517, + "raw_apical_rms": 0.15619299370344406, + "teaching_rms": 0.00048817587151015504, + "traffic_rms": 0.1561914335804525 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.844036448683255e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5339947055639765e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058057200265357834, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.0035868339443372354 + }, + { + "epoch": 158, + "eval_accuracy": 0.9332, + "eval_loss": 0.3618963806152344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075189054012299, + "mean_feedback_forward_cosine": 0.9989724245938387, + "mean_feedback_forward_relative_error": 0.044564094834707, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042458779175761664, + "instruction_rms": 0.00042458779165763, + "raw_apical_rms": 0.15619627830489188, + "teaching_rms": 0.00042458779175761664, + "traffic_rms": 0.1561949235736182 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.991372458742262e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536304876525867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058730891637487695, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0031154596315541615 + }, + { + "epoch": 159, + "eval_accuracy": 0.933, + "eval_loss": 0.3608845458984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707518681883812, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.04456403861668977, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004903175973574765, + "instruction_rms": 0.0004903175975856101, + "raw_apical_rms": 0.15619783319341446, + "teaching_rms": 0.0004903175973574765, + "traffic_rms": 0.15619598374444096 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.862629385930632e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5349964498701008e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005771215163845155, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.003495788243899329 + }, + { + "epoch": 160, + "eval_accuracy": 0.9338, + "eval_loss": 0.36179160766601565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075183838605881, + "mean_feedback_forward_cosine": 0.9989724289287221, + "mean_feedback_forward_relative_error": 0.04456398204307665, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004912375217397902, + "instruction_rms": 0.000491237521712185, + "raw_apical_rms": 0.15619367231437742, + "teaching_rms": 0.0004912375217397902, + "traffic_rms": 0.15619195051556656 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.83362699760437e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5367519236755603e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005842472803512416, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.003658046741928491 + }, + { + "epoch": 161, + "eval_accuracy": 0.9332, + "eval_loss": 0.36055188598632815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075180858373642, + "mean_feedback_forward_cosine": 0.9989724256775596, + "mean_feedback_forward_relative_error": 0.04456392541866411, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004898777905411831, + "instruction_rms": 0.0004898777904168807, + "raw_apical_rms": 0.15619185592653934, + "teaching_rms": 0.0004898777905411831, + "traffic_rms": 0.15619014678894416 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.207485370168797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5398887761782835e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811061620240841, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.003570969367016935 + }, + { + "epoch": 162, + "eval_accuracy": 0.933, + "eval_loss": 0.35955615844726563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075179368257523, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.04456387668509375, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00039760659487526136, + "instruction_rms": 0.0003976065949684868, + "raw_apical_rms": 0.1561934914480541, + "teaching_rms": 0.00039760659487526136, + "traffic_rms": 0.15619229634519607 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.611171616237698e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5381311905987076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005755785324461161, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.0027860320918675924 + }, + { + "epoch": 163, + "eval_accuracy": 0.9334, + "eval_loss": 0.3592337158203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075176388025284, + "mean_feedback_forward_cosine": 0.9989724386822093, + "mean_feedback_forward_relative_error": 0.044563823074779726, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004863620929113679, + "instruction_rms": 0.00048636209252903706, + "raw_apical_rms": 0.1561965802615731, + "teaching_rms": 0.0004863620929113679, + "traffic_rms": 0.1561948537767751 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.055869133409942e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5380353651053233e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00577728433299945, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.0035860527916759667 + }, + { + "epoch": 164, + "eval_accuracy": 0.9328, + "eval_loss": 0.3621196960449219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075174152851105, + "mean_feedback_forward_cosine": 0.9989724430170926, + "mean_feedback_forward_relative_error": 0.044563774375075645, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004124481990639766, + "instruction_rms": 0.0004124481989203935, + "raw_apical_rms": 0.1561922838465203, + "teaching_rms": 0.0004124481990639766, + "traffic_rms": 0.1561910573164225 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.283202473397978e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536488139497506e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00579818694022861, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.0028164971362799407 + }, + { + "epoch": 165, + "eval_accuracy": 0.932, + "eval_loss": 0.3607469970703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075171172618866, + "mean_feedback_forward_cosine": 0.9989724549380216, + "mean_feedback_forward_relative_error": 0.04456372271207246, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00044451223726630213, + "instruction_rms": 0.0004445122374880904, + "raw_apical_rms": 0.1561968047277056, + "teaching_rms": 0.00044451223726630213, + "traffic_rms": 0.15619524641360857 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.525871836833176e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5335534097653396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005788216628066704, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.0030991528306156396 + }, + { + "epoch": 166, + "eval_accuracy": 0.9334, + "eval_loss": 0.3609993591308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075169682502747, + "mean_feedback_forward_cosine": 0.9989724267612804, + "mean_feedback_forward_relative_error": 0.04456367130306634, + "min_feedback_forward_cosine": 0.9975121021270752 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003992931518194371, + "instruction_rms": 0.00039929315184112865, + "raw_apical_rms": 0.1561993358125229, + "teaching_rms": 0.0003992931518194371, + "traffic_rms": 0.15619810784919824 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.569396214852532e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535884391979886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005841754779233307, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.0029006872836707365 + }, + { + "epoch": 167, + "eval_accuracy": 0.9334, + "eval_loss": 0.3611225158691406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075167447328568, + "mean_feedback_forward_cosine": 0.9989724278450012, + "mean_feedback_forward_relative_error": 0.044563623636283654, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00040215078600330475, + "instruction_rms": 0.00040215078585350195, + "raw_apical_rms": 0.15619606264845814, + "teaching_rms": 0.00040215078600330475, + "traffic_rms": 0.15619477717082902 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.738526702245508e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536509695498972e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005817911546870289, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.0028576889017182918 + }, + { + "epoch": 168, + "eval_accuracy": 0.9326, + "eval_loss": 0.3629449462890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075164467096329, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.04456357481804761, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003998887652516597, + "instruction_rms": 0.0003998887651836683, + "raw_apical_rms": 0.15619790996399205, + "teaching_rms": 0.0003998887652516597, + "traffic_rms": 0.1561966645374295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.799268265604951e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5362143182145867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005799121166191996, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.002924299320164654 + }, + { + "epoch": 169, + "eval_accuracy": 0.9324, + "eval_loss": 0.36294425659179685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707516148686409, + "mean_feedback_forward_cosine": 0.9989724300124428, + "mean_feedback_forward_relative_error": 0.044563523358242076, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00044473610910030826, + "instruction_rms": 0.00044473610878450993, + "raw_apical_rms": 0.1562011833622537, + "teaching_rms": 0.00044473610910030826, + "traffic_rms": 0.15619974547398885 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.176765220468951e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538507105205204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00591410992160838, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.0030470485738996003 + }, + { + "epoch": 170, + "eval_accuracy": 0.933, + "eval_loss": 0.3620413635253906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707515999674797, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.044563474506139754, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042758043344175986, + "instruction_rms": 0.000427580433318418, + "raw_apical_rms": 0.15619711451121107, + "teaching_rms": 0.00042758043344175986, + "traffic_rms": 0.15619572164224343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.376089593615688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5369646367400142e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005832750105369693, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.0029799130838985242 + }, + { + "epoch": 171, + "eval_accuracy": 0.9328, + "eval_loss": 0.36404056396484374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075157761573792, + "mean_feedback_forward_cosine": 0.9989724484356967, + "mean_feedback_forward_relative_error": 0.04456342599270019, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004256224274186385, + "instruction_rms": 0.00042562242739830307, + "raw_apical_rms": 0.1562026264795551, + "teaching_rms": 0.0004256224274186385, + "traffic_rms": 0.1562012385918873 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.67029391113283e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535564534471448e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005755436385765845, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.003123174267510573 + }, + { + "epoch": 172, + "eval_accuracy": 0.9336, + "eval_loss": 0.3628458374023438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075154781341553, + "mean_feedback_forward_cosine": 0.998972440849651, + "mean_feedback_forward_relative_error": 0.044563368250700564, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004919505852206358, + "instruction_rms": 0.0004919505851019495, + "raw_apical_rms": 0.15619272672224851, + "teaching_rms": 0.0004919505852206358, + "traffic_rms": 0.15619057426172062 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.396567897581917e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5318478637538845e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005892030299470198, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.003438941920403805 + }, + { + "epoch": 173, + "eval_accuracy": 0.934, + "eval_loss": 0.361624853515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075151801109314, + "mean_feedback_forward_cosine": 0.9989724375984885, + "mean_feedback_forward_relative_error": 0.044563313929194756, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004732160200341671, + "instruction_rms": 0.0004732160199897636, + "raw_apical_rms": 0.15619125526362268, + "teaching_rms": 0.0004732160200341671, + "traffic_rms": 0.15618951902258618 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.914952876968528e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534674797093549e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005832087390802645, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.0033569188530039457 + }, + { + "epoch": 174, + "eval_accuracy": 0.9334, + "eval_loss": 0.3606834411621094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075148820877075, + "mean_feedback_forward_cosine": 0.99897245277058, + "mean_feedback_forward_relative_error": 0.04456325716931712, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000504879789185604, + "instruction_rms": 0.0005048797891479873, + "raw_apical_rms": 0.15618947941493014, + "teaching_rms": 0.000504879789185604, + "traffic_rms": 0.15618761728279687 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.534652723961699e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534835066469787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005824511633885487, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.003320137251706587 + }, + { + "epoch": 175, + "eval_accuracy": 0.9328, + "eval_loss": 0.3632712341308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075145095586777, + "mean_feedback_forward_cosine": 0.9989724516868591, + "mean_feedback_forward_relative_error": 0.04456320562484589, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004231173779611415, + "instruction_rms": 0.00042311737763738125, + "raw_apical_rms": 0.15618914205308448, + "teaching_rms": 0.0004231173779611415, + "traffic_rms": 0.1561877039875195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.140283031503714e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538623372146383e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005876878379789779, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.002933721011566619 + }, + { + "epoch": 176, + "eval_accuracy": 0.9332, + "eval_loss": 0.36194794311523437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075142860412598, + "mean_feedback_forward_cosine": 0.9989724419333718, + "mean_feedback_forward_relative_error": 0.04456315531649373, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00045991324012191817, + "instruction_rms": 0.0004599132399693753, + "raw_apical_rms": 0.15619017173639713, + "teaching_rms": 0.00045991324012191817, + "traffic_rms": 0.15618847591366844 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.06208717563095e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5391214217003264e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005789111506779168, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.00342260849043313 + }, + { + "epoch": 177, + "eval_accuracy": 0.9334, + "eval_loss": 0.3670292785644531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075142115354538, + "mean_feedback_forward_cosine": 0.9989724592729048, + "mean_feedback_forward_relative_error": 0.04456310263750228, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00040664537186046166, + "instruction_rms": 0.0004066453716585905, + "raw_apical_rms": 0.15619706280682688, + "teaching_rms": 0.00040664537186046166, + "traffic_rms": 0.1561957847530296 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.209095625488771e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534770102298202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005823902930677908, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.002844909175568157 + }, + { + "epoch": 178, + "eval_accuracy": 0.9332, + "eval_loss": 0.3643357971191406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707513839006424, + "mean_feedback_forward_cosine": 0.9989724495194175, + "mean_feedback_forward_relative_error": 0.044563054530458016, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004049486367224079, + "instruction_rms": 0.000404948636576789, + "raw_apical_rms": 0.15619552088355443, + "teaching_rms": 0.0004049486367224079, + "traffic_rms": 0.15619416135052885 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.023897552471681e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537750164351138e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058023818346295665, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.0029311306476593017 + }, + { + "epoch": 179, + "eval_accuracy": 0.9342, + "eval_loss": 0.36326700439453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707513689994812, + "mean_feedback_forward_cosine": 0.9989724386822093, + "mean_feedback_forward_relative_error": 0.04456300999630581, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003612708095619896, + "instruction_rms": 0.00036127080952046164, + "raw_apical_rms": 0.15619400731217237, + "teaching_rms": 0.0003612708095619896, + "traffic_rms": 0.15619300828769123 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.921884931898603e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360207882004023e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005841793443784828, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.0024818976380758815 + }, + { + "epoch": 180, + "eval_accuracy": 0.9336, + "eval_loss": 0.36097264404296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075133919715881, + "mean_feedback_forward_cosine": 0.9989724603566257, + "mean_feedback_forward_relative_error": 0.04456295865503224, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004035473568848594, + "instruction_rms": 0.00040354735642829235, + "raw_apical_rms": 0.15619308976432028, + "teaching_rms": 0.0004035473568848594, + "traffic_rms": 0.15619173579574833 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.915981742849908e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534677185287896e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005847928120514314, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.002834540325920615 + }, + { + "epoch": 181, + "eval_accuracy": 0.9336, + "eval_loss": 0.36409304809570314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075133174657822, + "mean_feedback_forward_cosine": 0.9989724506031383, + "mean_feedback_forward_relative_error": 0.04456291464580731, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003687424852903932, + "instruction_rms": 0.0003687424851625734, + "raw_apical_rms": 0.15620027226349922, + "teaching_rms": 0.0003687424852903932, + "traffic_rms": 0.1561991959946112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.412849273829124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5348776588620148e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005791613943938917, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.0028244014726537797 + }, + { + "epoch": 182, + "eval_accuracy": 0.933, + "eval_loss": 0.36588348388671876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075129449367523, + "mean_feedback_forward_cosine": 0.9989724571054632, + "mean_feedback_forward_relative_error": 0.04456286396492611, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004362118070622091, + "instruction_rms": 0.0004362118068422204, + "raw_apical_rms": 0.1561978659340271, + "teaching_rms": 0.0004362118070622091, + "traffic_rms": 0.1561963286981839 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.406312908462496e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5351237634552045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00577301282666261, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.003047733818511996 + }, + { + "epoch": 183, + "eval_accuracy": 0.9336, + "eval_loss": 0.3611108093261719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075125724077225, + "mean_feedback_forward_cosine": 0.9989724571054632, + "mean_feedback_forward_relative_error": 0.04456281140446663, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004148569442338212, + "instruction_rms": 0.0004148569440265609, + "raw_apical_rms": 0.15619684428622288, + "teaching_rms": 0.0004148569442338212, + "traffic_rms": 0.15619553491803845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.257590001723656e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5350059929103585e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005837293103899293, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.0029103908619326022 + }, + { + "epoch": 184, + "eval_accuracy": 0.9332, + "eval_loss": 0.36385486450195315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075124233961105, + "mean_feedback_forward_cosine": 0.9989724668589506, + "mean_feedback_forward_relative_error": 0.04456276637925343, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00039054727424933097, + "instruction_rms": 0.00039054727427044125, + "raw_apical_rms": 0.15619733470302585, + "teaching_rms": 0.00039054727424933097, + "traffic_rms": 0.15619622904411004 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.030456027403488e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535347866608352e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0057423141327328725, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.002705052987134291 + }, + { + "epoch": 185, + "eval_accuracy": 0.9336, + "eval_loss": 0.3632457275390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075121253728867, + "mean_feedback_forward_cosine": 0.9989724516868591, + "mean_feedback_forward_relative_error": 0.04456271618943323, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003811323365473934, + "instruction_rms": 0.0003811323366254667, + "raw_apical_rms": 0.15619683555499747, + "teaching_rms": 0.0003811323365473934, + "traffic_rms": 0.15619564626965682 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.183086836448638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5369952899846847e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005860040617343112, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.002638033893249101 + }, + { + "epoch": 186, + "eval_accuracy": 0.9334, + "eval_loss": 0.36154000244140627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075117528438568, + "mean_feedback_forward_cosine": 0.9989724451845342, + "mean_feedback_forward_relative_error": 0.044562668624249374, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042452841534813666, + "instruction_rms": 0.00042452841530739896, + "raw_apical_rms": 0.1561949806606766, + "teaching_rms": 0.00042452841534813666, + "traffic_rms": 0.15619353826659707 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.10155134697743e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355028719230432e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811353048002794, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.0030590131243069967 + }, + { + "epoch": 187, + "eval_accuracy": 0.935, + "eval_loss": 0.36507200927734373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075115293264389, + "mean_feedback_forward_cosine": 0.9989724625240672, + "mean_feedback_forward_relative_error": 0.04456262127919631, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042261649333655434, + "instruction_rms": 0.0004226164933721387, + "raw_apical_rms": 0.156190922916446, + "teaching_rms": 0.00042261649333655434, + "traffic_rms": 0.15618943955163558 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.915867618000448e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360549223900153e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005852771622581931, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.0030017898343917397 + }, + { + "epoch": 188, + "eval_accuracy": 0.9338, + "eval_loss": 0.36510807495117187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707511380314827, + "mean_feedback_forward_cosine": 0.9989724625240672, + "mean_feedback_forward_relative_error": 0.04456257711757313, + "min_feedback_forward_cosine": 0.9975124001502991 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003916084172831918, + "instruction_rms": 0.0003916084172182954, + "raw_apical_rms": 0.15619046241567305, + "teaching_rms": 0.0003916084172831918, + "traffic_rms": 0.15618924894105277 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.255608380773992e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5390736932889114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005833163123987753, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.0026285000529968077 + }, + { + "epoch": 189, + "eval_accuracy": 0.9338, + "eval_loss": 0.365090625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707511231303215, + "mean_feedback_forward_cosine": 0.9989724560217423, + "mean_feedback_forward_relative_error": 0.04456252743574706, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00040755755962995714, + "instruction_rms": 0.00040755755963706907, + "raw_apical_rms": 0.15619771865988386, + "teaching_rms": 0.00040755755962995714, + "traffic_rms": 0.15619639373621746 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.970876936269633e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538003063450597e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005825029032007757, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.002841304802708328 + }, + { + "epoch": 190, + "eval_accuracy": 0.9326, + "eval_loss": 0.36245419311523436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075109332799911, + "mean_feedback_forward_cosine": 0.9989724560217423, + "mean_feedback_forward_relative_error": 0.044562478295781395, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00038014948555528357, + "instruction_rms": 0.00038014948527374315, + "raw_apical_rms": 0.15619197940003895, + "teaching_rms": 0.00038014948555528357, + "traffic_rms": 0.15619094840030273 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.455015094137707e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355496912772462e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005919079384556868, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.0027392928045656944 + }, + { + "epoch": 191, + "eval_accuracy": 0.9334, + "eval_loss": 0.36492646484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075107097625732, + "mean_feedback_forward_cosine": 0.9989724495194175, + "mean_feedback_forward_relative_error": 0.044562429782341824, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004018901788420274, + "instruction_rms": 0.00040189017884162333, + "raw_apical_rms": 0.15619137479569864, + "teaching_rms": 0.0004018901788420274, + "traffic_rms": 0.15619018727307954 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.211887265157202e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535846275555213e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005815387216343496, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.0027275119118599428 + }, + { + "epoch": 192, + "eval_accuracy": 0.9324, + "eval_loss": 0.3650273254394531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075105607509613, + "mean_feedback_forward_cosine": 0.9989724679426714, + "mean_feedback_forward_relative_error": 0.04456238462166353, + "min_feedback_forward_cosine": 0.9975124001502991 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00038408126969583303, + "instruction_rms": 0.00038408126960202184, + "raw_apical_rms": 0.15618522479586572, + "teaching_rms": 0.00038408126969583303, + "traffic_rms": 0.1561841321759944 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.835953314521248e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5386043751217692e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005837757667829218, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.002694618460794704 + }, + { + "epoch": 193, + "eval_accuracy": 0.9328, + "eval_loss": 0.36529390869140627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075103372335434, + "mean_feedback_forward_cosine": 0.9989724722775546, + "mean_feedback_forward_relative_error": 0.0445623412897641, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003657145414101729, + "instruction_rms": 0.0003657145415227043, + "raw_apical_rms": 0.15618295268645463, + "teaching_rms": 0.0003657145414101729, + "traffic_rms": 0.1561818427703464 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.366949716924864e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5362354929076727e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00578793646396616, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.002424141924393674 + }, + { + "epoch": 194, + "eval_accuracy": 0.9334, + "eval_loss": 0.36309793090820314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075101137161255, + "mean_feedback_forward_cosine": 0.9989724668589506, + "mean_feedback_forward_relative_error": 0.04456229038875211, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004149085355985523, + "instruction_rms": 0.00041490853532579494, + "raw_apical_rms": 0.15619277628176287, + "teaching_rms": 0.0004149085355985523, + "traffic_rms": 0.15619127224170495 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.803163271091762e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5327822541004354e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005807142875516983, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.0029466939909590616 + }, + { + "epoch": 195, + "eval_accuracy": 0.933, + "eval_loss": 0.36418885498046877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075099647045135, + "mean_feedback_forward_cosine": 0.9989724841984835, + "mean_feedback_forward_relative_error": 0.044562242552638055, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003596243837953494, + "instruction_rms": 0.000359624383627786, + "raw_apical_rms": 0.15619140879149387, + "teaching_rms": 0.0003596243837953494, + "traffic_rms": 0.15619041368152756 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.270116159612371e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.532461871659178e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005761611191290388, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.0023758877529866168 + }, + { + "epoch": 196, + "eval_accuracy": 0.933, + "eval_loss": 0.3622114013671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075097411870956, + "mean_feedback_forward_cosine": 0.9989724657752297, + "mean_feedback_forward_relative_error": 0.04456219828941605, + "min_feedback_forward_cosine": 0.9975124001502991 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003788401002222492, + "instruction_rms": 0.00037884010010704724, + "raw_apical_rms": 0.15619391125265422, + "teaching_rms": 0.0003788401002222492, + "traffic_rms": 0.15619275072949035 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.44583045325453e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536320428224063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005971374289712673, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.002593078167364001 + }, + { + "epoch": 197, + "eval_accuracy": 0.9332, + "eval_loss": 0.3630668518066406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075095176696777, + "mean_feedback_forward_cosine": 0.9989724690263921, + "mean_feedback_forward_relative_error": 0.04456215097822926, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004019384723443367, + "instruction_rms": 0.00040193847233358605, + "raw_apical_rms": 0.1561901152440423, + "teaching_rms": 0.0004019384723443367, + "traffic_rms": 0.15618888084324176 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.272781405414207e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5356222845259446e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005798918022251434, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.003016370440295173 + }, + { + "epoch": 198, + "eval_accuracy": 0.934, + "eval_loss": 0.36500486450195313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075092196464539, + "mean_feedback_forward_cosine": 0.9989724636077881, + "mean_feedback_forward_relative_error": 0.044562104344367984, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004087684824610619, + "instruction_rms": 0.0004087684824830167, + "raw_apical_rms": 0.15618832771598362, + "teaching_rms": 0.0004087684824610619, + "traffic_rms": 0.15618686884398414 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.957885372888995e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537445699149463e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005813321412645424, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.002747606110189938 + }, + { + "epoch": 199, + "eval_accuracy": 0.9334, + "eval_loss": 0.36417661743164065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.070750892162323, + "mean_feedback_forward_cosine": 0.9989724657752297, + "mean_feedback_forward_relative_error": 0.04456205044619062, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004871525299883906, + "instruction_rms": 0.00048715252976229, + "raw_apical_rms": 0.15619343033867647, + "teaching_rms": 0.0004871525299883906, + "traffic_rms": 0.15619158905469818 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 8.859503708436023e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5383053457659095e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005776670791695986, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.003350807800433702 + }, + { + "epoch": 200, + "eval_accuracy": 0.9322, + "eval_loss": 0.3675199523925781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075086236000061, + "mean_feedback_forward_cosine": 0.9989724636077881, + "mean_feedback_forward_relative_error": 0.044561998918652534, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004177410640374182, + "instruction_rms": 0.0004177410639151842, + "raw_apical_rms": 0.15618223981341522, + "teaching_rms": 0.0004177410640374182, + "traffic_rms": 0.15618101233638496 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.024044007015618e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5338302460993663e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005836083964773954, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.002966017501304547 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9322, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.3675199523925781, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 3262279168, + "peak_memory_reserved_bytes": 5276434432, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 9.810454315584138e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 9.810454315584138e-08, + "min_residual_soma_slope": -8.382607319390445e-08, + "mse": 1.5715674817957412e-17, + "observations": 64, + "residual_traffic_rms_ratio": 7.224132997837907e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.224132997837907e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 60.00945520401001, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.3673684597015381, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 8344.301742315292, + "train_wall_s": 8280.477409601212 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.046645715832710266, + 0.036221180111169815, + 0.0219095591455698, + 0.024090727791190147, + 0.014862747862935066, + 0.017354732379317284, + 0.010909760370850563, + 0.012971695512533188, + 0.00854403991252184, + 0.010942773893475533, + 0.006431682035326958, + 0.007946319878101349, + 0.005390886217355728, + 0.007816178724169731, + 0.004525847267359495, + 0.0066139050759375095, + 0.003956272266805172, + 0.005918589886277914, + 0.0034256556537002325, + 0.00842861458659172, + 0.005142349284142256, + 0.006969384849071503, + 0.004458904732018709, + 0.005985064432024956, + 0.003961027599871159, + 0.005532111972570419, + 0.0035666259936988354, + 0.004960889462381601, + 0.0032047750428318977, + 0.004740030970424414, + 0.0028284969739615917, + 0.004207335412502289, + 0.00256044021807611, + 0.0038634713273495436, + 0.0023208747152239084, + 0.0035063333343714476, + 0.0021188128739595413, + 0.005200310610234737, + 0.003467726521193981, + 0.004882926121354103, + 0.0031611102167516947, + 0.00439144903793931, + 0.00288514606654644, + 0.003956384491175413, + 0.002680722391232848, + 0.003665727796033025, + 0.002519759815186262, + 0.003549956250935793, + 0.0023743261117488146, + 0.003312949789687991, + 0.0022739002015441656, + 0.0031900741159915924, + 0.002209193306043744, + 0.00286328187212348, + 0.0022419888991862535 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 3.999999761581421, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.24335139989852905, + 0.19567525386810303, + 0.07590705901384354, + 0.12499095499515533, + 0.039654169231653214, + 0.09477696567773819, + 0.02513667196035385, + 0.0682290568947792, + 0.017793802544474602, + 0.059991367161273956, + 0.012211106717586517, + 0.04119913652539253, + 0.009280920960009098, + 0.04255608469247818, + 0.0071248686872422695, + 0.03416771814227104, + 0.005649160128086805, + 0.03203295171260834, + 0.00472416402772069, + 0.046523354947566986, + 0.009258695878088474, + 0.03640424460172653, + 0.0073813083581626415, + 0.03129300847649574, + 0.006084667518734932, + 0.02907230518758297, + 0.005167320836335421, + 0.026048656553030014, + 0.004463230259716511, + 0.02523532323539257, + 0.0037100326735526323, + 0.022375741973519325, + 0.0031846852507442236, + 0.020636072382330894, + 0.002770467661321163, + 0.01875767856836319, + 0.0024420199915766716, + 0.027892472222447395, + 0.005269602406769991, + 0.026072902604937553, + 0.0044991569593548775, + 0.023685075342655182, + 0.0038742832839488983, + 0.020799512043595314, + 0.003461548825725913, + 0.019222838804125786, + 0.0030808867886662483, + 0.018855275586247444, + 0.0027911863289773464, + 0.017468731850385666, + 0.0025647676084190607, + 0.01688026450574398, + 0.002401146572083235, + 0.015071986243128777, + 0.0023585029412060976 + ], + "unscaled_traffic_rms": [ + 0.7667219638824463, + 0.7404345870018005, + 1.1545466184616089, + 0.7709590792655945, + 1.499236822128296, + 0.7324451804161072, + 1.7360707521438599, + 0.760479211807251, + 1.9206777811050415, + 0.7296232581138611, + 2.10683012008667, + 0.7715035080909729, + 2.323427200317383, + 0.7346708178520203, + 2.5408732891082764, + 0.7742869853973389, + 2.8013172149658203, + 0.7390626668930054, + 2.9005391597747803, + 0.7246781587600708, + 2.221630096435547, + 0.7657771706581116, + 2.4163222312927246, + 0.7650353908538818, + 2.60394024848938, + 0.7611521482467651, + 2.76090931892395, + 0.7617881298065186, + 2.872157573699951, + 0.751332700252533, + 3.049565553665161, + 0.7521244287490845, + 3.2159414291381836, + 0.748877227306366, + 3.3508780002593994, + 0.7477115988731384, + 3.470590591430664, + 0.7457654476165771, + 2.632249116897583, + 0.7491189241409302, + 2.8104023933410645, + 0.7416398525238037, + 2.9787662029266357, + 0.7608610391616821, + 3.097714424133301, + 0.7627859115600586, + 3.2714734077453613, + 0.7530955672264099, + 3.4026050567626953, + 0.7586011290550232, + 3.5463645458221436, + 0.7559298872947693, + 3.6802306175231934, + 0.7598950266838074, + 3.8023931980133057 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1125397954772992, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1125389952000000, + "local_weight_correlation_macs": 1129371264000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1129371264000000, + "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": 115016770519040, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 9000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 9000064, + "total_forward_equivalent_examples": 9000064, + "total_macs_estimate": 4509538465857536 + } +} diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json.manifest.json b/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json.manifest.json new file mode 100644 index 0000000..9cf067f --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json.manifest.json @@ -0,0 +1,150 @@ +{ + "architecture": "resnet56", + "command": [ + "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "experiments/conv_run.py", + "--mode", + "kp_traffic", + "--out", + "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json", + "--normalization", + "batchnorm", + "--a_scale", + "1", + "--alignment_probe", + "32", + "--device", + "cuda", + "--depth", + "56", + "--width", + "16", + "--seed", + "0", + "--loader_seed", + "0", + "--split_seed", + "2027", + "--batch_size", + "128", + "--epochs", + "200", + "--train_limit", + "0", + "--val_examples", + "5000", + "--eval_split", + "validation", + "--eval_every", + "1", + "--augment_train", + "1", + "--lr", + "0.1", + "--output_lr", + "0.1", + "--lr_schedule", + "step", + "--lr_milestones", + "100,150", + "--lr_gamma", + "0.1", + "--momentum", + "0.9", + "--weight_decay", + "1e-4", + "--traffic_rule", + "innovation", + "--predictor_mode", + "closed_form", + "--neutral_projection", + "1", + "--traffic_seed", + "5000", + "--traffic_ratio", + "4", + "--traffic_calibration_examples", + "64", + "--learn_P", + "1", + "--eta_P", + "0.1", + "--predictor_warmup_steps", + "1", + "--predictor_every", + "0" + ], + "completed_unix_time": 1787012775.9045618, + "cuda_visible_devices": "0", + "depth": 56, + "driver_wall_seconds": 8350.019015789032, + "epochs": 200, + "experiment_name": "resnet-r2-sdil-d56-lr0p1", + "hardware_lock": { + "cuda_visible_devices": "0", + "hardware_profile": "a6000", + "physical_gpu_index": 0, + "physical_gpu_name": "NVIDIA RTX A6000", + "physical_gpu_uuid": "GPU-71e7b0a5-bc0f-e6c0-6965-8a3e2ab76b0b", + "total_memory_mib": 49140 + }, + "lr_schedule": "step", + "method": "sdil", + "output": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json", + "output_exists": true, + "output_sha256": "84f1139c05218524c1be2807d0e2bacb54bc0a0bc0d0526a0dcf7d8a2de8f5f6", + "rate": 0.1, + "return_code": 0, + "selector": { + "p1_source": { + "git_commit": "19f3966fc2d15b1ed18862a779d91ee16ed56bfc", + "protocol_path": "/scratch/yurenh2/sdil-runs/resnet-p1/RESNET_CROSSOVER.md", + "protocol_sha256": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "python": "/home/yurenh2/miniconda3/envs/ep_pascal/bin/python3", + "tracked_files": { + "RESNET_CROSSOVER.md": "048dbaf06866bdd3959a637ba827c06063205f6d1cc693d89aa8fc9c1d24975d", + "experiments/analyze_resnet_crossover_p1.py": "6ff485a3fecc7b814dedf55d04abdb770a6ce904cdeb9e06a7dddcf26f996d5b", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/resnet_crossover_grid.py": "669c84846bc3115816ae3c9036be8722f21577612dc9d56a689d3321e837ca95", + "experiments/resnet_crossover_native.py": "c5e54324b000c37c53d495f7d3eb0d4a85dcf15e83ea3aab125a097ce61eff87", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "262031353b5952fb4d1b0e3c3d1e9b3186ae42b88601ccb89e1748ffd36ed36d" + } + }, + "path": "results/resnet_crossover/p1_selector.json", + "selected_rates": { + "bp": 0.1, + "clean_kp": 0.1, + "dfa": 0.01, + "dualprop": 0.05, + "ep": 0.0003, + "fa": 0.1, + "ff": 0.03, + "pepita": 0.003, + "sdil": 0.1 + }, + "sha256": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b" + }, + "source": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "protocol_path": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/RESNET_CROSSOVER.md", + "protocol_sha256": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "python": "/orion/u/oscarwan/scratch/miniconda3/envs/sdil-main/bin/python", + "tracked_files": { + "RESNET_CROSSOVER.md": "16d110298f88d383f528b9c9b81f20ba11920c89596534ebb4a5c1e057f97223", + "experiments/analyze_resnet_crossover_r2.py": "ddad2a41b45ef41b7f26ddde0d6fdb9255592307714351d65b98e109b73ca6d1", + "experiments/conv_run.py": "9fa2daea597d63981a78248fafb154da39775ccfae57aac9889e78d83dc4eb44", + "experiments/crossover_hardware.py": "f7fd5ab940d6790d1f60c1bf0ec2115db8db7163a4ee034b536908ef72ac025f", + "experiments/resnet_crossover_native.py": "c04a5311c2aa4f7820f0f6883aca6e678b1c3227085c5d88fef46c83ac65e83d", + "experiments/resnet_crossover_r2.py": "8fe2844211a0220873a11ad801d672db8e2869fcb914bc972396bfd24adb040f", + "results/resnet_crossover/p1_selector.json": "0f0fcdfcd0056b034a9cd95e57970782bb2c33785832ab1cbe27e5dbe212111b", + "sdil/conv.py": "09f87eef636aa5397b39a2a7a66909b8861640624507eabbdfaffe98577a5d2c", + "sdil/conv_crossover.py": "3c1373aa3555e77cce37f093dec7b7aaedca475f608b092d13706bfff1dbe1a8", + "sdil/data.py": "0e8496c3bbcaa090c280afc3dbdd8b4c9490eeaca3306afff8aea081dccde254" + } + }, + "stage": "r2", + "status": "completed", + "timeout_seconds": 172800 +} -- cgit v1.2.3