From cdc7d5ec262d3ce1b3ffa2012800ab159a7fc3d4 Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Sun, 26 Jul 2026 16:30:45 -0500 Subject: results: confirm standard-depth SDIL scaling --- results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json | 1993 ++++++++++++++++++++ 1 file changed, 1993 insertions(+) create mode 100644 results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json (limited to 'results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json') diff --git a/results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json b/results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json new file mode 100644 index 0000000..f0c00a1 --- /dev/null +++ b/results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json @@ -0,0 +1,1993 @@ +{ + "architecture": { + "adaptive_apical_parameters": 2260992, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 188416, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 376832, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 1884160 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "test", + "learn_P": 0, + "loader_seed": 14, + "lr": 0.03, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_dynamic_scaling_v2/dfa_d20_s14.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 14, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 0, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 10000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": -0.000274311537699153, + "innovation_negative_gradient_cosine": [ + -0.0004933382733725011, + 0.0003353674546815455, + -0.0010736181866377592, + 0.002177057322114706, + -0.001047421945258975, + -0.0015439155977219343, + 0.001516413874924183, + 0.0002001795801334083, + 0.0030826255679130554, + 0.00200518686324358, + 0.004862049128860235, + 0.002671668538823724, + 0.0019046046072617173, + 0.01546824537217617, + 0.01730044186115265, + 0.01902702823281288, + 0.017416751012206078, + 0.021910570561885834, + 0.03727034851908684 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0004933382733725011, + 0.0003353674546815455, + -0.0010736181866377592, + 0.002177057322114706, + -0.001047421945258975, + -0.0015439155977219343, + 0.001516413874924183, + 0.0002001795801334083, + 0.0030826255679130554, + 0.00200518686324358, + 0.004862049128860235, + 0.002671668538823724, + 0.0019046046072617173, + 0.01546824537217617, + 0.01730044186115265, + 0.01902702823281288, + 0.017416751012206078, + 0.021910570561885834, + 0.03727034851908684 + ], + "teaching_negative_gradient_cosine": [ + -0.0004933382733725011, + 0.0003353674546815455, + -0.0010736181866377592, + 0.002177057322114706, + -0.001047421945258975, + -0.0015439155977219343, + 0.001516413874924183, + 0.0002001795801334083, + 0.0030826255679130554, + 0.00200518686324358, + 0.004862049128860235, + 0.002671668538823724, + 0.0019046046072617173, + 0.01546824537217617, + 0.01730044186115265, + 0.01902702823281288, + 0.017416751012206078, + 0.021910570561885834, + 0.03727034851908684 + ], + "wall_s": 0.12834787368774414 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 391, + "train_examples": 50000, + "train_loss": 2.430827777557373 + }, + { + "epoch": 2, + "lr": 0.03, + "output_lr": 0.1, + "step": 782, + "train_examples": 50000, + "train_loss": 2.8695908666992187 + }, + { + "epoch": 3, + "lr": 0.03, + "output_lr": 0.1, + "step": 1173, + "train_examples": 50000, + "train_loss": 4.850728899841308 + }, + { + "epoch": 4, + "lr": 0.03, + "output_lr": 0.1, + "step": 1564, + "train_examples": 50000, + "train_loss": 8.524278156127929 + }, + { + "epoch": 5, + "lr": 0.03, + "output_lr": 0.1, + "step": 1955, + "train_examples": 50000, + "train_loss": 11.590560819702148 + }, + { + "epoch": 6, + "lr": 0.03, + "output_lr": 0.1, + "step": 2346, + "train_examples": 50000, + "train_loss": 15.759647039794922 + }, + { + "epoch": 7, + "lr": 0.03, + "output_lr": 0.1, + "step": 2737, + "train_examples": 50000, + "train_loss": 22.13480206237793 + }, + { + "epoch": 8, + "lr": 0.03, + "output_lr": 0.1, + "step": 3128, + "train_examples": 50000, + "train_loss": 26.629495576171873 + }, + { + "epoch": 9, + "lr": 0.03, + "output_lr": 0.1, + "step": 3519, + "train_examples": 50000, + "train_loss": 36.430580815429686 + }, + { + "epoch": 10, + "lr": 0.03, + "output_lr": 0.1, + "step": 3910, + "train_examples": 50000, + "train_loss": 44.20328767089844 + }, + { + "epoch": 11, + "lr": 0.03, + "output_lr": 0.1, + "step": 4301, + "train_examples": 50000, + "train_loss": 48.577757346191405 + }, + { + "epoch": 12, + "lr": 0.03, + "output_lr": 0.1, + "step": 4692, + "train_examples": 50000, + "train_loss": 56.446661684570316 + }, + { + "epoch": 13, + "lr": 0.03, + "output_lr": 0.1, + "step": 5083, + "train_examples": 50000, + "train_loss": 68.82190098266602 + }, + { + "epoch": 14, + "lr": 0.03, + "output_lr": 0.1, + "step": 5474, + "train_examples": 50000, + "train_loss": 73.90951887939453 + }, + { + "epoch": 15, + "lr": 0.03, + "output_lr": 0.1, + "step": 5865, + "train_examples": 50000, + "train_loss": 90.26532935302734 + }, + { + "epoch": 16, + "lr": 0.03, + "output_lr": 0.1, + "step": 6256, + "train_examples": 50000, + "train_loss": 111.07496459472657 + }, + { + "epoch": 17, + "lr": 0.03, + "output_lr": 0.1, + "step": 6647, + "train_examples": 50000, + "train_loss": 114.86494052490234 + }, + { + "epoch": 18, + "lr": 0.03, + "output_lr": 0.1, + "step": 7038, + "train_examples": 50000, + "train_loss": 142.34349526855468 + }, + { + "epoch": 19, + "lr": 0.03, + "output_lr": 0.1, + "step": 7429, + "train_examples": 50000, + "train_loss": 168.44054090087891 + }, + { + "epoch": 20, + "lr": 0.03, + "output_lr": 0.1, + "step": 7820, + "train_examples": 50000, + "train_loss": 179.05319467773438 + }, + { + "epoch": 21, + "lr": 0.03, + "output_lr": 0.1, + "step": 8211, + "train_examples": 50000, + "train_loss": 202.31314580078126 + }, + { + "epoch": 22, + "lr": 0.03, + "output_lr": 0.1, + "step": 8602, + "train_examples": 50000, + "train_loss": 205.62455334960939 + }, + { + "epoch": 23, + "lr": 0.03, + "output_lr": 0.1, + "step": 8993, + "train_examples": 50000, + "train_loss": 248.4839746875 + }, + { + "epoch": 24, + "lr": 0.03, + "output_lr": 0.1, + "step": 9384, + "train_examples": 50000, + "train_loss": 247.82219323242188 + }, + { + "epoch": 25, + "lr": 0.03, + "output_lr": 0.1, + "step": 9775, + "train_examples": 50000, + "train_loss": 279.3399982861328 + }, + { + "epoch": 26, + "lr": 0.03, + "output_lr": 0.1, + "step": 10166, + "train_examples": 50000, + "train_loss": 306.9608862109375 + }, + { + "epoch": 27, + "lr": 0.03, + "output_lr": 0.1, + "step": 10557, + "train_examples": 50000, + "train_loss": 346.06842751953127 + }, + { + "epoch": 28, + "lr": 0.03, + "output_lr": 0.1, + "step": 10948, + "train_examples": 50000, + "train_loss": 376.8380232617188 + }, + { + "epoch": 29, + "lr": 0.03, + "output_lr": 0.1, + "step": 11339, + "train_examples": 50000, + "train_loss": 432.8407747265625 + }, + { + "epoch": 30, + "lr": 0.03, + "output_lr": 0.1, + "step": 11730, + "train_examples": 50000, + "train_loss": 457.6128712109375 + }, + { + "epoch": 31, + "lr": 0.03, + "output_lr": 0.1, + "step": 12121, + "train_examples": 50000, + "train_loss": 466.56698500976563 + }, + { + "epoch": 32, + "lr": 0.03, + "output_lr": 0.1, + "step": 12512, + "train_examples": 50000, + "train_loss": 519.24828078125 + }, + { + "epoch": 33, + "lr": 0.03, + "output_lr": 0.1, + "step": 12903, + "train_examples": 50000, + "train_loss": 571.6328147070312 + }, + { + "epoch": 34, + "lr": 0.03, + "output_lr": 0.1, + "step": 13294, + "train_examples": 50000, + "train_loss": 572.9210324511719 + }, + { + "epoch": 35, + "lr": 0.03, + "output_lr": 0.1, + "step": 13685, + "train_examples": 50000, + "train_loss": 602.8327405859375 + }, + { + "epoch": 36, + "lr": 0.03, + "output_lr": 0.1, + "step": 14076, + "train_examples": 50000, + "train_loss": 670.990899296875 + }, + { + "epoch": 37, + "lr": 0.03, + "output_lr": 0.1, + "step": 14467, + "train_examples": 50000, + "train_loss": 746.16743203125 + }, + { + "epoch": 38, + "lr": 0.03, + "output_lr": 0.1, + "step": 14858, + "train_examples": 50000, + "train_loss": 728.067555703125 + }, + { + "epoch": 39, + "lr": 0.03, + "output_lr": 0.1, + "step": 15249, + "train_examples": 50000, + "train_loss": 754.1594594726563 + }, + { + "epoch": 40, + "lr": 0.03, + "output_lr": 0.1, + "step": 15640, + "train_examples": 50000, + "train_loss": 925.0366741796875 + }, + { + "epoch": 41, + "lr": 0.03, + "output_lr": 0.1, + "step": 16031, + "train_examples": 50000, + "train_loss": 838.8716380078125 + }, + { + "epoch": 42, + "lr": 0.03, + "output_lr": 0.1, + "step": 16422, + "train_examples": 50000, + "train_loss": 942.319340859375 + }, + { + "epoch": 43, + "lr": 0.03, + "output_lr": 0.1, + "step": 16813, + "train_examples": 50000, + "train_loss": 980.6203349609375 + }, + { + "epoch": 44, + "lr": 0.03, + "output_lr": 0.1, + "step": 17204, + "train_examples": 50000, + "train_loss": 1030.9713324414063 + }, + { + "epoch": 45, + "lr": 0.03, + "output_lr": 0.1, + "step": 17595, + "train_examples": 50000, + "train_loss": 1134.9342415234375 + }, + { + "epoch": 46, + "lr": 0.03, + "output_lr": 0.1, + "step": 17986, + "train_examples": 50000, + "train_loss": 1201.5530361328124 + }, + { + "epoch": 47, + "lr": 0.03, + "output_lr": 0.1, + "step": 18377, + "train_examples": 50000, + "train_loss": 1126.673959609375 + }, + { + "epoch": 48, + "lr": 0.03, + "output_lr": 0.1, + "step": 18768, + "train_examples": 50000, + "train_loss": 1320.0336044921876 + }, + { + "epoch": 49, + "lr": 0.03, + "output_lr": 0.1, + "step": 19159, + "train_examples": 50000, + "train_loss": 1355.6681817578126 + }, + { + "epoch": 50, + "lr": 0.03, + "output_lr": 0.1, + "step": 19550, + "train_examples": 50000, + "train_loss": 1343.8974589453126 + }, + { + "epoch": 51, + "lr": 0.03, + "output_lr": 0.1, + "step": 19941, + "train_examples": 50000, + "train_loss": 1293.3335163671875 + }, + { + "epoch": 52, + "lr": 0.03, + "output_lr": 0.1, + "step": 20332, + "train_examples": 50000, + "train_loss": 1492.314264765625 + }, + { + "epoch": 53, + "lr": 0.03, + "output_lr": 0.1, + "step": 20723, + "train_examples": 50000, + "train_loss": 1695.872830703125 + }, + { + "epoch": 54, + "lr": 0.03, + "output_lr": 0.1, + "step": 21114, + "train_examples": 50000, + "train_loss": 1602.5963324609374 + }, + { + "epoch": 55, + "lr": 0.03, + "output_lr": 0.1, + "step": 21505, + "train_examples": 50000, + "train_loss": 1713.265025546875 + }, + { + "epoch": 56, + "lr": 0.03, + "output_lr": 0.1, + "step": 21896, + "train_examples": 50000, + "train_loss": 1812.542683203125 + }, + { + "epoch": 57, + "lr": 0.03, + "output_lr": 0.1, + "step": 22287, + "train_examples": 50000, + "train_loss": 1860.6483206445312 + }, + { + "epoch": 58, + "lr": 0.03, + "output_lr": 0.1, + "step": 22678, + "train_examples": 50000, + "train_loss": 1952.0459861328125 + }, + { + "epoch": 59, + "lr": 0.03, + "output_lr": 0.1, + "step": 23069, + "train_examples": 50000, + "train_loss": 1966.81673078125 + }, + { + "epoch": 60, + "lr": 0.03, + "output_lr": 0.1, + "step": 23460, + "train_examples": 50000, + "train_loss": 2140.2526516796875 + }, + { + "epoch": 61, + "lr": 0.03, + "output_lr": 0.1, + "step": 23851, + "train_examples": 50000, + "train_loss": 2067.9922076171874 + }, + { + "epoch": 62, + "lr": 0.03, + "output_lr": 0.1, + "step": 24242, + "train_examples": 50000, + "train_loss": 2276.8285640234376 + }, + { + "epoch": 63, + "lr": 0.03, + "output_lr": 0.1, + "step": 24633, + "train_examples": 50000, + "train_loss": 2251.3169115625 + }, + { + "epoch": 64, + "lr": 0.03, + "output_lr": 0.1, + "step": 25024, + "train_examples": 50000, + "train_loss": 2529.34507828125 + }, + { + "epoch": 65, + "lr": 0.03, + "output_lr": 0.1, + "step": 25415, + "train_examples": 50000, + "train_loss": 2340.8230924609375 + }, + { + "epoch": 66, + "lr": 0.03, + "output_lr": 0.1, + "step": 25806, + "train_examples": 50000, + "train_loss": 2649.29321765625 + }, + { + "epoch": 67, + "lr": 0.03, + "output_lr": 0.1, + "step": 26197, + "train_examples": 50000, + "train_loss": 2679.8316846875 + }, + { + "epoch": 68, + "lr": 0.03, + "output_lr": 0.1, + "step": 26588, + "train_examples": 50000, + "train_loss": 2818.46077625 + }, + { + "epoch": 69, + "lr": 0.03, + "output_lr": 0.1, + "step": 26979, + "train_examples": 50000, + "train_loss": 3011.803917421875 + }, + { + "epoch": 70, + "lr": 0.03, + "output_lr": 0.1, + "step": 27370, + "train_examples": 50000, + "train_loss": 2959.873146484375 + }, + { + "epoch": 71, + "lr": 0.03, + "output_lr": 0.1, + "step": 27761, + "train_examples": 50000, + "train_loss": 3122.5627303125 + }, + { + "epoch": 72, + "lr": 0.03, + "output_lr": 0.1, + "step": 28152, + "train_examples": 50000, + "train_loss": 3118.783563828125 + }, + { + "epoch": 73, + "lr": 0.03, + "output_lr": 0.1, + "step": 28543, + "train_examples": 50000, + "train_loss": 3193.22344640625 + }, + { + "epoch": 74, + "lr": 0.03, + "output_lr": 0.1, + "step": 28934, + "train_examples": 50000, + "train_loss": 3433.978434609375 + }, + { + "epoch": 75, + "lr": 0.03, + "output_lr": 0.1, + "step": 29325, + "train_examples": 50000, + "train_loss": 3412.4790053125 + }, + { + "epoch": 76, + "lr": 0.03, + "output_lr": 0.1, + "step": 29716, + "train_examples": 50000, + "train_loss": 3107.52170140625 + }, + { + "epoch": 77, + "lr": 0.03, + "output_lr": 0.1, + "step": 30107, + "train_examples": 50000, + "train_loss": 3805.391980703125 + }, + { + "epoch": 78, + "lr": 0.03, + "output_lr": 0.1, + "step": 30498, + "train_examples": 50000, + "train_loss": 3394.3623228515626 + }, + { + "epoch": 79, + "lr": 0.03, + "output_lr": 0.1, + "step": 30889, + "train_examples": 50000, + "train_loss": 3846.74855015625 + }, + { + "epoch": 80, + "lr": 0.03, + "output_lr": 0.1, + "step": 31280, + "train_examples": 50000, + "train_loss": 3939.635757890625 + }, + { + "epoch": 81, + "lr": 0.03, + "output_lr": 0.1, + "step": 31671, + "train_examples": 50000, + "train_loss": 3881.772237421875 + }, + { + "epoch": 82, + "lr": 0.03, + "output_lr": 0.1, + "step": 32062, + "train_examples": 50000, + "train_loss": 3870.17540359375 + }, + { + "epoch": 83, + "lr": 0.03, + "output_lr": 0.1, + "step": 32453, + "train_examples": 50000, + "train_loss": 3981.95599703125 + }, + { + "epoch": 84, + "lr": 0.03, + "output_lr": 0.1, + "step": 32844, + "train_examples": 50000, + "train_loss": 4510.165597421875 + }, + { + "epoch": 85, + "lr": 0.03, + "output_lr": 0.1, + "step": 33235, + "train_examples": 50000, + "train_loss": 4949.16415359375 + }, + { + "epoch": 86, + "lr": 0.03, + "output_lr": 0.1, + "step": 33626, + "train_examples": 50000, + "train_loss": 4467.54481765625 + }, + { + "epoch": 87, + "lr": 0.03, + "output_lr": 0.1, + "step": 34017, + "train_examples": 50000, + "train_loss": 4262.94804125 + }, + { + "epoch": 88, + "lr": 0.03, + "output_lr": 0.1, + "step": 34408, + "train_examples": 50000, + "train_loss": 4779.1370278125 + }, + { + "epoch": 89, + "lr": 0.03, + "output_lr": 0.1, + "step": 34799, + "train_examples": 50000, + "train_loss": 4564.3299187890625 + }, + { + "epoch": 90, + "lr": 0.03, + "output_lr": 0.1, + "step": 35190, + "train_examples": 50000, + "train_loss": 4644.84135625 + }, + { + "epoch": 91, + "lr": 0.03, + "output_lr": 0.1, + "step": 35581, + "train_examples": 50000, + "train_loss": 4634.471859375 + }, + { + "epoch": 92, + "lr": 0.03, + "output_lr": 0.1, + "step": 35972, + "train_examples": 50000, + "train_loss": 5272.44126296875 + }, + { + "epoch": 93, + "lr": 0.03, + "output_lr": 0.1, + "step": 36363, + "train_examples": 50000, + "train_loss": 5398.831680546875 + }, + { + "epoch": 94, + "lr": 0.03, + "output_lr": 0.1, + "step": 36754, + "train_examples": 50000, + "train_loss": 5311.88040921875 + }, + { + "epoch": 95, + "lr": 0.03, + "output_lr": 0.1, + "step": 37145, + "train_examples": 50000, + "train_loss": 5491.19182453125 + }, + { + "epoch": 96, + "lr": 0.03, + "output_lr": 0.1, + "step": 37536, + "train_examples": 50000, + "train_loss": 5483.284246875 + }, + { + "epoch": 97, + "lr": 0.03, + "output_lr": 0.1, + "step": 37927, + "train_examples": 50000, + "train_loss": 5769.0386840625 + }, + { + "epoch": 98, + "lr": 0.03, + "output_lr": 0.1, + "step": 38318, + "train_examples": 50000, + "train_loss": 5507.14149140625 + }, + { + "epoch": 99, + "lr": 0.03, + "output_lr": 0.1, + "step": 38709, + "train_examples": 50000, + "train_loss": 5734.120633125 + }, + { + "epoch": 100, + "lr": 0.03, + "output_lr": 0.1, + "step": 39100, + "train_examples": 50000, + "train_loss": 6269.34827296875 + }, + { + "epoch": 101, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39491, + "train_examples": 50000, + "train_loss": 1487.558666796875 + }, + { + "epoch": 102, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 39882, + "train_examples": 50000, + "train_loss": 950.5722432617188 + }, + { + "epoch": 103, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40273, + "train_examples": 50000, + "train_loss": 832.749403671875 + }, + { + "epoch": 104, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 40664, + "train_examples": 50000, + "train_loss": 753.5799643359375 + }, + { + "epoch": 105, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41055, + "train_examples": 50000, + "train_loss": 705.256015078125 + }, + { + "epoch": 106, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41446, + "train_examples": 50000, + "train_loss": 674.2632983984375 + }, + { + "epoch": 107, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 41837, + "train_examples": 50000, + "train_loss": 637.6445683203125 + }, + { + "epoch": 108, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42228, + "train_examples": 50000, + "train_loss": 599.0539654101563 + }, + { + "epoch": 109, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 42619, + "train_examples": 50000, + "train_loss": 575.545034296875 + }, + { + "epoch": 110, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43010, + "train_examples": 50000, + "train_loss": 534.5734385546875 + }, + { + "epoch": 111, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43401, + "train_examples": 50000, + "train_loss": 547.174887890625 + }, + { + "epoch": 112, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 43792, + "train_examples": 50000, + "train_loss": 502.4583783984375 + }, + { + "epoch": 113, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44183, + "train_examples": 50000, + "train_loss": 489.33109234375 + }, + { + "epoch": 114, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44574, + "train_examples": 50000, + "train_loss": 460.824575859375 + }, + { + "epoch": 115, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 44965, + "train_examples": 50000, + "train_loss": 476.13343795898436 + }, + { + "epoch": 116, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45356, + "train_examples": 50000, + "train_loss": 473.103417109375 + }, + { + "epoch": 117, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 45747, + "train_examples": 50000, + "train_loss": 468.42028130859376 + }, + { + "epoch": 118, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46138, + "train_examples": 50000, + "train_loss": 451.5446798828125 + }, + { + "epoch": 119, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46529, + "train_examples": 50000, + "train_loss": 448.97799533203124 + }, + { + "epoch": 120, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 46920, + "train_examples": 50000, + "train_loss": 436.08307154296875 + }, + { + "epoch": 121, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47311, + "train_examples": 50000, + "train_loss": 452.778623984375 + }, + { + "epoch": 122, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 47702, + "train_examples": 50000, + "train_loss": 465.5214141796875 + }, + { + "epoch": 123, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48093, + "train_examples": 50000, + "train_loss": 458.1043487207031 + }, + { + "epoch": 124, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48484, + "train_examples": 50000, + "train_loss": 448.01370393554686 + }, + { + "epoch": 125, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 48875, + "train_examples": 50000, + "train_loss": 455.8718875292969 + }, + { + "epoch": 126, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49266, + "train_examples": 50000, + "train_loss": 441.31153591796874 + }, + { + "epoch": 127, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 49657, + "train_examples": 50000, + "train_loss": 471.42664501953124 + }, + { + "epoch": 128, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50048, + "train_examples": 50000, + "train_loss": 462.54126444335935 + }, + { + "epoch": 129, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50439, + "train_examples": 50000, + "train_loss": 478.6115515136719 + }, + { + "epoch": 130, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 50830, + "train_examples": 50000, + "train_loss": 469.2341409375 + }, + { + "epoch": 131, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51221, + "train_examples": 50000, + "train_loss": 454.92755802734376 + }, + { + "epoch": 132, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 51612, + "train_examples": 50000, + "train_loss": 441.7504991113281 + }, + { + "epoch": 133, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52003, + "train_examples": 50000, + "train_loss": 506.769214453125 + }, + { + "epoch": 134, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52394, + "train_examples": 50000, + "train_loss": 431.8849821191406 + }, + { + "epoch": 135, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 52785, + "train_examples": 50000, + "train_loss": 464.18567890625 + }, + { + "epoch": 136, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53176, + "train_examples": 50000, + "train_loss": 453.88623982421876 + }, + { + "epoch": 137, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53567, + "train_examples": 50000, + "train_loss": 453.206499375 + }, + { + "epoch": 138, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 53958, + "train_examples": 50000, + "train_loss": 476.1337943457031 + }, + { + "epoch": 139, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54349, + "train_examples": 50000, + "train_loss": 530.7403985253907 + }, + { + "epoch": 140, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 54740, + "train_examples": 50000, + "train_loss": 500.3454430273438 + }, + { + "epoch": 141, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55131, + "train_examples": 50000, + "train_loss": 441.26398439453124 + }, + { + "epoch": 142, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55522, + "train_examples": 50000, + "train_loss": 483.9563454199219 + }, + { + "epoch": 143, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 55913, + "train_examples": 50000, + "train_loss": 467.760279453125 + }, + { + "epoch": 144, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56304, + "train_examples": 50000, + "train_loss": 493.96308111328125 + }, + { + "epoch": 145, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 56695, + "train_examples": 50000, + "train_loss": 475.72507780273435 + }, + { + "epoch": 146, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57086, + "train_examples": 50000, + "train_loss": 453.8898660546875 + }, + { + "epoch": 147, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57477, + "train_examples": 50000, + "train_loss": 535.6345719726562 + }, + { + "epoch": 148, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 57868, + "train_examples": 50000, + "train_loss": 478.78393487304686 + }, + { + "epoch": 149, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58259, + "train_examples": 50000, + "train_loss": 532.946368515625 + }, + { + "epoch": 150, + "lr": 0.003, + "output_lr": 0.010000000000000002, + "step": 58650, + "train_examples": 50000, + "train_loss": 512.34838421875 + }, + { + "epoch": 151, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59041, + "train_examples": 50000, + "train_loss": 179.30683587402345 + }, + { + "epoch": 152, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59432, + "train_examples": 50000, + "train_loss": 152.58808825195314 + }, + { + "epoch": 153, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 59823, + "train_examples": 50000, + "train_loss": 140.42047415527344 + }, + { + "epoch": 154, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60214, + "train_examples": 50000, + "train_loss": 130.18362895507812 + }, + { + "epoch": 155, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60605, + "train_examples": 50000, + "train_loss": 123.7321546875 + }, + { + "epoch": 156, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 60996, + "train_examples": 50000, + "train_loss": 117.87116739746094 + }, + { + "epoch": 157, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61387, + "train_examples": 50000, + "train_loss": 113.00177812744141 + }, + { + "epoch": 158, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 61778, + "train_examples": 50000, + "train_loss": 107.92502445800781 + }, + { + "epoch": 159, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62169, + "train_examples": 50000, + "train_loss": 103.53310265136719 + }, + { + "epoch": 160, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62560, + "train_examples": 50000, + "train_loss": 100.67651490722656 + }, + { + "epoch": 161, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 62951, + "train_examples": 50000, + "train_loss": 96.55011247802734 + }, + { + "epoch": 162, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63342, + "train_examples": 50000, + "train_loss": 92.68933987060547 + }, + { + "epoch": 163, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 63733, + "train_examples": 50000, + "train_loss": 90.2942857080078 + }, + { + "epoch": 164, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64124, + "train_examples": 50000, + "train_loss": 86.75207002685546 + }, + { + "epoch": 165, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64515, + "train_examples": 50000, + "train_loss": 85.0101225024414 + }, + { + "epoch": 166, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 64906, + "train_examples": 50000, + "train_loss": 82.25027728271485 + }, + { + "epoch": 167, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65297, + "train_examples": 50000, + "train_loss": 80.04829027099609 + }, + { + "epoch": 168, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 65688, + "train_examples": 50000, + "train_loss": 79.29926852783203 + }, + { + "epoch": 169, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66079, + "train_examples": 50000, + "train_loss": 75.78352020751953 + }, + { + "epoch": 170, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66470, + "train_examples": 50000, + "train_loss": 73.82204443847657 + }, + { + "epoch": 171, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 66861, + "train_examples": 50000, + "train_loss": 72.57129687255859 + }, + { + "epoch": 172, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67252, + "train_examples": 50000, + "train_loss": 73.18166558715821 + }, + { + "epoch": 173, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 67643, + "train_examples": 50000, + "train_loss": 68.87338528564453 + }, + { + "epoch": 174, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68034, + "train_examples": 50000, + "train_loss": 68.01583039794922 + }, + { + "epoch": 175, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68425, + "train_examples": 50000, + "train_loss": 67.75869918701171 + }, + { + "epoch": 176, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 68816, + "train_examples": 50000, + "train_loss": 66.2078173413086 + }, + { + "epoch": 177, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69207, + "train_examples": 50000, + "train_loss": 66.47758010498048 + }, + { + "epoch": 178, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69598, + "train_examples": 50000, + "train_loss": 63.72786291503906 + }, + { + "epoch": 179, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 69989, + "train_examples": 50000, + "train_loss": 66.09409375976563 + }, + { + "epoch": 180, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70380, + "train_examples": 50000, + "train_loss": 67.07359419921875 + }, + { + "epoch": 181, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 70771, + "train_examples": 50000, + "train_loss": 63.96962786376953 + }, + { + "epoch": 182, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71162, + "train_examples": 50000, + "train_loss": 62.141143621826174 + }, + { + "epoch": 183, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71553, + "train_examples": 50000, + "train_loss": 62.4706698059082 + }, + { + "epoch": 184, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 71944, + "train_examples": 50000, + "train_loss": 60.51575397949219 + }, + { + "epoch": 185, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72335, + "train_examples": 50000, + "train_loss": 62.271037517089844 + }, + { + "epoch": 186, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 72726, + "train_examples": 50000, + "train_loss": 61.85910586181641 + }, + { + "epoch": 187, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73117, + "train_examples": 50000, + "train_loss": 62.088958939208986 + }, + { + "epoch": 188, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73508, + "train_examples": 50000, + "train_loss": 61.23988405395508 + }, + { + "epoch": 189, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 73899, + "train_examples": 50000, + "train_loss": 57.6482016003418 + }, + { + "epoch": 190, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74290, + "train_examples": 50000, + "train_loss": 60.53348528076172 + }, + { + "epoch": 191, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 74681, + "train_examples": 50000, + "train_loss": 57.87255747314453 + }, + { + "epoch": 192, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75072, + "train_examples": 50000, + "train_loss": 59.5084599609375 + }, + { + "epoch": 193, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75463, + "train_examples": 50000, + "train_loss": 57.39036342163086 + }, + { + "epoch": 194, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 75854, + "train_examples": 50000, + "train_loss": 59.12267486328125 + }, + { + "epoch": 195, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76245, + "train_examples": 50000, + "train_loss": 57.19819190673828 + }, + { + "epoch": 196, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 76636, + "train_examples": 50000, + "train_loss": 56.51260213256836 + }, + { + "epoch": 197, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77027, + "train_examples": 50000, + "train_loss": 57.803850379638675 + }, + { + "epoch": 198, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77418, + "train_examples": 50000, + "train_loss": 57.83694255371094 + }, + { + "epoch": 199, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 77809, + "train_examples": 50000, + "train_loss": 54.8499035168457 + }, + { + "epoch": 200, + "lr": 0.00030000000000000003, + "output_lr": 0.0010000000000000002, + "step": 78200, + "train_examples": 50000, + "train_loss": 55.30137119018555 + } + ], + "evaluation_protocol": { + "test_evaluations": 1, + "test_used_for_selection": false, + "validation_evaluations": 0 + }, + "final": { + "accuracy": 0.28, + "epoch": 200, + "evaluation_split": "test", + "finite": true, + "loss": 74.0652265625, + "step": 78200 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2169915904, + "peak_memory_reserved_bytes": 2810183680, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "test", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 14, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": null, + "test_examples": 10000, + "train_examples": 50000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": {}, + "validation_examples": 0, + "validation_index_sha256": null + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 0.7102277278900146, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 3641.387490272522, + "train_wall_s": 3640.65451335907 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 18841600000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 405510400000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 405510400000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 10000000, + "total_forward_equivalent_examples": 10000000, + "total_macs_estimate": 829862400000000 + } +} -- cgit v1.2.3