diff options
| author | Oscar Wan <oscarwan@orionp2.stanford.edu> | 2026-08-26 08:50:07 -0700 |
|---|---|---|
| committer | Oscar Wan <oscarwan@orionp2.stanford.edu> | 2026-08-26 08:50:07 -0700 |
| commit | 3c7c221ae6cfcdf6840da1d5262d94c75d56288b (patch) | |
| tree | e13adbf676e3b5e1dc57fcf26dcc153c7983abc4 /results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json | |
| parent | f55872beaee658ff76ac15fd421b73017820c320 (diff) | |
experiment: add completed A6000 crossover resultsa6000-x1-results
Diffstat (limited to 'results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json')
| -rw-r--r-- | results/resnet_crossover/r2/resnet-r2-dfa-d56-lr0p01.json | 2892 |
1 files changed, 2892 insertions, 0 deletions
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 + } +} |
