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-d32-lr0p01.json | |
| parent | f55872beaee658ff76ac15fd421b73017820c320 (diff) | |
experiment: add completed A6000 crossover resultsa6000-x1-results
Diffstat (limited to 'results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json')
| -rw-r--r-- | results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json | 2700 |
1 files changed, 2700 insertions, 0 deletions
diff --git a/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json b/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json new file mode 100644 index 0000000..3b96f18 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json @@ -0,0 +1,2700 @@ +{ + "architecture": { + "adaptive_apical_parameters": 3637248, + "adaptive_feedback_parameters": 0, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "residual_scale": 1.0, + "vectorizer_mode": "spatial_template", + "vectorizer_parameters": 3031040 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 32, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.01, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "lr": 0.01, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "dfa", + "momentum": 0.9, + "neutral_projection": 0, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-dfa-d32-lr0p01.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "nlms", + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "full_hidden_field", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 0, + "ordinary_examples": 9000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 0, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.00031910948309814556, + "innovation_negative_gradient_cosine": [ + 0.0026197603438049555, + 0.0006819418631494045, + 0.00017868289432954043, + 0.0009858061093837023, + -0.001170104369521141, + 0.0005506974994204938, + 0.00034609687281772494, + 5.504034925252199e-05, + -0.0002867505536414683, + -0.0007700761780142784, + 0.0017122221179306507, + 0.004392784088850021, + 0.0056973351165652275, + -0.00015289167640730739, + 0.0028711319901049137, + 0.003501438070088625, + 0.005120588466525078, + 0.003510239999741316, + 0.00014667602954432368, + 0.0016334315296262503, + 0.0033825868740677834, + 0.011951932683587074, + 0.016150185838341713, + 0.017887040972709656, + 0.010713891126215458, + 0.017346883192658424, + 0.025565773248672485, + 0.023414287716150284, + 0.025312259793281555, + 0.020356435328722, + 0.04579668864607811 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0026197603438049555, + 0.0006819418631494045, + 0.00017868289432954043, + 0.0009858061093837023, + -0.001170104369521141, + 0.0005506974994204938, + 0.00034609687281772494, + 5.504034925252199e-05, + -0.0002867505536414683, + -0.0007700761780142784, + 0.0017122221179306507, + 0.004392784088850021, + 0.0056973351165652275, + -0.00015289167640730739, + 0.0028711319901049137, + 0.003501438070088625, + 0.005120588466525078, + 0.003510239999741316, + 0.00014667602954432368, + 0.0016334315296262503, + 0.0033825868740677834, + 0.011951932683587074, + 0.016150185838341713, + 0.017887040972709656, + 0.010713891126215458, + 0.017346883192658424, + 0.025565773248672485, + 0.023414287716150284, + 0.025312259793281555, + 0.020356435328722, + 0.04579668864607811 + ], + "teaching_negative_gradient_cosine": [ + 0.0026197603438049555, + 0.0006819418631494045, + 0.00017868289432954043, + 0.0009858061093837023, + -0.001170104369521141, + 0.0005506974994204938, + 0.00034609687281772494, + 5.504034925252199e-05, + -0.0002867505536414683, + -0.0007700761780142784, + 0.0017122221179306507, + 0.004392784088850021, + 0.0056973351165652275, + -0.00015289167640730739, + 0.0028711319901049137, + 0.003501438070088625, + 0.005120588466525078, + 0.003510239999741316, + 0.00014667602954432368, + 0.0016334315296262503, + 0.0033825868740677834, + 0.011951932683587074, + 0.016150185838341713, + 0.017887040972709656, + 0.010713891126215458, + 0.017346883192658424, + 0.025565773248672485, + 0.023414287716150284, + 0.025312259793281555, + 0.020356435328722, + 0.04579668864607811 + ], + "wall_s": 0.1506354808807373 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.1968, + "eval_loss": 2.13564140625, + "eval_split": "validation", + "lr": 0.002, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.2802412104288736 + }, + { + "epoch": 2, + "eval_accuracy": 0.1762, + "eval_loss": 2.403104150390625, + "eval_split": "validation", + "lr": 0.004, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 2.3138541755676267 + }, + { + "epoch": 3, + "eval_accuracy": 0.166, + "eval_loss": 2.668672314453125, + "eval_split": "validation", + "lr": 0.006, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 3.1551799483829073 + }, + { + "epoch": 4, + "eval_accuracy": 0.1524, + "eval_loss": 3.85100947265625, + "eval_split": "validation", + "lr": 0.008, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 6.567851277033488 + }, + { + "epoch": 5, + "eval_accuracy": 0.1702, + "eval_loss": 9.0174177734375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 8.539597244093153 + }, + { + "epoch": 6, + "eval_accuracy": 0.2132, + "eval_loss": 6.2054705078125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 8.005350156317817 + }, + { + "epoch": 7, + "eval_accuracy": 0.2122, + "eval_loss": 6.59003876953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 8.283805978224013 + }, + { + "epoch": 8, + "eval_accuracy": 0.1796, + "eval_loss": 10.077484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 8.76988885184394 + }, + { + "epoch": 9, + "eval_accuracy": 0.194, + "eval_loss": 12.0967720703125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 9.657652395799424 + }, + { + "epoch": 10, + "eval_accuracy": 0.244, + "eval_loss": 5.8700173828125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 10.630738082038032 + }, + { + "epoch": 11, + "eval_accuracy": 0.229, + "eval_loss": 9.34055859375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 11.782854916551377 + }, + { + "epoch": 12, + "eval_accuracy": 0.257, + "eval_loss": 13.1749107421875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 12.472624915907119 + }, + { + "epoch": 13, + "eval_accuracy": 0.2512, + "eval_loss": 15.2414419921875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 16.400539314439563 + }, + { + "epoch": 14, + "eval_accuracy": 0.2382, + "eval_loss": 13.90708046875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 15.362716340637206 + }, + { + "epoch": 15, + "eval_accuracy": 0.1612, + "eval_loss": 20.7498, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 15.200875530666774 + }, + { + "epoch": 16, + "eval_accuracy": 0.1846, + "eval_loss": 27.910105078125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 18.039595056830514 + }, + { + "epoch": 17, + "eval_accuracy": 0.268, + "eval_loss": 27.628282421875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 19.268495307752822 + }, + { + "epoch": 18, + "eval_accuracy": 0.2842, + "eval_loss": 22.1950453125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 19.945682086859808 + }, + { + "epoch": 19, + "eval_accuracy": 0.2712, + "eval_loss": 20.5347578125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 22.335486112467446 + }, + { + "epoch": 20, + "eval_accuracy": 0.2604, + "eval_loss": 22.29314609375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 25.390615134006076 + }, + { + "epoch": 21, + "eval_accuracy": 0.1656, + "eval_loss": 40.85045546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 24.838533505588106 + }, + { + "epoch": 22, + "eval_accuracy": 0.2032, + "eval_loss": 59.94360859375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 30.458272143554687 + }, + { + "epoch": 23, + "eval_accuracy": 0.3094, + "eval_loss": 24.10346015625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 31.5929201107449 + }, + { + "epoch": 24, + "eval_accuracy": 0.2626, + "eval_loss": 27.9797484375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 35.06654516364203 + }, + { + "epoch": 25, + "eval_accuracy": 0.2232, + "eval_loss": 40.38759375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 34.452620131429036 + }, + { + "epoch": 26, + "eval_accuracy": 0.2136, + "eval_loss": 38.92077890625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 39.40891842041015 + }, + { + "epoch": 27, + "eval_accuracy": 0.2128, + "eval_loss": 36.42456640625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 42.401445976426864 + }, + { + "epoch": 28, + "eval_accuracy": 0.3382, + "eval_loss": 41.78907578125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 48.708959084743924 + }, + { + "epoch": 29, + "eval_accuracy": 0.3258, + "eval_loss": 47.034034375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 47.791174917263454 + }, + { + "epoch": 30, + "eval_accuracy": 0.2702, + "eval_loss": 50.88515546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 59.97448439670139 + }, + { + "epoch": 31, + "eval_accuracy": 0.2674, + "eval_loss": 62.69679375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 55.62923475884332 + }, + { + "epoch": 32, + "eval_accuracy": 0.2688, + "eval_loss": 91.7570109375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 56.54406637641059 + }, + { + "epoch": 33, + "eval_accuracy": 0.25, + "eval_loss": 53.91616796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 56.0292227613661 + }, + { + "epoch": 34, + "eval_accuracy": 0.2872, + "eval_loss": 108.42245, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 65.2930051079644 + }, + { + "epoch": 35, + "eval_accuracy": 0.3082, + "eval_loss": 52.419546875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 62.557755900065104 + }, + { + "epoch": 36, + "eval_accuracy": 0.21, + "eval_loss": 248.62119375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 71.82943644748264 + }, + { + "epoch": 37, + "eval_accuracy": 0.3456, + "eval_loss": 104.5305140625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 80.15177752956814 + }, + { + "epoch": 38, + "eval_accuracy": 0.2798, + "eval_loss": 63.1959390625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 74.20427698228625 + }, + { + "epoch": 39, + "eval_accuracy": 0.2526, + "eval_loss": 67.9075953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 86.0083361219618 + }, + { + "epoch": 40, + "eval_accuracy": 0.1372, + "eval_loss": 213.52529375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 90.54625876193576 + }, + { + "epoch": 41, + "eval_accuracy": 0.291, + "eval_loss": 85.5634140625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 97.03032708468967 + }, + { + "epoch": 42, + "eval_accuracy": 0.287, + "eval_loss": 132.8560890625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 102.40799818793403 + }, + { + "epoch": 43, + "eval_accuracy": 0.226, + "eval_loss": 219.480903125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 110.62748044162326 + }, + { + "epoch": 44, + "eval_accuracy": 0.3476, + "eval_loss": 139.70363125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 121.17438940836588 + }, + { + "epoch": 45, + "eval_accuracy": 0.3148, + "eval_loss": 115.74535625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 117.96440033772787 + }, + { + "epoch": 46, + "eval_accuracy": 0.2498, + "eval_loss": 161.649003125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 130.1123986111111 + }, + { + "epoch": 47, + "eval_accuracy": 0.2454, + "eval_loss": 309.7527625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 132.362695046658 + }, + { + "epoch": 48, + "eval_accuracy": 0.2128, + "eval_loss": 173.7661875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 141.5597231499566 + }, + { + "epoch": 49, + "eval_accuracy": 0.323, + "eval_loss": 87.1395671875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 139.6630462632921 + }, + { + "epoch": 50, + "eval_accuracy": 0.2816, + "eval_loss": 232.42523125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 163.6217903415256 + }, + { + "epoch": 51, + "eval_accuracy": 0.1946, + "eval_loss": 188.797478125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 164.3333604844835 + }, + { + "epoch": 52, + "eval_accuracy": 0.2566, + "eval_loss": 239.084840625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 178.35102162000868 + }, + { + "epoch": 53, + "eval_accuracy": 0.3502, + "eval_loss": 123.9788515625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 175.44256982150608 + }, + { + "epoch": 54, + "eval_accuracy": 0.3146, + "eval_loss": 206.992815625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 184.28907900390624 + }, + { + "epoch": 55, + "eval_accuracy": 0.3146, + "eval_loss": 163.399478125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 207.67702931043837 + }, + { + "epoch": 56, + "eval_accuracy": 0.2962, + "eval_loss": 223.243553125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 205.11963244628907 + }, + { + "epoch": 57, + "eval_accuracy": 0.3784, + "eval_loss": 137.992953125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 214.7449376844618 + }, + { + "epoch": 58, + "eval_accuracy": 0.2166, + "eval_loss": 476.17575, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 239.8989421766493 + }, + { + "epoch": 59, + "eval_accuracy": 0.2722, + "eval_loss": 290.41935625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 239.53743671875 + }, + { + "epoch": 60, + "eval_accuracy": 0.316, + "eval_loss": 164.1238, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 245.6346437608507 + }, + { + "epoch": 61, + "eval_accuracy": 0.1988, + "eval_loss": 365.0298, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 251.1119849609375 + }, + { + "epoch": 62, + "eval_accuracy": 0.2056, + "eval_loss": 389.20041875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 266.53731458333334 + }, + { + "epoch": 63, + "eval_accuracy": 0.3122, + "eval_loss": 268.056540625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 275.1557644124349 + }, + { + "epoch": 64, + "eval_accuracy": 0.22, + "eval_loss": 282.2733, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 276.21525709092884 + }, + { + "epoch": 65, + "eval_accuracy": 0.3348, + "eval_loss": 217.802796875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 293.0577929416233 + }, + { + "epoch": 66, + "eval_accuracy": 0.2684, + "eval_loss": 286.4739, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 333.1723864176432 + }, + { + "epoch": 67, + "eval_accuracy": 0.3034, + "eval_loss": 199.15364375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 319.33838850911457 + }, + { + "epoch": 68, + "eval_accuracy": 0.2108, + "eval_loss": 308.4932, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 284.5920985188802 + }, + { + "epoch": 69, + "eval_accuracy": 0.2392, + "eval_loss": 458.07015, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 337.78335944281685 + }, + { + "epoch": 70, + "eval_accuracy": 0.2808, + "eval_loss": 321.44930625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 315.11677765570744 + }, + { + "epoch": 71, + "eval_accuracy": 0.225, + "eval_loss": 421.11063125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 358.4162332628038 + }, + { + "epoch": 72, + "eval_accuracy": 0.325, + "eval_loss": 251.6319625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 358.8881578938802 + }, + { + "epoch": 73, + "eval_accuracy": 0.3256, + "eval_loss": 257.94384375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 403.11145792100695 + }, + { + "epoch": 74, + "eval_accuracy": 0.312, + "eval_loss": 222.5570625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 384.9969331814236 + }, + { + "epoch": 75, + "eval_accuracy": 0.2346, + "eval_loss": 340.03154375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 401.10929453396267 + }, + { + "epoch": 76, + "eval_accuracy": 0.2106, + "eval_loss": 1018.6606125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 399.87897292751734 + }, + { + "epoch": 77, + "eval_accuracy": 0.3102, + "eval_loss": 404.8272375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 429.4044539496528 + }, + { + "epoch": 78, + "eval_accuracy": 0.2868, + "eval_loss": 436.610575, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 457.54408018663196 + }, + { + "epoch": 79, + "eval_accuracy": 0.2612, + "eval_loss": 579.3421, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 483.91433331163194 + }, + { + "epoch": 80, + "eval_accuracy": 0.2456, + "eval_loss": 575.697925, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 465.56115231119793 + }, + { + "epoch": 81, + "eval_accuracy": 0.22, + "eval_loss": 1274.014525, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 491.67860532769095 + }, + { + "epoch": 82, + "eval_accuracy": 0.181, + "eval_loss": 725.602575, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 509.37582771809895 + }, + { + "epoch": 83, + "eval_accuracy": 0.2862, + "eval_loss": 433.20714375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 523.651078266059 + }, + { + "epoch": 84, + "eval_accuracy": 0.306, + "eval_loss": 331.74839375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 505.1366246799045 + }, + { + "epoch": 85, + "eval_accuracy": 0.2364, + "eval_loss": 566.8942875, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 579.8360815646702 + }, + { + "epoch": 86, + "eval_accuracy": 0.2528, + "eval_loss": 943.22735, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 574.1495292317709 + }, + { + "epoch": 87, + "eval_accuracy": 0.276, + "eval_loss": 709.0178, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 538.2599395833333 + }, + { + "epoch": 88, + "eval_accuracy": 0.2866, + "eval_loss": 380.433375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 624.6833575466579 + }, + { + "epoch": 89, + "eval_accuracy": 0.2842, + "eval_loss": 482.03155625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 608.3776178927951 + }, + { + "epoch": 90, + "eval_accuracy": 0.3342, + "eval_loss": 464.89378125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 647.8050713107639 + }, + { + "epoch": 91, + "eval_accuracy": 0.3126, + "eval_loss": 836.0266375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 613.0687445529514 + }, + { + "epoch": 92, + "eval_accuracy": 0.3, + "eval_loss": 486.60365, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 650.7298907063802 + }, + { + "epoch": 93, + "eval_accuracy": 0.298, + "eval_loss": 469.757075, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 620.019738937717 + }, + { + "epoch": 94, + "eval_accuracy": 0.2938, + "eval_loss": 415.58044375, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 746.7210650390625 + }, + { + "epoch": 95, + "eval_accuracy": 0.3562, + "eval_loss": 411.128125, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 742.6021716959635 + }, + { + "epoch": 96, + "eval_accuracy": 0.2962, + "eval_loss": 554.3056, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 746.4939790690104 + }, + { + "epoch": 97, + "eval_accuracy": 0.2416, + "eval_loss": 825.879775, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 697.7495639756944 + }, + { + "epoch": 98, + "eval_accuracy": 0.3696, + "eval_loss": 489.983625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 766.2687794487847 + }, + { + "epoch": 99, + "eval_accuracy": 0.249, + "eval_loss": 866.526825, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 707.4956386827257 + }, + { + "epoch": 100, + "eval_accuracy": 0.2272, + "eval_loss": 736.6153625, + "eval_split": "validation", + "lr": 0.01, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 860.6614454047309 + }, + { + "epoch": 101, + "eval_accuracy": 0.3856, + "eval_loss": 170.472615625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 196.49338107910157 + }, + { + "epoch": 102, + "eval_accuracy": 0.3778, + "eval_loss": 125.7411875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 141.02132633734809 + }, + { + "epoch": 103, + "eval_accuracy": 0.356, + "eval_loss": 110.767259375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 115.80608062608506 + }, + { + "epoch": 104, + "eval_accuracy": 0.343, + "eval_loss": 106.7759078125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 103.91751080322265 + }, + { + "epoch": 105, + "eval_accuracy": 0.3468, + "eval_loss": 96.4980328125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 96.93153265652127 + }, + { + "epoch": 106, + "eval_accuracy": 0.3528, + "eval_loss": 81.44060625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 92.24946930202908 + }, + { + "epoch": 107, + "eval_accuracy": 0.353, + "eval_loss": 77.130496875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 87.91589212239583 + }, + { + "epoch": 108, + "eval_accuracy": 0.3476, + "eval_loss": 75.27756875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 82.24052918836806 + }, + { + "epoch": 109, + "eval_accuracy": 0.3438, + "eval_loss": 74.3921109375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 80.62625283610026 + }, + { + "epoch": 110, + "eval_accuracy": 0.3346, + "eval_loss": 82.432540625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 74.16640807834202 + }, + { + "epoch": 111, + "eval_accuracy": 0.3414, + "eval_loss": 68.23063359375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 72.85556735432942 + }, + { + "epoch": 112, + "eval_accuracy": 0.3604, + "eval_loss": 60.68131328125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 71.44236545952691 + }, + { + "epoch": 113, + "eval_accuracy": 0.3438, + "eval_loss": 63.771634375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 66.29341225043403 + }, + { + "epoch": 114, + "eval_accuracy": 0.359, + "eval_loss": 59.54249375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 68.79256528998481 + }, + { + "epoch": 115, + "eval_accuracy": 0.3114, + "eval_loss": 64.330353125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 64.35555018920898 + }, + { + "epoch": 116, + "eval_accuracy": 0.2398, + "eval_loss": 98.6124921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 63.517505746799046 + }, + { + "epoch": 117, + "eval_accuracy": 0.3264, + "eval_loss": 54.6644921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 64.07564553019206 + }, + { + "epoch": 118, + "eval_accuracy": 0.3136, + "eval_loss": 73.9800765625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 63.32237302517361 + }, + { + "epoch": 119, + "eval_accuracy": 0.3248, + "eval_loss": 56.1960296875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 62.23008333536784 + }, + { + "epoch": 120, + "eval_accuracy": 0.3352, + "eval_loss": 76.2859921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 64.19782397460938 + }, + { + "epoch": 121, + "eval_accuracy": 0.325, + "eval_loss": 52.52334140625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 62.10008289048937 + }, + { + "epoch": 122, + "eval_accuracy": 0.3006, + "eval_loss": 64.4516078125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 69.5385602355957 + }, + { + "epoch": 123, + "eval_accuracy": 0.3222, + "eval_loss": 59.18459375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 68.50677893473308 + }, + { + "epoch": 124, + "eval_accuracy": 0.3454, + "eval_loss": 66.70076875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 64.05500425075955 + }, + { + "epoch": 125, + "eval_accuracy": 0.299, + "eval_loss": 96.0758625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 67.54195568440755 + }, + { + "epoch": 126, + "eval_accuracy": 0.3506, + "eval_loss": 50.27091875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 63.50760168999566 + }, + { + "epoch": 127, + "eval_accuracy": 0.3584, + "eval_loss": 42.83064453125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 61.460221661376956 + }, + { + "epoch": 128, + "eval_accuracy": 0.3492, + "eval_loss": 46.741225, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 68.67775972290039 + }, + { + "epoch": 129, + "eval_accuracy": 0.3366, + "eval_loss": 55.25401015625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 65.46874283650716 + }, + { + "epoch": 130, + "eval_accuracy": 0.3758, + "eval_loss": 35.30758203125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 60.136354113091365 + }, + { + "epoch": 131, + "eval_accuracy": 0.381, + "eval_loss": 46.256021875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 71.47179066297743 + }, + { + "epoch": 132, + "eval_accuracy": 0.3672, + "eval_loss": 39.7463421875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 62.817046925184464 + }, + { + "epoch": 133, + "eval_accuracy": 0.3436, + "eval_loss": 65.4880921875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 70.35553123440212 + }, + { + "epoch": 134, + "eval_accuracy": 0.3516, + "eval_loss": 61.6984265625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 64.2483411376953 + }, + { + "epoch": 135, + "eval_accuracy": 0.3028, + "eval_loss": 63.19440390625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 66.96016059977214 + }, + { + "epoch": 136, + "eval_accuracy": 0.3524, + "eval_loss": 51.108896875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 62.91782479654948 + }, + { + "epoch": 137, + "eval_accuracy": 0.216, + "eval_loss": 78.7760421875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 70.46476604546442 + }, + { + "epoch": 138, + "eval_accuracy": 0.368, + "eval_loss": 79.4366125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 72.66207418416342 + }, + { + "epoch": 139, + "eval_accuracy": 0.3388, + "eval_loss": 49.279284375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 61.06694605577257 + }, + { + "epoch": 140, + "eval_accuracy": 0.4082, + "eval_loss": 31.25261875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 59.71975162150065 + }, + { + "epoch": 141, + "eval_accuracy": 0.3726, + "eval_loss": 47.8163640625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 69.24733488498264 + }, + { + "epoch": 142, + "eval_accuracy": 0.2152, + "eval_loss": 96.3982875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 61.999896885850696 + }, + { + "epoch": 143, + "eval_accuracy": 0.418, + "eval_loss": 33.0312359375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 75.65242304958767 + }, + { + "epoch": 144, + "eval_accuracy": 0.356, + "eval_loss": 51.88513671875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 66.79744441663954 + }, + { + "epoch": 145, + "eval_accuracy": 0.3446, + "eval_loss": 57.6310875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 68.28875555216472 + }, + { + "epoch": 146, + "eval_accuracy": 0.3326, + "eval_loss": 53.0010203125, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 68.38004661187065 + }, + { + "epoch": 147, + "eval_accuracy": 0.4004, + "eval_loss": 39.793740625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 72.77143829074436 + }, + { + "epoch": 148, + "eval_accuracy": 0.3338, + "eval_loss": 62.11200625, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 63.826129070366754 + }, + { + "epoch": 149, + "eval_accuracy": 0.2826, + "eval_loss": 98.2671859375, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 71.38788469509548 + }, + { + "epoch": 150, + "eval_accuracy": 0.3802, + "eval_loss": 62.30995546875, + "eval_split": "validation", + "lr": 0.001, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 69.05508831041124 + }, + { + "epoch": 151, + "eval_accuracy": 0.4546, + "eval_loss": 23.298765625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 25.48181363050673 + }, + { + "epoch": 152, + "eval_accuracy": 0.4396, + "eval_loss": 21.12952734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 22.125638820054796 + }, + { + "epoch": 153, + "eval_accuracy": 0.4406, + "eval_loss": 19.194971875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 20.11697272983127 + }, + { + "epoch": 154, + "eval_accuracy": 0.4418, + "eval_loss": 17.68252578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 18.992026596917047 + }, + { + "epoch": 155, + "eval_accuracy": 0.438, + "eval_loss": 16.4560037109375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 17.603953161621092 + }, + { + "epoch": 156, + "eval_accuracy": 0.427, + "eval_loss": 15.984819140625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 16.884023179796007 + }, + { + "epoch": 157, + "eval_accuracy": 0.427, + "eval_loss": 14.9954673828125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 16.29205541856554 + }, + { + "epoch": 158, + "eval_accuracy": 0.4164, + "eval_loss": 14.2277224609375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 15.487207347954644 + }, + { + "epoch": 159, + "eval_accuracy": 0.4124, + "eval_loss": 14.00275390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 15.221728933207194 + }, + { + "epoch": 160, + "eval_accuracy": 0.3914, + "eval_loss": 14.5430076171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 14.397079717508952 + }, + { + "epoch": 161, + "eval_accuracy": 0.4068, + "eval_loss": 13.4264064453125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 14.152302168443468 + }, + { + "epoch": 162, + "eval_accuracy": 0.403, + "eval_loss": 12.5586671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 13.672727896457248 + }, + { + "epoch": 163, + "eval_accuracy": 0.391, + "eval_loss": 12.6078431640625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 13.379090177917481 + }, + { + "epoch": 164, + "eval_accuracy": 0.396, + "eval_loss": 11.770783203125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 12.801284927707249 + }, + { + "epoch": 165, + "eval_accuracy": 0.3832, + "eval_loss": 11.81651875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 12.41921732228597 + }, + { + "epoch": 166, + "eval_accuracy": 0.3774, + "eval_loss": 11.469639453125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 12.336118894958496 + }, + { + "epoch": 167, + "eval_accuracy": 0.403, + "eval_loss": 10.486908984375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 11.962089084540473 + }, + { + "epoch": 168, + "eval_accuracy": 0.3634, + "eval_loss": 11.5825072265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 11.826469014824761 + }, + { + "epoch": 169, + "eval_accuracy": 0.3782, + "eval_loss": 11.0532982421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 11.664914200168186 + }, + { + "epoch": 170, + "eval_accuracy": 0.3866, + "eval_loss": 10.220952734375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 11.301445455593534 + }, + { + "epoch": 171, + "eval_accuracy": 0.363, + "eval_loss": 11.0618185546875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 11.134456986490886 + }, + { + "epoch": 172, + "eval_accuracy": 0.3632, + "eval_loss": 11.0839998046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 10.998900364685058 + }, + { + "epoch": 173, + "eval_accuracy": 0.3562, + "eval_loss": 10.4976865234375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 10.807568714396158 + }, + { + "epoch": 174, + "eval_accuracy": 0.3702, + "eval_loss": 11.3861921875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 10.480875332811143 + }, + { + "epoch": 175, + "eval_accuracy": 0.3412, + "eval_loss": 10.180069140625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 10.594809376186795 + }, + { + "epoch": 176, + "eval_accuracy": 0.3464, + "eval_loss": 9.7542810546875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 10.234108125474718 + }, + { + "epoch": 177, + "eval_accuracy": 0.3656, + "eval_loss": 9.5753017578125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 10.208396461486817 + }, + { + "epoch": 178, + "eval_accuracy": 0.3656, + "eval_loss": 8.915300390625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 10.092519539896648 + }, + { + "epoch": 179, + "eval_accuracy": 0.3652, + "eval_loss": 8.97712861328125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 10.079463104248047 + }, + { + "epoch": 180, + "eval_accuracy": 0.2898, + "eval_loss": 10.8206451171875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 9.76496174299452 + }, + { + "epoch": 181, + "eval_accuracy": 0.3582, + "eval_loss": 9.1681966796875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 10.213150364345974 + }, + { + "epoch": 182, + "eval_accuracy": 0.365, + "eval_loss": 8.5061740234375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 9.497418525526259 + }, + { + "epoch": 183, + "eval_accuracy": 0.3328, + "eval_loss": 10.3511533203125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 9.881835655890571 + }, + { + "epoch": 184, + "eval_accuracy": 0.3122, + "eval_loss": 10.184031640625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 9.944425131225586 + }, + { + "epoch": 185, + "eval_accuracy": 0.384, + "eval_loss": 9.4912837890625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 9.53767821384006 + }, + { + "epoch": 186, + "eval_accuracy": 0.3214, + "eval_loss": 9.5392822265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 9.37409217291938 + }, + { + "epoch": 187, + "eval_accuracy": 0.3552, + "eval_loss": 8.4641732421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 9.708195000712077 + }, + { + "epoch": 188, + "eval_accuracy": 0.3766, + "eval_loss": 7.79647529296875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 9.330968306647407 + }, + { + "epoch": 189, + "eval_accuracy": 0.3472, + "eval_loss": 8.42413623046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 9.682495934889051 + }, + { + "epoch": 190, + "eval_accuracy": 0.3468, + "eval_loss": 8.903707421875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 9.455760409037271 + }, + { + "epoch": 191, + "eval_accuracy": 0.3394, + "eval_loss": 9.9961833984375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 9.311902354431153 + }, + { + "epoch": 192, + "eval_accuracy": 0.3286, + "eval_loss": 8.53327626953125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 9.558651632351346 + }, + { + "epoch": 193, + "eval_accuracy": 0.3126, + "eval_loss": 9.6648263671875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 9.197669339497883 + }, + { + "epoch": 194, + "eval_accuracy": 0.373, + "eval_loss": 7.50697080078125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 9.651352426656088 + }, + { + "epoch": 195, + "eval_accuracy": 0.3362, + "eval_loss": 9.22571796875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 8.665485123782688 + }, + { + "epoch": 196, + "eval_accuracy": 0.3336, + "eval_loss": 8.859921875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 8.966691216362847 + }, + { + "epoch": 197, + "eval_accuracy": 0.3446, + "eval_loss": 7.8034123046875, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 8.929122918023003 + }, + { + "epoch": 198, + "eval_accuracy": 0.2802, + "eval_loss": 12.0830345703125, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 9.355501737297907 + }, + { + "epoch": 199, + "eval_accuracy": 0.3612, + "eval_loss": 7.60638193359375, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 9.219078210618761 + }, + { + "epoch": 200, + "eval_accuracy": 0.3874, + "eval_loss": 6.95157197265625, + "eval_split": "validation", + "lr": 0.00010000000000000002, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 8.747222507137723 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.3874, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 6.95157197265625, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2338062336, + "peak_memory_reserved_bytes": 3078619136, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 34.20064926147461, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 1875.3221385478973, + "train_wall_s": 1841.099274635315 + }, + "work": { + "apical_macs_per_example": 3031040, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 27279360000000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 0, + "local_weight_correlation_macs": 619763328000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 619763328000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 0 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 0, + "forward_macs_per_example": 68862592, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 9000000, + "total_forward_equivalent_examples": 9000000, + "total_macs_estimate": 1266806016000000 + } +} |
