diff options
Diffstat (limited to 'results/oral_a_short/sdil_channel_gated_lr0.1.json')
| -rw-r--r-- | results/oral_a_short/sdil_channel_gated_lr0.1.json | 639 |
1 files changed, 639 insertions, 0 deletions
diff --git a/results/oral_a_short/sdil_channel_gated_lr0.1.json b/results/oral_a_short/sdil_channel_gated_lr0.1.json new file mode 100644 index 0000000..a546cab --- /dev/null +++ b/results/oral_a_short/sdil_channel_gated_lr0.1.json @@ -0,0 +1,639 @@ +{ + "apical_warmup": { + "last": { + "calibration_mse": 518.2959050717561, + "prediction_target_cosine": 0.00016553175127441687, + "target_power": 518.2958984375 + }, + "steps": 100 + }, + "architecture": { + "adaptive_apical_parameters": 390592, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_traffic_coefficients": 188416, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 376832, + "residual_scale": 1.0, + "vectorizer_mode": "channel_gated", + "vectorizer_parameters": 13760 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 100, + "alignment_probe": 32, + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda:0", + "epochs": 20, + "eta_A": 0.001, + "eta_P": 0.01, + "eval_every": 0, + "eval_split": "validation", + "learn_P": 0, + "loader_seed": 0, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "cosine", + "max_steps": 0, + "mode": "sdil", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_short/sdil_channel_gated_lr0.1.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "train_limit": 10000, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "channel_gated", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "counters": { + "apical_warmup_examples": 12688, + "calibration_event_examples": 62688, + "causal_scalar_observations": 990, + "logical_batch_loss_queries": 990, + "ordinary_examples": 200000, + "per_example_loss_terms": 125376, + "perturbation_events": 495, + "perturbation_forward_examples": 125376, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.08000707471122344, + "innovation_negative_gradient_cosine": [ + 0.1133790984749794, + 0.12308487296104431, + 0.09794244170188904, + 0.07138215750455856, + 0.05488227307796478, + 0.019371604546904564, + 0.015585479326546192, + 0.04312252625823021, + 0.006569968536496162, + 0.048957690596580505, + 0.041394077241420746, + -0.03040825016796589, + 0.023677144199609756, + 0.11618378013372421, + 0.185902938246727, + 0.21603626012802124, + 0.1565398871898651, + 0.16705910861492157, + -0.004543994553387165 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.1133790984749794, + 0.12308487296104431, + 0.09794244170188904, + 0.07138215750455856, + 0.05488227307796478, + 0.019371604546904564, + 0.015585479326546192, + 0.04312252625823021, + 0.006569968536496162, + 0.048957690596580505, + 0.041394077241420746, + -0.03040825016796589, + 0.023677144199609756, + 0.11618378013372421, + 0.185902938246727, + 0.21603626012802124, + 0.1565398871898651, + 0.16705910861492157, + -0.004543994553387165 + ], + "teaching_negative_gradient_cosine": [ + 0.1133790984749794, + 0.12308487296104431, + 0.09794244170188904, + 0.07138215750455856, + 0.05488227307796478, + 0.019371604546904564, + 0.015585479326546192, + 0.04312252625823021, + 0.006569968536496162, + 0.048957690596580505, + 0.041394077241420746, + -0.03040825016796589, + 0.023677144199609756, + 0.11618378013372421, + 0.185902938246727, + 0.21603626012802124, + 0.1565398871898651, + 0.16705910861492157, + -0.004543994553387165 + ], + "wall_s": 0.10671830177307129 + }, + "epochs": [ + { + "calibration": { + "calibration_mse": 60.43777185808058, + "prediction_target_cosine": -6.181264794266107e-05, + "target_power": 60.43776736048902 + }, + "epoch": 1, + "lr": 0.1, + "output_lr": 0.1, + "step": 79, + "train_examples": 10000, + "train_loss": 2.5334342124938964 + }, + { + "calibration": { + "calibration_mse": 23.54987083374619, + "prediction_target_cosine": -6.96036363511297e-05, + "target_power": 23.549866315204163 + }, + "epoch": 2, + "lr": 0.0993844170297569, + "output_lr": 0.0993844170297569, + "step": 158, + "train_examples": 10000, + "train_loss": 2.5487233268737794 + }, + { + "calibration": { + "calibration_mse": 11.827637325330516, + "prediction_target_cosine": -3.0372441458224223e-06, + "target_power": 11.827633745229358 + }, + "epoch": 3, + "lr": 0.09755282581475769, + "output_lr": 0.09755282581475769, + "step": 237, + "train_examples": 10000, + "train_loss": 2.461765655517578 + }, + { + "calibration": { + "calibration_mse": 30.754039637663446, + "prediction_target_cosine": 1.8163386607575518e-05, + "target_power": 30.754037120982627 + }, + "epoch": 4, + "lr": 0.09455032620941839, + "output_lr": 0.09455032620941839, + "step": 316, + "train_examples": 10000, + "train_loss": 2.3639930412292482 + }, + { + "calibration": { + "calibration_mse": 52.041269661478054, + "prediction_target_cosine": 7.097774633237941e-05, + "target_power": 52.041270009258184 + }, + "epoch": 5, + "lr": 0.09045084971874738, + "output_lr": 0.09045084971874738, + "step": 395, + "train_examples": 10000, + "train_loss": 2.3211933601379395 + }, + { + "calibration": { + "calibration_mse": 42.16324888252278, + "prediction_target_cosine": 7.073739299367894e-05, + "target_power": 42.16324767635568 + }, + "epoch": 6, + "lr": 0.08535533905932738, + "output_lr": 0.08535533905932738, + "step": 474, + "train_examples": 10000, + "train_loss": 2.2405354804992674 + }, + { + "calibration": { + "calibration_mse": 96.60961034550614, + "prediction_target_cosine": -2.6760619008438326e-05, + "target_power": 96.60960379281771 + }, + "epoch": 7, + "lr": 0.07938926261462366, + "output_lr": 0.07938926261462366, + "step": 553, + "train_examples": 10000, + "train_loss": 2.1257784160614013 + }, + { + "calibration": { + "calibration_mse": 95.43317693296298, + "prediction_target_cosine": -1.0208342923692714e-05, + "target_power": 95.43316866958713 + }, + "epoch": 8, + "lr": 0.07269952498697733, + "output_lr": 0.07269952498697733, + "step": 632, + "train_examples": 10000, + "train_loss": 2.2211479858398437 + }, + { + "calibration": { + "calibration_mse": 66.98153618496397, + "prediction_target_cosine": -1.2381250456731113e-05, + "target_power": 66.98153039761213 + }, + "epoch": 9, + "lr": 0.06545084971874737, + "output_lr": 0.06545084971874737, + "step": 711, + "train_examples": 10000, + "train_loss": 2.083836813926697 + }, + { + "calibration": { + "calibration_mse": 146.6930261971186, + "prediction_target_cosine": -5.327456506330552e-07, + "target_power": 146.69302486945753 + }, + "epoch": 10, + "lr": 0.05782172325201155, + "output_lr": 0.05782172325201155, + "step": 790, + "train_examples": 10000, + "train_loss": 1.9434806201934816 + }, + { + "calibration": { + "calibration_mse": 80.2759034222883, + "prediction_target_cosine": 4.8706224176594515e-05, + "target_power": 80.27590192608213 + }, + "epoch": 11, + "lr": 0.05, + "output_lr": 0.05, + "step": 869, + "train_examples": 10000, + "train_loss": 1.9414970417022706 + }, + { + "calibration": { + "calibration_mse": 96.45602903597789, + "prediction_target_cosine": -4.129046053586153e-05, + "target_power": 96.45602758309765 + }, + "epoch": 12, + "lr": 0.04217827674798845, + "output_lr": 0.04217827674798845, + "step": 948, + "train_examples": 10000, + "train_loss": 1.8372822145462036 + }, + { + "calibration": { + "calibration_mse": 176.84556970363073, + "prediction_target_cosine": -1.9519702580488654e-05, + "target_power": 176.84556127799596 + }, + "epoch": 13, + "lr": 0.03454915028125263, + "output_lr": 0.03454915028125263, + "step": 1027, + "train_examples": 10000, + "train_loss": 1.7941364185333253 + }, + { + "calibration": { + "calibration_mse": 75.1211485477286, + "prediction_target_cosine": 0.00010757188693459416, + "target_power": 75.12114817083045 + }, + "epoch": 14, + "lr": 0.027300475013022664, + "output_lr": 0.027300475013022664, + "step": 1106, + "train_examples": 10000, + "train_loss": 1.7450150983810424 + }, + { + "calibration": { + "calibration_mse": 114.3603014383627, + "prediction_target_cosine": -9.490821681384313e-07, + "target_power": 114.36030431791491 + }, + "epoch": 15, + "lr": 0.02061073738537635, + "output_lr": 0.02061073738537635, + "step": 1185, + "train_examples": 10000, + "train_loss": 1.6714060207366943 + }, + { + "calibration": { + "calibration_mse": 122.73489581566791, + "prediction_target_cosine": 4.2819634654651794e-05, + "target_power": 122.7348966235018 + }, + "epoch": 16, + "lr": 0.014644660940672627, + "output_lr": 0.014644660940672627, + "step": 1264, + "train_examples": 10000, + "train_loss": 1.6516524749755859 + }, + { + "calibration": { + "calibration_mse": 42.00022924316126, + "prediction_target_cosine": -4.6748324469624814e-05, + "target_power": 42.000227053771205 + }, + "epoch": 17, + "lr": 0.009549150281252633, + "output_lr": 0.009549150281252633, + "step": 1343, + "train_examples": 10000, + "train_loss": 1.6312512542724609 + }, + { + "calibration": { + "calibration_mse": 146.15481284242125, + "prediction_target_cosine": -1.823520416534038e-05, + "target_power": 146.15481366035732 + }, + "epoch": 18, + "lr": 0.005449673790581611, + "output_lr": 0.005449673790581611, + "step": 1422, + "train_examples": 10000, + "train_loss": 1.592088447189331 + }, + { + "calibration": { + "calibration_mse": 50.12404485339623, + "prediction_target_cosine": 7.522739697549936e-05, + "target_power": 50.12404224670939 + }, + "epoch": 19, + "lr": 0.0024471741852423235, + "output_lr": 0.0024471741852423235, + "step": 1501, + "train_examples": 10000, + "train_loss": 1.5661032293319703 + }, + { + "calibration": { + "calibration_mse": 81.33399665307944, + "prediction_target_cosine": 8.698869592416884e-05, + "target_power": 81.33399560536073 + }, + "epoch": 20, + "lr": 0.0006155829702431171, + "output_lr": 0.0006155829702431171, + "step": 1580, + "train_examples": 10000, + "train_loss": 1.5665943405151368 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.414, + "epoch": 20, + "evaluation_split": "validation", + "finite": true, + "loss": 1.597408447265625, + "step": 1580 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda:0", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 1733492736, + "peak_memory_reserved_bytes": 2858418176, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "a1d060e6d96d4ab973d16cd75149dd61899c7c95", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "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": 10000, + "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": 6.4902708530426025, + "evaluation_wall_s": 0.36696362495422363, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 100.41240739822388, + "train_wall_s": 93.55240631103516 + }, + "work": { + "apical_macs_per_example": 202176, + "causal_scalar_observations": 990, + "components": { + "apical_projection_macs": 43000409088, + "apical_regression_macs": 12674009088, + "bp_reverse_macs_estimate": 0, + "local_weight_correlation_macs": 8110208000000, + "ordinary_forward_macs": 8110208000000, + "perturbation_forward_macs": 5084127191040, + "warmup_clean_forward_macs": 514511595520 + }, + "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; elementwise nonlinearities and optimizer arithmetic excluded", + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 990, + "per_example_cross_entropy_terms": 125376, + "total_clean_forward_examples": 212688, + "total_forward_equivalent_examples": 338064, + "total_macs_estimate": 21874729204736 + } +} |