diff options
Diffstat (limited to 'results/oral_a_short/sdil_channel_gated_lr0.03.json')
| -rw-r--r-- | results/oral_a_short/sdil_channel_gated_lr0.03.json | 639 |
1 files changed, 639 insertions, 0 deletions
diff --git a/results/oral_a_short/sdil_channel_gated_lr0.03.json b/results/oral_a_short/sdil_channel_gated_lr0.03.json new file mode 100644 index 0000000..3932283 --- /dev/null +++ b/results/oral_a_short/sdil_channel_gated_lr0.03.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": 0.03, + "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_lr0.03.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": 0.08889565182228883, + "innovation_negative_gradient_cosine": [ + 0.05685567483305931, + 0.13134989142417908, + 0.08806467056274414, + 0.09720344841480255, + 0.06525854766368866, + 0.09464167803525925, + 0.02397942543029785, + 0.06854601949453354, + 0.07409851253032684, + 0.06666077673435211, + 0.07716609537601471, + 0.09574130177497864, + 0.034181758761405945, + 0.07528112828731537, + 0.1363137811422348, + 0.22182485461235046, + 0.13769981265068054, + 0.27749329805374146, + 0.11678922176361084 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.05685567483305931, + 0.13134989142417908, + 0.08806467056274414, + 0.09720344841480255, + 0.06525854766368866, + 0.09464167803525925, + 0.02397942543029785, + 0.06854601949453354, + 0.07409851253032684, + 0.06666077673435211, + 0.07716609537601471, + 0.09574130177497864, + 0.034181758761405945, + 0.07528112828731537, + 0.1363137811422348, + 0.22182485461235046, + 0.13769981265068054, + 0.27749329805374146, + 0.11678922176361084 + ], + "teaching_negative_gradient_cosine": [ + 0.05685567483305931, + 0.13134989142417908, + 0.08806467056274414, + 0.09720344841480255, + 0.06525854766368866, + 0.09464167803525925, + 0.02397942543029785, + 0.06854601949453354, + 0.07409851253032684, + 0.06666077673435211, + 0.07716609537601471, + 0.09574130177497864, + 0.034181758761405945, + 0.07528112828731537, + 0.1363137811422348, + 0.22182485461235046, + 0.13769981265068054, + 0.27749329805374146, + 0.11678922176361084 + ], + "wall_s": 0.10262227058410645 + }, + "epochs": [ + { + "calibration": { + "calibration_mse": 120.94099242469682, + "prediction_target_cosine": 1.506113768211827e-05, + "target_power": 120.94099207731854 + }, + "epoch": 1, + "lr": 0.03, + "output_lr": 0.1, + "step": 79, + "train_examples": 10000, + "train_loss": 2.7430426933288574 + }, + { + "calibration": { + "calibration_mse": 90.8829443671228, + "prediction_target_cosine": 4.748458338215727e-05, + "target_power": 90.88294743040333 + }, + "epoch": 2, + "lr": 0.029815325108927065, + "output_lr": 0.0993844170297569, + "step": 158, + "train_examples": 10000, + "train_loss": 2.6882332107543947 + }, + { + "calibration": { + "calibration_mse": 129.28391048064168, + "prediction_target_cosine": 2.1492919926159852e-05, + "target_power": 129.2838982463207 + }, + "epoch": 3, + "lr": 0.029265847744427302, + "output_lr": 0.09755282581475769, + "step": 237, + "train_examples": 10000, + "train_loss": 2.7864545795440674 + }, + { + "calibration": { + "calibration_mse": 114.01519549503767, + "prediction_target_cosine": 7.391323251578894e-05, + "target_power": 114.01519008284303 + }, + "epoch": 4, + "lr": 0.028365097862825515, + "output_lr": 0.09455032620941839, + "step": 316, + "train_examples": 10000, + "train_loss": 2.278718226623535 + }, + { + "calibration": { + "calibration_mse": 227.3724495573293, + "prediction_target_cosine": 4.533821080774298e-05, + "target_power": 227.37244714005323 + }, + "epoch": 5, + "lr": 0.02713525491562421, + "output_lr": 0.09045084971874738, + "step": 395, + "train_examples": 10000, + "train_loss": 2.3870798393249513 + }, + { + "calibration": { + "calibration_mse": 166.20845720914784, + "prediction_target_cosine": 6.765680817807511e-05, + "target_power": 166.20845642938562 + }, + "epoch": 6, + "lr": 0.02560660171779821, + "output_lr": 0.08535533905932738, + "step": 474, + "train_examples": 10000, + "train_loss": 2.2577471870422365 + }, + { + "calibration": { + "calibration_mse": 170.727521433827, + "prediction_target_cosine": -5.7050105486161436e-05, + "target_power": 170.72752845326198 + }, + "epoch": 7, + "lr": 0.023816778784387097, + "output_lr": 0.07938926261462366, + "step": 553, + "train_examples": 10000, + "train_loss": 2.12043060874939 + }, + { + "calibration": { + "calibration_mse": 209.52628865981265, + "prediction_target_cosine": -2.0228797603081682e-05, + "target_power": 209.52628848590894 + }, + "epoch": 8, + "lr": 0.0218098574960932, + "output_lr": 0.07269952498697733, + "step": 632, + "train_examples": 10000, + "train_loss": 2.059708542060852 + }, + { + "calibration": { + "calibration_mse": 141.29861236396044, + "prediction_target_cosine": -5.216942692074475e-06, + "target_power": 141.2986095050107 + }, + "epoch": 9, + "lr": 0.01963525491562421, + "output_lr": 0.06545084971874737, + "step": 711, + "train_examples": 10000, + "train_loss": 1.9873547559738158 + }, + { + "calibration": { + "calibration_mse": 252.29498001822952, + "prediction_target_cosine": 6.647677482561257e-05, + "target_power": 252.29497267383599 + }, + "epoch": 10, + "lr": 0.017346516975603462, + "output_lr": 0.05782172325201155, + "step": 790, + "train_examples": 10000, + "train_loss": 1.9427831512451172 + }, + { + "calibration": { + "calibration_mse": 112.07356150142402, + "prediction_target_cosine": -7.042171144742458e-05, + "target_power": 112.0735553999958 + }, + "epoch": 11, + "lr": 0.015, + "output_lr": 0.05, + "step": 869, + "train_examples": 10000, + "train_loss": 1.8960050271987916 + }, + { + "calibration": { + "calibration_mse": 275.5549939046497, + "prediction_target_cosine": 2.7840606439390747e-05, + "target_power": 275.55498504243127 + }, + "epoch": 12, + "lr": 0.012653483024396533, + "output_lr": 0.04217827674798845, + "step": 948, + "train_examples": 10000, + "train_loss": 1.7632465076446533 + }, + { + "calibration": { + "calibration_mse": 391.5784136378246, + "prediction_target_cosine": -4.5330952303582855e-05, + "target_power": 391.57838901851494 + }, + "epoch": 13, + "lr": 0.01036474508437579, + "output_lr": 0.03454915028125263, + "step": 1027, + "train_examples": 10000, + "train_loss": 1.7816886276245116 + }, + { + "calibration": { + "calibration_mse": 288.2343263507861, + "prediction_target_cosine": 9.16609312615192e-05, + "target_power": 288.23433236402013 + }, + "epoch": 14, + "lr": 0.0081901425039068, + "output_lr": 0.027300475013022664, + "step": 1106, + "train_examples": 10000, + "train_loss": 1.7592916584014893 + }, + { + "calibration": { + "calibration_mse": 162.82111501402179, + "prediction_target_cosine": 5.7469316424673235e-05, + "target_power": 162.82111272915546 + }, + "epoch": 15, + "lr": 0.006183221215612904, + "output_lr": 0.02061073738537635, + "step": 1185, + "train_examples": 10000, + "train_loss": 1.683985795021057 + }, + { + "calibration": { + "calibration_mse": 177.96064978913256, + "prediction_target_cosine": -4.3211184289433225e-05, + "target_power": 177.96064038461319 + }, + "epoch": 16, + "lr": 0.004393398282201788, + "output_lr": 0.014644660940672627, + "step": 1264, + "train_examples": 10000, + "train_loss": 1.6693348972320556 + }, + { + "calibration": { + "calibration_mse": 116.30543024762251, + "prediction_target_cosine": 5.207735526040399e-05, + "target_power": 116.30542734543103 + }, + "epoch": 17, + "lr": 0.00286474508437579, + "output_lr": 0.009549150281252633, + "step": 1343, + "train_examples": 10000, + "train_loss": 1.6360828407287598 + }, + { + "calibration": { + "calibration_mse": 224.51228453390218, + "prediction_target_cosine": 8.341093247343407e-05, + "target_power": 224.5122794786468 + }, + "epoch": 18, + "lr": 0.001634902137174483, + "output_lr": 0.005449673790581611, + "step": 1422, + "train_examples": 10000, + "train_loss": 1.614950588989258 + }, + { + "calibration": { + "calibration_mse": 227.45811103251967, + "prediction_target_cosine": 7.730374096652188e-05, + "target_power": 227.45810761075955 + }, + "epoch": 19, + "lr": 0.000734152255572697, + "output_lr": 0.0024471741852423235, + "step": 1501, + "train_examples": 10000, + "train_loss": 1.5980025550842285 + }, + { + "calibration": { + "calibration_mse": 233.88802596526028, + "prediction_target_cosine": 6.816316473038772e-06, + "target_power": 233.88802621876758 + }, + "epoch": 20, + "lr": 0.0001846748910729351, + "output_lr": 0.0006155829702431171, + "step": 1580, + "train_examples": 10000, + "train_loss": 1.598771397781372 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.4198, + "epoch": 20, + "evaluation_split": "validation", + "finite": true, + "loss": 1.60240986328125, + "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": 6.487434387207031, + "evaluation_wall_s": 0.372774600982666, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 100.67689514160156, + "train_wall_s": 93.81352663040161 + }, + "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 + } +} |