diff options
Diffstat (limited to 'results/oral_a_short/sdil_spatial_template_lr0.01.json')
| -rw-r--r-- | results/oral_a_short/sdil_spatial_template_lr0.01.json | 639 |
1 files changed, 639 insertions, 0 deletions
diff --git a/results/oral_a_short/sdil_spatial_template_lr0.01.json b/results/oral_a_short/sdil_spatial_template_lr0.01.json new file mode 100644 index 0000000..f040658 --- /dev/null +++ b/results/oral_a_short/sdil_spatial_template_lr0.01.json @@ -0,0 +1,639 @@ +{ + "apical_warmup": { + "last": { + "calibration_mse": 518.2958984375, + "prediction_target_cosine": 0.0001586056141805217, + "target_power": 518.2958984375 + }, + "steps": 100 + }, + "architecture": { + "adaptive_apical_parameters": 2260992, + "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": "spatial_template", + "vectorizer_parameters": 1884160 + }, + "args": { + "a_scale": 0.0, + "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": 0.01, + "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_spatial_template_lr0.01.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": "spatial_template", + "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": 0.001612375567977627, + "innovation_negative_gradient_cosine": [ + 0.001027052290737629, + -0.0022626847494393587, + 0.003048800863325596, + 0.0009945780038833618, + 0.003387338714674115, + 0.0034791682846844196, + 0.0036151609383523464, + 0.0016754134558141232, + -0.00018921078299172223, + 0.0040879761800169945, + 0.006074178963899612, + 0.0037323604337871075, + 0.002060003811493516, + 0.007457074709236622, + 0.008745873346924782, + 0.01761583611369133, + 0.004088421352207661, + 0.02299105003476143, + 0.015363558195531368 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.001027052290737629, + -0.0022626847494393587, + 0.003048800863325596, + 0.0009945780038833618, + 0.003387338714674115, + 0.0034791682846844196, + 0.0036151609383523464, + 0.0016754134558141232, + -0.00018921078299172223, + 0.0040879761800169945, + 0.006074178963899612, + 0.0037323604337871075, + 0.002060003811493516, + 0.007457074709236622, + 0.008745873346924782, + 0.01761583611369133, + 0.004088421352207661, + 0.02299105003476143, + 0.015363558195531368 + ], + "teaching_negative_gradient_cosine": [ + 0.001027052290737629, + -0.0022626847494393587, + 0.003048800863325596, + 0.0009945780038833618, + 0.003387338714674115, + 0.0034791682846844196, + 0.0036151609383523464, + 0.0016754134558141232, + -0.00018921078299172223, + 0.0040879761800169945, + 0.006074178963899612, + 0.0037323604337871075, + 0.002060003811493516, + 0.007457074709236622, + 0.008745873346924782, + 0.01761583611369133, + 0.004088421352207661, + 0.02299105003476143, + 0.015363558195531368 + ], + "wall_s": 0.1027383804321289 + }, + "epochs": [ + { + "calibration": { + "calibration_mse": 311.4595984723095, + "prediction_target_cosine": -2.2504479337353996e-05, + "target_power": 311.4595591177434 + }, + "epoch": 1, + "lr": 0.01, + "output_lr": 0.1, + "step": 79, + "train_examples": 10000, + "train_loss": 2.774099029159546 + }, + { + "calibration": { + "calibration_mse": 1526.5578844609468, + "prediction_target_cosine": 2.554489113023116e-06, + "target_power": 1526.557862022649 + }, + "epoch": 2, + "lr": 0.009938441702975689, + "output_lr": 0.0993844170297569, + "step": 158, + "train_examples": 10000, + "train_loss": 2.431015293121338 + }, + { + "calibration": { + "calibration_mse": 1242.6608512818873, + "prediction_target_cosine": -3.991147778175767e-05, + "target_power": 1242.6608249527821 + }, + "epoch": 3, + "lr": 0.009755282581475769, + "output_lr": 0.09755282581475769, + "step": 237, + "train_examples": 10000, + "train_loss": 2.466135598754883 + }, + { + "calibration": { + "calibration_mse": 634.9810205836453, + "prediction_target_cosine": -3.91370314985448e-06, + "target_power": 634.980954601372 + }, + "epoch": 4, + "lr": 0.00945503262094184, + "output_lr": 0.09455032620941839, + "step": 316, + "train_examples": 10000, + "train_loss": 2.6161225547790528 + }, + { + "calibration": { + "calibration_mse": 250.2808392957501, + "prediction_target_cosine": 4.677621205119292e-05, + "target_power": 250.28076969955282 + }, + "epoch": 5, + "lr": 0.009045084971874737, + "output_lr": 0.09045084971874738, + "step": 395, + "train_examples": 10000, + "train_loss": 2.443944037628174 + }, + { + "calibration": { + "calibration_mse": 280.3610884783708, + "prediction_target_cosine": 3.712464558796983e-08, + "target_power": 280.3610162495271 + }, + "epoch": 6, + "lr": 0.008535533905932738, + "output_lr": 0.08535533905932738, + "step": 474, + "train_examples": 10000, + "train_loss": 2.4743158672332766 + }, + { + "calibration": { + "calibration_mse": 397.6820018188137, + "prediction_target_cosine": -3.488966965419155e-05, + "target_power": 397.68190724253657 + }, + "epoch": 7, + "lr": 0.007938926261462366, + "output_lr": 0.07938926261462366, + "step": 553, + "train_examples": 10000, + "train_loss": 2.401729510116577 + }, + { + "calibration": { + "calibration_mse": 694.6581847718459, + "prediction_target_cosine": -1.5154622745509205e-05, + "target_power": 694.6580754621829 + }, + "epoch": 8, + "lr": 0.007269952498697734, + "output_lr": 0.07269952498697733, + "step": 632, + "train_examples": 10000, + "train_loss": 2.340212138366699 + }, + { + "calibration": { + "calibration_mse": 447.4696043644261, + "prediction_target_cosine": -5.656511341783755e-06, + "target_power": 447.4694925629574 + }, + "epoch": 9, + "lr": 0.006545084971874737, + "output_lr": 0.06545084971874737, + "step": 711, + "train_examples": 10000, + "train_loss": 2.2473524238586426 + }, + { + "calibration": { + "calibration_mse": 225.01660751981012, + "prediction_target_cosine": 2.417077360226098e-05, + "target_power": 225.016517061643 + }, + "epoch": 10, + "lr": 0.0057821723252011546, + "output_lr": 0.05782172325201155, + "step": 790, + "train_examples": 10000, + "train_loss": 2.1798657495498657 + }, + { + "calibration": { + "calibration_mse": 518.3563566866129, + "prediction_target_cosine": 1.0120506274893608e-05, + "target_power": 518.3562297463417 + }, + "epoch": 11, + "lr": 0.005, + "output_lr": 0.05, + "step": 869, + "train_examples": 10000, + "train_loss": 2.142610039138794 + }, + { + "calibration": { + "calibration_mse": 298.18500603163864, + "prediction_target_cosine": 2.516028858223128e-06, + "target_power": 298.1849143187552 + }, + "epoch": 12, + "lr": 0.004217827674798845, + "output_lr": 0.04217827674798845, + "step": 948, + "train_examples": 10000, + "train_loss": 2.0082340023040772 + }, + { + "calibration": { + "calibration_mse": 816.293761626534, + "prediction_target_cosine": -3.1195302557482935e-05, + "target_power": 816.2936095196268 + }, + "epoch": 13, + "lr": 0.003454915028125263, + "output_lr": 0.03454915028125263, + "step": 1027, + "train_examples": 10000, + "train_loss": 1.9961866901397705 + }, + { + "calibration": { + "calibration_mse": 557.8168947697982, + "prediction_target_cosine": 1.1375643767241642e-05, + "target_power": 557.8167687418669 + }, + "epoch": 14, + "lr": 0.0027300475013022664, + "output_lr": 0.027300475013022664, + "step": 1106, + "train_examples": 10000, + "train_loss": 1.9453593284606934 + }, + { + "calibration": { + "calibration_mse": 184.58966761069047, + "prediction_target_cosine": -4.041631141428753e-05, + "target_power": 184.58953192374145 + }, + "epoch": 15, + "lr": 0.0020610737385376348, + "output_lr": 0.02061073738537635, + "step": 1185, + "train_examples": 10000, + "train_loss": 1.9042009094238281 + }, + { + "calibration": { + "calibration_mse": 281.9687953793978, + "prediction_target_cosine": 2.328571111638906e-05, + "target_power": 281.9686650505601 + }, + "epoch": 16, + "lr": 0.0014644660940672626, + "output_lr": 0.014644660940672627, + "step": 1264, + "train_examples": 10000, + "train_loss": 1.8901697011947631 + }, + { + "calibration": { + "calibration_mse": 325.46421335831934, + "prediction_target_cosine": 5.358104560800165e-05, + "target_power": 325.4640667039415 + }, + "epoch": 17, + "lr": 0.0009549150281252633, + "output_lr": 0.009549150281252633, + "step": 1343, + "train_examples": 10000, + "train_loss": 1.8364084032058716 + }, + { + "calibration": { + "calibration_mse": 344.07036824645877, + "prediction_target_cosine": -6.079723779001693e-05, + "target_power": 344.07023295721285 + }, + "epoch": 18, + "lr": 0.0005449673790581611, + "output_lr": 0.005449673790581611, + "step": 1422, + "train_examples": 10000, + "train_loss": 1.816278349876404 + }, + { + "calibration": { + "calibration_mse": 371.9894568442003, + "prediction_target_cosine": -2.267388030846593e-05, + "target_power": 371.9893169559863 + }, + "epoch": 19, + "lr": 0.00024471741852423234, + "output_lr": 0.0024471741852423235, + "step": 1501, + "train_examples": 10000, + "train_loss": 1.7986092542648315 + }, + { + "calibration": { + "calibration_mse": 531.5486738582225, + "prediction_target_cosine": -2.485824704439809e-05, + "target_power": 531.548502974843 + }, + "epoch": 20, + "lr": 6.15582970243117e-05, + "output_lr": 0.0006155829702431171, + "step": 1580, + "train_examples": 10000, + "train_loss": 1.7901613536834717 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.3572, + "epoch": 20, + "evaluation_split": "validation", + "finite": true, + "loss": 1.7704051513671875, + "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": 1740965888, + "peak_memory_reserved_bytes": 2866806784, + "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": 6.035719156265259, + "evaluation_wall_s": 0.36167001724243164, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 95.64743661880493, + "train_wall_s": 89.2470269203186 + }, + "work": { + "apical_macs_per_example": 1884160, + "causal_scalar_observations": 990, + "components": { + "apical_projection_macs": 400738222080, + "apical_regression_macs": 118114222080, + "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": 22337907230720 + } +} |