diff options
Diffstat (limited to 'results/mirror_short/wm_lr0.1.json')
| -rw-r--r-- | results/mirror_short/wm_lr0.1.json | 732 |
1 files changed, 732 insertions, 0 deletions
diff --git a/results/mirror_short/wm_lr0.1.json b/results/mirror_short/wm_lr0.1.json new file mode 100644 index 0000000..e767b20 --- /dev/null +++ b/results/mirror_short/wm_lr0.1.json @@ -0,0 +1,732 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 267904, + "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_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "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": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "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": "/home/yurenh2/sdrn/data", + "depth": 20, + "device": "cuda", + "epochs": 20, + "eta_A": 0.01, + "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, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 20, + "mode": "wm", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/mirror_short/wm_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": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "local_parent_child_response", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 119, + "mirror_events": 119, + "mirror_readout_examples": 121856, + "ordinary_examples": 200000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.9392559031645457, + "feedback_forward_cosine": [ + 0.9959778785705566, + 0.995803713798523, + 0.9965077638626099, + 0.9960940480232239, + 0.9964685440063477, + 0.9963351488113403, + 0.9851020574569702, + 0.9706052541732788, + 0.971168041229248, + 0.9713930487632751, + 0.9713956117630005, + 0.9704664349555969, + 0.8957585096359253, + 0.8113258481025696, + 0.8151601552963257, + 0.813377857208252, + 0.8184781074523926, + 0.8133856058120728, + 0.9983060359954834 + ], + "feedback_forward_norm_ratio": [ + 1.0097280740737915, + 1.0134508609771729, + 0.9960123896598816, + 0.9936652779579163, + 1.0030072927474976, + 1.0111799240112305, + 1.010083794593811, + 1.0290576219558716, + 1.0363725423812866, + 1.0423928499221802, + 1.0202832221984863, + 1.0177212953567505, + 1.1110841035842896, + 1.2333585023880005, + 1.228413462638855, + 1.2249541282653809, + 1.2218670845031738, + 1.2358826398849487, + 0.9922027587890625 + ], + "innovation_negative_gradient_cosine": [ + 0.9430115818977356, + 0.9422318935394287, + 0.9404978156089783, + 0.9424664974212646, + 0.9352189302444458, + 0.9321087002754211, + 0.9271675944328308, + 0.9273618459701538, + 0.921048641204834, + 0.9095789194107056, + 0.9105615615844727, + 0.8976923823356628, + 0.8929047584533691, + 0.8215304613113403, + 0.9227415323257446, + 0.8373146057128906, + 0.9558568000793457, + 0.8506318926811218, + 0.9982843399047852 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.9430115818977356, + 0.9422318935394287, + 0.9404978156089783, + 0.9424664974212646, + 0.9352189302444458, + 0.9321087002754211, + 0.9271675944328308, + 0.9273618459701538, + 0.921048641204834, + 0.9095789194107056, + 0.9105615615844727, + 0.8976923823356628, + 0.8929047584533691, + 0.8215304613113403, + 0.9227415323257446, + 0.8373146057128906, + 0.9558568000793457, + 0.8506318926811218, + 0.9982843399047852 + ], + "teaching_negative_gradient_cosine": [ + 0.9430115818977356, + 0.9422318935394287, + 0.9404978156089783, + 0.9424664974212646, + 0.9352189302444458, + 0.9321087002754211, + 0.9271675944328308, + 0.9273618459701538, + 0.921048641204834, + 0.9095789194107056, + 0.9105615615844727, + 0.8976923823356628, + 0.8929047584533691, + 0.8215304613113403, + 0.9227415323257446, + 0.8373146057128906, + 0.9558568000793457, + 0.8506318926811218, + 0.9982843399047852 + ], + "wall_s": 0.10048437118530273 + }, + "epochs": [ + { + "epoch": 1, + "lr": 0.1, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17882412437844575, + "mirror_update_rms": 0.16884532120011128, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.1, + "step": 79, + "train_examples": 10000, + "train_loss": 2.8801190078735353 + }, + { + "epoch": 2, + "lr": 0.0993844170297569, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.18083112954536643, + "mirror_update_rms": 0.17052920790529394, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.0993844170297569, + "step": 158, + "train_examples": 10000, + "train_loss": 3.266864613342285 + }, + { + "epoch": 3, + "lr": 0.09755282581475769, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.18097631623416152, + "mirror_update_rms": 0.1705323497743055, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.09755282581475769, + "step": 237, + "train_examples": 10000, + "train_loss": 2.6971860549926756 + }, + { + "epoch": 4, + "lr": 0.09455032620941839, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17874982974141068, + "mirror_update_rms": 0.1682343265855635, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.09455032620941839, + "step": 316, + "train_examples": 10000, + "train_loss": 2.395349690246582 + }, + { + "epoch": 5, + "lr": 0.09045084971874738, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1804126488631345, + "mirror_update_rms": 0.1697212004375701, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.09045084971874738, + "step": 395, + "train_examples": 10000, + "train_loss": 2.2874154033660887 + }, + { + "epoch": 6, + "lr": 0.08535533905932738, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1793549907915837, + "mirror_update_rms": 0.1686189018588876, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.08535533905932738, + "step": 474, + "train_examples": 10000, + "train_loss": 2.1545730239868166 + }, + { + "epoch": 7, + "lr": 0.07938926261462366, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17956814557177364, + "mirror_update_rms": 0.16887362777752254, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.07938926261462366, + "step": 553, + "train_examples": 10000, + "train_loss": 2.048876251602173 + }, + { + "epoch": 8, + "lr": 0.07269952498697733, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.18047323995585826, + "mirror_update_rms": 0.1694450740766401, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.07269952498697733, + "step": 632, + "train_examples": 10000, + "train_loss": 2.0588326959609984 + }, + { + "epoch": 9, + "lr": 0.06545084971874737, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.18041623437220405, + "mirror_update_rms": 0.16956658080232978, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.06545084971874737, + "step": 711, + "train_examples": 10000, + "train_loss": 1.7462074737548827 + }, + { + "epoch": 10, + "lr": 0.05782172325201155, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.18025565517994876, + "mirror_update_rms": 0.16935577088897355, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.05782172325201155, + "step": 790, + "train_examples": 10000, + "train_loss": 1.7115148830413818 + }, + { + "epoch": 11, + "lr": 0.05, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1799140166082034, + "mirror_update_rms": 0.1689372311595095, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.05, + "step": 869, + "train_examples": 10000, + "train_loss": 1.585644446182251 + }, + { + "epoch": 12, + "lr": 0.04217827674798845, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17996184738981771, + "mirror_update_rms": 0.1688740446763439, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.04217827674798845, + "step": 948, + "train_examples": 10000, + "train_loss": 1.4297412275314332 + }, + { + "epoch": 13, + "lr": 0.03454915028125263, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17954532857647026, + "mirror_update_rms": 0.1683502653949776, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.03454915028125263, + "step": 1027, + "train_examples": 10000, + "train_loss": 1.3605230285644532 + }, + { + "epoch": 14, + "lr": 0.027300475013022664, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17886661544286583, + "mirror_update_rms": 0.16759217960599557, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.027300475013022664, + "step": 1106, + "train_examples": 10000, + "train_loss": 1.2885562606811523 + }, + { + "epoch": 15, + "lr": 0.02061073738537635, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17960541823280596, + "mirror_update_rms": 0.168391820135339, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.02061073738537635, + "step": 1185, + "train_examples": 10000, + "train_loss": 1.1695917907714843 + }, + { + "epoch": 16, + "lr": 0.014644660940672627, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.178230393488791, + "mirror_update_rms": 0.166971179061675, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.014644660940672627, + "step": 1264, + "train_examples": 10000, + "train_loss": 1.1232062786102295 + }, + { + "epoch": 17, + "lr": 0.009549150281252633, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17890263131667364, + "mirror_update_rms": 0.16779422421858642, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.009549150281252633, + "step": 1343, + "train_examples": 10000, + "train_loss": 1.062957594680786 + }, + { + "epoch": 18, + "lr": 0.005449673790581611, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.18061316378720993, + "mirror_update_rms": 0.16914562366260577, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.005449673790581611, + "step": 1422, + "train_examples": 10000, + "train_loss": 1.013933398246765 + }, + { + "epoch": 19, + "lr": 0.0024471741852423235, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1794060399647695, + "mirror_update_rms": 0.16813953202750714, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.0024471741852423235, + "step": 1501, + "train_examples": 10000, + "train_loss": 0.980167069530487 + }, + { + "epoch": 20, + "lr": 0.0006155829702431171, + "mirror": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.17921597333176215, + "mirror_update_rms": 0.16791399894923217, + "readout_batch_size": 1024.0 + }, + "output_lr": 0.0006155829702431171, + "step": 1580, + "train_examples": 10000, + "train_loss": 0.964193574333191 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.6404, + "epoch": 20, + "evaluation_split": "validation", + "finite": true, + "loss": 1.0210670532226562, + "step": 1580 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 1730831872, + "peak_memory_reserved_bytes": 2501902336, + "torch_version": "2.3.1+cu118" + }, + "mirror_warmup": { + "first": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.1821084153449638, + "mirror_update_rms": 0.19537643727527548, + "readout_batch_size": 1024 + }, + "last": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.18073508053859827, + "mirror_update_rms": 0.17050672066264347, + "readout_batch_size": 1024 + }, + "mean": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1793198434405597, + "mirror_update_rms": 0.17517745811581048, + "readout_batch_size": 1024.0 + }, + "steps": 20 + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "dba3eb29d5797c1b74d7e1e781523afad9faeb8e", + "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": 0.0, + "evaluation_wall_s": 0.36042237281799316, + "mirror_warmup_wall_s": 0.4621908664703369, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 90.53707838058472, + "train_wall_s": 89.71193861961365 + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 8021734400000, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "local_weight_correlation_macs": 8110208000000, + "mirror_local_correlation_macs": 4850843648, + "mirror_response_macs": 4850843648, + "ordinary_forward_macs": 8110208000000, + "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; elementwise nonlinearities and optimizer arithmetic excluded", + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 119, + "mirror_readout_probe_examples": 121856, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 200000, + "total_forward_equivalent_examples": 200000, + "total_macs_estimate": 24251852087296 + } +} |