diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 13:28:23 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 13:28:23 -0500 |
| commit | c27e0824a41e6c54dd9079f17c20c306c5f2e9c9 (patch) | |
| tree | 052d79f45fe1f8fc4e2de853fec6e592acdfed58 /results | |
| parent | d4a521072a25d8afdd62021f43538d7dad7586fd (diff) | |
results: close hierarchical task-scalar calibration gate
Diffstat (limited to 'results')
| -rw-r--r-- | results/oral_a_v4_calibration/fixed_hfa.json | 416 | ||||
| -rw-r--r-- | results/oral_a_v4_calibration/learned_hfa_etaA0.1.json | 437 | ||||
| -rw-r--r-- | results/oral_a_v4_calibration/learned_hfa_etaA1.0.json | 437 | ||||
| -rw-r--r-- | results/oral_a_v4_calibration/learned_hfa_etaA10.0.json | 437 | ||||
| -rw-r--r-- | results/oral_a_v4_calibration_gate.json | 130 |
5 files changed, 1857 insertions, 0 deletions
diff --git a/results/oral_a_v4_calibration/fixed_hfa.json b/results/oral_a_v4_calibration/fixed_hfa.json new file mode 100644 index 0000000..0cebb82 --- /dev/null +++ b/results/oral_a_v4_calibration/fixed_hfa.json @@ -0,0 +1,416 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 0, + "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": 267904, + "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": 64, + "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": 0, + "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": "constant", + "max_steps": 0, + "mode": "hfa", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_v4_calibration/fixed_hfa.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": null, + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "ordinary_examples": 0, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": -0.0002631608634449852, + "feedback_forward_cosine": [ + 0.0008809715509414673, + -0.002320552244782448, + 0.03730485588312149, + 0.02497659996151924, + -0.03438456356525421, + -0.001257734838873148, + 0.023045765236020088, + 0.010684739798307419, + -0.013966654427349567, + -0.03158111870288849, + 0.009583890438079834, + -9.433389641344547e-05, + 0.004490766208618879, + -0.004784357734024525, + 0.004104882013052702, + 0.004543168004602194, + -0.005818712059408426, + 0.0014950999757274985, + 0.003767170011997223 + ], + "feedback_forward_norm_ratio": [ + 0.9996462464332581, + 1.0550107955932617, + 0.9593725800514221, + 1.0017613172531128, + 0.999921441078186, + 0.9631232619285583, + 0.9992203712463379, + 0.9976226687431335, + 1.0018895864486694, + 1.0030434131622314, + 1.0180330276489258, + 1.0107296705245972, + 0.9962885975837708, + 0.9891077876091003, + 1.0009011030197144, + 1.0021207332611084, + 1.0043002367019653, + 1.0021206140518188, + 0.9992877244949341 + ], + "innovation_negative_gradient_cosine": [ + -0.0005718558095395565, + -0.0006265097763389349, + 0.0012021195143461227, + 3.764468419831246e-05, + 0.00048482639249414206, + -0.002105190185829997, + 0.0032924613915383816, + 0.0002729777479544282, + 0.007543372455984354, + -0.005317360162734985, + 0.009830432943999767, + -0.0010977099882438779, + 0.023688670247793198, + 0.007433340419083834, + 0.04359162598848343, + 0.0057456037029623985, + 0.039965368807315826, + 0.0030560996383428574, + 0.07787258923053741 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0005718558095395565, + -0.0006265097763389349, + 0.0012021195143461227, + 3.764468419831246e-05, + 0.00048482639249414206, + -0.002105190185829997, + 0.0032924613915383816, + 0.0002729777479544282, + 0.007543372455984354, + -0.005317360162734985, + 0.009830432943999767, + -0.0010977099882438779, + 0.023688670247793198, + 0.007433340419083834, + 0.04359162598848343, + 0.0057456037029623985, + 0.039965368807315826, + 0.0030560996383428574, + 0.07787258923053741 + ], + "teaching_negative_gradient_cosine": [ + -0.0005718558095395565, + -0.0006265097763389349, + 0.0012021195143461227, + 3.764468419831246e-05, + 0.00048482639249414206, + -0.002105190185829997, + 0.0032924613915383816, + 0.0002729777479544282, + 0.007543372455984354, + -0.005317360162734985, + 0.009830432943999767, + -0.0010977099882438779, + 0.023688670247793198, + 0.007433340419083834, + 0.04359162598848343, + 0.0057456037029623985, + 0.039965368807315826, + 0.0030560996383428574, + 0.07787258923053741 + ], + "wall_s": 0.20851445198059082 + }, + "epochs": [], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.1004, + "epoch": 0, + "evaluation_split": "validation", + "finite": true, + "loss": 16.593759375, + "step": 0 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 1730765824, + "peak_memory_reserved_bytes": 2168455168, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "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": 0.6183156967163086, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 0.6185517311096191, + "train_wall_s": 0.0 + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 0, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "local_weight_correlation_macs": 0, + "ordinary_forward_macs": 0, + "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, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 0, + "total_forward_equivalent_examples": 0, + "total_macs_estimate": 0 + } +} diff --git a/results/oral_a_v4_calibration/learned_hfa_etaA0.1.json b/results/oral_a_v4_calibration/learned_hfa_etaA0.1.json new file mode 100644 index 0000000..e92cc38 --- /dev/null +++ b/results/oral_a_v4_calibration/learned_hfa_etaA0.1.json @@ -0,0 +1,437 @@ +{ + "apical_warmup": { + "first": { + "calibration_mse": 4.429747588410153e-08, + "parameter_update_rms": 0.00021046965549480413, + "prediction_target_cosine": -0.02308727157707865, + "target_power": 4.4068150205784043e-08 + }, + "last": { + "calibration_mse": 7.57351844983796e-08, + "parameter_update_rms": 0.00027520026253326793, + "prediction_target_cosine": 0.008176687138101971, + "target_power": 7.566517331129459e-08 + }, + "mean": { + "calibration_mse": 1.2829968753120993e-06, + "parameter_update_rms": 0.0009033753133269748, + "prediction_target_cosine": 0.0006682040486695352, + "target_power": 1.2829480604683788e-06 + }, + "steps": 400 + }, + "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": 400, + "alignment_probe": 64, + "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": 0, + "eta_A": 0.1, + "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": "constant", + "max_steps": 0, + "mode": "lhfa", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_v4_calibration/learned_hfa_etaA0.1.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": "hierarchical_feedback_parameters", + "counters": { + "apical_warmup_examples": 50640, + "calibration_event_examples": 50640, + "causal_scalar_observations": 800, + "logical_batch_loss_queries": 800, + "ordinary_examples": 0, + "per_example_loss_terms": 101280, + "perturbation_events": 400, + "perturbation_forward_examples": 101280, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": -0.0001799722813302651, + "feedback_forward_cosine": [ + 0.00087751355022192, + -0.002306785434484482, + 0.03730355575680733, + 0.024973055347800255, + -0.03437461704015732, + -0.0012589972466230392, + 0.02304646745324135, + 0.010723255574703217, + -0.013957379385828972, + -0.03159216791391373, + 0.009621845558285713, + -8.752243593335152e-05, + 0.0044714659452438354, + -0.004768045619130135, + 0.004082047380506992, + 0.004587260074913502, + -0.005784230772405863, + 0.0014915686333552003, + 0.057022809982299805 + ], + "feedback_forward_norm_ratio": [ + 0.9984350204467773, + 1.0541573762893677, + 0.9591503143310547, + 1.0013585090637207, + 0.9998340606689453, + 0.9629143476486206, + 0.9992043972015381, + 0.9973520040512085, + 1.0017883777618408, + 1.0028769969940186, + 1.0179988145828247, + 1.0106147527694702, + 0.9962611794471741, + 0.9889401793479919, + 1.0009331703186035, + 1.00214684009552, + 1.0044093132019043, + 1.0020496845245361, + 1.0028640031814575 + ], + "innovation_negative_gradient_cosine": [ + -0.000455486006103456, + -0.00046591833233833313, + 0.0007518491474911571, + -0.00010768519132398069, + 0.0012330973986536264, + -0.0020356907043606043, + 0.004193619824945927, + 0.00030038447584956884, + 0.009274046868085861, + -0.004511808976531029, + 0.0123829934746027, + 0.0020061261020600796, + 0.024534553289413452, + -0.0031216288916766644, + 0.05966883897781372, + 0.010687511414289474, + 0.0716584175825119, + -0.002210831269621849, + 0.1456024944782257 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.000455486006103456, + -0.00046591833233833313, + 0.0007518491474911571, + -0.00010768519132398069, + 0.0012330973986536264, + -0.0020356907043606043, + 0.004193619824945927, + 0.00030038447584956884, + 0.009274046868085861, + -0.004511808976531029, + 0.0123829934746027, + 0.0020061261020600796, + 0.024534553289413452, + -0.0031216288916766644, + 0.05966883897781372, + 0.010687511414289474, + 0.0716584175825119, + -0.002210831269621849, + 0.1456024944782257 + ], + "teaching_negative_gradient_cosine": [ + -0.000455486006103456, + -0.00046591833233833313, + 0.0007518491474911571, + -0.00010768519132398069, + 0.0012330973986536264, + -0.0020356907043606043, + 0.004193619824945927, + 0.00030038447584956884, + 0.009274046868085861, + -0.004511808976531029, + 0.0123829934746027, + 0.0020061261020600796, + 0.024534553289413452, + -0.0031216288916766644, + 0.05966883897781372, + 0.010687511414289474, + 0.0716584175825119, + -0.002210831269621849, + 0.1456024944782257 + ], + "wall_s": 0.11083364486694336 + }, + "epochs": [], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.1004, + "epoch": 0, + "evaluation_split": "validation", + "finite": true, + "loss": 16.593759375, + "step": 0 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 1951938048, + "peak_memory_reserved_bytes": 2814377984, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "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": 30.135460138320923, + "evaluation_wall_s": 0.36730265617370605, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 30.503045558929443, + "train_wall_s": 0.0 + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 800, + "components": { + "apical_projection_macs": 2031103150080, + "apical_regression_macs": 4062206300160, + "bp_reverse_macs_estimate": 0, + "local_weight_correlation_macs": 0, + "ordinary_forward_macs": 0, + "perturbation_forward_macs": 4107009331200, + "warmup_clean_forward_macs": 2053504665600 + }, + "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": 800, + "per_example_cross_entropy_terms": 101280, + "total_clean_forward_examples": 50640, + "total_forward_equivalent_examples": 151920, + "total_macs_estimate": 12253823447040 + } +} diff --git a/results/oral_a_v4_calibration/learned_hfa_etaA1.0.json b/results/oral_a_v4_calibration/learned_hfa_etaA1.0.json new file mode 100644 index 0000000..92b3d5c --- /dev/null +++ b/results/oral_a_v4_calibration/learned_hfa_etaA1.0.json @@ -0,0 +1,437 @@ +{ + "apical_warmup": { + "first": { + "calibration_mse": 4.4297475884114785e-08, + "parameter_update_rms": 0.0002104696554948356, + "prediction_target_cosine": -0.023087271581171884, + "target_power": 4.4068150205784043e-08 + }, + "last": { + "calibration_mse": 0.0023870706237165916, + "parameter_update_rms": 0.04885765675630169, + "prediction_target_cosine": -0.006122446369815643, + "target_power": 0.0023838081348308516 + }, + "mean": { + "calibration_mse": 0.0004092785588329548, + "parameter_update_rms": 0.01223350082381715, + "prediction_target_cosine": -0.0005106259446759552, + "target_power": 0.0004088082577299622 + }, + "steps": 400 + }, + "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": 400, + "alignment_probe": 64, + "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": 0, + "eta_A": 1.0, + "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": "constant", + "max_steps": 0, + "mode": "lhfa", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_v4_calibration/learned_hfa_etaA1.0.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": "hierarchical_feedback_parameters", + "counters": { + "apical_warmup_examples": 50640, + "calibration_event_examples": 50640, + "causal_scalar_observations": 800, + "logical_batch_loss_queries": 800, + "ordinary_examples": 0, + "per_example_loss_terms": 101280, + "perturbation_events": 400, + "perturbation_forward_examples": 101280, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.0003479742251026134, + "feedback_forward_cosine": [ + -0.0009074704721570015, + -0.003265778534114361, + 0.03652038425207138, + 0.010221419855952263, + -0.031057430431246758, + 0.0008978573605418205, + 0.022362936288118362, + 0.018681205809116364, + -0.017771737650036812, + -0.02938218042254448, + 0.012845261953771114, + 0.004519296810030937, + 0.006171742919832468, + 0.0031000173185020685, + -0.004009199794381857, + -0.004137327894568443, + -0.004176586866378784, + -0.007538903504610062, + 0.0364728607237339 + ], + "feedback_forward_norm_ratio": [ + 0.6336489915847778, + 0.4933902621269226, + 0.8000262379646301, + 0.509473443031311, + 0.8807536959648132, + 0.5386269688606262, + 0.9606244564056396, + 0.38889551162719727, + 0.9032812118530273, + 0.3399011790752411, + 0.9401517510414124, + 0.34239858388900757, + 0.9191926717758179, + 0.629314124584198, + 1.5512702465057373, + 1.235514521598816, + 2.106657028198242, + 2.0514137744903564, + 58.47907257080078 + ], + "innovation_negative_gradient_cosine": [ + 0.0012116377474740148, + -0.002346375957131386, + 0.0019503228832036257, + 0.000998541247099638, + -0.0007515153847634792, + 0.0010252348147332668, + 0.0015550695825368166, + -0.002079828642308712, + 0.00542263500392437, + 0.004640556406229734, + 0.0045582870952785015, + 0.0007940567447803915, + 0.0036008423194289207, + 0.00043276697397232056, + 0.010686355642974377, + 0.0029150790069252253, + 0.0338863879442215, + 0.028322631493210793, + 0.05369921773672104 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.0012116377474740148, + -0.002346375957131386, + 0.0019503228832036257, + 0.000998541247099638, + -0.0007515153847634792, + 0.0010252348147332668, + 0.0015550695825368166, + -0.002079828642308712, + 0.00542263500392437, + 0.004640556406229734, + 0.0045582870952785015, + 0.0007940567447803915, + 0.0036008423194289207, + 0.00043276697397232056, + 0.010686355642974377, + 0.0029150790069252253, + 0.0338863879442215, + 0.028322631493210793, + 0.05369921773672104 + ], + "teaching_negative_gradient_cosine": [ + 0.0012116377474740148, + -0.002346375957131386, + 0.0019503228832036257, + 0.000998541247099638, + -0.0007515153847634792, + 0.0010252348147332668, + 0.0015550695825368166, + -0.002079828642308712, + 0.00542263500392437, + 0.004640556406229734, + 0.0045582870952785015, + 0.0007940567447803915, + 0.0036008423194289207, + 0.00043276697397232056, + 0.010686355642974377, + 0.0029150790069252253, + 0.0338863879442215, + 0.028322631493210793, + 0.05369921773672104 + ], + "wall_s": 0.11422944068908691 + }, + "epochs": [], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.1004, + "epoch": 0, + "evaluation_split": "validation", + "finite": true, + "loss": 16.593759375, + "step": 0 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 1951938048, + "peak_memory_reserved_bytes": 2814377984, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "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": 30.103167295455933, + "evaluation_wall_s": 0.3668816089630127, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 30.470245122909546, + "train_wall_s": 0.0 + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 800, + "components": { + "apical_projection_macs": 2031103150080, + "apical_regression_macs": 4062206300160, + "bp_reverse_macs_estimate": 0, + "local_weight_correlation_macs": 0, + "ordinary_forward_macs": 0, + "perturbation_forward_macs": 4107009331200, + "warmup_clean_forward_macs": 2053504665600 + }, + "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": 800, + "per_example_cross_entropy_terms": 101280, + "total_clean_forward_examples": 50640, + "total_forward_equivalent_examples": 151920, + "total_macs_estimate": 12253823447040 + } +} diff --git a/results/oral_a_v4_calibration/learned_hfa_etaA10.0.json b/results/oral_a_v4_calibration/learned_hfa_etaA10.0.json new file mode 100644 index 0000000..d05c0ae --- /dev/null +++ b/results/oral_a_v4_calibration/learned_hfa_etaA10.0.json @@ -0,0 +1,437 @@ +{ + "apical_warmup": { + "first": { + "calibration_mse": 4.429747588407501e-08, + "parameter_update_rms": 0.0002104696554947411, + "prediction_target_cosine": -0.02308727157965524, + "target_power": 4.4068150205784043e-08 + }, + "last": { + "calibration_mse": NaN, + "parameter_update_rms": NaN, + "prediction_target_cosine": NaN, + "target_power": NaN + }, + "mean": { + "calibration_mse": NaN, + "parameter_update_rms": NaN, + "prediction_target_cosine": NaN, + "target_power": NaN + }, + "steps": 400 + }, + "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": 400, + "alignment_probe": 64, + "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": 0, + "eta_A": 10.0, + "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": "constant", + "max_steps": 0, + "mode": "lhfa", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/oral_a_v4_calibration/learned_hfa_etaA10.0.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": "hierarchical_feedback_parameters", + "counters": { + "apical_warmup_examples": 50640, + "calibration_event_examples": 50640, + "causal_scalar_observations": 800, + "logical_batch_loss_queries": 800, + "ordinary_examples": 0, + "per_example_loss_terms": 101280, + "perturbation_events": 400, + "perturbation_forward_examples": 101280, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": NaN, + "feedback_forward_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "feedback_forward_norm_ratio": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "innovation_negative_gradient_cosine": [ + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "teaching_negative_gradient_cosine": [ + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + 0.0, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "wall_s": 0.11645221710205078 + }, + "epochs": [], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.1004, + "epoch": 0, + "evaluation_split": "validation", + "finite": true, + "loss": 16.593759375, + "step": 0 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 1951938048, + "peak_memory_reserved_bytes": 2814377984, + "torch_version": "2.3.1+cu118" + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "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": 30.35198736190796, + "evaluation_wall_s": 0.3776404857635498, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 30.72986936569214, + "train_wall_s": 0.0 + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 800, + "components": { + "apical_projection_macs": 2031103150080, + "apical_regression_macs": 4062206300160, + "bp_reverse_macs_estimate": 0, + "local_weight_correlation_macs": 0, + "ordinary_forward_macs": 0, + "perturbation_forward_macs": 4107009331200, + "warmup_clean_forward_macs": 2053504665600 + }, + "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": 800, + "per_example_cross_entropy_terms": 101280, + "total_clean_forward_examples": 50640, + "total_forward_equivalent_examples": 151920, + "total_macs_estimate": 12253823447040 + } +} diff --git a/results/oral_a_v4_calibration_gate.json b/results/oral_a_v4_calibration_gate.json new file mode 100644 index 0000000..0e1504d --- /dev/null +++ b/results/oral_a_v4_calibration_gate.json @@ -0,0 +1,130 @@ +{ + "checks": { + "all_four_records_finite": false, + "all_layer_at_least_0.10": false, + "candidate_selected": true, + "early_gain_over_fixed_hfa_at_least_0.04": false, + "early_third_at_least_0.05": false, + "feedback_norm_ratios_in_0.1_to_3": false + }, + "confirmation_test_seeds_touched": false, + "matched_fixed_hfa": { + "eta_A": null, + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.011278868802228795, + "early_third_alignment": -0.0002631608634449852, + "max_feedback_forward_norm_ratio": 1.0550107955932617, + "mean_feedback_forward_cosine": 0.0016142042954207252, + "min_feedback_forward_norm_ratio": 0.9593725800514221 + }, + "mode": "hfa", + "path": "results/oral_a_v4_calibration/fixed_hfa.json", + "source_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "total_macs": 0 + }, + "protocol": "oral_a_v4_hierarchical_causal_capture_v1", + "review_score_after": 5, + "review_score_before": 5, + "rows": [ + { + "eta_A": null, + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.011278868802228795, + "early_third_alignment": -0.0002631608634449852, + "max_feedback_forward_norm_ratio": 1.0550107955932617, + "mean_feedback_forward_cosine": 0.0016142042954207252, + "min_feedback_forward_norm_ratio": 0.9593725800514221 + }, + "mode": "hfa", + "path": "results/oral_a_v4_calibration/fixed_hfa.json", + "source_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "total_macs": 0 + }, + { + "eta_A": 0.1, + "finite": true, + "logical_batch_loss_queries": 800, + "metrics": { + "all_layer_alignment": 0.017336046508525646, + "early_third_alignment": -0.0001799722813302651, + "max_feedback_forward_norm_ratio": 1.0541573762893677, + "mean_calibration_mse": 1.2829968753120993e-06, + "mean_feedback_forward_cosine": 0.004424794705731696, + "mean_parameter_update_rms": 0.0009033753133269748, + "mean_prediction_target_cosine": 0.0006682040486695352, + "mean_target_power": 1.2829480604683788e-06, + "min_feedback_forward_norm_ratio": 0.9591503143310547 + }, + "mode": "lhfa", + "path": "results/oral_a_v4_calibration/learned_hfa_etaA0.1.json", + "source_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "total_macs": 12253823447040 + }, + { + "eta_A": 1.0, + "finite": true, + "logical_batch_loss_queries": 800, + "metrics": { + "all_layer_alignment": 0.007922205403079524, + "early_third_alignment": 0.0003479742251026134, + "max_feedback_forward_norm_ratio": 58.47907257080078, + "mean_calibration_mse": 0.0004092785588329548, + "mean_feedback_forward_cosine": 0.0026077035642964276, + "mean_parameter_update_rms": 0.01223350082381715, + "mean_prediction_target_cosine": -0.0005106259446759552, + "mean_target_power": 0.0004088082577299622, + "min_feedback_forward_norm_ratio": 0.3399011790752411 + }, + "mode": "lhfa", + "path": "results/oral_a_v4_calibration/learned_hfa_etaA1.0.json", + "source_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "total_macs": 12253823447040 + }, + { + "eta_A": 10.0, + "finite": false, + "logical_batch_loss_queries": 800, + "metrics": { + "all_layer_alignment": NaN, + "early_third_alignment": NaN, + "max_feedback_forward_norm_ratio": NaN, + "mean_calibration_mse": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_parameter_update_rms": NaN, + "mean_prediction_target_cosine": NaN, + "mean_target_power": NaN, + "min_feedback_forward_norm_ratio": NaN + }, + "mode": "lhfa", + "path": "results/oral_a_v4_calibration/learned_hfa_etaA10.0.json", + "source_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "total_macs": 12253823447040 + } + ], + "score_change_rule": "causal capture alone cannot raise score", + "selected_v4": { + "eta_A": 1.0, + "finite": true, + "logical_batch_loss_queries": 800, + "metrics": { + "all_layer_alignment": 0.007922205403079524, + "early_third_alignment": 0.0003479742251026134, + "max_feedback_forward_norm_ratio": 58.47907257080078, + "mean_calibration_mse": 0.0004092785588329548, + "mean_feedback_forward_cosine": 0.0026077035642964276, + "mean_parameter_update_rms": 0.01223350082381715, + "mean_prediction_target_cosine": -0.0005106259446759552, + "mean_target_power": 0.0004088082577299622, + "min_feedback_forward_norm_ratio": 0.3399011790752411 + }, + "mode": "lhfa", + "path": "results/oral_a_v4_calibration/learned_hfa_etaA1.0.json", + "source_commit": "d4a521072a25d8afdd62021f43538d7dad7586fd", + "total_macs": 12253823447040 + }, + "status": "failed" +} |
