diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 16:29:46 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-07-22 16:29:46 -0500 |
| commit | f310ad55b76ab96de488cdb4f2d05861ed4616f2 (patch) | |
| tree | 785a59950974949edfd5f2b5526d89c65f554e68 | |
| parent | 412314e3ed7bca666a41876980b479456f5e733d (diff) | |
results: record mixed-traffic short failure
| -rw-r--r-- | results/kp_innovation_short/innovation.json | 1051 | ||||
| -rw-r--r-- | results/kp_innovation_short/matched.json | 1051 | ||||
| -rw-r--r-- | results/kp_innovation_short/raw.json | 1051 | ||||
| -rw-r--r-- | results/kp_innovation_short_gate.json | 64 |
4 files changed, 3217 insertions, 0 deletions
diff --git a/results/kp_innovation_short/innovation.json b/results/kp_innovation_short/innovation.json new file mode 100644 index 0000000..6783166 --- /dev/null +++ b/results/kp_innovation_short/innovation.json @@ -0,0 +1,1051 @@ +{ + "architecture": { + "adaptive_apical_parameters": 376832, + "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": 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": 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": 0.1, + "eval_every": 0, + "eval_split": "validation", + "learn_P": 1, + "loader_seed": 0, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "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": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/kp_innovation_short/innovation.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 16, + "predictor_warmup_steps": 20, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 4000, + "train_limit": 0, + "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": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "ordinary_examples": 900000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 56320, + "predictor_warmup_examples": 2560, + "traffic_audit_examples": 64, + "traffic_calibration_examples": 64 + }, + "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 + ], + "feedback_forward_relative_error": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "innovation_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "instruction_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "matched_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "max_norm_match_direction_error": NaN, + "max_norm_match_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": NaN, + "raw_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "teaching_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "traffic_instruction_rms_ratio": NaN, + "used_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "wall_s": 0.1382739543914795 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 352, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 704, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1056, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1408, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": NaN + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.1, + "epoch": 20, + "evaluation_split": "validation", + "finite": false, + "loss": NaN, + "step": 7040 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2167159296, + "peak_memory_reserved_bytes": 3087007744, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "first_mse": 0.00028227133620732826, + "instruction_present": false, + "last_mse": 5.667926742440297e-06, + "mean_mse": 6.04510786383403e-05, + "post_warmup_traffic_residual_rms_ratio": 0.13994200329263365, + "steps": 20, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "412314e3ed7bca666a41876980b479456f5e733d", + "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": 45000, + "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": 0.3675510883331299, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.4973745346069336, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 660.3383731842041, + "train_wall_s": 658.9339711666107 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.008471464738249779, + 0.006617916747927666, + 0.003988116513937712, + 0.0043865954503417015, + 0.002713605063036084, + 0.003158062230795622, + 0.002004207344725728, + 0.003757704282179475, + 0.0027085733599960804, + 0.0030200635083019733, + 0.002056022873148322, + 0.002389939734712243, + 0.0016842411132529378, + 0.0034010226372629404, + 0.002549725119024515, + 0.0030542907770723104, + 0.0020918569061905146, + 0.002605808200314641, + 0.0018916188273578882 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.0443243607878685, + 0.035817068070173264, + 0.013778294436633587, + 0.022762222215533257, + 0.007214079145342112, + 0.017218582332134247, + 0.004616298712790012, + 0.02042512409389019, + 0.007327491883188486, + 0.01611751690506935, + 0.004703814629465342, + 0.01256294921040535, + 0.003324519144371152, + 0.01805141195654869, + 0.006063441745936871, + 0.016296451911330223, + 0.0044285994954407215, + 0.013874167576432228, + 0.0036007820162922144 + ], + "unscaled_traffic_rms": [ + 0.7644973993301392, + 0.7390796542167664, + 1.157796859741211, + 0.7708553671836853, + 1.504616141319275, + 0.7336404919624329, + 1.736635684967041, + 0.7358984351158142, + 1.478581428527832, + 0.7495108246803284, + 1.7483876943588257, + 0.7609485983848572, + 2.0264477729797363, + 0.7536302804946899, + 1.6820316314697266, + 0.7496824264526367, + 1.8894071578979492, + 0.7512690424919128, + 2.10134220123291 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 36100371755008, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 36097804800000, + "local_weight_correlation_macs": 36495936000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 36495936000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 109001195520 + }, + "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; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 1283917873152, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 902688, + "total_forward_equivalent_examples": 902688, + "total_macs_estimate": 145299049750528 + } +} diff --git a/results/kp_innovation_short/matched.json b/results/kp_innovation_short/matched.json new file mode 100644 index 0000000..1bcb956 --- /dev/null +++ b/results/kp_innovation_short/matched.json @@ -0,0 +1,1051 @@ +{ + "architecture": { + "adaptive_apical_parameters": 376832, + "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": 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": 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": 0.1, + "eval_every": 0, + "eval_split": "validation", + "learn_P": 1, + "loader_seed": 0, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "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": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/kp_innovation_short/matched.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 16, + "predictor_warmup_steps": 20, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "matched", + "traffic_seed": 4000, + "train_limit": 0, + "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": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "ordinary_examples": 900000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 56320, + "predictor_warmup_examples": 2560, + "traffic_audit_examples": 64, + "traffic_calibration_examples": 64 + }, + "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 + ], + "feedback_forward_relative_error": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "innovation_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "instruction_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "matched_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "max_norm_match_direction_error": NaN, + "max_norm_match_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": NaN, + "raw_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "teaching_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "traffic_instruction_rms_ratio": NaN, + "used_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "wall_s": 0.14116454124450684 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 352, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 704, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1056, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1408, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": NaN + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.1, + "epoch": 20, + "evaluation_split": "validation", + "finite": false, + "loss": NaN, + "step": 7040 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "7", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2167159296, + "peak_memory_reserved_bytes": 3170893824, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "first_mse": 0.0002822713352582903, + "instruction_present": false, + "last_mse": 5.667926707839955e-06, + "mean_mse": 6.0451078409730885e-05, + "post_warmup_traffic_residual_rms_ratio": 0.13994200341774746, + "steps": 20, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "412314e3ed7bca666a41876980b479456f5e733d", + "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": 45000, + "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": 0.36821722984313965, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.49567413330078125, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 671.1911971569061, + "train_wall_s": 669.7800099849701 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.008471464738249779, + 0.006617916747927666, + 0.003988116513937712, + 0.0043865954503417015, + 0.002713605063036084, + 0.003158062230795622, + 0.002004207344725728, + 0.003757704282179475, + 0.0027085731271654367, + 0.0030200635083019733, + 0.002056022873148322, + 0.0023899395018815994, + 0.0016842411132529378, + 0.0034010226372629404, + 0.002549725119024515, + 0.0030542907770723104, + 0.0020918569061905146, + 0.002605808200314641, + 0.0018916188273578882 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.0443243607878685, + 0.035817068070173264, + 0.013778294436633587, + 0.022762222215533257, + 0.007214079145342112, + 0.017218582332134247, + 0.004616298712790012, + 0.02042512409389019, + 0.007327491417527199, + 0.01611751690506935, + 0.004703814629465342, + 0.012562948279082775, + 0.003324519144371152, + 0.01805141195654869, + 0.006063441745936871, + 0.016296451911330223, + 0.0044285994954407215, + 0.013874167576432228, + 0.0036007820162922144 + ], + "unscaled_traffic_rms": [ + 0.7644973993301392, + 0.7390796542167664, + 1.157796859741211, + 0.7708553671836853, + 1.504616141319275, + 0.7336404919624329, + 1.736635684967041, + 0.7358984351158142, + 1.478581428527832, + 0.7495108246803284, + 1.7483876943588257, + 0.7609485983848572, + 2.0264477729797363, + 0.7536302804946899, + 1.6820316314697266, + 0.7496824264526367, + 1.8894071578979492, + 0.7512690424919128, + 2.10134220123291 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 36100371755008, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 36097804800000, + "local_weight_correlation_macs": 36495936000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 36495936000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 109001195520 + }, + "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; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 2131789873152, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 902688, + "total_forward_equivalent_examples": 902688, + "total_macs_estimate": 145299049750528 + } +} diff --git a/results/kp_innovation_short/raw.json b/results/kp_innovation_short/raw.json new file mode 100644 index 0000000..026fd07 --- /dev/null +++ b/results/kp_innovation_short/raw.json @@ -0,0 +1,1051 @@ +{ + "architecture": { + "adaptive_apical_parameters": 376832, + "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": 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": 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": 0.1, + "eval_every": 0, + "eval_split": "validation", + "learn_P": 1, + "loader_seed": 0, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "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": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/kp_innovation_short/raw.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 16, + "predictor_warmup_steps": 20, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "raw", + "traffic_seed": 4000, + "train_limit": 0, + "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": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "ordinary_examples": 900000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 56320, + "predictor_warmup_examples": 2560, + "traffic_audit_examples": 64, + "traffic_calibration_examples": 64 + }, + "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 + ], + "feedback_forward_relative_error": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "innovation_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "instruction_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "matched_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "max_norm_match_direction_error": NaN, + "max_norm_match_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": NaN, + "raw_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "teaching_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "traffic_instruction_rms_ratio": NaN, + "used_negative_gradient_cosine": [ + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN, + NaN + ], + "wall_s": 0.14948534965515137 + }, + "epochs": [ + { + "epoch": 1, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 352, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 2, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 704, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 3, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1056, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 4, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1408, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 5, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 6, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 7, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 8, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 9, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 10, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 11, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 12, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 13, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 14, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 15, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 16, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 17, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 18, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 19, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": NaN + }, + { + "epoch": 20, + "feedback_tracking": { + "max_feedback_forward_relative_error": NaN, + "mean_feedback_forward_cosine": NaN, + "mean_feedback_forward_relative_error": NaN, + "min_feedback_forward_cosine": NaN + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": NaN, + "instruction_rms": NaN, + "raw_apical_rms": NaN, + "teaching_rms": NaN, + "traffic_rms": NaN + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": NaN + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 1 + }, + "final": { + "accuracy": 0.1, + "epoch": 20, + "evaluation_split": "validation", + "finite": false, + "loss": NaN, + "step": 7040 + }, + "hardware": { + "cuda_device_name": "NVIDIA GeForce GTX 1080", + "cuda_visible_devices": "5", + "device": "cuda", + "device_total_memory_bytes": 8507949056, + "peak_memory_allocated_bytes": 2167159296, + "peak_memory_reserved_bytes": 3087007744, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "first_mse": 0.00028227133114579254, + "instruction_present": false, + "last_mse": 5.667926663353802e-06, + "mean_mse": 6.045107751728924e-05, + "post_warmup_traffic_residual_rms_ratio": 0.13994200354453204, + "steps": 20, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "412314e3ed7bca666a41876980b479456f5e733d", + "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": 45000, + "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": 0.3663215637207031, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.49329090118408203, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 658.3386063575745, + "train_wall_s": 656.9815294742584 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.008471464738249779, + 0.006617916747927666, + 0.003988116048276424, + 0.0043865954503417015, + 0.002713605063036084, + 0.003158062230795622, + 0.002004207344725728, + 0.003757704282179475, + 0.0027085733599960804, + 0.0030200635083019733, + 0.002056022873148322, + 0.002389939734712243, + 0.0016842411132529378, + 0.0034010226372629404, + 0.002549725119024515, + 0.0030542907770723104, + 0.0020918569061905146, + 0.002605808200314641, + 0.0018916188273578882 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.0443243607878685, + 0.035817068070173264, + 0.013778292573988438, + 0.022762222215533257, + 0.007214079145342112, + 0.017218582332134247, + 0.004616298712790012, + 0.02042512409389019, + 0.007327491883188486, + 0.01611751690506935, + 0.004703814629465342, + 0.01256294921040535, + 0.003324519144371152, + 0.01805141195654869, + 0.006063441745936871, + 0.016296451911330223, + 0.0044285994954407215, + 0.013874167576432228, + 0.0036007820162922144 + ], + "unscaled_traffic_rms": [ + 0.7644973993301392, + 0.7390796542167664, + 1.157796859741211, + 0.7708553671836853, + 1.504616141319275, + 0.7336404919624329, + 1.736635684967041, + 0.7358984351158142, + 1.478581428527832, + 0.7495108246803284, + 1.7483876943588257, + 0.7609485983848572, + 2.0264477729797363, + 0.7536302804946899, + 1.6820316314697266, + 0.7496824264526367, + 1.8894071578979492, + 0.7512690424919128, + 2.10134220123291 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 40108672, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 36100371755008, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 36097804800000, + "local_weight_correlation_macs": 36495936000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 36495936000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 109001195520 + }, + "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; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 1283917873152, + "forward_macs_per_example": 40551040, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 902688, + "total_forward_equivalent_examples": 902688, + "total_macs_estimate": 145299049750528 + } +} diff --git a/results/kp_innovation_short_gate.json b/results/kp_innovation_short_gate.json new file mode 100644 index 0000000..87cb02c --- /dev/null +++ b/results/kp_innovation_short_gate.json @@ -0,0 +1,64 @@ +{ + "checks": { + "each_affine_mac_count_at_most_1.40x_bp": true, + "elementwise_cost_reported": true, + "epoch11_to20_feedback_cosine_at_least_0.70": false, + "final_feedback_cosine_at_least_0.80": false, + "initial_layer_ratio_error_at_most_1e-5": true, + "innovation_accuracy_at_least_0.75": false, + "innovation_early_alignment_at_least_0.70": false, + "innovation_gain_over_matched_at_least_0.03": false, + "innovation_gain_over_raw_at_least_0.05": false, + "innovation_within_5_points_of_clean_kp": false, + "matched_norm_relative_error_at_most_1e-6": false, + "post_warmup_predictor_residual_ratio_at_most_0.25": true, + "records_trajectories_and_diagnostics_finite": false, + "same_network_alignment_gain_over_raw_at_least_0.15": false, + "zero_task_loss_queries": true + }, + "confirmation_test_seeds_touched": false, + "full_validation_opened": false, + "metrics": { + "accuracy": { + "innovation": 0.1, + "matched": 0.1, + "raw": 0.1 + }, + "clean_kp_accuracy": 0.8266, + "elementwise_operations_estimate": { + "innovation": 1283917873152, + "matched": 2131789873152, + "raw": 1283917873152 + }, + "epoch11_to20_feedback_forward_cosine": NaN, + "final_feedback_forward_cosine": NaN, + "innovation_early_third_alignment": NaN, + "innovation_gain_over_matched": 0.0, + "innovation_gain_over_raw": 0.0, + "logical_batch_loss_queries": { + "innovation": 0, + "matched": 0, + "raw": 0 + }, + "mac_ratio_to_bp": { + "innovation": 1.327079721520482, + "matched": 1.327079721520482, + "raw": 1.327079721520482 + }, + "matched_norm_relative_error": NaN, + "max_initial_layer_ratio_error": 4.76837158203125e-07, + "max_post_warmup_predictor_residual_rms_ratio": 0.13994200354453204, + "same_innovation_network_raw_early_third_alignment": NaN, + "source_commit": "412314e3ed7bca666a41876980b479456f5e733d", + "total_macs": { + "innovation": 145299049750528, + "matched": 145299049750528, + "raw": 145299049750528 + } + }, + "protocol": "kp_mixed_traffic_short_v1", + "review_score_after": 5, + "review_score_before": 5, + "score_change_rule": "a one-seed short mechanism screen cannot raise the review score", + "status": "failed" +} |