From 3f3045c7920d9c2d55ce3b4d428a06ebc89350fe Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Wed, 22 Jul 2026 13:37:49 -0500 Subject: results: pass normalized mirror causal-capture gate --- MIRROR_BASELINE.md | 16 + RESULTS.md | 23 ++ REVIEW_SCORECARD.md | 2 + ROADMAP.md | 8 + results/mirror_causal_capture/fixed_hfa.json | 430 +++++++++++++++++++++++ results/mirror_causal_capture/wm_etaM0.03.json | 451 +++++++++++++++++++++++++ results/mirror_causal_capture/wm_etaM0.1.json | 451 +++++++++++++++++++++++++ results/mirror_causal_capture/wm_etaM0.3.json | 451 +++++++++++++++++++++++++ results/mirror_causal_capture_gate.json | 129 +++++++ 9 files changed, 1961 insertions(+) create mode 100644 results/mirror_causal_capture/fixed_hfa.json create mode 100644 results/mirror_causal_capture/wm_etaM0.03.json create mode 100644 results/mirror_causal_capture/wm_etaM0.1.json create mode 100644 results/mirror_causal_capture/wm_etaM0.3.json create mode 100644 results/mirror_causal_capture_gate.json diff --git a/MIRROR_BASELINE.md b/MIRROR_BASELINE.md index c7ad348..0532955 100644 --- a/MIRROR_BASELINE.md +++ b/MIRROR_BASELINE.md @@ -84,3 +84,19 @@ cost at most 1.15x BP is considered a strong inherited baseline. It does not authorize test access; it instead opens design of a hierarchical innovation intervention on the same frozen recipe. +## Audited WM-1 outcome (2026-07-22) + +All four records are finite and every frozen capture check passes. The selected +rate is `eta_M=0.1`: + +| rule | early alignment | all-layer alignment | feedback/forward cosine | norm range | +|:--|--:|--:|--:|:--| +| fixed HFA | -0.000263 | 0.011279 | 0.001614 | [0.959, 1.055] | +| WM 0.03 | 0.075263 | 0.235197 | 0.625186 | [0.697, 0.787] | +| WM 0.10 | **0.446141** | **0.542159** | **0.915480** | **[0.889, 1.157]** | +| WM 0.30 | 0.112177 | 0.228114 | 0.853343 | [0.999, 1.697] | + +Each learned record uses 20 local observations, zero logical task-loss queries, +and estimated work `1.6305e9` MACs. WM-1 status is `passed`; WM-2 opens with +the predeclared `eta_M=0.1`. No validation accuracy was used for this selection +and CIFAR-10 test remains untouched. diff --git a/RESULTS.md b/RESULTS.md index 64596e5..2669dc1 100644 --- a/RESULTS.md +++ b/RESULTS.md @@ -783,6 +783,29 @@ The next comparator must change the information source rather than reopen the rate/query grid, for example local response-based learned FA or weight mirroring. No test endpoint was evaluated. +### Normalized local response mirror capture + +The inherited normalized response-mirror baseline changes exactly that +information source. A Gaussian parent probe passes through the local forward +synapses, while a separate Q/R update consumes only parent/child activities. +The frozen CIFAR-prefix WM-1 screen gives: + +| method/rate | early alignment | all-layer alignment | feedback/forward cosine | norm range | +|:--|--:|--:|--:|:--| +| fixed HFA | -0.000263 | 0.011279 | 0.001614 | [0.959, 1.055] | +| WM 0.03 | 0.075263 | 0.235197 | 0.625186 | [0.697, 0.787] | +| WM 0.10 | **0.446141** | **0.542159** | **0.915480** | **[0.889, 1.157]** | +| WM 0.30 | 0.112177 | 0.228114 | 0.853343 | [0.999, 1.697] | + +All five frozen checks pass at selected `eta_M=0.1`. Twenty batch-1 local +observations cost `1.6305e9` MACs and consume zero task-loss queries, only +`0.0067%` of the fixed-HFA short screen's `2.4242e13` MACs. This sharply +separates the previous failures: the hierarchical feedback family is usable +and cheap to identify from vector-valued local responses, whereas 800 global +task-scalar queries did not calibrate it. WM-2 therefore opens. The result is +credited entirely to an inherited weight-estimation baseline; it cannot raise +the SDIL novelty score by itself. + ## How to run `experiments/run.py --mode {bp,fa,dfa,sdil} --dataset {mnist,fmnist,cifar10} --depth D --residual {0,1} --act {tanh,gelu,silu,relu}` Batteries: `experiments/run_v2.sh "" "" `. diff --git a/REVIEW_SCORECARD.md b/REVIEW_SCORECARD.md index 03534a5..bc5b4ca 100644 --- a/REVIEW_SCORECARD.md +++ b/REVIEW_SCORECARD.md @@ -62,6 +62,7 @@ Every formal result report records: | Oral-A-v3 causal capture fails | 5 | Direct A/G perturbation lowers synthetic estimator variance but leaves real early alignment at 0.0071 | Coefficient regression is not the dominant early bottleneck; full scale remains closed | | Fixed hierarchical FA short gate fails | 5 | HFA reaches 43.52%, above DFA 37.16% and failed-v1 SDIL 41.98%, but below its frozen 50% full-run threshold | Spatial hierarchy helps; fixed random hierarchy remains insufficient and supplies no SDIL scale evidence | | Hierarchical task-scalar V4 fails | 5 | Stable calibration leaves early alignment near zero; higher rates explode to 58.5x norm or become nonfinite after 800 queries | Correct mechanics and the right spatial family are insufficient when one global scalar estimates 267,904 feedback parameters | +| Normalized response mirror capture passes | 5 | Twenty local observations reach 0.446 early alignment and 0.915 feedback/forward cosine at zero task-loss queries | Strongly resolves the engineering bottleneck, but all credit belongs to an inherited weight-estimation baseline until innovation is load-bearing | | Oral-A A4 | not opened | The prerequisite A3 gate failed | No oral-A confirmation claim is available | These are conditional reviewer forecasts, not promised scores. A failed stage leaves its negative @@ -82,6 +83,7 @@ retroactively reopened by success on standard vision benchmarks. | 2026-07-22 / hierarchical oracle | Held-out gated 3x3 maps from exact child error fields reach 0.9998 early alignment; ordinary activation context remains 0.0321 | 5 → 5 | Strongly localizes the missing information to spatial hierarchical error fields, but exact-gradient oracle inputs provide no evidence that the proposed local learner can obtain them | | 2026-07-22 / fixed HFA S1 | Three clean matched ResNet-20 records; selected HFA reaches 43.52% and 0.0404 early alignment but misses the frozen 50% gate | 5 → 5 | Establishes a stronger zero-query local baseline and confirms hierarchy helps, while closing an uncalibrated full run and leaving standard SDIL evidence absent | | 2026-07-22 / hierarchical V4-1 | Four frozen-forward records: etaA 0.1 leaves early alignment near zero, etaA 1 explodes one feedback norm to 58.5x, and etaA 10 is nonfinite | 5 → 5 | Closes global-task-scalar calibration of the full hierarchy at the fixed budget; richer local information is required before another accuracy endpoint | +| 2026-07-22 / response-mirror WM-1 | Four clean frozen-forward records; selected etaM 0.1 reaches 0.4461 early and 0.5422 all-layer alignment with zero task-loss queries | 5 → 5 | Opens a strong inherited-baseline accuracy test and isolates information source as the bottleneck, but supplies no Harnett-specific task evidence | Future rows are appended only after an audited frozen stage. A score staying flat is informative: engineering, theory exposition, or visualization may make the paper more defensible without diff --git a/ROADMAP.md b/ROADMAP.md index c84293d..19d159a 100644 --- a/ROADMAP.md +++ b/ROADMAP.md @@ -413,6 +413,14 @@ of the full 267,904-parameter feedback path at 400 events. The next branch must use richer local child-response information and include learned FA/weight mirroring as an explicit inherited baseline before adding innovation. +**Normalized response-mirror WM-1 status: passed.** Twenty batch-1 local +observations at selected etaM 0.1 raise early/all alignment from +`-0.000263/0.011279` to `0.446141/0.542159`, with feedback/forward cosine +`0.915480`, norm ratios `[0.889,1.157]`, zero task-loss queries, and only +`1.6305e9` MACs. All five gates pass. The bounded two-rate WM-2 accuracy screen +is opened. This is inherited baseline performance; the paper score changes +only if innovation later becomes load-bearing on top of the scalable mirror. + Prepare convolutional local-update primitives and ResNet-20/32/56 protocols early. Queue frozen runs opportunistically on authorized idle GPUs. Because BurstCCN already reports CIFAR-10 and ImageNet scaling, dataset scale alone is not novel. The oral-level target is a memorable joint diff --git a/results/mirror_causal_capture/fixed_hfa.json b/results/mirror_causal_capture/fixed_hfa.json new file mode 100644 index 0000000..358ab84 --- /dev/null +++ b/results/mirror_causal_capture/fixed_hfa.json @@ -0,0 +1,430 @@ +{ + "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, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "hfa", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/mirror_causal_capture/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, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "ordinary_examples": 0, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": -0.0002631632038780178, + "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": [ + -0.0005718562752008438, + -0.00062651711050421, + 0.001202113926410675, + 3.7641526432707906e-05, + 0.00048482773127034307, + -0.002105189021676779, + 0.0032924585975706577, + 0.0002729742554947734, + 0.007543369196355343, + -0.005317364819347858, + 0.009830433875322342, + -0.001097709173336625, + 0.02368866838514805, + 0.007433340419083834, + 0.04359162598848343, + 0.0057456037029623985, + 0.039965368807315826, + 0.0030560996383428574, + 0.07787258923053741 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + -0.0005718562752008438, + -0.00062651711050421, + 0.001202113926410675, + 3.7641526432707906e-05, + 0.00048482773127034307, + -0.002105189021676779, + 0.0032924585975706577, + 0.0002729742554947734, + 0.007543369196355343, + -0.005317364819347858, + 0.009830433875322342, + -0.001097709173336625, + 0.02368866838514805, + 0.007433340419083834, + 0.04359162598848343, + 0.0057456037029623985, + 0.039965368807315826, + 0.0030560996383428574, + 0.07787258923053741 + ], + "teaching_negative_gradient_cosine": [ + -0.0005718562752008438, + -0.00062651711050421, + 0.001202113926410675, + 3.7641526432707906e-05, + 0.00048482773127034307, + -0.002105189021676779, + 0.0032924585975706577, + 0.0002729742554947734, + 0.007543369196355343, + -0.005317364819347858, + 0.009830433875322342, + -0.001097709173336625, + 0.02368866838514805, + 0.007433340419083834, + 0.04359162598848343, + 0.0057456037029623985, + 0.039965368807315826, + 0.0030560996383428574, + 0.07787258923053741 + ], + "wall_s": 0.23655247688293457 + }, + "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": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "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": 0.6214184761047363, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 0.6216380596160889, + "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, + "mirror_local_correlation_macs": 0, + "mirror_response_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, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 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/mirror_causal_capture/wm_etaM0.03.json b/results/mirror_causal_capture/wm_etaM0.03.json new file mode 100644 index 0000000..730fc44 --- /dev/null +++ b/results/mirror_causal_capture/wm_etaM0.03.json @@ -0,0 +1,451 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 267904, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 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, + "mirror_batch_size": 1, + "mirror_eta": 0.03, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 20, + "mode": "wm", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/mirror_causal_capture/wm_etaM0.03.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "train_limit": 10000, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "local_parent_child_response", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 20, + "mirror_events": 20, + "mirror_readout_examples": 20480, + "ordinary_examples": 0, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.07526319908599059, + "feedback_forward_cosine": [ + 0.6423558592796326, + 0.6203256845474243, + 0.6713337302207947, + 0.6542323231697083, + 0.627842366695404, + 0.6573559045791626, + 0.6544321775436401, + 0.6422699689865112, + 0.6268530488014221, + 0.6174283027648926, + 0.6333785057067871, + 0.6304293870925903, + 0.6135519742965698, + 0.5899033546447754, + 0.5910322070121765, + 0.5885388851165771, + 0.5822848677635193, + 0.5901939272880554, + 0.6447985172271729 + ], + "feedback_forward_norm_ratio": [ + 0.7114052176475525, + 0.7314509749412537, + 0.7061368823051453, + 0.7213137149810791, + 0.7001281976699829, + 0.6969802379608154, + 0.7267898917198181, + 0.7221032977104187, + 0.7144089937210083, + 0.7062307596206665, + 0.7290550470352173, + 0.7246045470237732, + 0.7437252998352051, + 0.7757689356803894, + 0.7866175770759583, + 0.7793990969657898, + 0.7799720168113708, + 0.78629070520401, + 0.713131844997406 + ], + "innovation_negative_gradient_cosine": [ + 0.042569708079099655, + 0.04908004403114319, + 0.07656459510326385, + 0.0719727873802185, + 0.10851815342903137, + 0.10287390649318695, + 0.14432156085968018, + 0.12590278685092926, + 0.1938195526599884, + 0.16362953186035156, + 0.2534205913543701, + 0.21206563711166382, + 0.3097170889377594, + 0.2400786280632019, + 0.4135889410972595, + 0.3309183716773987, + 0.5393624305725098, + 0.4047758877277374, + 0.685565710067749 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.042569708079099655, + 0.04908004403114319, + 0.07656459510326385, + 0.0719727873802185, + 0.10851815342903137, + 0.10287390649318695, + 0.14432156085968018, + 0.12590278685092926, + 0.1938195526599884, + 0.16362953186035156, + 0.2534205913543701, + 0.21206563711166382, + 0.3097170889377594, + 0.2400786280632019, + 0.4135889410972595, + 0.3309183716773987, + 0.5393624305725098, + 0.4047758877277374, + 0.685565710067749 + ], + "teaching_negative_gradient_cosine": [ + 0.042569708079099655, + 0.04908004403114319, + 0.07656459510326385, + 0.0719727873802185, + 0.10851815342903137, + 0.10287390649318695, + 0.14432156085968018, + 0.12590278685092926, + 0.1938195526599884, + 0.16362953186035156, + 0.2534205913543701, + 0.21206563711166382, + 0.3097170889377594, + 0.2400786280632019, + 0.4135889410972595, + 0.3309183716773987, + 0.5393624305725098, + 0.4047758877277374, + 0.685565710067749 + ], + "wall_s": 0.1595914363861084 + }, + "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": 1730634752, + "peak_memory_reserved_bytes": 2181038080, + "torch_version": "2.3.1+cu118" + }, + "mirror_warmup": { + "first": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.1821084153449638, + "mirror_update_rms": 0.19537643727527548, + "readout_batch_size": 1024 + }, + "last": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.18073508053859827, + "mirror_update_rms": 0.17616583955544515, + "readout_batch_size": 1024 + }, + "mean": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1793198434405597, + "mirror_update_rms": 0.18231021973395675, + "readout_batch_size": 1024.0 + }, + "steps": 20 + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "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": 0.4319612979888916, + "mirror_warmup_wall_s": 0.6505701541900635, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 1.0828416347503662, + "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, + "mirror_local_correlation_macs": 815267840, + "mirror_response_macs": 815267840, + "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, + "mirror_probe_examples": 20, + "mirror_readout_probe_examples": 20480, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 0, + "total_forward_equivalent_examples": 0, + "total_macs_estimate": 1630535680 + } +} diff --git a/results/mirror_causal_capture/wm_etaM0.1.json b/results/mirror_causal_capture/wm_etaM0.1.json new file mode 100644 index 0000000..fe2506a --- /dev/null +++ b/results/mirror_causal_capture/wm_etaM0.1.json @@ -0,0 +1,451 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 267904, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 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, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 20, + "mode": "wm", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/mirror_causal_capture/wm_etaM0.1.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "train_limit": 10000, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "local_parent_child_response", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 20, + "mirror_events": 20, + "mirror_readout_examples": 20480, + "ordinary_examples": 0, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.44614077111085254, + "feedback_forward_cosine": [ + 0.9858509302139282, + 0.9850707054138184, + 0.9865717887878418, + 0.9863734841346741, + 0.9856951236724854, + 0.9871067404747009, + 0.9731586575508118, + 0.9550088047981262, + 0.9547361135482788, + 0.9553405046463013, + 0.954055905342102, + 0.9542233943939209, + 0.8649038672447205, + 0.7760670185089111, + 0.7740859985351562, + 0.7775933742523193, + 0.7737482190132141, + 0.7765145301818848, + 0.9880185127258301 + ], + "feedback_forward_norm_ratio": [ + 0.8923099637031555, + 0.8901821374893188, + 0.8893439173698425, + 0.8961004018783569, + 0.8912216424942017, + 0.8968877196311951, + 0.9173983335494995, + 0.9227779507637024, + 0.914923369884491, + 0.9082641005516052, + 0.9238913059234619, + 0.9232812523841858, + 1.0126720666885376, + 1.1361547708511353, + 1.1570484638214111, + 1.1266412734985352, + 1.137105941772461, + 1.1543664932250977, + 0.8931937217712402 + ], + "innovation_negative_gradient_cosine": [ + 0.43270143866539, + 0.43993276357650757, + 0.44664573669433594, + 0.44853127002716064, + 0.4532058835029602, + 0.455827534198761, + 0.4590722918510437, + 0.4557003974914551, + 0.47527873516082764, + 0.4777085483074188, + 0.5002998113632202, + 0.49793124198913574, + 0.5148465633392334, + 0.4711496829986572, + 0.6135821342468262, + 0.6229155659675598, + 0.7880423665046692, + 0.758634626865387, + 0.9890177249908447 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.43270143866539, + 0.43993276357650757, + 0.44664573669433594, + 0.44853127002716064, + 0.4532058835029602, + 0.455827534198761, + 0.4590722918510437, + 0.4557003974914551, + 0.47527873516082764, + 0.4777085483074188, + 0.5002998113632202, + 0.49793124198913574, + 0.5148465633392334, + 0.4711496829986572, + 0.6135821342468262, + 0.6229155659675598, + 0.7880423665046692, + 0.758634626865387, + 0.9890177249908447 + ], + "teaching_negative_gradient_cosine": [ + 0.43270143866539, + 0.43993276357650757, + 0.44664573669433594, + 0.44853127002716064, + 0.4532058835029602, + 0.455827534198761, + 0.4590722918510437, + 0.4557003974914551, + 0.47527873516082764, + 0.4777085483074188, + 0.5002998113632202, + 0.49793124198913574, + 0.5148465633392334, + 0.4711496829986572, + 0.6135821342468262, + 0.6229155659675598, + 0.7880423665046692, + 0.758634626865387, + 0.9890177249908447 + ], + "wall_s": 0.15237092971801758 + }, + "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": 1730634752, + "peak_memory_reserved_bytes": 2181038080, + "torch_version": "2.3.1+cu118" + }, + "mirror_warmup": { + "first": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.1821084153449638, + "mirror_update_rms": 0.19537643727527548, + "readout_batch_size": 1024 + }, + "last": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.18073508053859827, + "mirror_update_rms": 0.17050672066264347, + "readout_batch_size": 1024 + }, + "mean": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1793198434405597, + "mirror_update_rms": 0.17517745811581048, + "readout_batch_size": 1024.0 + }, + "steps": 20 + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "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": 0.4133622646331787, + "mirror_warmup_wall_s": 0.5381975173950195, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 0.9518508911132812, + "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, + "mirror_local_correlation_macs": 815267840, + "mirror_response_macs": 815267840, + "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, + "mirror_probe_examples": 20, + "mirror_readout_probe_examples": 20480, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 0, + "total_forward_equivalent_examples": 0, + "total_macs_estimate": 1630535680 + } +} diff --git a/results/mirror_causal_capture/wm_etaM0.3.json b/results/mirror_causal_capture/wm_etaM0.3.json new file mode 100644 index 0000000..5bdc1cb --- /dev/null +++ b/results/mirror_causal_capture/wm_etaM0.3.json @@ -0,0 +1,451 @@ +{ + "architecture": { + "adaptive_apical_parameters": 0, + "adaptive_feedback_parameters": 267904, + "base_width": 16, + "blocks_per_stage": 3, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 20, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 0, + "forward_parameters": 269722, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 0, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 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, + "mirror_batch_size": 1, + "mirror_eta": 0.3, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 20, + "mode": "wm", + "momentum": 0.9, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "results/mirror_causal_capture/wm_etaM0.3.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_warmup_steps": 0, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "train_limit": 10000, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 0, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "local_parent_child_response", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 20, + "mirror_events": 20, + "mirror_readout_examples": 20480, + "ordinary_examples": 0, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_warmup_examples": 0 + }, + "diagnostics": { + "early_third_mean": 0.11217667038242023, + "feedback_forward_cosine": [ + 0.9873843193054199, + 0.9871039390563965, + 0.9865926504135132, + 0.9879629611968994, + 0.9872314929962158, + 0.9884469509124756, + 0.9534262418746948, + 0.9105166792869568, + 0.9125481247901917, + 0.913772463798523, + 0.9096975326538086, + 0.9096673727035522, + 0.7419459819793701, + 0.6068660616874695, + 0.6080100536346436, + 0.6111626625061035, + 0.6065012216567993, + 0.6100532412528992, + 0.9946354031562805 + ], + "feedback_forward_norm_ratio": [ + 1.0173327922821045, + 1.0102897882461548, + 0.9990898966789246, + 1.0101393461227417, + 1.016475796699524, + 1.0270946025848389, + 1.0546412467956543, + 1.1024858951568604, + 1.0919255018234253, + 1.0828874111175537, + 1.1040380001068115, + 1.1013133525848389, + 1.355318546295166, + 1.649383306503296, + 1.6965118646621704, + 1.6202107667922974, + 1.6564159393310547, + 1.6852145195007324, + 1.0073041915893555 + ], + "innovation_negative_gradient_cosine": [ + 0.10878093540668488, + 0.11161622405052185, + 0.11394535005092621, + 0.11137086153030396, + 0.1137349009513855, + 0.11361175030469894, + 0.112947016954422, + 0.11172385513782501, + 0.11992612481117249, + 0.1252836287021637, + 0.13624073565006256, + 0.13715432584285736, + 0.145131915807724, + 0.1452513039112091, + 0.241938978433609, + 0.30763790011405945, + 0.48825663328170776, + 0.5947017669677734, + 0.9949096441268921 + ], + "normalization_state": "training_batch_stats_without_running_update", + "raw_negative_gradient_cosine": [ + 0.10878093540668488, + 0.11161622405052185, + 0.11394535005092621, + 0.11137086153030396, + 0.1137349009513855, + 0.11361175030469894, + 0.112947016954422, + 0.11172385513782501, + 0.11992612481117249, + 0.1252836287021637, + 0.13624073565006256, + 0.13715432584285736, + 0.145131915807724, + 0.1452513039112091, + 0.241938978433609, + 0.30763790011405945, + 0.48825663328170776, + 0.5947017669677734, + 0.9949096441268921 + ], + "teaching_negative_gradient_cosine": [ + 0.10878093540668488, + 0.11161622405052185, + 0.11394535005092621, + 0.11137086153030396, + 0.1137349009513855, + 0.11361175030469894, + 0.112947016954422, + 0.11172385513782501, + 0.11992612481117249, + 0.1252836287021637, + 0.13624073565006256, + 0.13715432584285736, + 0.145131915807724, + 0.1452513039112091, + 0.241938978433609, + 0.30763790011405945, + 0.48825663328170776, + 0.5947017669677734, + 0.9949096441268921 + ], + "wall_s": 0.14173507690429688 + }, + "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": 1730634752, + "peak_memory_reserved_bytes": 2181038080, + "torch_version": "2.3.1+cu118" + }, + "mirror_warmup": { + "first": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.1821084153449638, + "mirror_update_rms": 0.19537643727527548, + "readout_batch_size": 1024 + }, + "last": { + "conv_batch_size": 1, + "mirror_estimate_rms": 0.18073508053859827, + "mirror_update_rms": 0.1794914991970627, + "readout_batch_size": 1024 + }, + "mean": { + "conv_batch_size": 1.0, + "mirror_estimate_rms": 0.1793198434405597, + "mirror_update_rms": 0.179802025533114, + "readout_batch_size": 1024.0 + }, + "steps": 20 + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "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": 0.41222262382507324, + "mirror_warmup_wall_s": 0.47155046463012695, + "predictor_warmup_wall_s": 0.0, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 0.8840188980102539, + "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, + "mirror_local_correlation_macs": 815267840, + "mirror_response_macs": 815267840, + "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, + "mirror_probe_examples": 20, + "mirror_readout_probe_examples": 20480, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 0, + "total_forward_equivalent_examples": 0, + "total_macs_estimate": 1630535680 + } +} diff --git a/results/mirror_causal_capture_gate.json b/results/mirror_causal_capture_gate.json new file mode 100644 index 0000000..8023860 --- /dev/null +++ b/results/mirror_causal_capture_gate.json @@ -0,0 +1,129 @@ +{ + "checks": { + "all_four_records_finite": true, + "all_layer_at_least_0.50": true, + "candidate_selected": true, + "early_third_at_least_0.40": true, + "feedback_norm_ratios_in_0.5_to_1.5": true, + "mean_feedback_forward_cosine_at_least_0.85": true, + "zero_task_loss_queries": true + }, + "confirmation_test_seeds_touched": false, + "matched_fixed_hfa": { + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.01127886730950865, + "early_third_alignment": -0.0002631632038780178, + "max_feedback_forward_norm_ratio": 1.0550107955932617, + "mean_feedback_forward_cosine": 0.0016142042954207252, + "min_feedback_forward_norm_ratio": 0.9593725800514221 + }, + "mirror_eta": null, + "mirror_events": 0, + "mode": "hfa", + "path": "results/mirror_causal_capture/fixed_hfa.json", + "source_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "total_macs": 0 + }, + "protocol": "normalized_response_mirror_capture_v1", + "review_score_after": 5, + "review_score_before": 5, + "rows": [ + { + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.01127886730950865, + "early_third_alignment": -0.0002631632038780178, + "max_feedback_forward_norm_ratio": 1.0550107955932617, + "mean_feedback_forward_cosine": 0.0016142042954207252, + "min_feedback_forward_norm_ratio": 0.9593725800514221 + }, + "mirror_eta": null, + "mirror_events": 0, + "mode": "hfa", + "path": "results/mirror_causal_capture/fixed_hfa.json", + "source_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "total_macs": 0 + }, + { + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.23519715333455488, + "early_third_alignment": 0.07526319908599059, + "max_feedback_forward_norm_ratio": 0.7866175770759583, + "mean_feedback_forward_cosine": 0.6251863680387798, + "mean_mirror_estimate_rms": 0.1793198434405597, + "mean_mirror_update_rms": 0.18231021973395675, + "min_feedback_forward_norm_ratio": 0.6969802379608154 + }, + "mirror_eta": 0.03, + "mirror_events": 20, + "mode": "wm", + "path": "results/mirror_causal_capture/wm_etaM0.03.json", + "source_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "total_macs": 1630535680 + }, + { + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.5421591746179681, + "early_third_alignment": 0.44614077111085254, + "max_feedback_forward_norm_ratio": 1.1570484638214111, + "mean_feedback_forward_cosine": 0.9154801933388961, + "mean_mirror_estimate_rms": 0.1793198434405597, + "mean_mirror_update_rms": 0.17517745811581048, + "min_feedback_forward_norm_ratio": 0.8893439173698425 + }, + "mirror_eta": 0.1, + "mirror_events": 20, + "mode": "wm", + "path": "results/mirror_causal_capture/wm_etaM0.1.json", + "source_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "total_macs": 1630535680 + }, + { + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.22811388694926313, + "early_third_alignment": 0.11217667038242023, + "max_feedback_forward_norm_ratio": 1.6965118646621704, + "mean_feedback_forward_cosine": 0.8533434397295901, + "mean_mirror_estimate_rms": 0.1793198434405597, + "mean_mirror_update_rms": 0.179802025533114, + "min_feedback_forward_norm_ratio": 0.9990898966789246 + }, + "mirror_eta": 0.3, + "mirror_events": 20, + "mode": "wm", + "path": "results/mirror_causal_capture/wm_etaM0.3.json", + "source_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "total_macs": 1630535680 + } + ], + "score_change_rule": "inherited baseline capture cannot raise score", + "selected_wm": { + "finite": true, + "logical_batch_loss_queries": 0, + "metrics": { + "all_layer_alignment": 0.5421591746179681, + "early_third_alignment": 0.44614077111085254, + "max_feedback_forward_norm_ratio": 1.1570484638214111, + "mean_feedback_forward_cosine": 0.9154801933388961, + "mean_mirror_estimate_rms": 0.1793198434405597, + "mean_mirror_update_rms": 0.17517745811581048, + "min_feedback_forward_norm_ratio": 0.8893439173698425 + }, + "mirror_eta": 0.1, + "mirror_events": 20, + "mode": "wm", + "path": "results/mirror_causal_capture/wm_etaM0.1.json", + "source_commit": "37cf78b7f696c9eedc0745ebdaa656bff0ab2f81", + "total_macs": 1630535680 + }, + "status": "passed" +} -- cgit v1.2.3