From 0660bd1f6274e7982438506eec67870b105a0285 Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Wed, 22 Jul 2026 04:58:05 -0500 Subject: results: close calibration quality recovery --- RESULTS.md | 21 +++++++++++++++++++++ ROADMAP.md | 13 +++++++++++++ ...calibdev_v1_tent_l2_w8_context_k16_d1_t0_s0.json | 1 + ...calibdev_v1_tent_l2_w8_context_k16_d1_t0_s1.json | 1 + ...calibdev_v1_tent_l2_w8_context_k16_d1_t0_s2.json | 1 + ...calibdev_v1_tent_l2_w8_context_k16_d4_t0_s0.json | 1 + ...calibdev_v1_tent_l2_w8_context_k16_d4_t0_s1.json | 1 + ...calibdev_v1_tent_l2_w8_context_k16_d4_t0_s2.json | 1 + ..._calibdev_v1_tent_l2_w8_context_k4_d1_t0_s0.json | 1 + ..._calibdev_v1_tent_l2_w8_context_k4_d1_t0_s1.json | 1 + ..._calibdev_v1_tent_l2_w8_context_k4_d1_t0_s2.json | 1 + ..._calibdev_v1_tent_l2_w8_context_k4_d4_t0_s0.json | 1 + ..._calibdev_v1_tent_l2_w8_context_k4_d4_t0_s1.json | 1 + ..._calibdev_v1_tent_l2_w8_context_k4_d4_t0_s2.json | 1 + 14 files changed, 46 insertions(+) create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s0.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s1.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s2.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s0.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s1.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s2.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s0.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s1.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s2.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s0.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s1.json create mode 100644 results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s2.json diff --git a/RESULTS.md b/RESULTS.md index 2affc24..b0ac5dd 100644 --- a/RESULTS.md +++ b/RESULTS.md @@ -384,6 +384,27 @@ This is a diagnosis and upper-cost baseline, not the proposed scalable method. A `68.4x` ordinary training work, with `108.2` seconds mean wall time. The next recovery must retain the vectorizer and disclose how much of this query burden it avoids. +### Calibration-quality development screen (task seed 0 only) + +The first post-diagnosis development screen kept the context vectorizer and concurrent A/W updates +fixed, replacing simultaneous K1/e4 calibration by layerwise K4/e4 or K16/e4. The frozen rule +selected the higher mean d4 endpoint, preferring K4 within one point; all three model seeds and +both depths are retained. + +| directions | depth | validation acc (%) | depth gain | d4 lesion drop | d4 work / ordinary | +|-----------:|------:|-------------------:|-----------:|----------------:|-------------------:| +| K4 | 1 | 66.217 ± 5.264 | — | — | — | +| K4 | 4 | 65.633 ± 27.078 | -0.583 | +13.733 | 5.5x | +| K16 | 1 | 65.917 ± 4.753 | — | — | — | +| K16 | 4 | 65.567 ± 26.962 | -0.350 | +14.033 | 18.2x | + +Only one of three depth pairs and one of three d4 lesions are positive for either candidate. K16 +has one nonfinite d4 final loss and is ineligible; K4 is selected mechanically but is not advanced +to independent confirmation because its mean depth gain is negative. **Better per-event causal +targets do not by themselves cure the coupled A/W basin.** The next bounded development question +is temporal: calibrate A while W is held fixed, then open forward plasticity and maintain A with +the already selected low-query simultaneous schedule. + ## Paper-faithful non-backprop baselines `PETITE` in the initial request is treated as **PEPITA**. Implementations and hyperparameters were diff --git a/ROADMAP.md b/ROADMAP.md index 9a6f46f..60591c9 100644 --- a/ROADMAP.md +++ b/ROADMAP.md @@ -172,6 +172,19 @@ ineligible. No result from task seeds 3--5 may enter this selection. A selected frozen before a new independent panel on task seeds 6--8, and that panel must satisfy the original C2 gain/comparator/lesion conditions plus at least 10/15 positive depth gains and finite losses. +**Calibration-quality recovery status: failed development on 2026-07-22.** K4/e4 and K16/e4 +reached only `65.633 ± 27.078%` and `65.567 ± 26.962%` at d4, with mean depth gains of `-0.583` +and `-0.350` points. Each had only 1/3 positive depth gains and lesions; K16 also produced one +nonfinite d4 loss. The frozen rule mechanically selects finite K4, but its negative depth gain +prohibits spending untouched task seeds 6--8 on confirmation. Increasing target quality while A +and W learn concurrently is therefore closed as a recovery branch. + +One feedback-first timescale screen is allowed next on development task seed 0. It will use an +A-only calibration prefix with W and the output readout frozen, followed by the original +context-gated simultaneous K1/e4 joint phase. The prefix uses layerwise K4 targets and the same +`eta_A=0.01`; its length grid and selection rule must be committed before any run. This tests the +specific coupled-basin diagnosis without increasing vectorizer capacity or changing the task. + ### C3. Causal calibration is genuinely amortized **Status: passed on 2026-07-22.** The frozen five-seed confirmation retained 112.9% of the K16/e4 diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s0.json b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s0.json new file mode 100644 index 0000000..7c203a5 --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7099499106407166}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6415, "eval_loss": 0.613212158203125, "wall_s": 7.721480369567871, "val_acc": 0.6415, "val_loss": 0.613212158203125, "cos_r_negg": [0.9991190433502197], "cos_innovation_negg": [0.9991190433502197], "cos_apical_negg": [0.9991190433502197], "cos_Ac_negg": [0.9991190433502197], "r_norm": [3.28287410736084], "g_norm": [0.007435725536197424], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.64409065246582, "diagnostics_wall_s": 0.048326730728149414, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.64409065246582, "evaluation_wall_s": 0.024424076080322266}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 21120, "calibration_example_loss_evaluations": 5406720, "calibration_forward_equivalent_examples": 3659093.333333364, "training_forward_equivalent_examples": 4299093.333333364, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396736, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s1.json b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s1.json new file mode 100644 index 0000000..a6ef000 --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6922826766967773}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.623, "eval_loss": 0.6227610168457032, "wall_s": 9.846624851226807, "val_acc": 0.623, "val_loss": 0.6227610168457032, "cos_r_negg": [0.9995636940002441], "cos_innovation_negg": [0.9995636940002441], "cos_apical_negg": [0.9995636940002441], "cos_Ac_negg": [0.9995636940002441], "r_norm": [3.0956993103027344], "g_norm": [0.0070134601555764675], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 9.763425588607788, "diagnostics_wall_s": 0.052111148834228516, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 9.763425588607788, "evaluation_wall_s": 0.025647401809692383}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 21120, "calibration_example_loss_evaluations": 5406720, "calibration_forward_equivalent_examples": 3659093.333333364, "training_forward_equivalent_examples": 4299093.333333364, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396736, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s2.json b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s2.json new file mode 100644 index 0000000..d732197 --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k16_d1_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055632472038269}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.713, "eval_loss": 0.6081356506347656, "wall_s": 7.955592155456543, "val_acc": 0.713, "val_loss": 0.6081356506347656, "cos_r_negg": [0.9988279342651367], "cos_innovation_negg": [0.9988279342651367], "cos_apical_negg": [0.9988279342651367], "cos_Ac_negg": [0.9988279342651367], "r_norm": [2.9435062408447266], "g_norm": [0.006616460159420967], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 7.88253927230835, "diagnostics_wall_s": 0.04418826103210449, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 7.88253927230835, "evaluation_wall_s": 0.024120092391967773}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 21120, "calibration_example_loss_evaluations": 5406720, "calibration_forward_equivalent_examples": 3659093.333333364, "training_forward_equivalent_examples": 4299093.333333364, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396736, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s0.json b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s0.json new file mode 100644 index 0000000..48e1067 --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6903698444366455}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.967, "eval_loss": 0.0918966064453125, "wall_s": 37.54576015472412, "val_acc": 0.967, "val_loss": 0.0918966064453125, "cos_r_negg": [0.4613859951496124, 0.8076952695846558, 0.5814062356948853, 0.9952620267868042], "cos_innovation_negg": [0.4613859951496124, 0.8076952695846558, 0.5814062356948853, 0.9952620267868042], "cos_apical_negg": [0.4613859951496124, 0.8076952695846558, 0.5814062356948853, 0.9952620267868042], "cos_Ac_negg": [0.4613859951496124, 0.8076952695846558, 0.5814062356948853, 0.9952620267868042], "r_norm": [1.2774972915649414, 1.2854582071304321, 1.0173633098602295, 0.7485694289207458], "g_norm": [0.00627327524125576, 0.0048269121907651424, 0.0032183462753891945, 0.001905190758407116], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.7272084355354309, 1.06477952003479, 0.6488555669784546], "lesion_eval_acc": 0.546, "lesion_eval_loss": 5.615177001953125, "lesion_acc_drop": 0.42099999999999993}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 37.42564058303833, "diagnostics_wall_s": 0.075653076171875, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 37.42564058303833, "evaluation_wall_s": 0.039613962173461914}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 82560, "calibration_example_loss_evaluations": 21135360, "calibration_forward_equivalent_examples": 11037961.48148145, "training_forward_equivalent_examples": 11677961.48148145, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17590784, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s1.json b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s1.json new file mode 100644 index 0000000..aa9e19f --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7494125366210938}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5, "eval_loss": 9762.514, "wall_s": 35.32595610618591, "val_acc": 0.5, "val_loss": 9762.514, "cos_r_negg": [NaN, 0.16398262977600098, NaN, NaN], "cos_innovation_negg": [NaN, 0.16398262977600098, NaN, NaN], "cos_apical_negg": [NaN, 0.16398262977600098, NaN, NaN], "cos_Ac_negg": [NaN, 0.16398262977600098, NaN, NaN], "r_norm": [32761.859375, 6088.310546875, 5969.9365234375, 6006.544921875], "g_norm": [66.44831848144531, 66.77066802978516, 66.77066802978516, 66.77066802978516], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [212537345835008.0, 0.0, 0.0], "lesion_eval_acc": 0.5, "lesion_eval_loss": 9762.514, "lesion_acc_drop": 0.0}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 35.20574903488159, "diagnostics_wall_s": 0.07488751411437988, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 35.20574903488159, "evaluation_wall_s": 0.040342092514038086}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 82560, "calibration_example_loss_evaluations": 21135360, "calibration_forward_equivalent_examples": 11037961.48148145, "training_forward_equivalent_examples": 11677961.48148145, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17590784, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s2.json b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s2.json new file mode 100644 index 0000000..5ad2470 --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k16_d4_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6906555891036987}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5, "eval_loss": NaN, "wall_s": 37.2145893573761, "val_acc": 0.5, "val_loss": NaN, "cos_r_negg": [NaN, NaN, NaN, NaN], "cos_innovation_negg": [NaN, NaN, NaN, NaN], "cos_apical_negg": [NaN, NaN, NaN, NaN], "cos_Ac_negg": [NaN, NaN, NaN, NaN], "r_norm": [NaN, NaN, NaN, NaN], "g_norm": [NaN, NaN, NaN, NaN], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [NaN, NaN, NaN, NaN], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [NaN, NaN, NaN], "lesion_eval_acc": 0.5, "lesion_eval_loss": NaN, "lesion_acc_drop": 0.0}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 37.09890627861023, "diagnostics_wall_s": 0.06829667091369629, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 37.09890627861023, "evaluation_wall_s": 0.0421748161315918}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 82560, "calibration_example_loss_evaluations": 21135360, "calibration_forward_equivalent_examples": 11037961.48148145, "training_forward_equivalent_examples": 11677961.48148145, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17590784, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s0.json b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s0.json new file mode 100644 index 0000000..aa90d57 --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 4, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7099499106407166}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.6415, "eval_loss": 0.6132747192382813, "wall_s": 5.332623720169067, "val_acc": 0.6415, "val_loss": 0.6132747192382813, "cos_r_negg": [0.9991611242294312], "cos_innovation_negg": [0.9991611242294312], "cos_apical_negg": [0.9991611242294312], "cos_Ac_negg": [0.9991611242294312], "r_norm": [3.295236825942993], "g_norm": [0.007441789843142033], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.25883150100708, "diagnostics_wall_s": 0.04359292984008789, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.25883150100708, "evaluation_wall_s": 0.025421857833862305}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 5760, "calibration_example_loss_evaluations": 1474560, "calibration_forward_equivalent_examples": 1037653.3333333432, "training_forward_equivalent_examples": 1677653.3333333433, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396736, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s1.json b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s1.json new file mode 100644 index 0000000..ccb3b2d --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 4, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6922826766967773}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.623, "eval_loss": 0.6226553039550782, "wall_s": 5.299499273300171, "val_acc": 0.623, "val_loss": 0.6226553039550782, "cos_r_negg": [0.9995771646499634], "cos_innovation_negg": [0.9995771646499634], "cos_apical_negg": [0.9995771646499634], "cos_Ac_negg": [0.9995771646499634], "r_norm": [3.1129517555236816], "g_norm": [0.007015618495643139], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.221837520599365, "diagnostics_wall_s": 0.04796934127807617, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.221837520599365, "evaluation_wall_s": 0.02460312843322754}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 5760, "calibration_example_loss_evaluations": 1474560, "calibration_forward_equivalent_examples": 1037653.3333333432, "training_forward_equivalent_examples": 1677653.3333333433, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396736, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s2.json b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s2.json new file mode 100644 index 0000000..50b38c8 --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 1, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.0, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 4, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k4_d1_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7055632472038269}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.722, "eval_loss": 0.6086116333007813, "wall_s": 5.21956992149353, "val_acc": 0.722, "val_loss": 0.6086116333007813, "cos_r_negg": [0.9989438056945801], "cos_innovation_negg": [0.9989438056945801], "cos_apical_negg": [0.9989438056945801], "cos_Ac_negg": [0.9989438056945801], "r_norm": [2.9625072479248047], "g_norm": [0.006611068733036518], "traffic_norm": [0.0], "traffic_residual_norm": [0.0], "traffic_r2": [NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 5.144244432449341, "diagnostics_wall_s": 0.04676246643066406, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 5.144244432449341, "evaluation_wall_s": 0.024303197860717773}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 5760, "calibration_example_loss_evaluations": 1474560, "calibration_forward_equivalent_examples": 1037653.3333333432, "training_forward_equivalent_examples": 1677653.3333333433, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17396736, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s0.json b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s0.json new file mode 100644 index 0000000..878fdfd --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s0.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 4, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s0", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6903698444366455}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.969, "eval_loss": 0.08845006942749023, "wall_s": 17.606709480285645, "val_acc": 0.969, "val_loss": 0.08845006942749023, "cos_r_negg": [0.48526817560195923, 0.8153876066207886, 0.5922392010688782, 0.9952086210250854], "cos_innovation_negg": [0.48526817560195923, 0.8153876066207886, 0.5922392010688782, 0.9952086210250854], "cos_apical_negg": [0.48526817560195923, 0.8153876066207886, 0.5922392010688782, 0.9952086210250854], "cos_Ac_negg": [0.48526817560195923, 0.8153876066207886, 0.5922392010688782, 0.9952086210250854], "r_norm": [1.2112762928009033, 1.2245585918426514, 0.9780949950218201, 0.7144197225570679], "g_norm": [0.005882170982658863, 0.004561249166727066, 0.003075423650443554, 0.00182157673407346], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.7262412905693054, 1.0530037879943848, 0.6449858546257019], "lesion_eval_acc": 0.557, "lesion_eval_loss": 5.51016943359375, "lesion_acc_drop": 0.4119999999999999}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 17.487297296524048, "diagnostics_wall_s": 0.07443833351135254, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 17.487297296524048, "evaluation_wall_s": 0.040146827697753906}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 21120, "calibration_example_loss_evaluations": 5406720, "calibration_forward_equivalent_examples": 2882370.370370366, "training_forward_equivalent_examples": 3522370.370370366, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17590784, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s1.json b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s1.json new file mode 100644 index 0000000..9167b8f --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s1.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 4, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s1", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.7494125366210938}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5, "eval_loss": 93008.008, "wall_s": 24.880991458892822, "val_acc": 0.5, "val_loss": 93008.008, "cos_r_negg": [-0.9838489890098572, -1.0, NaN, NaN], "cos_innovation_negg": [-0.9838489890098572, -1.0, NaN, NaN], "cos_apical_negg": [-0.9838489890098572, -1.0, NaN, NaN], "cos_Ac_negg": [-0.9838489890098572, -1.0, NaN, NaN], "r_norm": [78.68692016601562, 73.67909240722656, 75.00726318359375, 75.61952209472656], "g_norm": [2.6812758445739746, 0.22072646021842957, 0.22072646021842957, 0.22072646021842957], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [24.448974609375, 0.0, 0.0], "lesion_eval_acc": 0.5, "lesion_eval_loss": 93008.008, "lesion_acc_drop": 0.0}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 24.758497714996338, "diagnostics_wall_s": 0.07221174240112305, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.758497714996338, "evaluation_wall_s": 0.04427909851074219}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 21120, "calibration_example_loss_evaluations": 5406720, "calibration_forward_equivalent_examples": 2882370.370370366, "training_forward_equivalent_examples": 3522370.370370366, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17590784, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file diff --git a/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s2.json b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s2.json new file mode 100644 index 0000000..a2aa83e --- /dev/null +++ b/results/c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s2.json @@ -0,0 +1 @@ +{"args": {"mode": "sdil", "dataset": "tentmap", "depth": 4, "width": 8, "act": "relu", "residual": 1, "residual_lesion_fraction": 0.3333333333333333, "epochs": 80, "batch_size": 256, "train_examples": 0, "val_examples": 2000, "split_seed": 2027, "eval_split": "validation", "task_seed": 0, "task_train_examples": 10000, "task_test_examples": 5000, "task_levels": 2, "task_n_in": 1, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.03, "eta_A": 0.01, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 4, "pert_mode": "layerwise", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "traffic_seed": 1234, "traffic_mode": "none", "predictor_mode": "diagonal", "vectorizer_mode": "context_gated", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c2_calibdev_v1_tent_l2_w8_context_k4_d4_t0_s2", "n_in": 1, "n_out": 2}, "split": {"dataset": "tentmap", "task_seed": 0, "train_examples": 8000, "test_examples": 5000, "evaluation_split": "validation", "synthetic_generator": true, "split_seed": 2027, "validation_examples": 2000, "validation_index_sha256": "f292d9bfeef1b6e79c80d4cf1c572ac12dee1be799a71a126390709d0386c6e4", "validation_class_counts": {"0": 1000, "1": 1000}, "split_from_training_only": true}, "provenance": {"git_commit": "9c99df512ba18217a7ec5f7d95175e490a6ba7f5", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 0.6906555891036987}, {"epoch_end": 0, "step": 32}, {"epoch_end": 1, "step": 64}, {"epoch_end": 2, "step": 96}, {"epoch_end": 3, "step": 128}, {"epoch_end": 4, "step": 160}, {"epoch_end": 5, "step": 192}, {"epoch_end": 6, "step": 224}, {"epoch_end": 7, "step": 256}, {"epoch_end": 8, "step": 288}, {"epoch_end": 9, "step": 320}, {"epoch_end": 10, "step": 352}, {"epoch_end": 11, "step": 384}, {"epoch_end": 12, "step": 416}, {"epoch_end": 13, "step": 448}, {"epoch_end": 14, "step": 480}, {"epoch_end": 15, "step": 512}, {"epoch_end": 16, "step": 544}, {"epoch_end": 17, "step": 576}, {"epoch_end": 18, "step": 608}, {"epoch_end": 19, "step": 640}, {"epoch_end": 20, "step": 672}, {"epoch_end": 21, "step": 704}, {"epoch_end": 22, "step": 736}, {"epoch_end": 23, "step": 768}, {"epoch_end": 24, "step": 800}, {"epoch_end": 25, "step": 832}, {"epoch_end": 26, "step": 864}, {"epoch_end": 27, "step": 896}, {"epoch_end": 28, "step": 928}, {"epoch_end": 29, "step": 960}, {"epoch_end": 30, "step": 992}, {"epoch_end": 31, "step": 1024}, {"epoch_end": 32, "step": 1056}, {"epoch_end": 33, "step": 1088}, {"epoch_end": 34, "step": 1120}, {"epoch_end": 35, "step": 1152}, {"epoch_end": 36, "step": 1184}, {"epoch_end": 37, "step": 1216}, {"epoch_end": 38, "step": 1248}, {"epoch_end": 39, "step": 1280}, {"epoch_end": 40, "step": 1312}, {"epoch_end": 41, "step": 1344}, {"epoch_end": 42, "step": 1376}, {"epoch_end": 43, "step": 1408}, {"epoch_end": 44, "step": 1440}, {"epoch_end": 45, "step": 1472}, {"epoch_end": 46, "step": 1504}, {"epoch_end": 47, "step": 1536}, {"epoch_end": 48, "step": 1568}, {"epoch_end": 49, "step": 1600}, {"epoch_end": 50, "step": 1632}, {"epoch_end": 51, "step": 1664}, {"epoch_end": 52, "step": 1696}, {"epoch_end": 53, "step": 1728}, {"epoch_end": 54, "step": 1760}, {"epoch_end": 55, "step": 1792}, {"epoch_end": 56, "step": 1824}, {"epoch_end": 57, "step": 1856}, {"epoch_end": 58, "step": 1888}, {"epoch_end": 59, "step": 1920}, {"epoch_end": 60, "step": 1952}, {"epoch_end": 61, "step": 1984}, {"epoch_end": 62, "step": 2016}, {"epoch_end": 63, "step": 2048}, {"epoch_end": 64, "step": 2080}, {"epoch_end": 65, "step": 2112}, {"epoch_end": 66, "step": 2144}, {"epoch_end": 67, "step": 2176}, {"epoch_end": 68, "step": 2208}, {"epoch_end": 69, "step": 2240}, {"epoch_end": 70, "step": 2272}, {"epoch_end": 71, "step": 2304}, {"epoch_end": 72, "step": 2336}, {"epoch_end": 73, "step": 2368}, {"epoch_end": 74, "step": 2400}, {"epoch_end": 75, "step": 2432}, {"epoch_end": 76, "step": 2464}, {"epoch_end": 77, "step": 2496}, {"epoch_end": 78, "step": 2528}, {"epoch_end": 79, "step": 2560}], "final": {"eval_split": "validation", "eval_acc": 0.5, "eval_loss": 9439962.112, "wall_s": 18.167455196380615, "val_acc": 0.5, "val_loss": 9439962.112, "cos_r_negg": [-0.8005666732788086, NaN, 1.0, NaN], "cos_innovation_negg": [-0.8005666732788086, NaN, 1.0, NaN], "cos_apical_negg": [-0.8005666732788086, NaN, 1.0, NaN], "cos_Ac_negg": [-0.8005666732788086, NaN, 1.0, NaN], "r_norm": [241.97695922851562, 469.93414306640625, 119.97386169433594, 106.78618621826172], "g_norm": [55.40341567993164, 55.40341567993164, 3.848010540008545, 3.848010540008545], "traffic_norm": [0.0, 0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN, NaN], "residual_lesion": {"interior_layers": [1, 2, 3], "lesioned_layers": [3], "branch_to_skip_rms": [0.0, 253.20391845703125, 0.0], "lesion_eval_acc": 0.5, "lesion_eval_loss": 9439962.112, "lesion_acc_drop": 0.0}}, "lesion_protocol": {"fraction_of_interior_blocks": 0.3333333333333333, "selection": "final_contiguous_interior_blocks", "probe_source": "training_prefix", "probe_examples": 512, "evaluation_split": "validation"}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 18.05645203590393, "diagnostics_wall_s": 0.06594657897949219, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 18.05645203590393, "evaluation_wall_s": 0.04004216194152832}, "cost": {"train_steps": 2560, "ordinary_training_forward_examples": 640000, "predictor_warmup_forward_examples": 0, "perturbation_events": 640, "calibration_batch_loss_evaluations": 21120, "calibration_example_loss_evaluations": 5406720, "calibration_forward_equivalent_examples": 2882370.370370366, "training_forward_equivalent_examples": 3522370.370370366, "max_perturbation_batch_expansion": 1}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "7", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 17590784, "peak_memory_reserved_bytes": 23068672}} \ No newline at end of file -- cgit v1.2.3