95 files changed, 95 insertions, 0 deletions

diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s0.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s0.json
new file mode 100644
index 0000000..dd8af4d
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.575254201889038}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9732, "eval_loss": 0.11138565406799317, "wall_s": 24.299553632736206, "test_acc": 0.9732, "test_loss": 0.11138565406799317, "cos_r_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_innovation_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_apical_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_Ac_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "r_norm": [0.011775060556828976, 0.01129802968353033, 0.011105283163487911], "g_norm": [2.9951132091809995e-05, 2.23430288315285e-05, 1.8669790733838454e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 24.1998610496521, "diagnostics_wall_s": 0.06180858612060547, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.1998610496521, "evaluation_wall_s": 0.03634524345397949}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251389952, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s1.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s1.json
new file mode 100644
index 0000000..69866cb
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3718113899230957}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9754, "eval_loss": 0.11091463012695313, "wall_s": 23.494471311569214, "test_acc": 0.9754, "test_loss": 0.11091463012695313, "cos_r_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_innovation_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_apical_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_Ac_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "r_norm": [0.006153065711259842, 0.005907644052058458, 0.006034198682755232], "g_norm": [1.6483892977703363e-05, 1.2261286428838503e-05, 9.706364835437853e-06], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 23.40929365158081, "diagnostics_wall_s": 0.043768882751464844, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.40929365158081, "evaluation_wall_s": 0.039998531341552734}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "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": 251389952, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s2.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s2.json
new file mode 100644
index 0000000..20ae215
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4491982460021973}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9737, "eval_loss": 0.11859232807159424, "wall_s": 25.446815490722656, "test_acc": 0.9737, "test_loss": 0.11859232807159424, "cos_r_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_innovation_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_apical_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_Ac_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "r_norm": [0.00880928710103035, 0.00863414816558361, 0.00877653993666172], "g_norm": [2.5090203052968718e-05, 1.8679147615330294e-05, 1.479045477026375e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 25.374990940093994, "diagnostics_wall_s": 0.04135560989379883, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.374990940093994, "evaluation_wall_s": 0.029029130935668945}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251389952, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s3.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s3.json
new file mode 100644
index 0000000..a9cee3c
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "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", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.5783634185791016}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9754, "eval_loss": 0.11722123413085937, "wall_s": 26.30913019180298, "test_acc": 0.9754, "test_loss": 0.11722123413085937, "cos_r_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_innovation_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_apical_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_Ac_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "r_norm": [0.013960935175418854, 0.013710178434848785, 0.013982996344566345], "g_norm": [3.613009539549239e-05, 2.8648952138610184e-05, 2.2737141989637166e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 26.226622581481934, "diagnostics_wall_s": 0.048334598541259766, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 26.226622581481934, "evaluation_wall_s": 0.0327146053314209}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "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": 251389952, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s4.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s4.json
new file mode 100644
index 0000000..b0ec683
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "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", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_none_rho0_t1234_matched_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4974002838134766}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.976, "eval_loss": 0.10724757919311523, "wall_s": 24.116638898849487, "test_acc": 0.976, "test_loss": 0.10724757919311523, "cos_r_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_innovation_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_apical_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_Ac_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "r_norm": [0.01008433848619461, 0.010013793595135212, 0.009999487549066544], "g_norm": [2.774084896373097e-05, 2.0131683413637802e-05, 1.640784103074111e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 24.032322645187378, "diagnostics_wall_s": 0.04549741744995117, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.032322645187378, "evaluation_wall_s": 0.037337541580200195}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251389952, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s0.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s0.json
new file mode 100644
index 0000000..f4d027f
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.575254201889038}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9732, "eval_loss": 0.11138565406799317, "wall_s": 30.42645573616028, "test_acc": 0.9732, "test_loss": 0.11138565406799317, "cos_r_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_innovation_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_apical_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_Ac_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "r_norm": [0.011775060556828976, 0.01129802968353033, 0.011105283163487911], "g_norm": [2.9951132091809995e-05, 2.23430288315285e-05, 1.8669790733838454e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 30.32719349861145, "diagnostics_wall_s": 0.06258225440979004, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 30.32719349861145, "evaluation_wall_s": 0.03512859344482422}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s1.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s1.json
new file mode 100644
index 0000000..16cd084
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3718113899230957}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9754, "eval_loss": 0.11091463012695313, "wall_s": 22.51677370071411, "test_acc": 0.9754, "test_loss": 0.11091463012695313, "cos_r_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_innovation_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_apical_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_Ac_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "r_norm": [0.006153065711259842, 0.005907644052058458, 0.006034198682755232], "g_norm": [1.6483892977703363e-05, 1.2261286428838503e-05, 9.706364835437853e-06], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 22.414365768432617, "diagnostics_wall_s": 0.06711697578430176, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.414365768432617, "evaluation_wall_s": 0.03384542465209961}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "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": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s2.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s2.json
new file mode 100644
index 0000000..ce22e30
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4491982460021973}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9737, "eval_loss": 0.11859232807159424, "wall_s": 21.845251083374023, "test_acc": 0.9737, "test_loss": 0.11859232807159424, "cos_r_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_innovation_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_apical_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_Ac_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "r_norm": [0.00880928710103035, 0.00863414816558361, 0.00877653993666172], "g_norm": [2.5090203052968718e-05, 1.8679147615330294e-05, 1.479045477026375e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 21.7434024810791, "diagnostics_wall_s": 0.06464219093322754, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.7434024810791, "evaluation_wall_s": 0.03583121299743652}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s3.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s3.json
new file mode 100644
index 0000000..09b55ee
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "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", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.5783634185791016}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9754, "eval_loss": 0.11722123413085937, "wall_s": 31.394969940185547, "test_acc": 0.9754, "test_loss": 0.11722123413085937, "cos_r_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_innovation_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_apical_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_Ac_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "r_norm": [0.013960935175418854, 0.013710178434848785, 0.013982996344566345], "g_norm": [3.613009539549239e-05, 2.8648952138610184e-05, 2.2737141989637166e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 31.290912628173828, "diagnostics_wall_s": 0.06588125228881836, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 31.290912628173828, "evaluation_wall_s": 0.03657650947570801}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "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": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s4.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s4.json
new file mode 100644
index 0000000..c5af2f2
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "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", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_none_rho0_t1234_raw_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4974002838134766}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.976, "eval_loss": 0.10724757919311523, "wall_s": 23.384228467941284, "test_acc": 0.976, "test_loss": 0.10724757919311523, "cos_r_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_innovation_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_apical_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_Ac_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "r_norm": [0.01008433848619461, 0.010013793595135212, 0.009999487549066544], "g_norm": [2.774084896373097e-05, 2.0131683413637802e-05, 1.640784103074111e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 23.291709899902344, "diagnostics_wall_s": 0.061714887619018555, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.291709899902344, "evaluation_wall_s": 0.029337167739868164}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s0.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s0.json
new file mode 100644
index 0000000..f62543e
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.575254201889038}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9732, "eval_loss": 0.11138565406799317, "wall_s": 22.438310861587524, "test_acc": 0.9732, "test_loss": 0.11138565406799317, "cos_r_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_innovation_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_apical_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "cos_Ac_negg": [0.6306580901145935, 0.3576743006706238, 0.3238198161125183], "r_norm": [0.011775060556828976, 0.01129802968353033, 0.011105283163487911], "g_norm": [2.9951132091809995e-05, 2.23430288315285e-05, 1.8669790733838454e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 22.33100128173828, "diagnostics_wall_s": 0.0720365047454834, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.33100128173828, "evaluation_wall_s": 0.03364872932434082}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s1.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s1.json
new file mode 100644
index 0000000..9bf96a4
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3718113899230957}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9754, "eval_loss": 0.11091463012695313, "wall_s": 21.980138540267944, "test_acc": 0.9754, "test_loss": 0.11091463012695313, "cos_r_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_innovation_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_apical_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "cos_Ac_negg": [0.6636902689933777, 0.35125532746315, 0.3263653814792633], "r_norm": [0.006153065711259842, 0.005907644052058458, 0.006034198682755232], "g_norm": [1.6483892977703363e-05, 1.2261286428838503e-05, 9.706364835437853e-06], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 21.88034701347351, "diagnostics_wall_s": 0.062250614166259766, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.88034701347351, "evaluation_wall_s": 0.0360715389251709}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "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": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s2.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s2.json
new file mode 100644
index 0000000..97ac5af
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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", "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": "c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4491982460021973}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9737, "eval_loss": 0.11859232807159424, "wall_s": 22.313729524612427, "test_acc": 0.9737, "test_loss": 0.11859232807159424, "cos_r_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_innovation_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_apical_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "cos_Ac_negg": [0.6204646229743958, 0.3519253730773926, 0.32553520798683167], "r_norm": [0.00880928710103035, 0.00863414816558361, 0.00877653993666172], "g_norm": [2.5090203052968718e-05, 1.8679147615330294e-05, 1.479045477026375e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 22.20801615715027, "diagnostics_wall_s": 0.0666956901550293, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.20801615715027, "evaluation_wall_s": 0.0376126766204834}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s3.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s3.json
new file mode 100644
index 0000000..f43076b
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.5783634185791016}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9754, "eval_loss": 0.11722123413085937, "wall_s": 24.89414405822754, "test_acc": 0.9754, "test_loss": 0.11722123413085937, "cos_r_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_innovation_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_apical_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "cos_Ac_negg": [0.6545129418373108, 0.35001340508461, 0.3532850444316864], "r_norm": [0.013960935175418854, 0.013710178434848785, 0.013982996344566345], "g_norm": [3.613009539549239e-05, 2.8648952138610184e-05, 2.2737141989637166e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 24.79232168197632, "diagnostics_wall_s": 0.06327342987060547, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.79232168197632, "evaluation_wall_s": 0.03712940216064453}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "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": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s4.json b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s4.json
new file mode 100644
index 0000000..084a697
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_none_rho0_t1234_residual_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4974002838134766}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.976, "eval_loss": 0.10724757919311523, "wall_s": 21.939162969589233, "test_acc": 0.976, "test_loss": 0.10724757919311523, "cos_r_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_innovation_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_apical_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "cos_Ac_negg": [0.6415334939956665, 0.38924384117126465, 0.3096104562282562], "r_norm": [0.01008433848619461, 0.010013793595135212, 0.009999487549066544], "g_norm": [2.774084896373097e-05, 2.0131683413637802e-05, 1.640784103074111e-05], "traffic_norm": [0.0, 0.0, 0.0], "traffic_residual_norm": [0.0, 0.0, 0.0], "traffic_r2": [NaN, NaN, NaN]}, "timing": {"warmup_wall_s": 0.0, "training_loop_wall_s": 21.836637258529663, "diagnostics_wall_s": 0.06574678421020508, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.836637258529663, "evaluation_wall_s": 0.03518795967102051}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 0, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1575072.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 250861568, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s0.json
new file mode 100644
index 0000000..91696e3
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9384, "eval_loss": 0.28771597747802735, "wall_s": 25.519243240356445, "test_acc": 0.9384, "test_loss": 0.28771597747802735, "cos_r_negg": [0.008832640014588833, -0.0018819477409124374, 0.16430285573005676], "cos_innovation_negg": [0.6748199462890625, 0.3530857563018799, 0.30060461163520813], "cos_apical_negg": [0.008832637220621109, -0.0018819477409124374, 0.16430285573005676], "cos_Ac_negg": [0.8954887390136719, 0.5546722412109375, 0.35940009355545044], "r_norm": [0.37099215388298035, 0.318928986787796, 0.32940080761909485], "g_norm": [0.0010478491894900799, 0.0010434838477522135, 0.0010429946705698967], "traffic_norm": [8.329248428344727, 12.515583038330078, 16.629772186279297], "traffic_residual_norm": [2.0436277736735065e-06, 2.7863206923939288e-06, 3.084647232753923e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.40581250190734863, "training_loop_wall_s": 25.02034044265747, "diagnostics_wall_s": 0.05865335464477539, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.02034044265747, "evaluation_wall_s": 0.03301739692687988}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s1.json
new file mode 100644
index 0000000..5f7ce95
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9446, "eval_loss": 0.24227369766235352, "wall_s": 27.600069522857666, "test_acc": 0.9446, "test_loss": 0.24227369766235352, "cos_r_negg": [-0.06036503240466118, -0.05939137935638428, -0.08853970468044281], "cos_innovation_negg": [0.6483702659606934, 0.43442729115486145, 0.25312304496765137], "cos_apical_negg": [-0.06036502867937088, -0.05939137935638428, -0.08853968977928162], "cos_Ac_negg": [0.8865984678268433, 0.6900924444198608, 0.47889870405197144], "r_norm": [0.37710124254226685, 0.35658106207847595, 0.3585524559020996], "g_norm": [0.0010855398140847683, 0.0010805416386574507, 0.0010804599151015282], "traffic_norm": [8.331944465637207, 12.595000267028809, 16.7486629486084], "traffic_residual_norm": [2.250018951599486e-06, 3.3797728065110277e-06, 4.045399691676721e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3910202980041504, "training_loop_wall_s": 27.13391399383545, "diagnostics_wall_s": 0.0409853458404541, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 27.13391399383545, "evaluation_wall_s": 0.03275442123413086}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s2.json
new file mode 100644
index 0000000..675709a
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9447, "eval_loss": 0.24914219970703125, "wall_s": 24.80102777481079, "test_acc": 0.9447, "test_loss": 0.24914219970703125, "cos_r_negg": [0.029729261994361877, -0.06665410846471786, -0.046575769782066345], "cos_innovation_negg": [0.7170541286468506, 0.4534969925880432, 0.26807981729507446], "cos_apical_negg": [0.02972925454378128, -0.06665411591529846, -0.04657576233148575], "cos_Ac_negg": [0.8764669895172119, 0.5911966562271118, 0.4236738383769989], "r_norm": [0.3285461366176605, 0.3007344603538513, 0.3156193494796753], "g_norm": [0.0009545443463139236, 0.0009542822954244912, 0.0009539557504467666], "traffic_norm": [8.333220481872559, 12.633472442626953, 16.791173934936523], "traffic_residual_norm": [2.0188185771985445e-06, 2.3796387722541112e-06, 2.9248667487991042e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.418071985244751, "training_loop_wall_s": 24.299387216567993, "diagnostics_wall_s": 0.046390533447265625, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.299387216567993, "evaluation_wall_s": 0.035596609115600586}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s3.json
new file mode 100644
index 0000000..b5664e0
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9543, "eval_loss": 0.2171243667602539, "wall_s": 28.12531042098999, "test_acc": 0.9543, "test_loss": 0.2171243667602539, "cos_r_negg": [-0.02884913608431816, 0.055912554264068604, -0.06897787004709244], "cos_innovation_negg": [0.6546174883842468, 0.2905173599720001, 0.42574450373649597], "cos_apical_negg": [-0.02884913980960846, 0.05591254681348801, -0.06897787749767303], "cos_Ac_negg": [0.885455846786499, 0.3929228186607361, 0.4887555241584778], "r_norm": [0.31261566281318665, 0.2838948369026184, 0.2912977337837219], "g_norm": [0.0009173100697807968, 0.0009115543216466904, 0.0009115543216466904], "traffic_norm": [8.331897735595703, 12.580266952514648, 16.77628517150879], "traffic_residual_norm": [2.077749059026246e-06, 2.5801964511629194e-06, 3.050315854125074e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4315519332885742, "training_loop_wall_s": 27.604475736618042, "diagnostics_wall_s": 0.04403424263000488, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 27.604475736618042, "evaluation_wall_s": 0.04315686225891113}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s4.json
new file mode 100644
index 0000000..2a6bfd9
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_matched_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9487, "eval_loss": 0.23967629318237305, "wall_s": 25.855849027633667, "test_acc": 0.9487, "test_loss": 0.23967629318237305, "cos_r_negg": [-0.02158479206264019, -0.06811972707509995, -0.1290549784898758], "cos_innovation_negg": [0.6875743865966797, 0.3550911545753479, 0.21561986207962036], "cos_apical_negg": [-0.02158479019999504, -0.06811972707509995, -0.1290549635887146], "cos_Ac_negg": [0.9077825546264648, 0.5756210088729858, 0.4791038930416107], "r_norm": [0.3078910708427429, 0.2726975083351135, 0.29029199481010437], "g_norm": [0.000872611184604466, 0.0008609532378613949, 0.0008593454258516431], "traffic_norm": [8.328157424926758, 12.652183532714844, 16.756868362426758], "traffic_residual_norm": [2.172870836147922e-06, 2.797886736516375e-06, 3.723160261870362e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.37734389305114746, "training_loop_wall_s": 25.401981353759766, "diagnostics_wall_s": 0.042177438735961914, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.401981353759766, "evaluation_wall_s": 0.03278493881225586}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s0.json
new file mode 100644
index 0000000..6cdf113
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9391, "eval_loss": 0.2799006774902344, "wall_s": 23.822447538375854, "test_acc": 0.9391, "test_loss": 0.2799006774902344, "cos_r_negg": [-0.05449733883142471, -0.07957816123962402, 0.1536504179239273], "cos_innovation_negg": [0.6028393507003784, 0.2781147360801697, 0.5405070185661316], "cos_apical_negg": [-0.05449733883142471, -0.07957816123962402, 0.1536504179239273], "cos_Ac_negg": [0.7412353754043579, 0.4476010203361511, 0.6699270606040955], "r_norm": [8.465667724609375, 12.76370620727539, 16.760169982910156], "g_norm": [0.0011132706422358751, 0.0011062131961807609, 0.0011062121484428644], "traffic_norm": [8.352486610412598, 12.704788208007812, 16.73007583618164], "traffic_residual_norm": [2.4232836040027905e-06, 3.023187900907942e-06, 3.226175067538861e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4128270149230957, "training_loop_wall_s": 23.306333541870117, "diagnostics_wall_s": 0.062105417251586914, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.306333541870117, "evaluation_wall_s": 0.03969597816467285}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s1.json
new file mode 100644
index 0000000..fab9f72
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9317, "eval_loss": 0.3392356079101562, "wall_s": 23.868990898132324, "test_acc": 0.9317, "test_loss": 0.3392356079101562, "cos_r_negg": [-0.027143694460392, 0.18751823902130127, 0.030136264860630035], "cos_innovation_negg": [0.6832994222640991, 0.5199635624885559, 0.4485171437263489], "cos_apical_negg": [-0.027143694460392, 0.18751823902130127, 0.030136264860630035], "cos_Ac_negg": [0.758114755153656, 0.6052322387695312, 0.4402357041835785], "r_norm": [8.431733131408691, 12.711181640625, 16.858783721923828], "g_norm": [0.0009231159347109497, 0.0009228074923157692, 0.0009228008566424251], "traffic_norm": [8.351934432983398, 12.664560317993164, 16.828163146972656], "traffic_residual_norm": [2.5815943445195444e-06, 3.3177598197653424e-06, 3.246299684178666e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.5034172534942627, "training_loop_wall_s": 23.26886487007141, "diagnostics_wall_s": 0.062026262283325195, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.26886487007141, "evaluation_wall_s": 0.033202409744262695}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s2.json
new file mode 100644
index 0000000..2c3f57c
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9394, "eval_loss": 0.3063594421386719, "wall_s": 23.8635036945343, "test_acc": 0.9394, "test_loss": 0.3063594421386719, "cos_r_negg": [0.027045585215091705, -0.06143486499786377, -0.1715303212404251], "cos_innovation_negg": [0.6767744421958923, 0.4215810298919678, 0.16852208971977234], "cos_apical_negg": [0.027045585215091705, -0.06143486499786377, -0.1715303212404251], "cos_Ac_negg": [0.798451840877533, 0.5794470310211182, 0.251015841960907], "r_norm": [8.46950626373291, 12.666690826416016, 16.80787467956543], "g_norm": [0.0011514323996379972, 0.0011514330981299281, 0.001151431119069457], "traffic_norm": [8.352810859680176, 12.61552619934082, 16.760875701904297], "traffic_residual_norm": [2.3430782221112167e-06, 2.563356247264892e-06, 2.868708634196082e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.42588138580322266, "training_loop_wall_s": 23.325332641601562, "diagnostics_wall_s": 0.06626081466674805, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.325332641601562, "evaluation_wall_s": 0.04383587837219238}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s3.json
new file mode 100644
index 0000000..cd37a4e
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9387, "eval_loss": 0.2842404655456543, "wall_s": 23.279388666152954, "test_acc": 0.9387, "test_loss": 0.2842404655456543, "cos_r_negg": [-0.001771169831044972, 0.20777902007102966, 0.13769841194152832], "cos_innovation_negg": [0.6828497648239136, 0.44723325967788696, 0.5529673099517822], "cos_apical_negg": [-0.001771169831044972, 0.20777902007102966, 0.13769841194152832], "cos_Ac_negg": [0.8054697513580322, 0.3791911005973816, 0.5615983009338379], "r_norm": [8.460217475891113, 12.605305671691895, 16.4954833984375], "g_norm": [0.0012134473072364926, 0.001213288283906877, 0.001213288283906877], "traffic_norm": [8.352582931518555, 12.547894477844238, 16.449684143066406], "traffic_residual_norm": [2.384957269896404e-06, 2.8890185603813734e-06, 2.8671981908701127e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4160728454589844, "training_loop_wall_s": 22.76474928855896, "diagnostics_wall_s": 0.06191301345825195, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.76474928855896, "evaluation_wall_s": 0.03522968292236328}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s4.json
new file mode 100644
index 0000000..f85d9b4
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_raw_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9284, "eval_loss": 0.3447878829956055, "wall_s": 23.02985978126526, "test_acc": 0.9284, "test_loss": 0.3447878829956055, "cos_r_negg": [0.03912627696990967, 0.06158579885959625, 0.11182344704866409], "cos_innovation_negg": [0.7227728366851807, 0.5899728536605835, 0.40535834431648254], "cos_apical_negg": [0.03912627696990967, 0.06158579885959625, 0.11182344704866409], "cos_Ac_negg": [0.7895934581756592, 0.6133448481559753, 0.34795278310775757], "r_norm": [8.490508079528809, 12.737724304199219, 16.803142547607422], "g_norm": [0.001291685737669468, 0.0012906290357932448, 0.0012906290357932448], "traffic_norm": [8.351887702941895, 12.67103385925293, 16.748966217041016], "traffic_residual_norm": [2.4463743102387525e-06, 3.320532414363697e-06, 3.4206291275040712e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3780324459075928, "training_loop_wall_s": 22.549708366394043, "diagnostics_wall_s": 0.06091642379760742, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.549708366394043, "evaluation_wall_s": 0.039571285247802734}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s0.json
new file mode 100644
index 0000000..9d99652
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.975, "eval_loss": 0.1107682333946228, "wall_s": 23.864957809448242, "test_acc": 0.975, "test_loss": 0.1107682333946228, "cos_r_negg": [0.40649521350860596, 0.2343309372663498, 0.21780982613563538], "cos_innovation_negg": [0.40649521350860596, 0.2343309372663498, 0.21780982613563538], "cos_apical_negg": [-0.006061049643903971, 0.04211680218577385, 0.0668993592262268], "cos_Ac_negg": [0.6221227645874023, 0.3515009880065918, 0.2978827953338623], "r_norm": [0.015230732038617134, 0.014827835373580456, 0.014884503558278084], "g_norm": [3.613426451920532e-05, 2.8192658646730706e-05, 2.3867145500844344e-05], "traffic_norm": [7.990795135498047, 8.985038757324219, 9.6758394241333], "traffic_residual_norm": [5.1598253776319325e-06, 5.853671609656885e-06, 6.657435733359307e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4013376235961914, "training_loop_wall_s": 23.35324192047119, "diagnostics_wall_s": 0.07262706756591797, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.35324192047119, "evaluation_wall_s": 0.03607916831970215}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s1.json
new file mode 100644
index 0000000..0dc3ad0
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9751, "eval_loss": 0.11871002101898194, "wall_s": 22.655405521392822, "test_acc": 0.9751, "test_loss": 0.11871002101898194, "cos_r_negg": [0.36715877056121826, 0.1995827853679657, 0.19979479908943176], "cos_innovation_negg": [0.36715877056121826, 0.1995827853679657, 0.19979479908943176], "cos_apical_negg": [-0.00112900510430336, 0.021674158051609993, 0.05385264754295349], "cos_Ac_negg": [0.6502986550331116, 0.3535965383052826, 0.3093958795070648], "r_norm": [0.01420152559876442, 0.013605762273073196, 0.013946533203125], "g_norm": [3.561765333870426e-05, 2.8250537070562132e-05, 2.2729578631697223e-05], "traffic_norm": [7.983911514282227, 8.878379821777344, 9.71786880493164], "traffic_residual_norm": [5.205261004448403e-06, 5.9350104493205436e-06, 6.71843190502841e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.40143799781799316, "training_loop_wall_s": 22.146238565444946, "diagnostics_wall_s": 0.07197284698486328, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.146238565444946, "evaluation_wall_s": 0.03431129455566406}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s2.json
new file mode 100644
index 0000000..dff235d
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9732, "eval_loss": 0.11439530372619629, "wall_s": 21.83359384536743, "test_acc": 0.9732, "test_loss": 0.11439530372619629, "cos_r_negg": [0.3970538377761841, 0.22283419966697693, 0.22080056369304657], "cos_innovation_negg": [0.3970538377761841, 0.22283419966697693, 0.22080056369304657], "cos_apical_negg": [-0.017070356756448746, 0.03650590032339096, 0.053913015872240067], "cos_Ac_negg": [0.6470867395401001, 0.3505892753601074, 0.3380177617073059], "r_norm": [0.01056949608027935, 0.010258040390908718, 0.010548199526965618], "g_norm": [2.854992271750234e-05, 2.1322772226994857e-05, 1.7368030967190862e-05], "traffic_norm": [7.99361515045166, 9.070598602294922, 9.833293914794922], "traffic_residual_norm": [5.290669832902495e-06, 5.8965633797924966e-06, 6.7772207330563106e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.39454007148742676, "training_loop_wall_s": 21.343998670578003, "diagnostics_wall_s": 0.06367087364196777, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.343998670578003, "evaluation_wall_s": 0.03002023696899414}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s3.json
new file mode 100644
index 0000000..b02e82e
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9748, "eval_loss": 0.12563528475761412, "wall_s": 22.94641351699829, "test_acc": 0.9748, "test_loss": 0.12563528475761412, "cos_r_negg": [0.36988508701324463, 0.20127764344215393, 0.21127942204475403], "cos_innovation_negg": [0.36988508701324463, 0.20127764344215393, 0.21127942204475403], "cos_apical_negg": [-0.00523132411763072, 0.036218978464603424, 0.04524986073374748], "cos_Ac_negg": [0.6499233841896057, 0.3453362286090851, 0.3546518087387085], "r_norm": [0.011109747923910618, 0.010918039828538895, 0.011081942357122898], "g_norm": [2.9655280741280876e-05, 2.2530344722326845e-05, 1.8494665710022673e-05], "traffic_norm": [7.971855163574219, 8.960420608520508, 9.720222473144531], "traffic_residual_norm": [5.441151643026387e-06, 6.009920980432071e-06, 6.864996066724416e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.39499521255493164, "training_loop_wall_s": 22.455888509750366, "diagnostics_wall_s": 0.06260108947753906, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.455888509750366, "evaluation_wall_s": 0.031087875366210938}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s4.json
new file mode 100644
index 0000000..1216fa4
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_residual_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9756, "eval_loss": 0.12336151428222657, "wall_s": 32.37203931808472, "test_acc": 0.9756, "test_loss": 0.12336151428222657, "cos_r_negg": [0.343009352684021, 0.2202245593070984, 0.2114848494529724], "cos_innovation_negg": [0.343009352684021, 0.2202245593070984, 0.2114848494529724], "cos_apical_negg": [-0.020809464156627655, 0.033554933965206146, 0.0629345178604126], "cos_Ac_negg": [0.6337064504623413, 0.37473535537719727, 0.319587767124176], "r_norm": [0.010424706153571606, 0.010282013565301895, 0.0104760080575943], "g_norm": [2.8133361411164515e-05, 2.0877358110737987e-05, 1.6686952221789397e-05], "traffic_norm": [7.979086399078369, 8.89151382446289, 9.892303466796875], "traffic_residual_norm": [5.435612365545239e-06, 5.828165285493014e-06, 6.91844661560026e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.37706589698791504, "training_loop_wall_s": 31.828230381011963, "diagnostics_wall_s": 0.12075543403625488, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 31.828230381011963, "evaluation_wall_s": 0.04422640800476074}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s0.json
new file mode 100644
index 0000000..d72101b
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9763, "eval_loss": 0.10766794590950013, "wall_s": 21.593028783798218, "test_acc": 0.9763, "test_loss": 0.10766794590950013, "cos_r_negg": [0.11072160303592682, 0.06893026828765869, 0.02502078004181385], "cos_innovation_negg": [0.11072160303592682, 0.06893026828765869, 0.02502078004181385], "cos_apical_negg": [-0.005082389339804649, 0.06711208075284958, 0.08699697256088257], "cos_Ac_negg": [0.6613782048225403, 0.3728935718536377, 0.34934496879577637], "r_norm": [0.03579896688461304, 0.04280184581875801, 0.024701014161109924], "g_norm": [2.8734584702760912e-05, 2.2360143702826463e-05, 1.802581755327992e-05], "traffic_norm": [8.02535343170166, 8.535879135131836, 9.110616683959961], "traffic_residual_norm": [0.026223231106996536, 0.03398709371685982, 0.014764643274247646], "traffic_r2": [0.9999893307685852, 0.9999841451644897, 0.9999973773956299]}, "timing": {"warmup_wall_s": 0.4292867183685303, "training_loop_wall_s": 21.057815313339233, "diagnostics_wall_s": 0.07333636283874512, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.057815313339233, "evaluation_wall_s": 0.031096458435058594}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s1.json
new file mode 100644
index 0000000..b37d7be
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.973, "eval_loss": 0.11291258239746094, "wall_s": 24.534424304962158, "test_acc": 0.973, "test_loss": 0.11291258239746094, "cos_r_negg": [0.05433207005262375, 0.04470684006810188, 0.05635334923863411], "cos_innovation_negg": [0.05433207005262375, 0.04470684006810188, 0.05635334923863411], "cos_apical_negg": [-0.015993623062968254, 0.04313649237155914, 0.07549088448286057], "cos_Ac_negg": [0.6639877557754517, 0.3679823875427246, 0.356120765209198], "r_norm": [0.061860308051109314, 0.030491210520267487, 0.012517457827925682], "g_norm": [3.069898230023682e-05, 2.3626780603080988e-05, 1.8184804503107443e-05], "traffic_norm": [8.025636672973633, 8.495809555053711, 9.219423294067383], "traffic_residual_norm": [0.05411946028470993, 0.021769192069768906, 0.002195447450503707], "traffic_r2": [0.9999545216560364, 0.9999934434890747, 0.9999999403953552]}, "timing": {"warmup_wall_s": 0.4903535842895508, "training_loop_wall_s": 23.95119881629944, "diagnostics_wall_s": 0.060378313064575195, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.95119881629944, "evaluation_wall_s": 0.03112173080444336}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s2.json
new file mode 100644
index 0000000..ce68996
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9753, "eval_loss": 0.11365967063903809, "wall_s": 21.04224443435669, "test_acc": 0.9753, "test_loss": 0.11365967063903809, "cos_r_negg": [0.13764429092407227, 0.07627443969249725, 0.060289863497018814], "cos_innovation_negg": [0.13764429092407227, 0.07627443969249725, 0.060289863497018814], "cos_apical_negg": [-0.0036570471711456776, 0.06257910281419754, 0.07077328860759735], "cos_Ac_negg": [0.6769694089889526, 0.34781140089035034, 0.33678799867630005], "r_norm": [0.04965744912624359, 0.043415963649749756, 0.022615700960159302], "g_norm": [4.837669985136017e-05, 3.6696954339277e-05, 2.9547798476414755e-05], "traffic_norm": [8.037530899047852, 8.658622741699219, 9.034192085266113], "traffic_residual_norm": [0.03472704440355301, 0.028455331921577454, 0.005953298415988684], "traffic_r2": [0.9999813437461853, 0.9999892115592957, 0.9999995827674866]}, "timing": {"warmup_wall_s": 0.3882114887237549, "training_loop_wall_s": 20.558480501174927, "diagnostics_wall_s": 0.06005525588989258, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 20.558480501174927, "evaluation_wall_s": 0.03405904769897461}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s3.json
new file mode 100644
index 0000000..199278c
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9757, "eval_loss": 0.10749600176811218, "wall_s": 20.999139070510864, "test_acc": 0.9757, "test_loss": 0.10749600176811218, "cos_r_negg": [0.10514326393604279, 0.053552597761154175, 0.05368883162736893], "cos_innovation_negg": [0.10514326393604279, 0.053552597761154175, 0.05368883162736893], "cos_apical_negg": [-0.011545569635927677, 0.0667373314499855, 0.0911838486790657], "cos_Ac_negg": [0.6535895466804504, 0.33521080017089844, 0.3739749789237976], "r_norm": [0.05023074150085449, 0.07859623432159424, 0.04855361580848694], "g_norm": [4.264892777428031e-05, 3.3459553378634155e-05, 2.623140244395472e-05], "traffic_norm": [8.044744491577148, 8.694783210754395, 9.116923332214355], "traffic_residual_norm": [0.0366254597902298, 0.06647418439388275, 0.03510233387351036], "traffic_r2": [0.9999792575836182, 0.9999415874481201, 0.9999852180480957]}, "timing": {"warmup_wall_s": 0.41413354873657227, "training_loop_wall_s": 20.48820161819458, "diagnostics_wall_s": 0.06426548957824707, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 20.48820161819458, "evaluation_wall_s": 0.031026840209960938}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s4.json
new file mode 100644
index 0000000..beb3b24
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 1234, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t1234_taskfit_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9728, "eval_loss": 0.11457855453491211, "wall_s": 21.89788293838501, "test_acc": 0.9728, "test_loss": 0.11457855453491211, "cos_r_negg": [0.11268746852874756, 0.09148629754781723, 0.05826069414615631], "cos_innovation_negg": [0.11268746852874756, 0.09148629754781723, 0.05826069414615631], "cos_apical_negg": [-0.0049586063250899315, 0.07400916516780853, 0.09508149325847626], "cos_Ac_negg": [0.6753122806549072, 0.3872530460357666, 0.33360719680786133], "r_norm": [0.04516848921775818, 0.04618894308805466, 0.0190668236464262], "g_norm": [3.830218338407576e-05, 2.8545688110170886e-05, 2.2967848053667694e-05], "traffic_norm": [8.023269653320312, 8.866741180419922, 9.34913444519043], "traffic_residual_norm": [0.03351593017578125, 0.034918274730443954, 0.005822678096592426], "traffic_r2": [0.9999825358390808, 0.9999845027923584, 0.9999995827674866]}, "timing": {"warmup_wall_s": 0.427626371383667, "training_loop_wall_s": 21.372203826904297, "diagnostics_wall_s": 0.06580519676208496, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.372203826904297, "evaluation_wall_s": 0.030589580535888672}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s0.json
new file mode 100644
index 0000000..08b864c
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9449, "eval_loss": 0.2653144721984863, "wall_s": 26.912935495376587, "test_acc": 0.9449, "test_loss": 0.2653144721984863, "cos_r_negg": [-0.005616891197860241, -0.03258371725678444, 0.1308538317680359], "cos_innovation_negg": [0.6908656358718872, 0.49458107352256775, 0.67838454246521], "cos_apical_negg": [-0.005616891197860241, -0.032583706080913544, 0.1308538317680359], "cos_Ac_negg": [0.8771350383758545, 0.6284027099609375, 0.8099914789199829], "r_norm": [0.3894351124763489, 0.3394854664802551, 0.35024407505989075], "g_norm": [0.0010940675856545568, 0.0010924478992819786, 0.0010924472007900476], "traffic_norm": [8.34655475616455, 12.953167915344238, 17.00919532775879], "traffic_residual_norm": [2.076473265333334e-06, 2.846996721928008e-06, 3.3116953090939205e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.42833852767944336, "training_loop_wall_s": 26.403773307800293, "diagnostics_wall_s": 0.047145843505859375, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 26.403773307800293, "evaluation_wall_s": 0.03218650817871094}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s1.json
new file mode 100644
index 0000000..028c19e
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9375, "eval_loss": 0.2846288284301758, "wall_s": 24.612092971801758, "test_acc": 0.9375, "test_loss": 0.2846288284301758, "cos_r_negg": [-0.04484473913908005, -0.042697563767433167, -0.07335079461336136], "cos_innovation_negg": [0.6408790349960327, 0.3284810185432434, 0.36388805508613586], "cos_apical_negg": [-0.04484473168849945, -0.042697563767433167, -0.07335079461336136], "cos_Ac_negg": [0.8818541169166565, 0.5614533424377441, 0.6310142874717712], "r_norm": [0.3467492461204529, 0.31706586480140686, 0.32342439889907837], "g_norm": [0.0009757609223015606, 0.0009711544262245297, 0.0009711264865472913], "traffic_norm": [8.33980941772461, 12.871767044067383, 16.93987464904785], "traffic_residual_norm": [2.272890469612321e-06, 3.276829147580429e-06, 4.131054993194994e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.41860532760620117, "training_loop_wall_s": 24.112335443496704, "diagnostics_wall_s": 0.04682016372680664, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.112335443496704, "evaluation_wall_s": 0.03290867805480957}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s2.json
new file mode 100644
index 0000000..c2981a7
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9416, "eval_loss": 0.2807616226196289, "wall_s": 25.696439266204834, "test_acc": 0.9416, "test_loss": 0.2807616226196289, "cos_r_negg": [-0.04002371430397034, -0.1802777349948883, -0.20198321342468262], "cos_innovation_negg": [0.6571562886238098, 0.3197261393070221, 0.23755508661270142], "cos_apical_negg": [-0.04002371430397034, -0.1802777349948883, -0.20198321342468262], "cos_Ac_negg": [0.8892356753349304, 0.5594062805175781, 0.46483880281448364], "r_norm": [0.3929564952850342, 0.35543787479400635, 0.3721754252910614], "g_norm": [0.001136169070377946, 0.00113583798520267, 0.0011357192415744066], "traffic_norm": [8.347539901733398, 12.980802536010742, 17.160221099853516], "traffic_residual_norm": [2.1527105218410725e-06, 2.6093819087691372e-06, 3.160766254950431e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3646535873413086, "training_loop_wall_s": 25.246201753616333, "diagnostics_wall_s": 0.04504227638244629, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.246201753616333, "evaluation_wall_s": 0.0390777587890625}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s3.json
new file mode 100644
index 0000000..94c1245
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9444, "eval_loss": 0.25304798278808593, "wall_s": 26.783438205718994, "test_acc": 0.9444, "test_loss": 0.25304798278808593, "cos_r_negg": [0.06716900318861008, 0.11799640208482742, 0.24995282292366028], "cos_innovation_negg": [0.7435568571090698, 0.3666471838951111, 0.3544878363609314], "cos_apical_negg": [0.06716899573802948, 0.11799639463424683, 0.2499527931213379], "cos_Ac_negg": [0.9023409485816956, 0.43007487058639526, 0.4529529809951782], "r_norm": [0.5117753744125366, 0.45328983664512634, 0.46601298451423645], "g_norm": [0.0014653841499239206, 0.0014639315195381641, 0.001463931635953486], "traffic_norm": [8.343515396118164, 12.876875877380371, 16.94000244140625], "traffic_residual_norm": [2.2070175873523112e-06, 2.597020284156315e-06, 3.293756208222476e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.42239928245544434, "training_loop_wall_s": 26.269617795944214, "diagnostics_wall_s": 0.054862260818481445, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 26.269617795944214, "evaluation_wall_s": 0.035085439682006836}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s4.json
new file mode 100644
index 0000000..310901d
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_matched_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9421, "eval_loss": 0.26484778747558596, "wall_s": 25.178852558135986, "test_acc": 0.9421, "test_loss": 0.26484778747558596, "cos_r_negg": [-0.016409248113632202, 0.04584228992462158, 0.03140558674931526], "cos_innovation_negg": [0.6999280452728271, 0.24368780851364136, 0.3717809319496155], "cos_apical_negg": [-0.016409248113632202, 0.04584228992462158, 0.03140559047460556], "cos_Ac_negg": [0.9021148085594177, 0.3322249948978424, 0.5264469385147095], "r_norm": [0.4231308698654175, 0.38380783796310425, 0.41192835569381714], "g_norm": [0.001223342027515173, 0.0012201304780319333, 0.0012201309436932206], "traffic_norm": [8.345022201538086, 12.949339866638184, 16.912776947021484], "traffic_residual_norm": [1.9672911548695993e-06, 2.7341295663063647e-06, 3.3799869925132953e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3819863796234131, "training_loop_wall_s": 24.715914487838745, "diagnostics_wall_s": 0.04752397537231445, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.715914487838745, "evaluation_wall_s": 0.03169369697570801}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251792384, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s0.json
new file mode 100644
index 0000000..fac7e3e
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9472, "eval_loss": 0.24270529251098633, "wall_s": 22.897919178009033, "test_acc": 0.9472, "test_loss": 0.24270529251098633, "cos_r_negg": [-6.90720698912628e-05, -0.022983646020293236, 0.2705652117729187], "cos_innovation_negg": [0.6054344177246094, 0.39441365003585815, 0.46361085772514343], "cos_apical_negg": [-6.90720698912628e-05, -0.022983646020293236, 0.2705652117729187], "cos_Ac_negg": [0.7064036726951599, 0.4745027422904968, 0.5400671362876892], "r_norm": [8.440542221069336, 13.026922225952148, 17.071672439575195], "g_norm": [0.0008239856688305736, 0.0008235256536863744, 0.0008235268178395927], "traffic_norm": [8.364140510559082, 12.986955642700195, 17.044044494628906], "traffic_residual_norm": [2.6322643407183932e-06, 2.9238392471597763e-06, 2.9365842237893958e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.41277599334716797, "training_loop_wall_s": 22.3798565864563, "diagnostics_wall_s": 0.0702812671661377, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.3798565864563, "evaluation_wall_s": 0.03371381759643555}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s1.json
new file mode 100644
index 0000000..5a33b41
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8947, "eval_loss": 0.5388675506591797, "wall_s": 31.346553564071655, "test_acc": 0.8947, "test_loss": 0.5388675506591797, "cos_r_negg": [0.015948526561260223, 0.10499942302703857, 0.07888168841600418], "cos_innovation_negg": [0.6588118076324463, 0.434018075466156, 0.14215905964374542], "cos_apical_negg": [0.015948526561260223, 0.10499942302703857, 0.07888168841600418], "cos_Ac_negg": [0.7685883641242981, 0.5631532669067383, 0.16325610876083374], "r_norm": [8.501237869262695, 12.992756843566895, 17.025592803955078], "g_norm": [0.001441394560970366, 0.0014402545057237148, 0.0014401886146515608], "traffic_norm": [8.3642578125, 12.915250778198242, 16.978796005249023], "traffic_residual_norm": [2.478574515407672e-06, 3.1292561288864817e-06, 3.4141994547098875e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.40084218978881836, "training_loop_wall_s": 30.84887170791626, "diagnostics_wall_s": 0.06290054321289062, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 30.84887170791626, "evaluation_wall_s": 0.032354116439819336}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s2.json
new file mode 100644
index 0000000..3e72e9e
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9378, "eval_loss": 0.30999940490722655, "wall_s": 23.175837993621826, "test_acc": 0.9378, "test_loss": 0.30999940490722655, "cos_r_negg": [-0.005226381588727236, 0.12293691188097, -0.04440073296427727], "cos_innovation_negg": [0.739219605922699, 0.32075250148773193, 0.35279524326324463], "cos_apical_negg": [-0.005226381588727236, 0.12293691188097, -0.04440073296427727], "cos_Ac_negg": [0.7734048366546631, 0.32790637016296387, 0.36244791746139526], "r_norm": [8.446846961975098, 12.973806381225586, 17.032562255859375], "g_norm": [0.0008459198870696127, 0.0008459139498881996, 0.0008459139498881996], "traffic_norm": [8.365156173706055, 12.93982982635498, 17.008987426757812], "traffic_residual_norm": [2.5557048957125517e-06, 2.6968950805894565e-06, 2.9753637136309408e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3944053649902344, "training_loop_wall_s": 22.678145170211792, "diagnostics_wall_s": 0.06880378723144531, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.678145170211792, "evaluation_wall_s": 0.032645225524902344}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s3.json
new file mode 100644
index 0000000..3b6ab80
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9374, "eval_loss": 0.3218420837402344, "wall_s": 22.084923267364502, "test_acc": 0.9374, "test_loss": 0.3218420837402344, "cos_r_negg": [-0.009849442169070244, -0.21520501375198364, 0.1557636559009552], "cos_innovation_negg": [0.7191661596298218, 0.5167700052261353, 0.2776455283164978], "cos_apical_negg": [-0.009849442169070244, -0.21520501375198364, 0.1557636559009552], "cos_Ac_negg": [0.8133797645568848, 0.6442769765853882, 0.3287287950515747], "r_norm": [8.440735816955566, 12.977214813232422, 17.0805721282959], "g_norm": [0.0008997489931061864, 0.0008996519027277827, 0.0008996519027277827], "traffic_norm": [8.365768432617188, 12.943485260009766, 17.04888153076172], "traffic_residual_norm": [2.3861257432145067e-06, 2.7966202651441563e-06, 2.891911208280362e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.37070274353027344, "training_loop_wall_s": 21.6171395778656, "diagnostics_wall_s": 0.06451630592346191, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.6171395778656, "evaluation_wall_s": 0.03107142448425293}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s4.json
new file mode 100644
index 0000000..8db4725
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_raw_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9342, "eval_loss": 0.3230666595458984, "wall_s": 22.68609356880188, "test_acc": 0.9342, "test_loss": 0.3230666595458984, "cos_r_negg": [0.018833991140127182, -0.10359469056129456, 0.008554721251130104], "cos_innovation_negg": [0.6650756597518921, 0.4537888765335083, 0.4049416780471802], "cos_apical_negg": [0.018833991140127182, -0.10359469056129456, 0.008554721251130104], "cos_Ac_negg": [0.7649593353271484, 0.5013107657432556, 0.4447886347770691], "r_norm": [8.481805801391602, 13.030004501342773, 16.95760154724121], "g_norm": [0.0011589701753109694, 0.0011580400168895721, 0.0011579864658415318], "traffic_norm": [8.363784790039062, 12.977391242980957, 16.931068420410156], "traffic_residual_norm": [2.8231409032741794e-06, 3.1253300676326035e-06, 3.4356387459411053e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.38440823554992676, "training_loop_wall_s": 22.189126014709473, "diagnostics_wall_s": 0.07609033584594727, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.189126014709473, "evaluation_wall_s": 0.0346834659576416}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s0.json
new file mode 100644
index 0000000..aa6cb90
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9753, "eval_loss": 0.11156675567626953, "wall_s": 22.97356653213501, "test_acc": 0.9753, "test_loss": 0.11156675567626953, "cos_r_negg": [0.40024876594543457, 0.2172940969467163, 0.19159981608390808], "cos_innovation_negg": [0.40024876594543457, 0.2172940969467163, 0.19159981608390808], "cos_apical_negg": [-0.024542074650526047, 0.03382657840847969, 0.03991641104221344], "cos_Ac_negg": [0.6214439868927002, 0.3514292240142822, 0.30384644865989685], "r_norm": [0.013064531609416008, 0.01268321368843317, 0.01270569209009409], "g_norm": [3.176351310685277e-05, 2.4828423192957416e-05, 2.0556512026814744e-05], "traffic_norm": [8.006340026855469, 9.090728759765625, 9.86185073852539], "traffic_residual_norm": [5.2943437367503066e-06, 5.807882189401425e-06, 6.805684279242996e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4516143798828125, "training_loop_wall_s": 22.42755389213562, "diagnostics_wall_s": 0.06242728233337402, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.42755389213562, "evaluation_wall_s": 0.030637741088867188}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s1.json
new file mode 100644
index 0000000..d05af75
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9745, "eval_loss": 0.11184067821502686, "wall_s": 23.010007858276367, "test_acc": 0.9745, "test_loss": 0.11184067821502686, "cos_r_negg": [0.3482290506362915, 0.19860529899597168, 0.1998586654663086], "cos_innovation_negg": [0.3482290506362915, 0.19860529899597168, 0.1998586654663086], "cos_apical_negg": [-0.01462691929191351, 0.013561777770519257, 0.053916625678539276], "cos_Ac_negg": [0.6409331560134888, 0.3578130304813385, 0.3111937940120697], "r_norm": [0.011118384078145027, 0.010921696200966835, 0.011138053610920906], "g_norm": [2.7770565793616697e-05, 2.134642272721976e-05, 1.724167486827355e-05], "traffic_norm": [7.9990644454956055, 9.046333312988281, 9.884603500366211], "traffic_residual_norm": [5.323567165760323e-06, 5.999334007356083e-06, 6.865524483146146e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.427065372467041, "training_loop_wall_s": 22.484235286712646, "diagnostics_wall_s": 0.0639045238494873, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.484235286712646, "evaluation_wall_s": 0.033260345458984375}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s2.json
new file mode 100644
index 0000000..c742626
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9722, "eval_loss": 0.1216939709663391, "wall_s": 22.47914981842041, "test_acc": 0.9722, "test_loss": 0.1216939709663391, "cos_r_negg": [0.40605276823043823, 0.22561293840408325, 0.22984668612480164], "cos_innovation_negg": [0.40605276823043823, 0.22561293840408325, 0.22984668612480164], "cos_apical_negg": [-0.0067420112900435925, 0.045456189662218094, 0.06318053603172302], "cos_Ac_negg": [0.665642261505127, 0.3539874255657196, 0.3550703525543213], "r_norm": [0.009378659538924694, 0.009186340495944023, 0.009411070495843887], "g_norm": [2.4638042305014096e-05, 1.8807508240570314e-05, 1.5384006474050693e-05], "traffic_norm": [8.005279541015625, 9.134838104248047, 9.970107078552246], "traffic_residual_norm": [5.289411092235241e-06, 5.918814167671371e-06, 6.783165190427098e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3928873538970947, "training_loop_wall_s": 21.99056911468506, "diagnostics_wall_s": 0.06225228309631348, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.99056911468506, "evaluation_wall_s": 0.031911373138427734}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s3.json
new file mode 100644
index 0000000..a7a6506
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9748, "eval_loss": 0.12321665916442871, "wall_s": 24.27957057952881, "test_acc": 0.9748, "test_loss": 0.12321665916442871, "cos_r_negg": [0.3740592896938324, 0.18684445321559906, 0.20470204949378967], "cos_innovation_negg": [0.3740592896938324, 0.18684445321559906, 0.20470204949378967], "cos_apical_negg": [-0.010669438168406487, 0.02583073079586029, 0.04441595450043678], "cos_Ac_negg": [0.6589754819869995, 0.319297730922699, 0.33950603008270264], "r_norm": [0.011183940805494785, 0.011044985614717007, 0.011132434941828251], "g_norm": [2.9891194571973756e-05, 2.2706517484039068e-05, 1.8320410163141787e-05], "traffic_norm": [7.9823899269104, 9.130853652954102, 9.90524959564209], "traffic_residual_norm": [5.438227617560187e-06, 6.208496870385716e-06, 6.812108040321618e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3889448642730713, "training_loop_wall_s": 23.779377698898315, "diagnostics_wall_s": 0.07552194595336914, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.779377698898315, "evaluation_wall_s": 0.0342411994934082}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s4.json
new file mode 100644
index 0000000..6dba9e7
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_residual_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9727, "eval_loss": 0.12106650485992432, "wall_s": 22.72337055206299, "test_acc": 0.9727, "test_loss": 0.12106650485992432, "cos_r_negg": [0.3574386537075043, 0.22825950384140015, 0.2319839596748352], "cos_innovation_negg": [0.3574386537075043, 0.22825950384140015, 0.2319839596748352], "cos_apical_negg": [-0.030896756798028946, 0.034093134105205536, 0.07533615082502365], "cos_Ac_negg": [0.6357108354568481, 0.3790915906429291, 0.337469220161438], "r_norm": [0.010120055638253689, 0.009905066341161728, 0.010113190859556198], "g_norm": [2.7074032914242707e-05, 2.0331854102551006e-05, 1.632649582461454e-05], "traffic_norm": [7.985693454742432, 9.118696212768555, 10.007740020751953], "traffic_residual_norm": [5.389824764279183e-06, 6.053356173651991e-06, 7.049691703286953e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.38509178161621094, "training_loop_wall_s": 22.244410276412964, "diagnostics_wall_s": 0.06220436096191406, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.244410276412964, "evaluation_wall_s": 0.03014659881591797}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s0.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s0.json
new file mode 100644
index 0000000..8e709cc
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.975, "eval_loss": 0.1062827712059021, "wall_s": 21.29588532447815, "test_acc": 0.975, "test_loss": 0.1062827712059021, "cos_r_negg": [0.07139585167169571, 0.04545552283525467, 0.0368279330432415], "cos_innovation_negg": [0.07139585167169571, 0.04545552283525467, 0.0368279330432415], "cos_apical_negg": [-0.010667459107935429, 0.060603752732276917, 0.09176652878522873], "cos_Ac_negg": [0.6516298055648804, 0.3659258484840393, 0.3480594754219055], "r_norm": [0.059630461037158966, 0.04982667788863182, 0.012674720026552677], "g_norm": [2.505168959032744e-05, 1.9286877432023175e-05, 1.5699373761890456e-05], "traffic_norm": [8.041580200195312, 8.89494800567627, 9.392997741699219], "traffic_residual_norm": [0.05254768207669258, 0.04283016175031662, 0.0038638929836452007], "traffic_r2": [0.9999573230743408, 0.9999768137931824, 0.9999998211860657]}, "timing": {"warmup_wall_s": 0.3865323066711426, "training_loop_wall_s": 20.77860713005066, "diagnostics_wall_s": 0.09168648719787598, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 20.77860713005066, "evaluation_wall_s": 0.03766012191772461}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s1.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s1.json
new file mode 100644
index 0000000..6a628b0
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9742, "eval_loss": 0.10817494134902954, "wall_s": 20.808370113372803, "test_acc": 0.9742, "test_loss": 0.10817494134902954, "cos_r_negg": [0.08485221862792969, 0.03912276774644852, 0.07091297209262848], "cos_innovation_negg": [0.08485221862792969, 0.03912276774644852, 0.07091297209262848], "cos_apical_negg": [-0.0027848128229379654, 0.03917714208364487, 0.0921391025185585], "cos_Ac_negg": [0.6704590320587158, 0.36065202951431274, 0.3649185597896576], "r_norm": [0.0477873757481575, 0.03189641237258911, 0.01602151431143284], "g_norm": [4.0244463889393955e-05, 3.117135202046484e-05, 2.421604767732788e-05], "traffic_norm": [8.040409088134766, 8.695037841796875, 9.315973281860352], "traffic_residual_norm": [0.03594830632209778, 0.01958788000047207, 0.0019564079120755196], "traffic_r2": [0.9999800324440002, 0.9999949336051941, 0.9999999403953552]}, "timing": {"warmup_wall_s": 0.37747716903686523, "training_loop_wall_s": 20.337913751602173, "diagnostics_wall_s": 0.05857205390930176, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 20.337913751602173, "evaluation_wall_s": 0.0330500602722168}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s2.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s2.json
new file mode 100644
index 0000000..36b195d
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9741, "eval_loss": 0.12146714277267456, "wall_s": 22.39625382423401, "test_acc": 0.9741, "test_loss": 0.12146714277267456, "cos_r_negg": [0.11920889467000961, 0.055413950234651566, 0.06188645213842392], "cos_innovation_negg": [0.11920889467000961, 0.055413950234651566, 0.06188645213842392], "cos_apical_negg": [-0.01888185180723667, 0.06126725673675537, 0.09225860238075256], "cos_Ac_negg": [0.6841520071029663, 0.35505548119544983, 0.35110336542129517], "r_norm": [0.04843105748295784, 0.05346192419528961, 0.06583494693040848], "g_norm": [4.460390846361406e-05, 3.380070847924799e-05, 2.7108293579658493e-05], "traffic_norm": [8.061336517333984, 8.680484771728516, 8.981460571289062], "traffic_residual_norm": [0.03485754132270813, 0.04051923751831055, 0.05311163514852524], "traffic_r2": [0.9999812841415405, 0.999978244304657, 0.9999650716781616]}, "timing": {"warmup_wall_s": 0.5587372779846191, "training_loop_wall_s": 21.735727548599243, "diagnostics_wall_s": 0.06713128089904785, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.735727548599243, "evaluation_wall_s": 0.033026933670043945}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s3.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s3.json
new file mode 100644
index 0000000..034e18f
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9741, "eval_loss": 0.11768623809814453, "wall_s": 22.652867078781128, "test_acc": 0.9741, "test_loss": 0.11768623809814453, "cos_r_negg": [0.085869699716568, 0.0526663176715374, 0.044147659093141556], "cos_innovation_negg": [0.085869699716568, 0.0526663176715374, 0.044147659093141556], "cos_apical_negg": [-0.013621192425489426, 0.05604138970375061, 0.08363503217697144], "cos_Ac_negg": [0.6645025014877319, 0.3466097116470337, 0.3638889491558075], "r_norm": [0.03337634354829788, 0.23520736396312714, 0.052339810878038406], "g_norm": [3.7898949813097715e-05, 2.9457762138918042e-05, 2.3462100216420367e-05], "traffic_norm": [8.054317474365234, 8.827223777770996, 9.226447105407715], "traffic_residual_norm": [0.019953230395913124, 0.22608835995197296, 0.03984665870666504], "traffic_r2": [0.9999938607215881, 0.9993445873260498, 0.9999814033508301]}, "timing": {"warmup_wall_s": 0.3820981979370117, "training_loop_wall_s": 22.15985894203186, "diagnostics_wall_s": 0.07477235794067383, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.15985894203186, "evaluation_wall_s": 0.034601449966430664}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s4.json b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s4.json
new file mode 100644
index 0000000..ac469d1
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "traffic_seed": 5678, "traffic_mode": "soma", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_soma_rho0p5_t5678_taskfit_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9751, "eval_loss": 0.11406267356872558, "wall_s": 27.308403730392456, "test_acc": 0.9751, "test_loss": 0.11406267356872558, "cos_r_negg": [0.09274396300315857, 0.08347947150468826, 0.05967649072408676], "cos_innovation_negg": [0.09274396300315857, 0.08347947150468826, 0.05967649072408676], "cos_apical_negg": [-0.004067207220941782, 0.06454106420278549, 0.07815012335777283], "cos_Ac_negg": [0.668519139289856, 0.37243449687957764, 0.3502269983291626], "r_norm": [0.06798107177019119, 0.04883084073662758, 0.017440903931856155], "g_norm": [3.0404025892494246e-05, 2.3247936042025685e-05, 1.881461685115937e-05], "traffic_norm": [8.025266647338867, 9.126158714294434, 9.523771286010742], "traffic_residual_norm": [0.059495702385902405, 0.03999023139476776, 0.006739976815879345], "traffic_r2": [0.9999450445175171, 0.9999808073043823, 0.9999995231628418]}, "timing": {"warmup_wall_s": 0.5252945423126221, "training_loop_wall_s": 26.68479609489441, "diagnostics_wall_s": 0.06497311592102051, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 26.68479609489441, "evaluation_wall_s": 0.03160452842712402}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251264000, "peak_memory_reserved_bytes": 255852544}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s0.json
new file mode 100644
index 0000000..d03e429
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9374, "eval_loss": 0.36395843505859377, "wall_s": 26.402500867843628, "test_acc": 0.9374, "test_loss": 0.36395843505859377, "cos_r_negg": [-0.11995599418878555, 0.0421837642788887, 0.12121343612670898], "cos_innovation_negg": [0.40009063482284546, 0.32023096084594727, 0.44239911437034607], "cos_apical_negg": [-0.11995600163936615, 0.0421837642788887, 0.12121343612670898], "cos_Ac_negg": [0.8881239891052246, 0.573719322681427, 0.6563642024993896], "r_norm": [0.5375386476516724, 0.48322105407714844, 0.47901666164398193], "g_norm": [0.0016289378982037306, 0.0016289371997117996, 0.001628936966881156], "traffic_norm": [7.258359909057617, 7.188303470611572, 7.124587535858154], "traffic_residual_norm": [0.00011412724416004494, 5.399487417889759e-05, 1.2480227269406896e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.40468478202819824, "training_loop_wall_s": 25.92184019088745, "diagnostics_wall_s": 0.0424497127532959, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.92184019088745, "evaluation_wall_s": 0.031743526458740234}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s1.json
new file mode 100644
index 0000000..1c8f2aa
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9422, "eval_loss": 0.3187836166381836, "wall_s": 25.362569570541382, "test_acc": 0.9422, "test_loss": 0.3187836166381836, "cos_r_negg": [-0.2356729954481125, -0.1757695972919464, -0.023581936955451965], "cos_innovation_negg": [0.43207263946533203, 0.25264492630958557, 0.36304593086242676], "cos_apical_negg": [-0.23567301034927368, -0.1757695972919464, -0.023581933230161667], "cos_Ac_negg": [0.8693699240684509, 0.4694439768791199, 0.6709126234054565], "r_norm": [0.4917760193347931, 0.4423219561576843, 0.4445633292198181], "g_norm": [0.0014223186299204826, 0.0014223186299204826, 0.0014223186299204826], "traffic_norm": [7.260921478271484, 7.195919036865234, 7.130514144897461], "traffic_residual_norm": [0.00010064143862109631, 3.138545071124099e-05, 1.5518437521677697e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3914661407470703, "training_loop_wall_s": 24.89373469352722, "diagnostics_wall_s": 0.04354095458984375, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.89373469352722, "evaluation_wall_s": 0.03244972229003906}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s2.json
new file mode 100644
index 0000000..661c5a1
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.94, "eval_loss": 0.35168352355957033, "wall_s": 26.244162797927856, "test_acc": 0.94, "test_loss": 0.35168352355957033, "cos_r_negg": [-0.07493103295564651, 0.1721763014793396, -0.16233405470848083], "cos_innovation_negg": [0.6342378258705139, 0.4099453091621399, 0.38120603561401367], "cos_apical_negg": [-0.07493102550506592, 0.172176331281662, -0.16233405470848083], "cos_Ac_negg": [0.8297814130783081, 0.5752111077308655, 0.5955493450164795], "r_norm": [0.6678394079208374, 0.6028999090194702, 0.5877668857574463], "g_norm": [0.0020150248892605305, 0.0020150248892605305, 0.0020150248892605305], "traffic_norm": [7.253799915313721, 7.17811918258667, 7.12187385559082], "traffic_residual_norm": [6.949315138626844e-05, 2.532080361561384e-05, 1.6005637917260174e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3780324459075928, "training_loop_wall_s": 25.7858784198761, "diagnostics_wall_s": 0.045160770416259766, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.7858784198761, "evaluation_wall_s": 0.03337287902832031}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s3.json
new file mode 100644
index 0000000..6a43e7b
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9243, "eval_loss": 0.4747757049560547, "wall_s": 25.32297968864441, "test_acc": 0.9243, "test_loss": 0.4747757049560547, "cos_r_negg": [-0.3405349850654602, -0.006475353613495827, 0.04238409921526909], "cos_innovation_negg": [0.6151122450828552, 0.32139530777931213, 0.355514258146286], "cos_apical_negg": [-0.3405349552631378, -0.006475353613495827, 0.04238409921526909], "cos_Ac_negg": [0.9047661423683167, 0.6119104027748108, 0.49519112706184387], "r_norm": [0.8841969966888428, 0.7979073524475098, 0.8237870931625366], "g_norm": [0.002440979238599539, 0.002440978540107608, 0.0024409787729382515], "traffic_norm": [7.260605812072754, 7.185789108276367, 7.130100727081299], "traffic_residual_norm": [6.749574095010757e-05, 4.185823490843177e-05, 1.5494468925680849e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3981184959411621, "training_loop_wall_s": 24.848467111587524, "diagnostics_wall_s": 0.041005611419677734, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.848467111587524, "evaluation_wall_s": 0.033867597579956055}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s4.json
new file mode 100644
index 0000000..72b1f3b
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_matched_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.919, "eval_loss": 0.4502234115600586, "wall_s": 27.96734619140625, "test_acc": 0.919, "test_loss": 0.4502234115600586, "cos_r_negg": [-0.2048216015100479, 0.17576193809509277, -0.1610853672027588], "cos_innovation_negg": [0.6471096277236938, 0.3692294955253601, 0.4800722002983093], "cos_apical_negg": [-0.20482158660888672, 0.17576195299625397, -0.1610853672027588], "cos_Ac_negg": [0.8459927439689636, 0.5130444765090942, 0.6096118688583374], "r_norm": [0.6481717824935913, 0.5746652483940125, 0.5833587050437927], "g_norm": [0.001908272155560553, 0.0019082720391452312, 0.0019082720391452312], "traffic_norm": [7.254999160766602, 7.1894378662109375, 7.118302345275879], "traffic_residual_norm": [0.00032382679637521505, 6.807304453104734e-05, 1.4200627447280567e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.3950061798095703, "training_loop_wall_s": 27.498600006103516, "diagnostics_wall_s": 0.04103684425354004, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 27.498600006103516, "evaluation_wall_s": 0.03098464012145996}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s0.json
new file mode 100644
index 0000000..7a17525
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8819, "eval_loss": 1.0573158721923828, "wall_s": 21.746095418930054, "test_acc": 0.8819, "test_loss": 1.0573158721923828, "cos_r_negg": [-0.11890213936567307, -0.11463772505521774, -0.1681540310382843], "cos_innovation_negg": [0.5563148856163025, 0.37651437520980835, 0.5558599829673767], "cos_apical_negg": [-0.11890213936567307, -0.11463772505521774, -0.1681540310382843], "cos_Ac_negg": [0.7267563343048096, 0.8758476376533508, 0.699513852596283], "r_norm": [8.38715934753418, 8.299959182739258, 8.254770278930664], "g_norm": [0.004805026110261679, 0.004805026110261679, 0.004805026110261679], "traffic_norm": [7.277026653289795, 7.207949638366699, 7.152407646179199], "traffic_residual_norm": [8.912661724025384e-05, 9.468111966270953e-05, 5.274063141769147e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.39913129806518555, "training_loop_wall_s": 21.24267339706421, "diagnostics_wall_s": 0.07142877578735352, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.24267339706421, "evaluation_wall_s": 0.031439781188964844}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s1.json
new file mode 100644
index 0000000..4ab47c3
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.7088, "eval_loss": 2.0055474487304688, "wall_s": 23.520283460617065, "test_acc": 0.7088, "test_loss": 2.0055474487304688, "cos_r_negg": [-0.017453845590353012, 0.2416442334651947, 0.42573806643486023], "cos_innovation_negg": [0.47104695439338684, 0.5552130937576294, 0.6405587792396545], "cos_apical_negg": [-0.017453845590353012, 0.2416442334651947, 0.42573806643486023], "cos_Ac_negg": [0.6046013832092285, 0.6404615640640259, 0.828432023525238], "r_norm": [10.673662185668945, 10.47724437713623, 10.39781665802002], "g_norm": [0.014000448398292065, 0.014000448398292065, 0.014000448398292065], "traffic_norm": [7.2825164794921875, 7.222955226898193, 7.156313896179199], "traffic_residual_norm": [3.051201201742515e-05, 2.622320425871294e-05, 5.239501206233399e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4539151191711426, "training_loop_wall_s": 22.96756863594055, "diagnostics_wall_s": 0.06472349166870117, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.96756863594055, "evaluation_wall_s": 0.032459259033203125}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s2.json
new file mode 100644
index 0000000..a3d24f5
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.6846, "eval_loss": 2.3730392578125, "wall_s": 23.810314416885376, "test_acc": 0.6846, "test_loss": 2.3730392578125, "cos_r_negg": [-0.03460131213068962, -0.07163257896900177, 0.3990454375743866], "cos_innovation_negg": [0.5587286949157715, 0.5452563166618347, 0.7492662668228149], "cos_apical_negg": [-0.03460131213068962, -0.07163257896900177, 0.3990454375743866], "cos_Ac_negg": [0.7297882437705994, 0.6798570156097412, 0.9168288707733154], "r_norm": [9.130071640014648, 8.876184463500977, 8.89448356628418], "g_norm": [0.008711210452020168, 0.008711207658052444, 0.008711207658052444], "traffic_norm": [7.230881690979004, 7.201321125030518, 7.12637996673584], "traffic_residual_norm": [5.69226176594384e-05, 4.4266736949793994e-05, 6.122847935330356e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.37047672271728516, "training_loop_wall_s": 23.341272592544556, "diagnostics_wall_s": 0.06346702575683594, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.341272592544556, "evaluation_wall_s": 0.03344130516052246}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s3.json
new file mode 100644
index 0000000..0c7cc5f
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8725, "eval_loss": 1.002587060546875, "wall_s": 23.872297048568726, "test_acc": 0.8725, "test_loss": 1.002587060546875, "cos_r_negg": [-0.200788676738739, 0.19021935760974884, 0.16198401153087616], "cos_innovation_negg": [0.7552520632743835, 0.7572027444839478, 0.43548670411109924], "cos_apical_negg": [-0.200788676738739, 0.19021935760974884, 0.16198401153087616], "cos_Ac_negg": [0.8502153158187866, 0.9149423837661743, 0.51751309633255], "r_norm": [8.260965347290039, 8.139471054077148, 8.09028148651123], "g_norm": [0.004538854118436575, 0.004538854118436575, 0.004538854118436575], "traffic_norm": [7.25691032409668, 7.216630935668945, 7.151445388793945], "traffic_residual_norm": [7.413216371787712e-05, 0.00011341478966642171, 6.766840101590788e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.41014981269836426, "training_loop_wall_s": 23.344248056411743, "diagnostics_wall_s": 0.06675958633422852, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.344248056411743, "evaluation_wall_s": 0.04956769943237305}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s4.json
new file mode 100644
index 0000000..0c64da1
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_raw_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8359, "eval_loss": 1.260870684814453, "wall_s": 23.964892148971558, "test_acc": 0.8359, "test_loss": 1.260870684814453, "cos_r_negg": [-0.101652130484581, 0.3881489038467407, 0.17948991060256958], "cos_innovation_negg": [0.6955921649932861, 0.7255971431732178, 0.5941733121871948], "cos_apical_negg": [-0.101652130484581, 0.3881489038467407, 0.17948991060256958], "cos_Ac_negg": [0.8036449551582336, 0.852196991443634, 0.7483469247817993], "r_norm": [8.592475891113281, 8.492074966430664, 8.443746566772461], "g_norm": [0.005307844839990139, 0.005307844839990139, 0.005307844839990139], "traffic_norm": [7.268642902374268, 7.219735145568848, 7.155752182006836], "traffic_residual_norm": [0.00026886974228546023, 5.626009442494251e-05, 4.84871236494655e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4264798164367676, "training_loop_wall_s": 23.435991287231445, "diagnostics_wall_s": 0.06897592544555664, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.435991287231445, "evaluation_wall_s": 0.031873464584350586}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s0.json
new file mode 100644
index 0000000..41e6f0b
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9451, "eval_loss": 0.21419091873168947, "wall_s": 30.487451553344727, "test_acc": 0.9451, "test_loss": 0.21419091873168947, "cos_r_negg": [0.13965576887130737, 0.14264404773712158, 0.42591720819473267], "cos_innovation_negg": [0.13965576887130737, 0.14264404773712158, 0.42591720819473267], "cos_apical_negg": [0.018366429954767227, -0.07961118966341019, 0.0006677049677819014], "cos_Ac_negg": [0.8313585519790649, 0.4737814664840698, 0.4473564624786377], "r_norm": [1.7187482118606567, 3.125070095062256, 0.2268524020910263], "g_norm": [0.000771180959418416, 0.0007456038147211075, 0.0007264381274580956], "traffic_norm": [5.5503692626953125, 5.4763336181640625, 5.508896827697754], "traffic_residual_norm": [1.551558494567871, 3.0304622650146484, 1.5491693829972064e-06], "traffic_r2": [0.921751081943512, 0.6937336921691895, 1.0]}, "timing": {"warmup_wall_s": 0.3985898494720459, "training_loop_wall_s": 29.985090017318726, "diagnostics_wall_s": 0.06983637809753418, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 29.985090017318726, "evaluation_wall_s": 0.032305002212524414}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s1.json
new file mode 100644
index 0000000..6f2cb89
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9471, "eval_loss": 0.20374724502563477, "wall_s": 24.4136643409729, "test_acc": 0.9471, "test_loss": 0.20374724502563477, "cos_r_negg": [0.14502492547035217, 0.0900474414229393, 0.4225185513496399], "cos_innovation_negg": [0.14502492547035217, 0.0900474414229393, 0.4225185513496399], "cos_apical_negg": [-0.014185112901031971, 0.014600119553506374, 0.0743662565946579], "cos_Ac_negg": [0.8292077779769897, 0.49440857768058777, 0.4312993884086609], "r_norm": [0.8862433433532715, 0.9833095669746399, 0.24149370193481445], "g_norm": [0.0008547571487724781, 0.0008258850430138409, 0.0008041338296607137], "traffic_norm": [5.398778438568115, 5.5092926025390625, 5.517882347106934], "traffic_residual_norm": [0.6633012890815735, 0.8215587139129639, 1.5637555179637275e-06], "traffic_r2": [0.9846965670585632, 0.9776671528816223, 1.0]}, "timing": {"warmup_wall_s": 0.4069099426269531, "training_loop_wall_s": 23.903870105743408, "diagnostics_wall_s": 0.0709388256072998, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.903870105743408, "evaluation_wall_s": 0.030460119247436523}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s2.json
new file mode 100644
index 0000000..5b4d991
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9529, "eval_loss": 0.18999507369995117, "wall_s": 24.426897287368774, "test_acc": 0.9529, "test_loss": 0.18999507369995117, "cos_r_negg": [0.09906965494155884, 0.07154213637113571, 0.364967942237854], "cos_innovation_negg": [0.09906965494155884, 0.07154213637113571, 0.364967942237854], "cos_apical_negg": [0.013267232105135918, -0.04054811969399452, -0.03852660581469536], "cos_Ac_negg": [0.8237807750701904, 0.4561780095100403, 0.45065563917160034], "r_norm": [0.947187066078186, 1.8443055152893066, 0.22157973051071167], "g_norm": [0.0007304180762730539, 0.0007126069394871593, 0.0007040700875222683], "traffic_norm": [5.608591079711914, 5.647092342376709, 5.625067710876465], "traffic_residual_norm": [0.7465074062347412, 1.7139729261398315, 1.682780066403211e-06], "traffic_r2": [0.9820652604103088, 0.9077444672584534, 1.0]}, "timing": {"warmup_wall_s": 0.52158522605896, "training_loop_wall_s": 23.80042839050293, "diagnostics_wall_s": 0.06604862213134766, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.80042839050293, "evaluation_wall_s": 0.03730344772338867}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s3.json
new file mode 100644
index 0000000..688bc43
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9489, "eval_loss": 0.20095864639282227, "wall_s": 24.43290662765503, "test_acc": 0.9489, "test_loss": 0.20095864639282227, "cos_r_negg": [0.09850500524044037, 0.13039235770702362, 0.40143001079559326], "cos_innovation_negg": [0.09850500524044037, 0.13039235770702362, 0.40143001079559326], "cos_apical_negg": [-0.09685540199279785, -0.08691652864217758, 0.04926358908414841], "cos_Ac_negg": [0.8050340414047241, 0.35376012325286865, 0.4094179570674896], "r_norm": [1.4125397205352783, 2.1520965099334717, 0.23173299431800842], "g_norm": [0.0007456367602571845, 0.0007348887156695127, 0.0007067525293678045], "traffic_norm": [5.472574234008789, 5.292076110839844, 5.381652355194092], "traffic_residual_norm": [1.2265489101409912, 2.0305442810058594, 1.4824006484559504e-06], "traffic_r2": [0.9493032097816467, 0.8523202538490295, 1.0]}, "timing": {"warmup_wall_s": 0.4269390106201172, "training_loop_wall_s": 23.905134201049805, "diagnostics_wall_s": 0.06761002540588379, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.905134201049805, "evaluation_wall_s": 0.03171539306640625}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s4.json
new file mode 100644
index 0000000..291d718
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_residual_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9449, "eval_loss": 0.2301503028869629, "wall_s": 31.71674346923828, "test_acc": 0.9449, "test_loss": 0.2301503028869629, "cos_r_negg": [0.13322201371192932, 0.05153956264257431, 0.3626507520675659], "cos_innovation_negg": [0.13322201371192932, 0.05153956264257431, 0.3626507520675659], "cos_apical_negg": [-0.025270704180002213, -0.02715746872127056, 0.08674274384975433], "cos_Ac_negg": [0.8083757162094116, 0.45907530188560486, 0.4242839515209198], "r_norm": [0.8999630212783813, 1.4482049942016602, 0.2597443461418152], "g_norm": [0.0008502369746565819, 0.0008327558171004057, 0.0008152241352945566], "traffic_norm": [5.708662986755371, 5.661407470703125, 5.440089702606201], "traffic_residual_norm": [0.656178891658783, 1.2684688568115234, 1.5313034964492545e-06], "traffic_r2": [0.9865016937255859, 0.9497414827346802, 1.0]}, "timing": {"warmup_wall_s": 0.4716794490814209, "training_loop_wall_s": 31.132710933685303, "diagnostics_wall_s": 0.07224321365356445, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 31.132710933685303, "evaluation_wall_s": 0.03849220275878906}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s0.json
new file mode 100644
index 0000000..06a8f44
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9439, "eval_loss": 0.20404349822998047, "wall_s": 21.669994354248047, "test_acc": 0.9439, "test_loss": 0.20404349822998047, "cos_r_negg": [0.18332037329673767, 0.08745530247688293, 0.14948982000350952], "cos_innovation_negg": [0.18332037329673767, 0.08745530247688293, 0.14948982000350952], "cos_apical_negg": [-0.01747150346636772, 0.15819931030273438, 0.12332531064748764], "cos_Ac_negg": [0.78850257396698, 0.42859870195388794, 0.5869132876396179], "r_norm": [0.8720560073852539, 1.420119285583496, 1.8671869039535522], "g_norm": [0.0008312328718602657, 0.0008176489500328898, 0.000815921404864639], "traffic_norm": [4.798291206359863, 4.735294818878174, 4.673320770263672], "traffic_residual_norm": [0.640842080116272, 1.2669792175292969, 1.7251646518707275], "traffic_r2": [0.9821474552154541, 0.928424596786499, 0.863762378692627]}, "timing": {"warmup_wall_s": 0.38118863105773926, "training_loop_wall_s": 21.188705444335938, "diagnostics_wall_s": 0.06740045547485352, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.188705444335938, "evaluation_wall_s": 0.031232118606567383}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s1.json
new file mode 100644
index 0000000..c4a2386
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9393, "eval_loss": 0.21552113571166992, "wall_s": 22.883296012878418, "test_acc": 0.9393, "test_loss": 0.21552113571166992, "cos_r_negg": [0.05526755005121231, 0.08757957816123962, 0.14373072981834412], "cos_innovation_negg": [0.05526755005121231, 0.08757957816123962, 0.14373072981834412], "cos_apical_negg": [0.005546543747186661, 0.09588900953531265, 0.11208465695381165], "cos_Ac_negg": [0.7355006337165833, 0.41952717304229736, 0.45110589265823364], "r_norm": [1.3038424253463745, 1.7835543155670166, 2.0772836208343506], "g_norm": [0.0009597803000360727, 0.0009146032389253378, 0.0009045974584296346], "traffic_norm": [4.711557388305664, 4.69498348236084, 4.704967021942139], "traffic_residual_norm": [1.0467817783355713, 1.6010355949401855, 1.9046320915222168], "traffic_r2": [0.9505427479743958, 0.8836418390274048, 0.8361632227897644]}, "timing": {"warmup_wall_s": 0.3823857307434082, "training_loop_wall_s": 22.41043519973755, "diagnostics_wall_s": 0.05745506286621094, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.41043519973755, "evaluation_wall_s": 0.031633853912353516}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s2.json
new file mode 100644
index 0000000..f425568
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9416, "eval_loss": 0.21297063827514648, "wall_s": 26.268155336380005, "test_acc": 0.9416, "test_loss": 0.21297063827514648, "cos_r_negg": [0.15929771959781647, 0.08694209158420563, 0.09771256148815155], "cos_innovation_negg": [0.15929771959781647, 0.08694209158420563, 0.09771256148815155], "cos_apical_negg": [-0.011843657121062279, 0.12377423048019409, -0.06581348180770874], "cos_Ac_negg": [0.794185221195221, 0.4123392701148987, 0.44737911224365234], "r_norm": [1.2410669326782227, 1.1401857137680054, 1.6759381294250488], "g_norm": [0.0008636899292469025, 0.000839787651784718, 0.00083386484766379], "traffic_norm": [4.750250816345215, 4.713382244110107, 4.716704368591309], "traffic_residual_norm": [1.031112790107727, 0.9594378471374512, 1.5172045230865479], "traffic_r2": [0.9524432420730591, 0.9584637880325317, 0.896543562412262]}, "timing": {"warmup_wall_s": 0.4715080261230469, "training_loop_wall_s": 25.699090480804443, "diagnostics_wall_s": 0.06408262252807617, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.699090480804443, "evaluation_wall_s": 0.03175473213195801}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s3.json
new file mode 100644
index 0000000..dec0d1f
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9388, "eval_loss": 0.22583512115478516, "wall_s": 25.21292471885681, "test_acc": 0.9388, "test_loss": 0.22583512115478516, "cos_r_negg": [0.07182647287845612, 0.07071448117494583, 0.08539074659347534], "cos_innovation_negg": [0.07182647287845612, 0.07071448117494583, 0.08539074659347534], "cos_apical_negg": [0.01488976739346981, 0.06992676109075546, 0.15211902558803558], "cos_Ac_negg": [0.7032680511474609, 0.48914018273353577, 0.47150689363479614], "r_norm": [1.634055256843567, 2.4011483192443848, 1.898813009262085], "g_norm": [0.0008784089586697519, 0.0008503547287546098, 0.0008457821095362306], "traffic_norm": [4.884483814239502, 5.002573013305664, 4.827104568481445], "traffic_residual_norm": [1.422595500946045, 2.255345582962036, 1.7285417318344116], "traffic_r2": [0.9151604771614075, 0.796796441078186, 0.8714852333068848]}, "timing": {"warmup_wall_s": 0.4762122631072998, "training_loop_wall_s": 24.625603437423706, "diagnostics_wall_s": 0.06892061233520508, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.625603437423706, "evaluation_wall_s": 0.040261268615722656}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s4.json
new file mode 100644
index 0000000..080f393
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 1234, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t1234_taskfit_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.939, "eval_loss": 0.22383403015136719, "wall_s": 21.07701063156128, "test_acc": 0.939, "test_loss": 0.22383403015136719, "cos_r_negg": [0.12303165346384048, 0.038294680416584015, 0.13789823651313782], "cos_innovation_negg": [0.12303165346384048, 0.038294680416584015, 0.13789823651313782], "cos_apical_negg": [0.016671277582645416, 0.14093156158924103, 0.16091731190681458], "cos_Ac_negg": [0.755894660949707, 0.3823791444301605, 0.4353826940059662], "r_norm": [1.431825876235962, 1.4205665588378906, 1.8946917057037354], "g_norm": [0.0008226179052144289, 0.0008044852875173092, 0.0007986994460225105], "traffic_norm": [4.898571014404297, 4.894956111907959, 4.935544490814209], "traffic_residual_norm": [1.2309315204620361, 1.25641930103302, 1.7511460781097412], "traffic_r2": [0.9368518590927124, 0.9341278672218323, 0.8741442561149597]}, "timing": {"warmup_wall_s": 0.4278371334075928, "training_loop_wall_s": 20.553102016448975, "diagnostics_wall_s": 0.06059455871582031, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 20.553102016448975, "evaluation_wall_s": 0.03399229049682617}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s0.json
new file mode 100644
index 0000000..62b0913
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9394, "eval_loss": 0.36469058227539064, "wall_s": 29.72229313850403, "test_acc": 0.9394, "test_loss": 0.36469058227539064, "cos_r_negg": [-0.18458425998687744, -0.00875612162053585, -0.10877688974142075], "cos_innovation_negg": [0.40155768394470215, 0.327457994222641, -0.014385282061994076], "cos_apical_negg": [-0.18458425998687744, -0.008756123483181, -0.10877688974142075], "cos_Ac_negg": [0.8263881206512451, 0.7317373156547546, 0.43226590752601624], "r_norm": [0.6835544109344482, 0.6088284254074097, 0.6047548055648804], "g_norm": [0.0019828907679766417, 0.0019828907679766417, 0.001982895191758871], "traffic_norm": [7.430420875549316, 7.461974143981934, 7.280427932739258], "traffic_residual_norm": [0.00023300026077777147, 9.980670438380912e-05, 1.6248488918790827e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.49945807456970215, "training_loop_wall_s": 29.146092653274536, "diagnostics_wall_s": 0.04321146011352539, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 29.146092653274536, "evaluation_wall_s": 0.0319066047668457}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s1.json
new file mode 100644
index 0000000..74add39
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9399, "eval_loss": 0.3419333251953125, "wall_s": 27.896985054016113, "test_acc": 0.9399, "test_loss": 0.3419333251953125, "cos_r_negg": [-0.24893954396247864, -0.09210536628961563, 0.17535686492919922], "cos_innovation_negg": [0.4837198555469513, 0.24332106113433838, 0.4203285574913025], "cos_apical_negg": [-0.24893954396247864, -0.09210537374019623, 0.17535685002803802], "cos_Ac_negg": [0.8914418816566467, 0.32740330696105957, 0.6763486862182617], "r_norm": [0.5702698230743408, 0.5157419443130493, 0.5148005485534668], "g_norm": [0.0016570452135056257, 0.0016570452135056257, 0.001657045679166913], "traffic_norm": [7.4069623947143555, 7.464835166931152, 7.28193998336792], "traffic_residual_norm": [0.000285520451143384, 5.074025830253959e-05, 1.3841092822985956e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4698505401611328, "training_loop_wall_s": 27.34854555130005, "diagnostics_wall_s": 0.045871734619140625, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 27.34854555130005, "evaluation_wall_s": 0.03096914291381836}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s2.json
new file mode 100644
index 0000000..d8738ca
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9088, "eval_loss": 0.5678901092529297, "wall_s": 25.763311862945557, "test_acc": 0.9088, "test_loss": 0.5678901092529297, "cos_r_negg": [-0.20870761573314667, -0.021330995485186577, -0.06700399518013], "cos_innovation_negg": [0.5858175754547119, 0.4132708013057709, 0.4315761625766754], "cos_apical_negg": [-0.20870760083198547, -0.021330995485186577, -0.06700399518013], "cos_Ac_negg": [0.8954030275344849, 0.5962347388267517, 0.889598548412323], "r_norm": [0.9347981810569763, 0.836817741394043, 0.8430970907211304], "g_norm": [0.0026609590277075768, 0.0026609590277075768, 0.0026609585620462894], "traffic_norm": [7.41218376159668, 7.463861465454102, 7.27644157409668], "traffic_residual_norm": [7.056762115098536e-05, 3.2245767215499654e-05, 1.508040213593631e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.43422961235046387, "training_loop_wall_s": 25.254053592681885, "diagnostics_wall_s": 0.040718793869018555, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.254053592681885, "evaluation_wall_s": 0.03274130821228027}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s3.json
new file mode 100644
index 0000000..8a9b305
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8752, "eval_loss": 0.759413818359375, "wall_s": 26.772302389144897, "test_acc": 0.8752, "test_loss": 0.759413818359375, "cos_r_negg": [-0.16312775015830994, 0.055848751217126846, 0.12166592478752136], "cos_innovation_negg": [0.4842345118522644, 0.3963109254837036, 0.311221718788147], "cos_apical_negg": [-0.16312776505947113, 0.05584874376654625, 0.12166592478752136], "cos_Ac_negg": [0.7544422149658203, 0.7541813850402832, 0.6927931308746338], "r_norm": [1.3919694423675537, 1.2916438579559326, 1.2924094200134277], "g_norm": [0.003703216090798378, 0.003703216090798378, 0.003703216090798378], "traffic_norm": [7.426485538482666, 7.4785590171813965, 7.265680313110352], "traffic_residual_norm": [0.00017284868226852268, 0.0001476680627092719, 1.3942760688223643e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4068152904510498, "training_loop_wall_s": 26.286672115325928, "diagnostics_wall_s": 0.045949697494506836, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 26.286672115325928, "evaluation_wall_s": 0.031069517135620117}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s4.json
new file mode 100644
index 0000000..5213979
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_matched_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8987, "eval_loss": 0.6056966400146484, "wall_s": 25.89843201637268, "test_acc": 0.8987, "test_loss": 0.6056966400146484, "cos_r_negg": [-0.18516398966312408, -0.09135419875383377, 0.01518169790506363], "cos_innovation_negg": [0.5252408385276794, 0.5771122574806213, 0.37934282422065735], "cos_apical_negg": [-0.18516398966312408, -0.09135419875383377, 0.015181698836386204], "cos_Ac_negg": [0.8990577459335327, 0.8434588313102722, 0.5518466830253601], "r_norm": [1.1204187870025635, 1.004320502281189, 0.9979392290115356], "g_norm": [0.0032422258518636227, 0.003242219565436244, 0.0032422193326056004], "traffic_norm": [7.400125503540039, 7.43695068359375, 7.258433818817139], "traffic_residual_norm": [0.00027675862656906247, 0.00015587943198624998, 1.5558557606709655e-06], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4367222785949707, "training_loop_wall_s": 25.384114265441895, "diagnostics_wall_s": 0.04347848892211914, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 25.384114265441895, "evaluation_wall_s": 0.03258872032165527}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 252310528, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s0.json
new file mode 100644
index 0000000..4f873ad
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.7831, "eval_loss": 1.5132845947265625, "wall_s": 23.266320943832397, "test_acc": 0.7831, "test_loss": 1.5132845947265625, "cos_r_negg": [-0.09743721783161163, 0.5071524977684021, 0.41641947627067566], "cos_innovation_negg": [0.7708443999290466, 0.7779732942581177, 0.6197921633720398], "cos_apical_negg": [-0.09743721783161163, 0.5071524977684021, 0.41641947627067566], "cos_Ac_negg": [0.8034586906433105, 0.8994665145874023, 0.7209811210632324], "r_norm": [9.751852035522461, 9.627442359924316, 9.425800323486328], "g_norm": [0.009441126137971878, 0.009441126137971878, 0.009441126137971878], "traffic_norm": [7.449576377868652, 7.492090225219727, 7.306649208068848], "traffic_residual_norm": [0.00014322652714326978, 7.21431351848878e-05, 5.33831666871265e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.43585658073425293, "training_loop_wall_s": 22.731841802597046, "diagnostics_wall_s": 0.06401801109313965, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.731841802597046, "evaluation_wall_s": 0.03318381309509277}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s1.json
new file mode 100644
index 0000000..7c840a7
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8487, "eval_loss": 1.0184347900390625, "wall_s": 24.79096031188965, "test_acc": 0.8487, "test_loss": 1.0184347900390625, "cos_r_negg": [-0.15518751740455627, 0.0013080942444503307, 0.325840562582016], "cos_innovation_negg": [0.6180557012557983, 0.49730974435806274, 0.7229243516921997], "cos_apical_negg": [-0.15518751740455627, 0.0013080942444503307, 0.325840562582016], "cos_Ac_negg": [0.8530423045158386, 0.7729367613792419, 0.7693204879760742], "r_norm": [8.649240493774414, 8.648488998413086, 8.475626945495605], "g_norm": [0.0048919799737632275, 0.0048919799737632275, 0.00489197950810194], "traffic_norm": [7.427854537963867, 7.4985151290893555, 7.302495002746582], "traffic_residual_norm": [3.861152799800038e-05, 3.641301009338349e-05, 4.7391529278684175e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.49753355979919434, "training_loop_wall_s": 24.181016206741333, "diagnostics_wall_s": 0.07856297492980957, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.181016206741333, "evaluation_wall_s": 0.03203773498535156}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s2.json
new file mode 100644
index 0000000..c1761df
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8684, "eval_loss": 0.9066820587158203, "wall_s": 23.074899196624756, "test_acc": 0.8684, "test_loss": 0.9066820587158203, "cos_r_negg": [-0.23015455901622772, 0.1590893566608429, -0.2147156298160553], "cos_innovation_negg": [0.5460265278816223, 0.5958513617515564, 0.2238433063030243], "cos_apical_negg": [-0.23015455901622772, 0.1590893566608429, -0.2147156298160553], "cos_Ac_negg": [0.8191403746604919, 0.7846303582191467, 0.38217559456825256], "r_norm": [8.497917175292969, 8.483491897583008, 8.307286262512207], "g_norm": [0.0043309591710567474, 0.0043309591710567474, 0.0043309591710567474], "traffic_norm": [7.453058242797852, 7.505092144012451, 7.308140277862549], "traffic_residual_norm": [4.988698856323026e-05, 4.926814654027112e-05, 4.135496283197426e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.38953089714050293, "training_loop_wall_s": 22.58379340171814, "diagnostics_wall_s": 0.06833505630493164, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.58379340171814, "evaluation_wall_s": 0.03162837028503418}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s3.json
new file mode 100644
index 0000000..46a3515
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.8696, "eval_loss": 1.1050154357910156, "wall_s": 23.678353309631348, "test_acc": 0.8696, "test_loss": 1.1050154357910156, "cos_r_negg": [-0.16098260879516602, -0.040091339498758316, 0.06577572226524353], "cos_innovation_negg": [0.6416236758232117, 0.5932443141937256, 0.2930694818496704], "cos_apical_negg": [-0.16098260879516602, -0.040091339498758316, 0.06577572226524353], "cos_Ac_negg": [0.7917026281356812, 0.8602825403213501, 0.25545158982276917], "r_norm": [8.372699737548828, 8.364789962768555, 8.184026718139648], "g_norm": [0.003809974528849125, 0.003809973131865263, 0.003809973131865263], "traffic_norm": [7.447551727294922, 7.4955034255981445, 7.297910690307617], "traffic_residual_norm": [0.00010005550575442612, 8.87800706550479e-05, 6.098250651120907e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.40567493438720703, "training_loop_wall_s": 23.175222158432007, "diagnostics_wall_s": 0.06318235397338867, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.175222158432007, "evaluation_wall_s": 0.03273367881774902}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s4.json
new file mode 100644
index 0000000..2960948
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_raw_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.4716, "eval_loss": 8.6107765625, "wall_s": 23.422959089279175, "test_acc": 0.4716, "test_loss": 8.6107765625, "cos_r_negg": [0.37820175290107727, 0.7199957966804504, 0.1417844146490097], "cos_innovation_negg": [0.7329291701316833, 0.8051186800003052, 0.7668689489364624], "cos_apical_negg": [0.37820175290107727, 0.7199957966804504, 0.1417844146490097], "cos_Ac_negg": [0.7841273546218872, 0.9409587979316711, 0.7613928914070129], "r_norm": [13.817055702209473, 13.751391410827637, 13.406961441040039], "g_norm": [0.023938056081533432, 0.023938048630952835, 0.023938048630952835], "traffic_norm": [7.433984756469727, 7.482303619384766, 7.295555114746094], "traffic_residual_norm": [0.00010348523937864229, 6.03110202064272e-05, 4.5440765461535193e-07], "traffic_r2": [1.0, 1.0, 1.0]}, "timing": {"warmup_wall_s": 0.4004383087158203, "training_loop_wall_s": 22.904294967651367, "diagnostics_wall_s": 0.08301115036010742, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.904294967651367, "evaluation_wall_s": 0.03366255760192871}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s0.json
new file mode 100644
index 0000000..430ad43
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9493, "eval_loss": 0.19811741485595702, "wall_s": 23.437824964523315, "test_acc": 0.9493, "test_loss": 0.19811741485595702, "cos_r_negg": [0.12373015284538269, 0.09362047910690308, 0.41498199105262756], "cos_innovation_negg": [0.12373015284538269, 0.09362047910690308, 0.41498199105262756], "cos_apical_negg": [-0.0802716612815857, 0.04821352660655975, -0.015287727117538452], "cos_Ac_negg": [0.8610398173332214, 0.38702309131622314, 0.4479701817035675], "r_norm": [0.9015616178512573, 0.9342986345291138, 0.20927976071834564], "g_norm": [0.0006880770088173449, 0.0006679360521957278, 0.0006539606838487089], "traffic_norm": [5.538880825042725, 5.664424896240234, 5.5203351974487305], "traffic_residual_norm": [0.7144427299499512, 0.7898406982421875, 1.5910126194285112e-06], "traffic_r2": [0.983143150806427, 0.9803016781806946, 1.0]}, "timing": {"warmup_wall_s": 0.37246012687683105, "training_loop_wall_s": 22.963003158569336, "diagnostics_wall_s": 0.07035422325134277, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 22.963003158569336, "evaluation_wall_s": 0.030454158782958984}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s1.json
new file mode 100644
index 0000000..ab7ed27
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9447, "eval_loss": 0.21659343795776367, "wall_s": 24.882244348526, "test_acc": 0.9447, "test_loss": 0.21659343795776367, "cos_r_negg": [0.1397830694913864, 0.10718908905982971, 0.327843576669693], "cos_innovation_negg": [0.1397830694913864, 0.10718908905982971, 0.327843576669693], "cos_apical_negg": [-0.07277929782867432, 0.057858631014823914, 0.08742552995681763], "cos_Ac_negg": [0.8486078977584839, 0.41449272632598877, 0.3476944863796234], "r_norm": [1.3091695308685303, 1.2179452180862427, 0.27224141359329224], "g_norm": [0.000999947777017951, 0.0009553098352625966, 0.0009440375724807382], "traffic_norm": [5.735688209533691, 5.880033493041992, 5.63303279876709], "traffic_residual_norm": [1.0707035064697266, 1.011474609375, 1.3801020486425841e-06], "traffic_r2": [0.964876651763916, 0.9702267646789551, 1.0]}, "timing": {"warmup_wall_s": 0.41986560821533203, "training_loop_wall_s": 24.358300924301147, "diagnostics_wall_s": 0.06535601615905762, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.358300924301147, "evaluation_wall_s": 0.034676551818847656}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s2.json
new file mode 100644
index 0000000..c932ca6
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9465, "eval_loss": 0.22191392059326173, "wall_s": 29.935991764068604, "test_acc": 0.9465, "test_loss": 0.22191392059326173, "cos_r_negg": [0.1340917944908142, 0.12599967420101166, 0.40774816274642944], "cos_innovation_negg": [0.1340917944908142, 0.12599967420101166, 0.40774816274642944], "cos_apical_negg": [-0.02876896969974041, 0.016543813049793243, -0.012833161279559135], "cos_Ac_negg": [0.8124484419822693, 0.43747401237487793, 0.4348751902580261], "r_norm": [0.7935860753059387, 2.369819164276123, 0.23606204986572266], "g_norm": [0.0007939150091260672, 0.0007607018924318254, 0.0007520347135141492], "traffic_norm": [5.904628753662109, 6.078125, 5.8825788497924805], "traffic_residual_norm": [0.5675947070121765, 2.2450692653656006, 1.5767341210448649e-06], "traffic_r2": [0.9901782274246216, 0.8635675311088562, 1.0]}, "timing": {"warmup_wall_s": 0.40016961097717285, "training_loop_wall_s": 29.43533945083618, "diagnostics_wall_s": 0.06141972541809082, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 29.43533945083618, "evaluation_wall_s": 0.037375688552856445}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s3.json
new file mode 100644
index 0000000..39ef3a1
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9469, "eval_loss": 0.21639195404052736, "wall_s": 24.216378450393677, "test_acc": 0.9469, "test_loss": 0.21639195404052736, "cos_r_negg": [0.1532067358493805, 0.13010060787200928, 0.4209510385990143], "cos_innovation_negg": [0.1532067358493805, 0.13010060787200928, 0.4209510385990143], "cos_apical_negg": [-0.003763415850698948, 0.019061636179685593, 0.0051312861032783985], "cos_Ac_negg": [0.8264986276626587, 0.4897308945655823, 0.44737061858177185], "r_norm": [0.8803632855415344, 1.2989137172698975, 0.2743544578552246], "g_norm": [0.0009238200727850199, 0.0009008155902847648, 0.0008913272758945823], "traffic_norm": [5.7799530029296875, 5.913604259490967, 5.704232215881348], "traffic_residual_norm": [0.6126716136932373, 1.1064293384552002, 1.4838212791801197e-06], "traffic_r2": [0.9885541796684265, 0.9648581743240356, 1.0]}, "timing": {"warmup_wall_s": 0.40581798553466797, "training_loop_wall_s": 23.71332859992981, "diagnostics_wall_s": 0.06071972846984863, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.71332859992981, "evaluation_wall_s": 0.0348362922668457}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s4.json
new file mode 100644
index 0000000..f9eadd8
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "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": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_residual_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9224, "eval_loss": 0.3122957092285156, "wall_s": 24.3182590007782, "test_acc": 0.9224, "test_loss": 0.3122957092285156, "cos_r_negg": [0.17593039572238922, 0.1243385374546051, 0.3746523857116699], "cos_innovation_negg": [0.17593039572238922, 0.1243385374546051, 0.3746523857116699], "cos_apical_negg": [-0.056353963911533356, 0.010803273878991604, -0.005878431256860495], "cos_Ac_negg": [0.7922127842903137, 0.4836106300354004, 0.3779418170452118], "r_norm": [1.3733118772506714, 2.8807106018066406, 0.3779491186141968], "g_norm": [0.001204206608235836, 0.0011841937666758895, 0.001171146403066814], "traffic_norm": [5.66041374206543, 5.774425506591797, 5.574830055236816], "traffic_residual_norm": [1.056479573249817, 2.708658218383789, 1.5299608548957622e-06], "traffic_r2": [0.9648914933204651, 0.7794899940490723, 1.0]}, "timing": {"warmup_wall_s": 0.5032100677490234, "training_loop_wall_s": 23.72182846069336, "diagnostics_wall_s": 0.05964374542236328, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 23.72182846069336, "evaluation_wall_s": 0.0316615104675293}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s0.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s0.json
new file mode 100644
index 0000000..b7410a5
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s0.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s0", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4198756217956543}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9462, "eval_loss": 0.19011031494140626, "wall_s": 22.439016580581665, "test_acc": 0.9462, "test_loss": 0.19011031494140626, "cos_r_negg": [0.12363502383232117, 0.04051904380321503, 0.15849155187606812], "cos_innovation_negg": [0.12363502383232117, 0.04051904380321503, 0.15849155187606812], "cos_apical_negg": [-0.006018919870257378, 0.056936055421829224, 0.022556520998477936], "cos_Ac_negg": [0.7686668038368225, 0.22741201519966125, 0.5276838541030884], "r_norm": [0.8834772706031799, 1.0749695301055908, 1.4904228448867798], "g_norm": [0.0009852718794718385, 0.0009658317430876195, 0.0009594411822035909], "traffic_norm": [4.897823333740234, 4.859535217285156, 4.87021541595459], "traffic_residual_norm": [0.579185426235199, 0.8425225019454956, 1.2832462787628174], "traffic_r2": [0.9859632253646851, 0.9699428081512451, 0.9305841326713562]}, "timing": {"warmup_wall_s": 0.4306635856628418, "training_loop_wall_s": 21.90067720413208, "diagnostics_wall_s": 0.06675434112548828, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.90067720413208, "evaluation_wall_s": 0.03916621208190918}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s1.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s1.json
new file mode 100644
index 0000000..419c918
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s1.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s1", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.420098066329956}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9399, "eval_loss": 0.21745218811035155, "wall_s": 20.879685640335083, "test_acc": 0.9399, "test_loss": 0.21745218811035155, "cos_r_negg": [0.15627393126487732, 0.09437164664268494, 0.10686609148979187], "cos_innovation_negg": [0.15627393126487732, 0.09437164664268494, 0.10686609148979187], "cos_apical_negg": [-0.08171367645263672, 0.036267880350351334, 0.2763538658618927], "cos_Ac_negg": [0.804373562335968, 0.3928840458393097, 0.22661301493644714], "r_norm": [1.19974684715271, 1.3968892097473145, 1.5219535827636719], "g_norm": [0.0008564904564991593, 0.0008368249982595444, 0.0008356499020010233], "traffic_norm": [5.162819862365723, 5.16898775100708, 5.12668514251709], "traffic_residual_norm": [0.9851971864700317, 1.2363965511322021, 1.3629871606826782], "traffic_r2": [0.9635688662528992, 0.9428133368492126, 0.9293545484542847]}, "timing": {"warmup_wall_s": 0.3899383544921875, "training_loop_wall_s": 20.379945278167725, "diagnostics_wall_s": 0.06414270401000977, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 20.379945278167725, "evaluation_wall_s": 0.0332179069519043}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s2.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s2.json
new file mode 100644
index 0000000..9ecda89
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s2.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "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": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s2", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.453664779663086}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9381, "eval_loss": 0.21647214965820313, "wall_s": 27.904258966445923, "test_acc": 0.9381, "test_loss": 0.21647214965820313, "cos_r_negg": [0.15562331676483154, 0.12706689536571503, 0.11186005175113678], "cos_innovation_negg": [0.15562331676483154, 0.12706689536571503, 0.11186005175113678], "cos_apical_negg": [-0.0067375171929597855, 0.038256000727415085, 0.13064783811569214], "cos_Ac_negg": [0.7916405200958252, 0.4632820188999176, 0.5846660137176514], "r_norm": [1.1705858707427979, 1.9456396102905273, 3.077195882797241], "g_norm": [0.0009003219311125576, 0.0008812692249193788, 0.000878743187058717], "traffic_norm": [5.1894330978393555, 5.250396728515625, 5.028636455535889], "traffic_residual_norm": [0.9391981363296509, 1.796920895576477, 2.959989547729492], "traffic_r2": [0.9672163724899292, 0.8829179406166077, 0.6536843180656433]}, "timing": {"warmup_wall_s": 0.4264869689941406, "training_loop_wall_s": 27.368278741836548, "diagnostics_wall_s": 0.07294893264770508, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 27.368278741836548, "evaluation_wall_s": 0.03468632698059082}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s3.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s3.json
new file mode 100644
index 0000000..481f4d8
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s3.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s3", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.526390790939331}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9438, "eval_loss": 0.19909090728759765, "wall_s": 21.725806951522827, "test_acc": 0.9438, "test_loss": 0.19909090728759765, "cos_r_negg": [0.1560942679643631, 0.019476432353258133, 0.183937668800354], "cos_innovation_negg": [0.1560942679643631, 0.019476432353258133, 0.183937668800354], "cos_apical_negg": [0.028684357181191444, 0.10628005862236023, 0.02715000882744789], "cos_Ac_negg": [0.796181321144104, 0.3163127303123474, 0.6579781770706177], "r_norm": [0.9978978633880615, 1.452843427658081, 1.6201298236846924], "g_norm": [0.0006829906487837434, 0.0006746909348294139, 0.0006681550876237452], "traffic_norm": [4.543017864227295, 4.721316337585449, 4.647101879119873], "traffic_residual_norm": [0.804266631603241, 1.3129496574401855, 1.480965256690979], "traffic_r2": [0.9686335325241089, 0.9226823449134827, 0.8984634280204773]}, "timing": {"warmup_wall_s": 0.4219987392425537, "training_loop_wall_s": 21.202532291412354, "diagnostics_wall_s": 0.06644463539123535, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 21.202532291412354, "evaluation_wall_s": 0.03332400321960449}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "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": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file
diff --git a/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s4.json b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s4.json
new file mode 100644
index 0000000..62cbef1
--- /dev/null
+++ b/results/c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s4.json
@@ -0,0 +1 @@
+{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 1, "epochs": 15, "batch_size": 128, "train_examples": 0, "val_examples": 0, "split_seed": 2027, "eval_split": "test", "task_seed": 0, "task_train_examples": 50000, "task_test_examples": 10000, "task_levels": 8, "task_n_in": 128, "task_classes": 10, "teacher_depth": 8, "teacher_width": 64, "teacher_residual": 1, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.05, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "feedback_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 1, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 0, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "traffic_seed": 5678, "traffic_mode": "topdown", "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100000, "diagnostics": "alignment", "diagnostics_schedule": "final", "eval_every": 0, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "c1_confirm_v1_mnist_topdown_rho0p2_t5678_taskfit_d3_s4", "n_in": 784, "n_out": 10}, "split": {"dataset": "mnist", "split_seed": null, "validation_examples": 0, "validation_index_sha256": null, "validation_class_counts": {}, "train_examples": 60000, "test_examples": 10000, "split_from_training_only": true, "evaluation_split": "test"}, "provenance": {"git_commit": "18c9c1c1d1ec31560c78d41f4ff6957339c1e7c7", "git_dirty": false}, "diagnostic_protocol": {"probe_source": "training_prefix", "probe_examples": 512, "schedule": "final"}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4526076316833496}, {"epoch_end": 0, "step": 469}, {"epoch_end": 1, "step": 938}, {"epoch_end": 2, "step": 1407}, {"epoch_end": 3, "step": 1876}, {"epoch_end": 4, "step": 2345}, {"epoch_end": 5, "step": 2814}, {"epoch_end": 6, "step": 3283}, {"epoch_end": 7, "step": 3752}, {"epoch_end": 8, "step": 4221}, {"epoch_end": 9, "step": 4690}, {"epoch_end": 10, "step": 5159}, {"epoch_end": 11, "step": 5628}, {"epoch_end": 12, "step": 6097}, {"epoch_end": 13, "step": 6566}, {"epoch_end": 14, "step": 7035}], "final": {"eval_split": "test", "eval_acc": 0.9312, "eval_loss": 0.25196285247802735, "wall_s": 24.71712899208069, "test_acc": 0.9312, "test_loss": 0.25196285247802735, "cos_r_negg": [0.14276087284088135, 0.12869684398174286, 0.17692109942436218], "cos_innovation_negg": [0.14276087284088135, 0.12869684398174286, 0.17692109942436218], "cos_apical_negg": [-0.015038731507956982, 0.09145880490541458, 0.1499987244606018], "cos_Ac_negg": [0.7806680798530579, 0.40483251214027405, 0.5848004817962646], "r_norm": [1.159287691116333, 1.4185460805892944, 1.8101997375488281], "g_norm": [0.001015752088278532, 0.000988541403785348, 0.000983846839517355], "traffic_norm": [5.12418794631958, 5.169732093811035, 4.960330009460449], "traffic_residual_norm": [0.895977795124054, 1.2146010398864746, 1.623690128326416], "traffic_r2": [0.9693496227264404, 0.9448230862617493, 0.8929054737091064]}, "timing": {"warmup_wall_s": 0.4121577739715576, "training_loop_wall_s": 24.19232416152954, "diagnostics_wall_s": 0.0781867504119873, "inline_diagnostics_wall_s": 0.0, "optimizer_wall_s_excluding_inline_diagnostics": 24.19232416152954, "evaluation_wall_s": 0.03239250183105469}, "cost": {"train_steps": 7035, "ordinary_training_forward_examples": 900000, "predictor_warmup_forward_examples": 25600, "perturbation_events": 1759, "calibration_batch_loss_evaluations": 3518, "calibration_example_loss_evaluations": 450048, "calibration_forward_equivalent_examples": 675072.0, "training_forward_equivalent_examples": 1600672.0, "max_perturbation_batch_expansion": 2}, "hardware": {"device": "cuda", "torch_version": "2.3.1+cu118", "cuda_device_name": "NVIDIA GeForce GTX 1080", "cuda_visible_devices": "5", "device_total_memory_bytes": 8507949056, "peak_memory_allocated_bytes": 251786240, "peak_memory_reserved_bytes": 257949696}}
\ No newline at end of file