diff options
| author | Oscar Wan <oscarwan@orionp2.stanford.edu> | 2026-08-26 08:50:07 -0700 |
|---|---|---|
| committer | Oscar Wan <oscarwan@orionp2.stanford.edu> | 2026-08-26 08:50:07 -0700 |
| commit | 3c7c221ae6cfcdf6840da1d5262d94c75d56288b (patch) | |
| tree | e13adbf676e3b5e1dc57fcf26dcc153c7983abc4 /results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json | |
| parent | f55872beaee658ff76ac15fd421b73017820c320 (diff) | |
experiment: add completed A6000 crossover resultsa6000-x1-results
Diffstat (limited to 'results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json')
| -rw-r--r-- | results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json | 7904 |
1 files changed, 7904 insertions, 0 deletions
diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json new file mode 100644 index 0000000..f742aaa --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json @@ -0,0 +1,7904 @@ +{ + "architecture": { + "adaptive_apical_parameters": 1064960, + "adaptive_feedback_parameters": 848512, + "base_width": 16, + "blocks_per_stage": 9, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 56, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 532480, + "forward_parameters": 853018, + "hidden_shapes": [ + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 16, + 32, + 32 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 32, + 16, + 16 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ], + [ + 64, + 8, + 8 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 1064960, + "residual_scale": 1.0, + "vectorizer_mode": null, + "vectorizer_parameters": 0 + }, + "args": { + "a_scale": 1.0, + "a_warmup_steps": 0, + "alignment_probe": 32, + "apical_calibration_mode": "unit_targets", + "apical_seed": null, + "augment_train": 1, + "batch_size": 128, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "data_dir": "/orion/u/oscarwan/scratch/sdil-data", + "depth": 56, + "device": "cuda", + "epochs": 200, + "eta_A": 0.01, + "eta_P": 0.1, + "eval_every": 1, + "eval_split": "validation", + "learn_P": 1, + "loader_seed": 0, + "lr": 0.1, + "lr_gamma": 0.1, + "lr_milestones": "100,150", + "lr_schedule": "step", + "max_steps": 0, + "mirror_batch_size": 1, + "mirror_eta": 0.1, + "mirror_every": 16, + "mirror_noise_std": 1.0, + "mirror_seed": 3000, + "mirror_warmup_steps": 0, + "mode": "kp_traffic", + "momentum": 0.9, + "neutral_projection": 1, + "normalization": "batchnorm", + "nuisance_scale": 0.0, + "out": "/orion/u/oscarwan/scratch/sdil-work/sdil-a6000-full-rerun/results/resnet_crossover/r2/resnet-r2-sdil-d56-lr0p1.json", + "output_lr": 0.1, + "pert_directions": 1, + "pert_every": 4, + "pert_sigma": 0.01, + "perturb_seed": 1000, + "predictor_every": 0, + "predictor_mode": "closed_form", + "predictor_warmup_steps": 1, + "residual_scale": null, + "seed": 0, + "split_seed": 2027, + "traffic_calibration_examples": 64, + "traffic_ratio": 4.0, + "traffic_rule": "innovation", + "traffic_seed": 5000, + "train_limit": 0, + "use_residual": 1, + "val_examples": 5000, + "vectorizer_mode": "spatial_template", + "warmup_epochs": 5, + "weight_decay": 0.0001, + "weight_scale": 1.0, + "width": 16 + }, + "calibration_metric_space": "reciprocal_local_activity_products_with_mixed_apical_traffic", + "counters": { + "apical_warmup_examples": 0, + "calibration_event_examples": 0, + "causal_scalar_observations": 0, + "logical_batch_loss_queries": 0, + "mirror_conv_examples": 0, + "mirror_events": 0, + "mirror_readout_examples": 0, + "neutral_projection_examples": 9000000, + "ordinary_examples": 9000000, + "per_example_loss_terms": 0, + "perturbation_events": 0, + "perturbation_forward_examples": 0, + "predictor_update_examples": 64, + "predictor_warmup_examples": 0, + "traffic_audit_examples": 0, + "traffic_calibration_examples": 64 + }, + "diagnostics": { + "early_third_mean": 0.9988019201490614, + "feedback_forward_cosine": [ + 0.9994082450866699, + 0.9993633031845093, + 0.9988594055175781, + 0.9988021850585938, + 0.9989034533500671, + 0.9987430572509766, + 0.9988512992858887, + 0.9986526966094971, + 0.9989784359931946, + 0.9988847374916077, + 0.9990218281745911, + 0.998627781867981, + 0.9988900423049927, + 0.9984332323074341, + 0.9992363452911377, + 0.9988893270492554, + 0.9993957281112671, + 0.9991062879562378, + 0.9990890622138977, + 0.9993386268615723, + 0.9988206624984741, + 0.9987505674362183, + 0.9989955425262451, + 0.9988095164299011, + 0.9991083741188049, + 0.9988147616386414, + 0.9991130828857422, + 0.9988299012184143, + 0.9990316033363342, + 0.9987291693687439, + 0.9991141557693481, + 0.998677134513855, + 0.9991635084152222, + 0.9987846612930298, + 0.9991999864578247, + 0.9988572001457214, + 0.999275267124176, + 0.9995086193084717, + 0.9993034601211548, + 0.9992854595184326, + 0.9993381500244141, + 0.9992458820343018, + 0.999390721321106, + 0.9992026090621948, + 0.999376654624939, + 0.9990814924240112, + 0.9993203282356262, + 0.998900294303894, + 0.9990503191947937, + 0.9984703063964844, + 0.998579740524292, + 0.9977589845657349, + 0.9986475706100464, + 0.9975122213363647, + 0.9999625086784363 + ], + "feedback_forward_norm_ratio": [ + 0.9990415573120117, + 1.000943660736084, + 0.9984080791473389, + 1.0028012990951538, + 0.9976716041564941, + 1.0041753053665161, + 0.9988983869552612, + 1.0034799575805664, + 0.9985147714614868, + 1.0039784908294678, + 0.9979335069656372, + 1.0033996105194092, + 0.9982885718345642, + 1.0039857625961304, + 0.9969147443771362, + 1.0046236515045166, + 0.9986658096313477, + 1.0027612447738647, + 0.9974470138549805, + 1.001051664352417, + 0.9977295398712158, + 1.0032198429107666, + 0.9985910058021545, + 1.0023927688598633, + 0.9979374408721924, + 1.00351881980896, + 0.9980901479721069, + 1.0027828216552734, + 0.9979441165924072, + 1.0024582147598267, + 0.9984496235847473, + 1.002616047859192, + 0.9983158707618713, + 1.003697156906128, + 0.998077929019928, + 1.0036767721176147, + 0.9981355667114258, + 1.0010948181152344, + 0.9994342923164368, + 1.0016738176345825, + 0.9992212057113647, + 1.0019851922988892, + 0.9994403719902039, + 1.0023409128189087, + 0.9992360472679138, + 1.00275719165802, + 0.9991130232810974, + 1.0027894973754883, + 0.9987571239471436, + 1.0030573606491089, + 0.998371958732605, + 1.0036388635635376, + 0.9987108111381531, + 1.0035512447357178, + 1.0061359405517578 + ], + "feedback_forward_relative_error": [ + 0.0343991219997406, + 0.035711634904146194, + 0.04774899035692215, + 0.04909480735659599, + 0.046834658831357956, + 0.05041529983282089, + 0.047919176518917084, + 0.05211520567536354, + 0.04519075155258179, + 0.04748785123229027, + 0.044232290238142014, + 0.05258563905954361, + 0.047105684876441956, + 0.056232575327157974, + 0.039142630994319916, + 0.047465842217206955, + 0.03476492688059807, + 0.04242744669318199, + 0.04270471632480621, + 0.0364033468067646, + 0.048563480377197266, + 0.05017266422510147, + 0.04481198638677597, + 0.0489116869866848, + 0.04223593324422836, + 0.04889996349811554, + 0.042119480669498444, + 0.04852398484945297, + 0.04401214048266411, + 0.05053481459617615, + 0.04208802059292793, + 0.051570046693086624, + 0.04090111702680588, + 0.049531225115060806, + 0.04000690579414368, + 0.04803706333041191, + 0.03808150812983513, + 0.03138519451022148, + 0.037317704409360886, + 0.03787294030189514, + 0.03637689724564552, + 0.038922250270843506, + 0.034905411303043365, + 0.04005312919616699, + 0.035302672535181046, + 0.04300840571522713, + 0.03686448559165001, + 0.04704458639025688, + 0.04357275739312172, + 0.055480051785707474, + 0.05327729135751724, + 0.06716600060462952, + 0.05199092626571655, + 0.07075086236000061, + 0.01063375361263752 + ], + "innovation_negative_gradient_cosine": [ + 0.9989402890205383, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248114585876, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985315203666687, + 0.9990233182907104, + 0.9983853697776794, + 0.9990407824516296, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990947842597961, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991479516029358, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.999323308467865, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992249011993408, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.999476432800293, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8107086420059204 + ], + "instruction_negative_gradient_cosine": [ + 0.9989403486251831, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248710632324, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985314607620239, + 0.9990233182907104, + 0.9983853101730347, + 0.9990407228469849, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990948438644409, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991480112075806, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.9993233680725098, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992247819900513, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.9994763731956482, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8192939758300781 + ], + "matched_negative_gradient_cosine": [ + 0.008659751154482365, + 0.004593271762132645, + 0.006168784573674202, + 0.019933126866817474, + -0.006501247175037861, + -0.002974468283355236, + -0.019822947680950165, + 0.006686114706099033, + -0.026584379374980927, + 0.013955975882709026, + -0.031510911881923676, + -0.02690306305885315, + -0.035326167941093445, + -0.009890604764223099, + -0.031980231404304504, + 0.007654354441910982, + -0.03439664840698242, + -0.022561632096767426, + -0.02662527561187744, + -0.017831280827522278, + -0.057944610714912415, + -0.009949561208486557, + -0.057219497859478, + 0.010901489295065403, + -0.05287568271160126, + 0.041184812784194946, + -0.0532805360853672, + 0.03105335868895054, + -0.06220395117998123, + 0.038169000297784805, + -0.06922180950641632, + 0.04855819046497345, + -0.07716953754425049, + 0.03764723241329193, + -0.05986480787396431, + 0.061009179800748825, + -0.04746749624609947, + 0.04781028628349304, + -0.06197467818856239, + 0.07869354635477066, + -0.06947663426399231, + 0.08068685978651047, + -0.06836137920618057, + 0.07235347479581833, + -0.04931625351309776, + 0.10434074699878693, + -0.05912315845489502, + 0.0692649632692337, + -0.04305483400821686, + 0.05201111361384392, + -0.027310216799378395, + -0.010442962870001793, + 0.01267608255147934, + 0.03919082134962082, + 0.11070923507213593 + ], + "max_norm_match_direction_error": 2.384185791015625e-07, + "max_norm_match_relative_error": 2.003661876415208e-07, + "mean_feedback_forward_cosine": 0.9989724636077881, + "mean_feedback_forward_relative_error": 0.044561998918652534, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.15488560497760773, + "max_absolute_post_projection_soma_slope": 2.4402833531667056e-08, + "max_absolute_pre_projection_soma_slope": 0.15488560497760773, + "max_positive_post_projection_soma_slope": 2.331008275291424e-08, + "min_post_projection_soma_slope": -2.4402833531667056e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.4101872854046475e-08, + "pre_projection_traffic_rms_ratio": 0.00682154858224028 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.006821548582191776, + "raw_negative_gradient_cosine": [ + 0.008659751154482365, + 0.004593271762132645, + 0.006168785504996777, + 0.019933126866817474, + -0.006501247175037861, + -0.0029744685161858797, + -0.019822947680950165, + 0.006686114706099033, + -0.026584379374980927, + 0.013955974951386452, + -0.031510911881923676, + -0.02690306305885315, + -0.035326167941093445, + -0.009890605695545673, + -0.031980231404304504, + 0.00765435304492712, + -0.03439664840698242, + -0.022561632096767426, + -0.02662527561187744, + -0.017831280827522278, + -0.057944610714912415, + -0.009949559345841408, + -0.057219497859478, + 0.010901490226387978, + -0.05287568271160126, + 0.041184812784194946, + -0.053280532360076904, + 0.03105335868895054, + -0.06220395117998123, + 0.038169000297784805, + -0.06922180950641632, + 0.04855819791555405, + -0.07716953754425049, + 0.03764723241329193, + -0.05986480787396431, + 0.061009179800748825, + -0.047467492520809174, + 0.04781028628349304, + -0.06197467818856239, + 0.07869355380535126, + -0.06947663426399231, + 0.08068685978651047, + -0.06836137920618057, + 0.07235348224639893, + -0.04931625351309776, + 0.10434074699878693, + -0.05912315100431442, + 0.0692649632692337, + -0.04305483028292656, + 0.05201110988855362, + -0.027310214936733246, + -0.010442961938679218, + 0.012676079757511616, + 0.03919082134962082, + 0.11070923507213593 + ], + "teaching_negative_gradient_cosine": [ + 0.9989402890205383, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248114585876, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985315203666687, + 0.9990233182907104, + 0.9983853697776794, + 0.9990407824516296, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990947842597961, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991479516029358, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.999323308467865, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992249011993408, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.999476432800293, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8107086420059204 + ], + "traffic_instruction_rms_ratio": 215.5495647029216, + "used_negative_gradient_cosine": [ + 0.9989402890205383, + 0.9988391995429993, + 0.9989385604858398, + 0.9986226558685303, + 0.9989287853240967, + 0.9985703229904175, + 0.9989549517631531, + 0.9983248114585876, + 0.998971164226532, + 0.9985275268554688, + 0.9990020990371704, + 0.9985315203666687, + 0.9990233182907104, + 0.9983853697776794, + 0.9990407824516296, + 0.9988095760345459, + 0.99907386302948, + 0.9989497661590576, + 0.9990947842597961, + 0.9989767670631409, + 0.9991223216056824, + 0.9985553622245789, + 0.9991262555122375, + 0.9986892938613892, + 0.9991479516029358, + 0.9986973404884338, + 0.9991680383682251, + 0.9988492727279663, + 0.9991772174835205, + 0.9987500905990601, + 0.9992104768753052, + 0.9989532232284546, + 0.9992344379425049, + 0.9989112019538879, + 0.9992313385009766, + 0.9989949464797974, + 0.9992384314537048, + 0.999323308467865, + 0.9992855787277222, + 0.9990972876548767, + 0.999295711517334, + 0.9991129636764526, + 0.9993433952331543, + 0.9992249011993408, + 0.9994274377822876, + 0.999358594417572, + 0.999487042427063, + 0.9994292259216309, + 0.999476432800293, + 0.9994724988937378, + 0.9995529055595398, + 0.9995627403259277, + 0.9997342824935913, + 0.99985271692276, + 0.8107086420059204 + ], + "wall_s": 0.4923238754272461 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.3038, + "eval_loss": 1.9626150634765624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.4722906351089478, + "mean_feedback_forward_cosine": 0.08352781883017583, + "mean_feedback_forward_relative_error": 1.3621288386258212, + "min_feedback_forward_cosine": 0.01647607609629631 + }, + "lr": 0.02, + "mixed_apical": { + "innovation_rms": 0.004356151012226876, + "instruction_rms": 0.004356151012217689, + "raw_apical_rms": 0.06061163826323626, + "teaching_rms": 0.004356151012226876, + "traffic_rms": 0.060407860278224694 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.1548856645822525, + "maximum_absolute_post_projection_soma_slope": 9.257554722807981e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.039406479428281e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006108313938754286, + "minimum_observations": 72 + }, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.133220109324985 + }, + { + "epoch": 2, + "eval_accuracy": 0.4618, + "eval_loss": 1.5612975830078124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3664660453796387, + "mean_feedback_forward_cosine": 0.16610823449763384, + "mean_feedback_forward_relative_error": 1.2985886638814752, + "min_feedback_forward_cosine": 0.07316796481609344 + }, + "lr": 0.04, + "mixed_apical": { + "innovation_rms": 0.0034584715151784293, + "instruction_rms": 0.0034584715164519913, + "raw_apical_rms": 0.06749097762423845, + "teaching_rms": 0.0034584715151784293, + "traffic_rms": 0.0674012550651796 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.701681425942297e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.9294788966921167e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007481774138670017, + "minimum_observations": 72 + }, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.689423344315423 + }, + { + "epoch": 3, + "eval_accuracy": 0.4822, + "eval_loss": 1.8836912109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.335602045059204, + "mean_feedback_forward_cosine": 0.26583654568953946, + "mean_feedback_forward_relative_error": 1.218086308782751, + "min_feedback_forward_cosine": 0.11645860224962234 + }, + "lr": 0.06000000000000001, + "mixed_apical": { + "innovation_rms": 0.004173249144323395, + "instruction_rms": 0.004173249145743122, + "raw_apical_rms": 0.07839940296300797, + "teaching_rms": 0.004173249144323395, + "traffic_rms": 0.07828697291035139 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.905030940946745e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8459908782882565e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008971926468541658, + "minimum_observations": 72 + }, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.3186583264244927 + }, + { + "epoch": 4, + "eval_accuracy": 0.6074, + "eval_loss": 1.1148765869140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3030064105987549, + "mean_feedback_forward_cosine": 0.3486238040707328, + "mean_feedback_forward_relative_error": 1.1469927928664467, + "min_feedback_forward_cosine": 0.16100770235061646 + }, + "lr": 0.08, + "mixed_apical": { + "innovation_rms": 0.0040261518638995695, + "instruction_rms": 0.004026151863866751, + "raw_apical_rms": 0.08739736235297962, + "teaching_rms": 0.0040261518638995695, + "traffic_rms": 0.08730287845580852 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.533971313478105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7700298915347967e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008616507541436013, + "minimum_observations": 72 + }, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 1.0482054900699191 + }, + { + "epoch": 5, + "eval_accuracy": 0.634, + "eval_loss": 1.18164111328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.274701714515686, + "mean_feedback_forward_cosine": 0.4180870847268538, + "mean_feedback_forward_relative_error": 1.0841255122965032, + "min_feedback_forward_cosine": 0.20132866501808167 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00344417015497597, + "instruction_rms": 0.00344417015405666, + "raw_apical_rms": 0.0936506411050976, + "teaching_rms": 0.00344417015497597, + "traffic_rms": 0.09358633788170477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.826502385592903e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.725172805987594e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007545408892908994, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.8692952327728272 + }, + { + "epoch": 6, + "eval_accuracy": 0.7196, + "eval_loss": 0.8544650756835938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2515367269515991, + "mean_feedback_forward_cosine": 0.46626698022538965, + "mean_feedback_forward_relative_error": 1.0380515185269443, + "min_feedback_forward_cosine": 0.23207521438598633 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003258748170606174, + "instruction_rms": 0.003258748168094511, + "raw_apical_rms": 0.09746526204424157, + "teaching_rms": 0.003258748170606174, + "traffic_rms": 0.09741016939437787 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.529219286401712e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7000357580493475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007613073673005167, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7370866859753926 + }, + { + "epoch": 7, + "eval_accuracy": 0.7446, + "eval_loss": 0.7813277587890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2309913635253906, + "mean_feedback_forward_cosine": 0.5051967512477528, + "mean_feedback_forward_relative_error": 0.9995049790902572, + "min_feedback_forward_cosine": 0.2589109241962433 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0031316636807330396, + "instruction_rms": 0.0031316636808242466, + "raw_apical_rms": 0.10071810745913976, + "teaching_rms": 0.0031316636807330396, + "traffic_rms": 0.10066887091447983 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.269455520348856e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6890729415319725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00749110066189653, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6620292295561896 + }, + { + "epoch": 8, + "eval_accuracy": 0.7572, + "eval_loss": 0.7254913330078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2107770442962646, + "mean_feedback_forward_cosine": 0.5387616542252627, + "mean_feedback_forward_relative_error": 0.965096912058917, + "min_feedback_forward_cosine": 0.2846664488315582 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0030568417743375748, + "instruction_rms": 0.003056841775456467, + "raw_apical_rms": 0.10333989084849575, + "teaching_rms": 0.0030568417743375748, + "traffic_rms": 0.10329416930041671 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.03358714101887e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6730053206974266e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007363044872899511, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.6020555562443204 + }, + { + "epoch": 9, + "eval_accuracy": 0.8042, + "eval_loss": 0.6033475341796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1921597719192505, + "mean_feedback_forward_cosine": 0.5684559215198863, + "mean_feedback_forward_relative_error": 0.9336268262429671, + "min_feedback_forward_cosine": 0.30742526054382324 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003028564959114619, + "instruction_rms": 0.003028564959126992, + "raw_apical_rms": 0.10490871755052607, + "teaching_rms": 0.003028564959114619, + "traffic_rms": 0.10486444962000695 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.068348312737726e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6725913714204886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007720522522345244, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.551194996770223 + }, + { + "epoch": 10, + "eval_accuracy": 0.7772, + "eval_loss": 0.6904501342773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1731219291687012, + "mean_feedback_forward_cosine": 0.5953169167041779, + "mean_feedback_forward_relative_error": 0.9040230079130693, + "min_feedback_forward_cosine": 0.3304404616355896 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00298673036466182, + "instruction_rms": 0.002986730365661981, + "raw_apical_rms": 0.10737435769335861, + "teaching_rms": 0.00298673036466182, + "traffic_rms": 0.10733233793075171 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.138749597539572e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6607941445358757e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006803165530875269, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5172687877972921 + }, + { + "epoch": 11, + "eval_accuracy": 0.8256, + "eval_loss": 0.5396714233398437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.154249906539917, + "mean_feedback_forward_cosine": 0.61987811868841, + "mean_feedback_forward_relative_error": 0.8761465267701583, + "min_feedback_forward_cosine": 0.3531190752983093 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029536718346400105, + "instruction_rms": 0.002953671834104379, + "raw_apical_rms": 0.11032681417903577, + "teaching_rms": 0.0029536718346400105, + "traffic_rms": 0.11028671176853683 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.04118576777546e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.648240257180984e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006627346222083808, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.4821384444978502 + }, + { + "epoch": 12, + "eval_accuracy": 0.7906, + "eval_loss": 0.629708203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1347694396972656, + "mean_feedback_forward_cosine": 0.6426574105566198, + "mean_feedback_forward_relative_error": 0.8493387200615623, + "min_feedback_forward_cosine": 0.375612735748291 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029327823604470378, + "instruction_rms": 0.00293278236035286, + "raw_apical_rms": 0.11178448445325447, + "teaching_rms": 0.0029327823604470378, + "traffic_rms": 0.11174537583442526 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.192748003286397e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.645749663223053e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006768195241669802, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.46439163663652205 + }, + { + "epoch": 13, + "eval_accuracy": 0.7176, + "eval_loss": 1.2164780517578124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.116695761680603, + "mean_feedback_forward_cosine": 0.6637622697786851, + "mean_feedback_forward_relative_error": 0.8237342596054077, + "min_feedback_forward_cosine": 0.3959151804447174 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002959575737588152, + "instruction_rms": 0.0029595757375393207, + "raw_apical_rms": 0.11238785921812094, + "teaching_rms": 0.002959575737588152, + "traffic_rms": 0.11234834424661272 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.434101061993715e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639734164071579e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007556765464320654, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.431485446272956 + }, + { + "epoch": 14, + "eval_accuracy": 0.8306, + "eval_loss": 0.527476123046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0985679626464844, + "mean_feedback_forward_cosine": 0.6832012095234611, + "mean_feedback_forward_relative_error": 0.7993083840066736, + "min_feedback_forward_cosine": 0.41602540016174316 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029475493765237, + "instruction_rms": 0.002947549376620449, + "raw_apical_rms": 0.11334492394345563, + "teaching_rms": 0.0029475493765237, + "traffic_rms": 0.11330598076956266 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.384948392224032e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.635391588293572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007641804239876833, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.4132346541987525 + }, + { + "epoch": 15, + "eval_accuracy": 0.8368, + "eval_loss": 0.5055118041992187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0807214975357056, + "mean_feedback_forward_cosine": 0.7016383577476848, + "mean_feedback_forward_relative_error": 0.7754036990079013, + "min_feedback_forward_cosine": 0.43487748503685 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029272191084330516, + "instruction_rms": 0.002927219109138898, + "raw_apical_rms": 0.11455419807492699, + "teaching_rms": 0.0029272191084330516, + "traffic_rms": 0.11451626827829946 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.27533154329285e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6350265657651034e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0076884928417303685, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.39594793183803556 + }, + { + "epoch": 16, + "eval_accuracy": 0.8376, + "eval_loss": 0.49911519165039064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.062043309211731, + "mean_feedback_forward_cosine": 0.7190157613971017, + "mean_feedback_forward_relative_error": 0.7521753657947887, + "min_feedback_forward_cosine": 0.45529162883758545 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029609705366583886, + "instruction_rms": 0.0029609705357800196, + "raw_apical_rms": 0.11580640097877634, + "teaching_rms": 0.0029609705366583886, + "traffic_rms": 0.11576787682414112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.146004238871683e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6201921493682125e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0073920124447960185, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.37886198110050623 + }, + { + "epoch": 17, + "eval_accuracy": 0.8362, + "eval_loss": 0.5104703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0426470041275024, + "mean_feedback_forward_cosine": 0.7352717670527371, + "mean_feedback_forward_relative_error": 0.7296695384112272, + "min_feedback_forward_cosine": 0.47560685873031616 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029236920534005634, + "instruction_rms": 0.0029236920541952276, + "raw_apical_rms": 0.11692784126086204, + "teaching_rms": 0.0029236920534005634, + "traffic_rms": 0.11689065791646244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.37444310061619e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6217122107274594e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007021743952473259, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.36305745289060803 + }, + { + "epoch": 18, + "eval_accuracy": 0.8484, + "eval_loss": 0.4793226806640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0218993425369263, + "mean_feedback_forward_cosine": 0.7508972482247787, + "mean_feedback_forward_relative_error": 0.7074883742765947, + "min_feedback_forward_cosine": 0.4965720772743225 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029090364970267677, + "instruction_rms": 0.0029090364967422036, + "raw_apical_rms": 0.11792658249965413, + "teaching_rms": 0.0029090364970267677, + "traffic_rms": 0.11788996664729536 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.450660888252969e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6166336038137776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006906754591517689, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.3581241923438178 + }, + { + "epoch": 19, + "eval_accuracy": 0.8336, + "eval_loss": 0.5050219970703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0014246702194214, + "mean_feedback_forward_cosine": 0.7655846194787459, + "mean_feedback_forward_relative_error": 0.686012976007028, + "min_feedback_forward_cosine": 0.5174472332000732 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029182402393648416, + "instruction_rms": 0.0029182402380410533, + "raw_apical_rms": 0.11888263616236855, + "teaching_rms": 0.0029182402393648416, + "traffic_rms": 0.11884613260719862 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.689170106710662e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6154687636359724e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006872142968705691, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.33987536052068074 + }, + { + "epoch": 20, + "eval_accuracy": 0.843, + "eval_loss": 0.4968329345703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9816786646842957, + "mean_feedback_forward_cosine": 0.7795224471525712, + "mean_feedback_forward_relative_error": 0.6649386021223935, + "min_feedback_forward_cosine": 0.5363730192184448 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002928576557955807, + "instruction_rms": 0.0029285765574235023, + "raw_apical_rms": 0.11968668469044214, + "teaching_rms": 0.002928576557955807, + "traffic_rms": 0.1196502353407163 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.13406427873997e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.605554074264612e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0073758731777270985, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3248981048266093 + }, + { + "epoch": 21, + "eval_accuracy": 0.8438, + "eval_loss": 0.4893141540527344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9613377451896667, + "mean_feedback_forward_cosine": 0.7925911849195307, + "mean_feedback_forward_relative_error": 0.6446035880934108, + "min_feedback_forward_cosine": 0.5557418465614319 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002895096260092414, + "instruction_rms": 0.00289509625882239, + "raw_apical_rms": 0.12005545551222437, + "teaching_rms": 0.002895096260092414, + "traffic_rms": 0.12001988512162164 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.405740376498215e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.604683605215343e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0070417378300881295, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.3175357684188419 + }, + { + "epoch": 22, + "eval_accuracy": 0.8296, + "eval_loss": 0.54222138671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9419049620628357, + "mean_feedback_forward_cosine": 0.8045841130343351, + "mean_feedback_forward_relative_error": 0.6253614441914992, + "min_feedback_forward_cosine": 0.5741333961486816 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002856555226427493, + "instruction_rms": 0.002856555226869535, + "raw_apical_rms": 0.12087110037492259, + "teaching_rms": 0.002856555226427493, + "traffic_rms": 0.12083667601829956 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.764066557389924e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6102840511883144e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007370012425247404, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.30489753291871813 + }, + { + "epoch": 23, + "eval_accuracy": 0.8574, + "eval_loss": 0.4504137390136719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9202107787132263, + "mean_feedback_forward_cosine": 0.8164499846371738, + "mean_feedback_forward_relative_error": 0.6057515130801634, + "min_feedback_forward_cosine": 0.593589186668396 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028725876334469535, + "instruction_rms": 0.00287258763288266, + "raw_apical_rms": 0.12143231882816564, + "teaching_rms": 0.0028725876334469535, + "traffic_rms": 0.12139765691039796 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.967339689050277e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5969387802185478e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0071395836607705205, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.30050436871846514 + }, + { + "epoch": 24, + "eval_accuracy": 0.8556, + "eval_loss": 0.4592766662597656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8989444971084595, + "mean_feedback_forward_cosine": 0.8273687156763944, + "mean_feedback_forward_relative_error": 0.5870982256802646, + "min_feedback_forward_cosine": 0.6122702360153198 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002864436829320474, + "instruction_rms": 0.002864436829603412, + "raw_apical_rms": 0.12229681651152781, + "teaching_rms": 0.002864436829320474, + "traffic_rms": 0.12226248775156293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.430216442116944e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5990416529969212e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007536012273628919, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.2934852758381102 + }, + { + "epoch": 25, + "eval_accuracy": 0.868, + "eval_loss": 0.44546570434570315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8786854147911072, + "mean_feedback_forward_cosine": 0.8374691377986562, + "mean_feedback_forward_relative_error": 0.5693387931043451, + "min_feedback_forward_cosine": 0.6295708417892456 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002829651196747579, + "instruction_rms": 0.0028296511963980223, + "raw_apical_rms": 0.12340358394249921, + "teaching_rms": 0.002829651196747579, + "traffic_rms": 0.12337042871511392 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.942688912658923e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5987336739498122e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007212634374511028, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.28005119054052563 + }, + { + "epoch": 26, + "eval_accuracy": 0.8544, + "eval_loss": 0.47900587768554687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8583404421806335, + "mean_feedback_forward_cosine": 0.8471909403800965, + "mean_feedback_forward_relative_error": 0.5516635892066089, + "min_feedback_forward_cosine": 0.6465649604797363 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002848069834085287, + "instruction_rms": 0.0028480698337839737, + "raw_apical_rms": 0.1239141001535537, + "teaching_rms": 0.002848069834085287, + "traffic_rms": 0.12388064220186878 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.506996896860073e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6022283644044384e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007315140923491566, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.27400048888524375 + }, + { + "epoch": 27, + "eval_accuracy": 0.8618, + "eval_loss": 0.4623237609863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8366420269012451, + "mean_feedback_forward_cosine": 0.8568273132497614, + "mean_feedback_forward_relative_error": 0.5336897641420364, + "min_feedback_forward_cosine": 0.6642788648605347 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0029013843485941938, + "instruction_rms": 0.0029013843489158237, + "raw_apical_rms": 0.12409414597047937, + "teaching_rms": 0.0029013843485941938, + "traffic_rms": 0.12405940981755702 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.39095185053884e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5923831749284995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007635893560403707, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.2738130042764876 + }, + { + "epoch": 28, + "eval_accuracy": 0.8622, + "eval_loss": 0.4267572326660156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8168511390686035, + "mean_feedback_forward_cosine": 0.8653341174125672, + "mean_feedback_forward_relative_error": 0.5172499830072577, + "min_feedback_forward_cosine": 0.6802014112472534 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00283451752625823, + "instruction_rms": 0.002834517526025941, + "raw_apical_rms": 0.1251018315361786, + "teaching_rms": 0.00283451752625823, + "traffic_rms": 0.12506885912587118 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.042048366656672e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.584903010045293e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0072322718528165875, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.255333635351393 + }, + { + "epoch": 29, + "eval_accuracy": 0.8388, + "eval_loss": 0.5879559814453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.795055091381073, + "mean_feedback_forward_cosine": 0.8737094499848106, + "mean_feedback_forward_relative_error": 0.5007094104181636, + "min_feedback_forward_cosine": 0.6969889402389526 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028299987552350097, + "instruction_rms": 0.002829998755785549, + "raw_apical_rms": 0.12658011802634273, + "teaching_rms": 0.0028299987552350097, + "traffic_rms": 0.12654780600309287 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.592021861706598e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5873349930203038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0077517706763474605, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.25626799415747326 + }, + { + "epoch": 30, + "eval_accuracy": 0.8718, + "eval_loss": 0.39867880859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7741381525993347, + "mean_feedback_forward_cosine": 0.8812859730287032, + "mean_feedback_forward_relative_error": 0.4851856117898768, + "min_feedback_forward_cosine": 0.7126460075378418 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027692479169138754, + "instruction_rms": 0.002769247917051474, + "raw_apical_rms": 0.12651338429739048, + "teaching_rms": 0.0027692479169138754, + "traffic_rms": 0.1264823177734938 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.423312098353563e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.57817413257678e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007284180781127167, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.2529096601009369 + }, + { + "epoch": 31, + "eval_accuracy": 0.8496, + "eval_loss": 0.5267428649902344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7539872527122498, + "mean_feedback_forward_cosine": 0.888480608029799, + "mean_feedback_forward_relative_error": 0.47003339474851435, + "min_feedback_forward_cosine": 0.7276049852371216 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002777148944164583, + "instruction_rms": 0.0027771489435388587, + "raw_apical_rms": 0.1268157266566673, + "teaching_rms": 0.002777148944164583, + "traffic_rms": 0.12678454579159426 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.542652724623622e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5803557519019572e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007124422477869925, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.24340678247875638 + }, + { + "epoch": 32, + "eval_accuracy": 0.8584, + "eval_loss": 0.4592197570800781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7332122325897217, + "mean_feedback_forward_cosine": 0.8954121329567649, + "mean_feedback_forward_relative_error": 0.4549131179397756, + "min_feedback_forward_cosine": 0.7420244216918945 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0028007994985260643, + "instruction_rms": 0.0028007994988742528, + "raw_apical_rms": 0.12702853135736303, + "teaching_rms": 0.0028007994985260643, + "traffic_rms": 0.12699686934374244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.398566026495246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.580052024833378e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007594909668813672, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.23412086204952665 + }, + { + "epoch": 33, + "eval_accuracy": 0.8736, + "eval_loss": 0.4201716491699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7124688625335693, + "mean_feedback_forward_cosine": 0.9018361796032298, + "mean_feedback_forward_relative_error": 0.4405053669756109, + "min_feedback_forward_cosine": 0.7564120292663574 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027640386068428814, + "instruction_rms": 0.0027640386061699306, + "raw_apical_rms": 0.12794259905336494, + "teaching_rms": 0.0027640386068428814, + "traffic_rms": 0.1279120062439201 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.957049774631741e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5846975676409703e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007508806525524007, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.2334535021940867 + }, + { + "epoch": 34, + "eval_accuracy": 0.8588, + "eval_loss": 0.4472315795898438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6925833821296692, + "mean_feedback_forward_cosine": 0.9078588420694524, + "mean_feedback_forward_relative_error": 0.42656055174090646, + "min_feedback_forward_cosine": 0.7698280811309814 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027950519476158046, + "instruction_rms": 0.002795051948121846, + "raw_apical_rms": 0.12850303163694832, + "teaching_rms": 0.0027950519476158046, + "traffic_rms": 0.12847181246060527 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.994501854431292e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.58093701741168e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007139756940525537, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.23220217074288263 + }, + { + "epoch": 35, + "eval_accuracy": 0.8782, + "eval_loss": 0.4101513122558594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6739795804023743, + "mean_feedback_forward_cosine": 0.9134018529545177, + "mean_feedback_forward_relative_error": 0.41330894218249753, + "min_feedback_forward_cosine": 0.7817182540893555 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002732587766544747, + "instruction_rms": 0.0027325877671195275, + "raw_apical_rms": 0.12855414815159502, + "teaching_rms": 0.002732587766544747, + "traffic_rms": 0.128524331035323 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.077385160731865e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5769118836785654e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007295633148785567, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.218170282819536 + }, + { + "epoch": 36, + "eval_accuracy": 0.8558, + "eval_loss": 0.4931694885253906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6537808775901794, + "mean_feedback_forward_cosine": 0.9188592899929393, + "mean_feedback_forward_relative_error": 0.3998900077559731, + "min_feedback_forward_cosine": 0.7946034669876099 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002769010866746977, + "instruction_rms": 0.0027690108679081483, + "raw_apical_rms": 0.12890115929377205, + "teaching_rms": 0.002769010866746977, + "traffic_rms": 0.12887049951636173 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.622905767197153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5788079860832783e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007345657039752453, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.22459270853466456 + }, + { + "epoch": 37, + "eval_accuracy": 0.8776, + "eval_loss": 0.40465287475585937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.634764552116394, + "mean_feedback_forward_cosine": 0.9238546707413413, + "mean_feedback_forward_relative_error": 0.38725803738290615, + "min_feedback_forward_cosine": 0.8060342073440552 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027250066728778913, + "instruction_rms": 0.002725006674491068, + "raw_apical_rms": 0.1296990459836724, + "teaching_rms": 0.0027250066728778913, + "traffic_rms": 0.12966971913174946 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.402741341091314e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5756175122683314e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007059058562824513, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.21901262276437547 + }, + { + "epoch": 38, + "eval_accuracy": 0.8746, + "eval_loss": 0.4201908935546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6174896359443665, + "mean_feedback_forward_cosine": 0.9285104047168385, + "mean_feedback_forward_relative_error": 0.37506215965205975, + "min_feedback_forward_cosine": 0.816362202167511 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002758986976064875, + "instruction_rms": 0.0027589869757443514, + "raw_apical_rms": 0.13022396191701172, + "teaching_rms": 0.002758986976064875, + "traffic_rms": 0.13019391394380037 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.348520369987455e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.581195620728288e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006967136351535924, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.20924033000204298 + }, + { + "epoch": 39, + "eval_accuracy": 0.8458, + "eval_loss": 0.5435376342773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5990682244300842, + "mean_feedback_forward_cosine": 0.9329551046544855, + "mean_feedback_forward_relative_error": 0.36302519400011407, + "min_feedback_forward_cosine": 0.8269726037979126 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027273322335747492, + "instruction_rms": 0.0027273322343506355, + "raw_apical_rms": 0.13095048467476908, + "teaching_rms": 0.0027273322335747492, + "traffic_rms": 0.13092118149951398 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.58836825095932e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5812156990171665e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007481262871491552, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.20766454303529527 + }, + { + "epoch": 40, + "eval_accuracy": 0.874, + "eval_loss": 0.43536727294921873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5813534259796143, + "mean_feedback_forward_cosine": 0.9371668880636042, + "mean_feedback_forward_relative_error": 0.3513018066232855, + "min_feedback_forward_cosine": 0.8368985056877136 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027336582577564207, + "instruction_rms": 0.002733658257960806, + "raw_apical_rms": 0.13118291598587228, + "teaching_rms": 0.0027336582577564207, + "traffic_rms": 0.13115348568521756 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.182926964333092e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.579920515301132e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007629555184683061, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.20535460352367824 + }, + { + "epoch": 41, + "eval_accuracy": 0.8628, + "eval_loss": 0.4404171691894531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5636888742446899, + "mean_feedback_forward_cosine": 0.9411124727942727, + "mean_feedback_forward_relative_error": 0.3399224884130738, + "min_feedback_forward_cosine": 0.8466179966926575 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002744078526469126, + "instruction_rms": 0.002744078525429673, + "raw_apical_rms": 0.1316687178999931, + "teaching_rms": 0.002744078526469126, + "traffic_rms": 0.1316392812613315 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.20359870140419e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5820787800352437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008055984411047367, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.2067696696360906 + }, + { + "epoch": 42, + "eval_accuracy": 0.8614, + "eval_loss": 0.4805859619140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5467633605003357, + "mean_feedback_forward_cosine": 0.9447931029579856, + "mean_feedback_forward_relative_error": 0.32900360971689224, + "min_feedback_forward_cosine": 0.8555374145507812 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002707756478686592, + "instruction_rms": 0.002707756479250177, + "raw_apical_rms": 0.13231939173287324, + "teaching_rms": 0.002707756478686592, + "traffic_rms": 0.13229083411503598 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.411892883306791e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5674920046896586e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007553069431776005, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.19649066227674483 + }, + { + "epoch": 43, + "eval_accuracy": 0.883, + "eval_loss": 0.38886524658203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5288952589035034, + "mean_feedback_forward_cosine": 0.9484201853925531, + "mean_feedback_forward_relative_error": 0.31789819408546793, + "min_feedback_forward_cosine": 0.864727795124054 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002765419179195218, + "instruction_rms": 0.002765419179768991, + "raw_apical_rms": 0.1326336693797406, + "teaching_rms": 0.002765419179195218, + "traffic_rms": 0.13260393202195417 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.192547712975284e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5708784935248106e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007264604335689377, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.20149312126371596 + }, + { + "epoch": 44, + "eval_accuracy": 0.8612, + "eval_loss": 0.50747626953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5118293762207031, + "mean_feedback_forward_cosine": 0.9517511595379222, + "mean_feedback_forward_relative_error": 0.3073600098490715, + "min_feedback_forward_cosine": 0.8732614517211914 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027203200955642414, + "instruction_rms": 0.0027203200957793164, + "raw_apical_rms": 0.13295848846007663, + "teaching_rms": 0.0027203200955642414, + "traffic_rms": 0.13292969522240394 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.60637206553838e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5711413034686096e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007564231186813065, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.19517155153486465 + }, + { + "epoch": 45, + "eval_accuracy": 0.8878, + "eval_loss": 0.36498590087890626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4960921108722687, + "mean_feedback_forward_cosine": 0.9547883597287264, + "mean_feedback_forward_relative_error": 0.2974040485241196, + "min_feedback_forward_cosine": 0.8808695077896118 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026858554466063487, + "instruction_rms": 0.002685855445555374, + "raw_apical_rms": 0.13319069534527686, + "teaching_rms": 0.0026858554466063487, + "traffic_rms": 0.13316264565196495 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.15449451402128e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5824865636116507e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0071554614644181004, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.18736503223843046 + }, + { + "epoch": 46, + "eval_accuracy": 0.8796, + "eval_loss": 0.4100537841796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.479801744222641, + "mean_feedback_forward_cosine": 0.9577374501661821, + "mean_feedback_forward_relative_error": 0.28743237392468884, + "min_feedback_forward_cosine": 0.8884043097496033 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002689226776063514, + "instruction_rms": 0.0026892267768847474, + "raw_apical_rms": 0.13383122359073962, + "teaching_rms": 0.002689226776063514, + "traffic_rms": 0.1338033331991483 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.937294554949403e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5711693143053502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007326037402498839, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.1950026453812917 + }, + { + "epoch": 47, + "eval_accuracy": 0.8474, + "eval_loss": 0.5450638427734374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4652605950832367, + "mean_feedback_forward_cosine": 0.9603951844302091, + "mean_feedback_forward_relative_error": 0.2781396528536623, + "min_feedback_forward_cosine": 0.895006537437439 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027059804394138154, + "instruction_rms": 0.0027059804383819715, + "raw_apical_rms": 0.13411446537974409, + "teaching_rms": 0.0027059804394138154, + "traffic_rms": 0.13408619038165967 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.11744689851912e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.575908293843505e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007480411684179155, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.18229786556031968 + }, + { + "epoch": 48, + "eval_accuracy": 0.881, + "eval_loss": 0.40165243530273437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.45054587721824646, + "mean_feedback_forward_cosine": 0.9629461017521945, + "mean_feedback_forward_relative_error": 0.26894587508656764, + "min_feedback_forward_cosine": 0.9014789462089539 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002693908686589515, + "instruction_rms": 0.0026939086853265126, + "raw_apical_rms": 0.13502132911330866, + "teaching_rms": 0.002693908686589515, + "traffic_rms": 0.1349935960286189 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.1770195129656713e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.570783537931346e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006955519458433687, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.1800456293159061 + }, + { + "epoch": 49, + "eval_accuracy": 0.8728, + "eval_loss": 0.4312097717285156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.43543577194213867, + "mean_feedback_forward_cosine": 0.9654507127675143, + "mean_feedback_forward_relative_error": 0.25960782170295715, + "min_feedback_forward_cosine": 0.9078564643859863 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027744577659100825, + "instruction_rms": 0.0027744577652871103, + "raw_apical_rms": 0.13506830415447602, + "teaching_rms": 0.0027744577659100825, + "traffic_rms": 0.1350388678008716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.2062105270160828e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5670809307587052e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00705050694589141, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.187842915391922 + }, + { + "epoch": 50, + "eval_accuracy": 0.882, + "eval_loss": 0.4146776062011719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.42082518339157104, + "mean_feedback_forward_cosine": 0.9676603544842113, + "mean_feedback_forward_relative_error": 0.25113575214689426, + "min_feedback_forward_cosine": 0.9138859510421753 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026742389896612447, + "instruction_rms": 0.0026742389904405466, + "raw_apical_rms": 0.13622161122472387, + "teaching_rms": 0.0026742389896612447, + "traffic_rms": 0.13619444996311894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.755938469244938e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5681843587762813e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006928600617987082, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.17939616974724665 + }, + { + "epoch": 51, + "eval_accuracy": 0.8604, + "eval_loss": 0.5345984741210937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.40693342685699463, + "mean_feedback_forward_cosine": 0.9697972904552113, + "mean_feedback_forward_relative_error": 0.24261356687003915, + "min_feedback_forward_cosine": 0.9194050431251526 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027662623067010294, + "instruction_rms": 0.002766262307855727, + "raw_apical_rms": 0.1360560502837404, + "teaching_rms": 0.0027662623067010294, + "traffic_rms": 0.13602696343853785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.263002999025048e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5683109799004856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007084063566015009, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.1764916455878152 + }, + { + "epoch": 52, + "eval_accuracy": 0.8872, + "eval_loss": 0.4156570983886719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3940022587776184, + "mean_feedback_forward_cosine": 0.9716921968893572, + "mean_feedback_forward_relative_error": 0.23481411662968724, + "min_feedback_forward_cosine": 0.9243811964988708 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002697988098239205, + "instruction_rms": 0.0026979880970540156, + "raw_apical_rms": 0.13647367571385105, + "teaching_rms": 0.002697988098239205, + "traffic_rms": 0.13644610343628633 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.969654802764126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5684261488329718e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007448526446113501, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.17460551758077408 + }, + { + "epoch": 53, + "eval_accuracy": 0.8586, + "eval_loss": 0.523772998046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.38133642077445984, + "mean_feedback_forward_cosine": 0.9734930721196261, + "mean_feedback_forward_relative_error": 0.2271831222555854, + "min_feedback_forward_cosine": 0.929129958152771 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026592711490205072, + "instruction_rms": 0.0026592711492970256, + "raw_apical_rms": 0.1371275600132018, + "teaching_rms": 0.0026592711490205072, + "traffic_rms": 0.13710088693649733 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.430827426129326e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5650635465185947e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007512666166711919, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.16919431093004014 + }, + { + "epoch": 54, + "eval_accuracy": 0.8722, + "eval_loss": 0.4497177734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3682323396205902, + "mean_feedback_forward_cosine": 0.9752173001115972, + "mean_feedback_forward_relative_error": 0.21963099959221752, + "min_feedback_forward_cosine": 0.9339085817337036 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026435040710614346, + "instruction_rms": 0.002643504071121708, + "raw_apical_rms": 0.13807241382641705, + "teaching_rms": 0.0026435040710614346, + "traffic_rms": 0.138046214003334 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.264454010031841e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5631190660242787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00658986963550856, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.17247383966445923 + }, + { + "epoch": 55, + "eval_accuracy": 0.8752, + "eval_loss": 0.4191963256835938, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3554491400718689, + "mean_feedback_forward_cosine": 0.9768868413838473, + "mean_feedback_forward_relative_error": 0.21204523864117536, + "min_feedback_forward_cosine": 0.9383438229560852 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027582438894246095, + "instruction_rms": 0.0027582438892282783, + "raw_apical_rms": 0.13812332969784827, + "teaching_rms": 0.0027582438894246095, + "traffic_rms": 0.13809479552881931 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.316208521819135e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5671207016380378e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0070906782215834656, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.17254678353203667 + }, + { + "epoch": 56, + "eval_accuracy": 0.8906, + "eval_loss": 0.37485194091796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.34314805269241333, + "mean_feedback_forward_cosine": 0.9784383351152593, + "mean_feedback_forward_relative_error": 0.20477277054028079, + "min_feedback_forward_cosine": 0.9425045251846313 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027295839618689363, + "instruction_rms": 0.0027295839617653022, + "raw_apical_rms": 0.13859104716160184, + "teaching_rms": 0.0027295839618689363, + "traffic_rms": 0.13856313390866049 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.156334109164163e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.566586840540628e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0073859677514521915, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.17252226725286907 + }, + { + "epoch": 57, + "eval_accuracy": 0.8628, + "eval_loss": 0.516157080078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33128440380096436, + "mean_feedback_forward_cosine": 0.9798409548672763, + "mean_feedback_forward_relative_error": 0.19796063107523051, + "min_feedback_forward_cosine": 0.9463651180267334 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026615503012947686, + "instruction_rms": 0.0026615503021120116, + "raw_apical_rms": 0.13962380587554718, + "teaching_rms": 0.0026615503012947686, + "traffic_rms": 0.1395975968215392 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.387510774809925e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.568659816810336e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0068918217166773765, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.16815749957296583 + }, + { + "epoch": 58, + "eval_accuracy": 0.8842, + "eval_loss": 0.39102216796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3196627199649811, + "mean_feedback_forward_cosine": 0.9811797521331094, + "mean_feedback_forward_relative_error": 0.19124092947353016, + "min_feedback_forward_cosine": 0.950006902217865 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026910145468295055, + "instruction_rms": 0.002691014546299131, + "raw_apical_rms": 0.14000806357468948, + "teaching_rms": 0.0026910145468295055, + "traffic_rms": 0.13998117499928586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.236707233460038e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5617678302335342e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007050061817657493, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.16205002730157642 + }, + { + "epoch": 59, + "eval_accuracy": 0.8536, + "eval_loss": 0.5629090026855469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.30899161100387573, + "mean_feedback_forward_cosine": 0.9823998147791082, + "mean_feedback_forward_relative_error": 0.18491454686630857, + "min_feedback_forward_cosine": 0.9532514810562134 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026542196189168255, + "instruction_rms": 0.0026542196190562158, + "raw_apical_rms": 0.1408215516346445, + "teaching_rms": 0.0026542196189168255, + "traffic_rms": 0.14079564388851343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.042845050544202e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5614803441729725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006744619242398327, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.16429483944310083 + }, + { + "epoch": 60, + "eval_accuracy": 0.8662, + "eval_loss": 0.4899179992675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.298313170671463, + "mean_feedback_forward_cosine": 0.9836080356077714, + "mean_feedback_forward_relative_error": 0.17843473235314541, + "min_feedback_forward_cosine": 0.9563993811607361 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027398928836518495, + "instruction_rms": 0.0027398928831306704, + "raw_apical_rms": 0.1412928126177813, + "teaching_rms": 0.0027398928836518495, + "traffic_rms": 0.14126510981437895 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.259701900215077e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5647205410476997e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006542335349927483, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.16467524908648598 + }, + { + "epoch": 61, + "eval_accuracy": 0.8658, + "eval_loss": 0.4761692321777344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2877059280872345, + "mean_feedback_forward_cosine": 0.9847212650559165, + "mean_feedback_forward_relative_error": 0.17225712991573594, + "min_feedback_forward_cosine": 0.9594355821609497 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026823350079127613, + "instruction_rms": 0.002682335009338564, + "raw_apical_rms": 0.14184140255649313, + "teaching_rms": 0.0026823350079127613, + "traffic_rms": 0.14181507741524693 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.723926387370739e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5598152529847118e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006406189595088024, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.1637429113732444 + }, + { + "epoch": 62, + "eval_accuracy": 0.874, + "eval_loss": 0.4491427978515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27832525968551636, + "mean_feedback_forward_cosine": 0.985716106674888, + "mean_feedback_forward_relative_error": 0.16652044640346006, + "min_feedback_forward_cosine": 0.9620136022567749 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026883491568320092, + "instruction_rms": 0.0026883491571807086, + "raw_apical_rms": 0.14177491808751969, + "teaching_rms": 0.0026883491568320092, + "traffic_rms": 0.14174836359817858 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.302907789608071e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5672505030602204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0063795182501372415, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.15938696756495369 + }, + { + "epoch": 63, + "eval_accuracy": 0.8836, + "eval_loss": 0.416558642578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26862141489982605, + "mean_feedback_forward_cosine": 0.9866406364874406, + "mean_feedback_forward_relative_error": 0.16102071370590818, + "min_feedback_forward_cosine": 0.9645927548408508 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026748116780336787, + "instruction_rms": 0.0026748116780000463, + "raw_apical_rms": 0.14231261999482483, + "teaching_rms": 0.0026748116780336787, + "traffic_rms": 0.1422866051708621 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.027869771519363e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5610342012518305e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006332642947062814, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.1564018759330114 + }, + { + "epoch": 64, + "eval_accuracy": 0.8804, + "eval_loss": 0.4190369567871094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2594032287597656, + "mean_feedback_forward_cosine": 0.9875170274214311, + "mean_feedback_forward_relative_error": 0.1556335460055958, + "min_feedback_forward_cosine": 0.9669644832611084 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00265454344144199, + "instruction_rms": 0.002654543442255734, + "raw_apical_rms": 0.14241026046077243, + "teaching_rms": 0.00265454344144199, + "traffic_rms": 0.14238456670964295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.0995398724844563e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558226096933054e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006739131120864999, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.15596432549556097 + }, + { + "epoch": 65, + "eval_accuracy": 0.876, + "eval_loss": 0.43032020874023436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.25037944316864014, + "mean_feedback_forward_cosine": 0.9883387251333757, + "mean_feedback_forward_relative_error": 0.15041881521994419, + "min_feedback_forward_cosine": 0.9692103862762451 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002660427249889889, + "instruction_rms": 0.0026604272507897045, + "raw_apical_rms": 0.14271273509772173, + "teaching_rms": 0.002660427249889889, + "traffic_rms": 0.14268685941881587 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.739331036622389e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5603208316482917e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00676777397236648, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.15521818123393588 + }, + { + "epoch": 66, + "eval_accuracy": 0.8644, + "eval_loss": 0.4574156127929688, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.241669699549675, + "mean_feedback_forward_cosine": 0.9891077453439886, + "mean_feedback_forward_relative_error": 0.14536555781960486, + "min_feedback_forward_cosine": 0.9712958931922913 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026611669122147048, + "instruction_rms": 0.0026611669117068064, + "raw_apical_rms": 0.14344285344871494, + "teaching_rms": 0.0026611669122147048, + "traffic_rms": 0.14341728438724502 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.2062049847827438e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5584886618607892e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006897932759860844, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.15296029583613077 + }, + { + "epoch": 67, + "eval_accuracy": 0.871, + "eval_loss": 0.470939794921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.23326486349105835, + "mean_feedback_forward_cosine": 0.9898286559365013, + "mean_feedback_forward_relative_error": 0.14045636931603606, + "min_feedback_forward_cosine": 0.9732484221458435 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002687069164880757, + "instruction_rms": 0.0026870691648951677, + "raw_apical_rms": 0.14413292472521066, + "teaching_rms": 0.002687069164880757, + "traffic_rms": 0.14410686004612444 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.190936367531322e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5566121694916376e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006414959937737721, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.1511115888701545 + }, + { + "epoch": 68, + "eval_accuracy": 0.8824, + "eval_loss": 0.4170122924804687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22500966489315033, + "mean_feedback_forward_cosine": 0.9904924251816489, + "mean_feedback_forward_relative_error": 0.13578086393800648, + "min_feedback_forward_cosine": 0.975089430809021 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002664870463922071, + "instruction_rms": 0.0026648704640862457, + "raw_apical_rms": 0.14456652877016374, + "teaching_rms": 0.002664870463922071, + "traffic_rms": 0.14454092368457447 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.0229710767362121e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5550262469623638e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006032912033059801, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.14787748815748428 + }, + { + "epoch": 69, + "eval_accuracy": 0.8874, + "eval_loss": 0.408225830078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2171240746974945, + "mean_feedback_forward_cosine": 0.9911184993657198, + "mean_feedback_forward_relative_error": 0.13122503445907072, + "min_feedback_forward_cosine": 0.9767956137657166 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002645522793587049, + "instruction_rms": 0.002645522793048388, + "raw_apical_rms": 0.14505173493671394, + "teaching_rms": 0.002645522793587049, + "traffic_rms": 0.14502663517708173 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.356143237937431e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5635902284659227e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006305431332244356, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.15232352664205764 + }, + { + "epoch": 70, + "eval_accuracy": 0.882, + "eval_loss": 0.4383751708984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2094569355249405, + "mean_feedback_forward_cosine": 0.9917126774787903, + "mean_feedback_forward_relative_error": 0.12675295553424143, + "min_feedback_forward_cosine": 0.9783931374549866 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026780029889819277, + "instruction_rms": 0.002678002988411314, + "raw_apical_rms": 0.14585905041703728, + "teaching_rms": 0.0026780029889819277, + "traffic_rms": 0.14583342579872963 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.433038007320647e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.558632909576437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005996440704448436, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.1509448115931617 + }, + { + "epoch": 71, + "eval_accuracy": 0.8922, + "eval_loss": 0.3718404113769531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20150041580200195, + "mean_feedback_forward_cosine": 0.9922957691279325, + "mean_feedback_forward_relative_error": 0.12221148996190591, + "min_feedback_forward_cosine": 0.9799900054931641 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027130463732828367, + "instruction_rms": 0.0027130463740779698, + "raw_apical_rms": 0.14632378821722586, + "teaching_rms": 0.0027130463732828367, + "traffic_rms": 0.14629759574932538 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.710679573667221e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5560421661491046e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005826156288919894, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.1526694540924496 + }, + { + "epoch": 72, + "eval_accuracy": 0.8636, + "eval_loss": 0.4955664001464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19431686401367188, + "mean_feedback_forward_cosine": 0.9928037383339622, + "mean_feedback_forward_relative_error": 0.1181078931147402, + "min_feedback_forward_cosine": 0.9813801646232605 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002664071129739347, + "instruction_rms": 0.0026640711295466517, + "raw_apical_rms": 0.14673525991584466, + "teaching_rms": 0.002664071129739347, + "traffic_rms": 0.1467101764499398 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.126009731768136e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5553409037084792e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006670596422769175, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.1505700844446818 + }, + { + "epoch": 73, + "eval_accuracy": 0.8908, + "eval_loss": 0.39359422607421873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1880732625722885, + "mean_feedback_forward_cosine": 0.9932585954666138, + "mean_feedback_forward_relative_error": 0.11430438337000934, + "min_feedback_forward_cosine": 0.9825577735900879 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025850056044761024, + "instruction_rms": 0.0025850056042038927, + "raw_apical_rms": 0.14750154417055503, + "teaching_rms": 0.0025850056044761024, + "traffic_rms": 0.14747790681118195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.290473291732269e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.547891947157436e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005991140922600184, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.13924342036247253 + }, + { + "epoch": 74, + "eval_accuracy": 0.8754, + "eval_loss": 0.4568603271484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1814851313829422, + "mean_feedback_forward_cosine": 0.9937111085111444, + "mean_feedback_forward_relative_error": 0.1103922130370682, + "min_feedback_forward_cosine": 0.983745813369751 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002688761067319657, + "instruction_rms": 0.0026887610679484295, + "raw_apical_rms": 0.1482457856465738, + "teaching_rms": 0.002688761067319657, + "traffic_rms": 0.14822035964027308 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.460766937583685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5501059713435917e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006072493714815955, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.14499519442982145 + }, + { + "epoch": 75, + "eval_accuracy": 0.8836, + "eval_loss": 0.44175254516601564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17547009885311127, + "mean_feedback_forward_cosine": 0.9941201881928877, + "mean_feedback_forward_relative_error": 0.10673196173527023, + "min_feedback_forward_cosine": 0.9847981929779053 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026194071485565852, + "instruction_rms": 0.0026194071484548567, + "raw_apical_rms": 0.14859383211916885, + "teaching_rms": 0.0026194071485565852, + "traffic_rms": 0.14856960693123666 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.422455183814236e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5488001708791423e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006249918872554391, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.14262372442086538 + }, + { + "epoch": 76, + "eval_accuracy": 0.8904, + "eval_loss": 0.4010156982421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16894270479679108, + "mean_feedback_forward_cosine": 0.9945245125076988, + "mean_feedback_forward_relative_error": 0.10299218760972673, + "min_feedback_forward_cosine": 0.9859023690223694 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026902458711936606, + "instruction_rms": 0.002690245872698505, + "raw_apical_rms": 0.14890981101763492, + "teaching_rms": 0.0026902458711936606, + "traffic_rms": 0.14888451675335065 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.025010345136252e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5514415971110283e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006046222514356379, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.14856343885262807 + }, + { + "epoch": 77, + "eval_accuracy": 0.888, + "eval_loss": 0.3846547119140625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16314925253391266, + "mean_feedback_forward_cosine": 0.9948757789351723, + "mean_feedback_forward_relative_error": 0.09962514144453136, + "min_feedback_forward_cosine": 0.9868456125259399 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026074787589298602, + "instruction_rms": 0.0026074787585614817, + "raw_apical_rms": 0.1487536503573009, + "teaching_rms": 0.0026074787589298602, + "traffic_rms": 0.14872976845351318 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.563444105405324e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.550309463471458e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006040926408161601, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.14225712541474236 + }, + { + "epoch": 78, + "eval_accuracy": 0.8872, + "eval_loss": 0.42608485717773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15741144120693207, + "mean_feedback_forward_cosine": 0.99521865086122, + "mean_feedback_forward_relative_error": 0.09623410363088955, + "min_feedback_forward_cosine": 0.9877543449401855 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026644436601364232, + "instruction_rms": 0.002664443659893001, + "raw_apical_rms": 0.1491940605432092, + "teaching_rms": 0.0026644436601364232, + "traffic_rms": 0.14916935816296806 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.823434534590888e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545706453746411e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006045228585124519, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.14104981551302803 + }, + { + "epoch": 79, + "eval_accuracy": 0.8934, + "eval_loss": 0.3808540100097656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15193703770637512, + "mean_feedback_forward_cosine": 0.9955377231944691, + "mean_feedback_forward_relative_error": 0.09295748018405654, + "min_feedback_forward_cosine": 0.9885869026184082 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002669457632122175, + "instruction_rms": 0.0026694576316966968, + "raw_apical_rms": 0.14943006841902376, + "teaching_rms": 0.002669457632122175, + "traffic_rms": 0.149405098505395 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.857086987110961e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5490430314861486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006168012162918069, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.1423822466108534 + }, + { + "epoch": 80, + "eval_accuracy": 0.8898, + "eval_loss": 0.3982021240234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14650283753871918, + "mean_feedback_forward_cosine": 0.9958310755816373, + "mean_feedback_forward_relative_error": 0.08984928002411668, + "min_feedback_forward_cosine": 0.9893860220909119 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026474612464264867, + "instruction_rms": 0.0026474612460560456, + "raw_apical_rms": 0.14947968514396548, + "teaching_rms": 0.0026474612464264867, + "traffic_rms": 0.1494552083984354 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.022184433935763e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.543926067696769e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006218834414476274, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.14141934985054863 + }, + { + "epoch": 81, + "eval_accuracy": 0.8836, + "eval_loss": 0.4241302490234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1411639302968979, + "mean_feedback_forward_cosine": 0.9961040648547086, + "mean_feedback_forward_relative_error": 0.08685493939979509, + "min_feedback_forward_cosine": 0.9901441335678101 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025987194917833226, + "instruction_rms": 0.0025987194914536575, + "raw_apical_rms": 0.14968001705164785, + "teaching_rms": 0.0025987194917833226, + "traffic_rms": 0.14965637338464843 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.208247865264639e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545970864669793e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0060768839228992275, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.13652204445468055 + }, + { + "epoch": 82, + "eval_accuracy": 0.8654, + "eval_loss": 0.4978619323730469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13596899807453156, + "mean_feedback_forward_cosine": 0.9963705561377786, + "mean_feedback_forward_relative_error": 0.08382239555093375, + "min_feedback_forward_cosine": 0.9908521771430969 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027191995463910993, + "instruction_rms": 0.0027191995464462345, + "raw_apical_rms": 0.15058060300650536, + "teaching_rms": 0.0027191995463910993, + "traffic_rms": 0.1505549825375809 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.142924118852534e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5464224443863883e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0064172513863293805, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.1410555468744702 + }, + { + "epoch": 83, + "eval_accuracy": 0.8796, + "eval_loss": 0.45452017211914064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13104906678199768, + "mean_feedback_forward_cosine": 0.9966188224879178, + "mean_feedback_forward_relative_error": 0.08089822991327807, + "min_feedback_forward_cosine": 0.9914987087249756 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002680861010178563, + "instruction_rms": 0.0026808610101687782, + "raw_apical_rms": 0.15046814068250927, + "teaching_rms": 0.002680861010178563, + "traffic_rms": 0.15044312746630847 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.218634496235609e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541083891368073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006708994528578518, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.14174490228758918 + }, + { + "epoch": 84, + "eval_accuracy": 0.8776, + "eval_loss": 0.43775721435546877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12613120675086975, + "mean_feedback_forward_cosine": 0.9968532724813981, + "mean_feedback_forward_relative_error": 0.0780410936948928, + "min_feedback_forward_cosine": 0.9921197891235352 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002686031433754745, + "instruction_rms": 0.0026860314337306635, + "raw_apical_rms": 0.150985361330108, + "teaching_rms": 0.002686031433754745, + "traffic_rms": 0.15096039520575474 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.15800885839235e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5473791648249723e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006400952485962795, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.14036392624113295 + }, + { + "epoch": 85, + "eval_accuracy": 0.8952, + "eval_loss": 0.39643720703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12179127335548401, + "mean_feedback_forward_cosine": 0.9970596129244024, + "mean_feedback_forward_relative_error": 0.07543538016351786, + "min_feedback_forward_cosine": 0.9926533699035645 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002612883120352127, + "instruction_rms": 0.0026128831216062705, + "raw_apical_rms": 0.15135610358838467, + "teaching_rms": 0.002612883120352127, + "traffic_rms": 0.15133266316972901 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.039478250841057e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541004753195001e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006462480140216263, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.13974663161171808 + }, + { + "epoch": 86, + "eval_accuracy": 0.8948, + "eval_loss": 0.4052803955078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1174011081457138, + "mean_feedback_forward_cosine": 0.997258413921703, + "mean_feedback_forward_relative_error": 0.0728431259366599, + "min_feedback_forward_cosine": 0.993171751499176 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026489839610354844, + "instruction_rms": 0.0026489839610946936, + "raw_apical_rms": 0.15199564036918278, + "teaching_rms": 0.0026489839610354844, + "traffic_rms": 0.15197157663331626 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.204347778928422e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541520011626562e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006599853107754889, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.1359469562371572 + }, + { + "epoch": 87, + "eval_accuracy": 0.899, + "eval_loss": 0.35590679931640623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11299395561218262, + "mean_feedback_forward_cosine": 0.9974545836448669, + "mean_feedback_forward_relative_error": 0.07018705796111714, + "min_feedback_forward_cosine": 0.9936717748641968 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0027426366580564406, + "instruction_rms": 0.002742636658924247, + "raw_apical_rms": 0.15216527045753545, + "teaching_rms": 0.0027426366580564406, + "traffic_rms": 0.15213959259253293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.208154784166254e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5491023232544303e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006589823912851985, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.1432716718620724 + }, + { + "epoch": 88, + "eval_accuracy": 0.8828, + "eval_loss": 0.44636400146484373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10919094830751419, + "mean_feedback_forward_cosine": 0.9976166118275035, + "mean_feedback_forward_relative_error": 0.06791295161978765, + "min_feedback_forward_cosine": 0.9940903782844543 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002568962072145562, + "instruction_rms": 0.002568962072070146, + "raw_apical_rms": 0.15291764332156205, + "teaching_rms": 0.002568962072145562, + "traffic_rms": 0.15289512250203224 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.191164286268759e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539541243301297e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006509948402681217, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.13177943440013462 + }, + { + "epoch": 89, + "eval_accuracy": 0.8844, + "eval_loss": 0.4189477661132813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10539348423480988, + "mean_feedback_forward_cosine": 0.9977782498706471, + "mean_feedback_forward_relative_error": 0.06556682921946048, + "min_feedback_forward_cosine": 0.9944922924041748 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026700113278570257, + "instruction_rms": 0.0026700113274014236, + "raw_apical_rms": 0.15335840010050059, + "teaching_rms": 0.0026700113278570257, + "traffic_rms": 0.15333401514771836 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.212346986307239e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5447497966286252e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006545180418373123, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.13200987011061774 + }, + { + "epoch": 90, + "eval_accuracy": 0.8722, + "eval_loss": 0.47785639038085936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10174806416034698, + "mean_feedback_forward_cosine": 0.9979229591109536, + "mean_feedback_forward_relative_error": 0.06339765506034548, + "min_feedback_forward_cosine": 0.9948654770851135 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0025989833391281162, + "instruction_rms": 0.0025989833395016213, + "raw_apical_rms": 0.15412977703443345, + "teaching_rms": 0.0025989833391281162, + "traffic_rms": 0.15410691832897805 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.776132721597605e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5438698962622395e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006579018614455897, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.13670137802229987 + }, + { + "epoch": 91, + "eval_accuracy": 0.8918, + "eval_loss": 0.4050358825683594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09814438968896866, + "mean_feedback_forward_cosine": 0.998060323975303, + "mean_feedback_forward_relative_error": 0.061264423179355536, + "min_feedback_forward_cosine": 0.9952216148376465 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002605491879054624, + "instruction_rms": 0.0026054918786173427, + "raw_apical_rms": 0.15449997791814127, + "teaching_rms": 0.002605491879054624, + "traffic_rms": 0.15447700282163823 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.898832083379602e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5388043103855907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006240335922778175, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.1353029336505466 + }, + { + "epoch": 92, + "eval_accuracy": 0.8886, + "eval_loss": 0.41280927734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0946253091096878, + "mean_feedback_forward_cosine": 0.9981919646263122, + "mean_feedback_forward_relative_error": 0.059149937975135715, + "min_feedback_forward_cosine": 0.9955571889877319 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026225778220454157, + "instruction_rms": 0.0026225778222808077, + "raw_apical_rms": 0.1546422124371648, + "teaching_rms": 0.0026225778220454157, + "traffic_rms": 0.15461872216771375 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.40321084435891e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5417959049635595e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006154486073687765, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.13231551391018762 + }, + { + "epoch": 93, + "eval_accuracy": 0.8864, + "eval_loss": 0.440756201171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09125298261642456, + "mean_feedback_forward_cosine": 0.998314782706174, + "mean_feedback_forward_relative_error": 0.05710795244032686, + "min_feedback_forward_cosine": 0.9958686828613281 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026295626089894945, + "instruction_rms": 0.0026295626091631798, + "raw_apical_rms": 0.15497635629214854, + "teaching_rms": 0.0026295626089894945, + "traffic_rms": 0.15495296632301486 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.132918173307189e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.545857133853126e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0060287912510769, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.13290769317944845 + }, + { + "epoch": 94, + "eval_accuracy": 0.8788, + "eval_loss": 0.4635826843261719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08786870539188385, + "mean_feedback_forward_cosine": 0.9984281691637906, + "mean_feedback_forward_relative_error": 0.0551531354134733, + "min_feedback_forward_cosine": 0.9961684346199036 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026328994008423196, + "instruction_rms": 0.0026328994005010184, + "raw_apical_rms": 0.1553290348678801, + "teaching_rms": 0.0026328994008423196, + "traffic_rms": 0.15530565884898515 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.298514421132495e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5375114678351604e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006197718110068988, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.13446046793195937 + }, + { + "epoch": 95, + "eval_accuracy": 0.8726, + "eval_loss": 0.4548491027832031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08478537946939468, + "mean_feedback_forward_cosine": 0.998533051664179, + "mean_feedback_forward_relative_error": 0.053281935693865475, + "min_feedback_forward_cosine": 0.9964317083358765 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002580716866748122, + "instruction_rms": 0.0025807168673490683, + "raw_apical_rms": 0.15564059803859528, + "teaching_rms": 0.002580716866748122, + "traffic_rms": 0.15561814308990682 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.592698381131413e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.530378163367741e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.006073440122871595, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.1281175126393636 + }, + { + "epoch": 96, + "eval_accuracy": 0.8854, + "eval_loss": 0.4196091125488281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0817914828658104, + "mean_feedback_forward_cosine": 0.9986320441419428, + "mean_feedback_forward_relative_error": 0.05145530422980135, + "min_feedback_forward_cosine": 0.9966788291931152 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002601581738691615, + "instruction_rms": 0.0026015817386345575, + "raw_apical_rms": 0.1561557088080724, + "teaching_rms": 0.002601581738691615, + "traffic_rms": 0.1561330561640619 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.298388737808637e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534340281688561e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0061203401449506445, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.1325615198334058 + }, + { + "epoch": 97, + "eval_accuracy": 0.8898, + "eval_loss": 0.38385213623046877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07893867790699005, + "mean_feedback_forward_cosine": 0.9987242817878723, + "mean_feedback_forward_relative_error": 0.04968901854008436, + "min_feedback_forward_cosine": 0.9969061017036438 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026407954787508905, + "instruction_rms": 0.00264079547831246, + "raw_apical_rms": 0.156541361440323, + "teaching_rms": 0.0026407954787508905, + "traffic_rms": 0.15651807622354477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.303297960310374e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534870101288664e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0059349221765024164, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.13027635041210386 + }, + { + "epoch": 98, + "eval_accuracy": 0.8778, + "eval_loss": 0.4473033386230469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07644806802272797, + "mean_feedback_forward_cosine": 0.998805943402377, + "mean_feedback_forward_relative_error": 0.048069786178794774, + "min_feedback_forward_cosine": 0.997097373008728 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002568411945691171, + "instruction_rms": 0.0025684119463123905, + "raw_apical_rms": 0.15654332971914975, + "teaching_rms": 0.002568411945691171, + "traffic_rms": 0.15652118693536912 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.238367689093138e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360332500682992e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005915503647992677, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.12253387283749051 + }, + { + "epoch": 99, + "eval_accuracy": 0.8902, + "eval_loss": 0.39856865234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07376903295516968, + "mean_feedback_forward_cosine": 0.9988860639658841, + "mean_feedback_forward_relative_error": 0.046428847753188826, + "min_feedback_forward_cosine": 0.9972964525222778 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0026522896668071372, + "instruction_rms": 0.0026522896669382012, + "raw_apical_rms": 0.15687143960009156, + "teaching_rms": 0.0026522896668071372, + "traffic_rms": 0.1568479856923433 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.49559711468828e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535734913191686e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005805982409407321, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.12809759814474317 + }, + { + "epoch": 100, + "eval_accuracy": 0.8754, + "eval_loss": 0.47369862060546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07100122421979904, + "mean_feedback_forward_cosine": 0.9989642349156467, + "mean_feedback_forward_relative_error": 0.044769683361730796, + "min_feedback_forward_cosine": 0.9974948167800903 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.002654077659886959, + "instruction_rms": 0.0026540776599525647, + "raw_apical_rms": 0.15676781284296182, + "teaching_rms": 0.002654077659886959, + "traffic_rms": 0.15674415038878697 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.196233925948036e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.53599456264671e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005805873848419814, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.13096666544874508 + }, + { + "epoch": 101, + "eval_accuracy": 0.9234, + "eval_loss": 0.27488249816894533, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0709770992398262, + "mean_feedback_forward_cosine": 0.9989649100737138, + "mean_feedback_forward_relative_error": 0.04475421941077167, + "min_feedback_forward_cosine": 0.9974966049194336 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018681300327095065, + "instruction_rms": 0.0018681300327021025, + "raw_apical_rms": 0.15710599554554727, + "teaching_rms": 0.0018681300327095065, + "traffic_rms": 0.15709385198047787 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.328763729219645e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541821079967624e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005917019782837112, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.07250560165014532 + }, + { + "epoch": 102, + "eval_accuracy": 0.9236, + "eval_loss": 0.27401832275390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07096787542104721, + "mean_feedback_forward_cosine": 0.9989651679992676, + "mean_feedback_forward_relative_error": 0.044747015613723885, + "min_feedback_forward_cosine": 0.9974972009658813 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015019077679723524, + "instruction_rms": 0.0015019077677623576, + "raw_apical_rms": 0.1570744741692994, + "teaching_rms": 0.0015019077679723524, + "traffic_rms": 0.1570666584012238 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.297803333334741e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534955971026108e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00589227916419364, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.04508370224568579 + }, + { + "epoch": 103, + "eval_accuracy": 0.9272, + "eval_loss": 0.2832565734863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07095963507890701, + "mean_feedback_forward_cosine": 0.9989653944969177, + "mean_feedback_forward_relative_error": 0.04474038678136739, + "min_feedback_forward_cosine": 0.9974977970123291 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001402311703568493, + "instruction_rms": 0.001402311703717671, + "raw_apical_rms": 0.1571234752168091, + "teaching_rms": 0.001402311703568493, + "traffic_rms": 0.15711630915522729 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 6.596710733219879e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360707500546335e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005898395072908181, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.03774840233557754 + }, + { + "epoch": 104, + "eval_accuracy": 0.925, + "eval_loss": 0.2829547271728516, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07095282524824142, + "mean_feedback_forward_cosine": 0.9989656166596846, + "mean_feedback_forward_relative_error": 0.04473473756489429, + "min_feedback_forward_cosine": 0.9974982738494873 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012750670215814861, + "instruction_rms": 0.0012750670214078238, + "raw_apical_rms": 0.1570559243132205, + "teaching_rms": 0.0012750670215814861, + "traffic_rms": 0.15704995341997907 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.518710193859079e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535662505794863e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00590715919836829, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.030839167262448206 + }, + { + "epoch": 105, + "eval_accuracy": 0.9252, + "eval_loss": 0.2876938934326172, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07094645500183105, + "mean_feedback_forward_cosine": 0.9989657987247814, + "mean_feedback_forward_relative_error": 0.04472943408922716, + "min_feedback_forward_cosine": 0.9974987506866455 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012315131502262538, + "instruction_rms": 0.0012315131500998835, + "raw_apical_rms": 0.15701118973065295, + "teaching_rms": 0.0012315131502262538, + "traffic_rms": 0.1570055661357571 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.324130990582489e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538040215241406e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005888449746465611, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.027694817138049336 + }, + { + "epoch": 106, + "eval_accuracy": 0.9274, + "eval_loss": 0.28754229736328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07094030827283859, + "mean_feedback_forward_cosine": 0.9989660057154569, + "mean_feedback_forward_relative_error": 0.044724249264056035, + "min_feedback_forward_cosine": 0.9974992275238037 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012348020577495457, + "instruction_rms": 0.0012348020577501633, + "raw_apical_rms": 0.1569274317159094, + "teaching_rms": 0.0012348020577495457, + "traffic_rms": 0.15692170682291784 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.317851213883841e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538755021751766e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005776124519394298, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.025074613791704177 + }, + { + "epoch": 107, + "eval_accuracy": 0.9274, + "eval_loss": 0.28696969299316405, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.070933997631073, + "mean_feedback_forward_cosine": 0.9989662062038075, + "mean_feedback_forward_relative_error": 0.04471908270974051, + "min_feedback_forward_cosine": 0.9974997043609619 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012253141640255732, + "instruction_rms": 0.0012253141635396715, + "raw_apical_rms": 0.1568674549808664, + "teaching_rms": 0.0012253141640255732, + "traffic_rms": 0.15686162558067718 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.282807246156153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536815027355969e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058119740785779524, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.023852526133590273 + }, + { + "epoch": 108, + "eval_accuracy": 0.9278, + "eval_loss": 0.2932923095703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07092788070440292, + "mean_feedback_forward_cosine": 0.9989663926037875, + "mean_feedback_forward_relative_error": 0.044714076359840954, + "min_feedback_forward_cosine": 0.9975001215934753 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001164947536507392, + "instruction_rms": 0.0011649475362220188, + "raw_apical_rms": 0.15685390294079102, + "teaching_rms": 0.001164947536507392, + "traffic_rms": 0.15684856122585833 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.801549000736486e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537907331784123e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005743717111929405, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.02110563539125853 + }, + { + "epoch": 109, + "eval_accuracy": 0.9278, + "eval_loss": 0.3009987487792969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07092272490262985, + "mean_feedback_forward_cosine": 0.9989665714177218, + "mean_feedback_forward_relative_error": 0.04470948527821086, + "min_feedback_forward_cosine": 0.9975005388259888 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010989609109914053, + "instruction_rms": 0.0010989609108569018, + "raw_apical_rms": 0.15679209054773383, + "teaching_rms": 0.0010989609109914053, + "traffic_rms": 0.15678733926444163 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.114888944670383e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355600671949557e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005833617276694973, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.018469695284962655 + }, + { + "epoch": 110, + "eval_accuracy": 0.9322, + "eval_loss": 0.2989709991455078, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07091721147298813, + "mean_feedback_forward_cosine": 0.9989667556502603, + "mean_feedback_forward_relative_error": 0.0447049082680182, + "min_feedback_forward_cosine": 0.9975008964538574 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011106760717449825, + "instruction_rms": 0.001110676071854277, + "raw_apical_rms": 0.15674526207655107, + "teaching_rms": 0.0011106760717449825, + "traffic_rms": 0.15674021156413928 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.926139661800335e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5407929228254788e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005785295681242412, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.017986075497666995 + }, + { + "epoch": 111, + "eval_accuracy": 0.9312, + "eval_loss": 0.2975237548828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07091204077005386, + "mean_feedback_forward_cosine": 0.9989669182083823, + "mean_feedback_forward_relative_error": 0.04470046184618365, + "min_feedback_forward_cosine": 0.9975012540817261 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010458979697199353, + "instruction_rms": 0.0010458979695164385, + "raw_apical_rms": 0.15675384503128525, + "teaching_rms": 0.0010458979697199353, + "traffic_rms": 0.15674930876130302 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.771234239977275e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5387045144860748e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005776015332794581, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.01620114773346318 + }, + { + "epoch": 112, + "eval_accuracy": 0.9324, + "eval_loss": 0.3009730224609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07090695202350616, + "mean_feedback_forward_cosine": 0.9989670731804587, + "mean_feedback_forward_relative_error": 0.04469619986008514, + "min_feedback_forward_cosine": 0.9975014925003052 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010423431884891887, + "instruction_rms": 0.001042343188609069, + "raw_apical_rms": 0.1567646603551989, + "teaching_rms": 0.0010423431884891887, + "traffic_rms": 0.15676002837858788 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 6.636144433969093e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5359185499159353e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00577949730688871, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.01542122282071246 + }, + { + "epoch": 113, + "eval_accuracy": 0.9304, + "eval_loss": 0.3087217346191406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07090183347463608, + "mean_feedback_forward_cosine": 0.9989672541618347, + "mean_feedback_forward_relative_error": 0.04469187412072312, + "min_feedback_forward_cosine": 0.9975018501281738 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010069105057670844, + "instruction_rms": 0.0010069105051730399, + "raw_apical_rms": 0.15674903436380308, + "teaching_rms": 0.0010069105057670844, + "traffic_rms": 0.15674476255955913 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.054693185182259e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5373920009531483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005810685867104819, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.014235611286262672 + }, + { + "epoch": 114, + "eval_accuracy": 0.931, + "eval_loss": 0.314695849609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07089726626873016, + "mean_feedback_forward_cosine": 0.9989674113013528, + "mean_feedback_forward_relative_error": 0.044688034074550326, + "min_feedback_forward_cosine": 0.9975022077560425 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009729803436573065, + "instruction_rms": 0.000972980343533662, + "raw_apical_rms": 0.15672653995108682, + "teaching_rms": 0.0009729803436573065, + "traffic_rms": 0.1567224886940586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.398914275176139e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538041464532142e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005762569430221804, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.012990544721484185 + }, + { + "epoch": 115, + "eval_accuracy": 0.9284, + "eval_loss": 0.3111776794433594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07089243829250336, + "mean_feedback_forward_cosine": 0.9989675565199418, + "mean_feedback_forward_relative_error": 0.044684005274691366, + "min_feedback_forward_cosine": 0.9975023865699768 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009499882285774866, + "instruction_rms": 0.0009499882290781333, + "raw_apical_rms": 0.15675961693258456, + "teaching_rms": 0.0009499882285774866, + "traffic_rms": 0.15675556098779417 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.990103346675824e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5365512890623903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005769194975041045, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.012409936531053649 + }, + { + "epoch": 116, + "eval_accuracy": 0.9318, + "eval_loss": 0.3148053466796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07088766992092133, + "mean_feedback_forward_cosine": 0.9989677049896933, + "mean_feedback_forward_relative_error": 0.04467999024147337, + "min_feedback_forward_cosine": 0.997502863407135 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009559801532536852, + "instruction_rms": 0.0009559801530943921, + "raw_apical_rms": 0.15673376741355344, + "teaching_rms": 0.0009559801532536852, + "traffic_rms": 0.1567297328532247 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.778653809964453e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5335779749294922e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005744965806277495, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.012501899681488673 + }, + { + "epoch": 117, + "eval_accuracy": 0.9322, + "eval_loss": 0.3190792907714844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07088316231966019, + "mean_feedback_forward_cosine": 0.9989678827199069, + "mean_feedback_forward_relative_error": 0.04467604162002152, + "min_feedback_forward_cosine": 0.9975031018257141 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009494193280651492, + "instruction_rms": 0.0009494193279696085, + "raw_apical_rms": 0.15675959832923328, + "teaching_rms": 0.0009494193280651492, + "traffic_rms": 0.15675568312690336 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.358942528095213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.533411126948754e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005877179956535044, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.011608204885323842 + }, + { + "epoch": 118, + "eval_accuracy": 0.9298, + "eval_loss": 0.32249583129882814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087850570678711, + "mean_feedback_forward_cosine": 0.9989680203524502, + "mean_feedback_forward_relative_error": 0.044672052206640896, + "min_feedback_forward_cosine": 0.9975035190582275 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009212186690548354, + "instruction_rms": 0.0009212186691952662, + "raw_apical_rms": 0.1567560234107983, + "teaching_rms": 0.0009212186690548354, + "traffic_rms": 0.15675203733659918 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 8.317160649085054e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5406398805912062e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005816009318486093, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.011224714832069974 + }, + { + "epoch": 119, + "eval_accuracy": 0.932, + "eval_loss": 0.32551557006835935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087486982345581, + "mean_feedback_forward_cosine": 0.9989681460640647, + "mean_feedback_forward_relative_error": 0.0446686068888415, + "min_feedback_forward_cosine": 0.9975037574768066 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008470678397713774, + "instruction_rms": 0.0008470678396755314, + "raw_apical_rms": 0.15666118073060328, + "teaching_rms": 0.0008470678397713774, + "traffic_rms": 0.15665776704803078 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.599363354733214e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.541024069794903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005867086167817335, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.009133812613785267 + }, + { + "epoch": 120, + "eval_accuracy": 0.931, + "eval_loss": 0.3299600830078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07087042182683945, + "mean_feedback_forward_cosine": 0.9989683086221869, + "mean_feedback_forward_relative_error": 0.044664719616147604, + "min_feedback_forward_cosine": 0.9975039958953857 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008949044177462138, + "instruction_rms": 0.0008949044175458051, + "raw_apical_rms": 0.15663499626042995, + "teaching_rms": 0.0008949044177462138, + "traffic_rms": 0.15663112010010294 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.459373480538488e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5368161104063114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005843225887937226, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.009784037462208007 + }, + { + "epoch": 121, + "eval_accuracy": 0.9292, + "eval_loss": 0.32630562744140623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07086643576622009, + "mean_feedback_forward_cosine": 0.9989684527570551, + "mean_feedback_forward_relative_error": 0.044661184095523575, + "min_feedback_forward_cosine": 0.9975043535232544 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008145382890291357, + "instruction_rms": 0.0008145382889972386, + "raw_apical_rms": 0.1566612766499781, + "teaching_rms": 0.0008145382890291357, + "traffic_rms": 0.15665788889761117 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 6.135365993031883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5382620985560022e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005812800528565876, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.009073666857431332 + }, + { + "epoch": 122, + "eval_accuracy": 0.9302, + "eval_loss": 0.32952420654296877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07086223363876343, + "mean_feedback_forward_cosine": 0.9989685979756442, + "mean_feedback_forward_relative_error": 0.044657568464224986, + "min_feedback_forward_cosine": 0.9975045919418335 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000828141355639425, + "instruction_rms": 0.0008281413559018742, + "raw_apical_rms": 0.15664252699002104, + "teaching_rms": 0.000828141355639425, + "traffic_rms": 0.15663918014748687 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 5.9446421118991566e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5356018937259457e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005891065101916085, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.009004947647555835 + }, + { + "epoch": 123, + "eval_accuracy": 0.9304, + "eval_loss": 0.33173782958984377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07085803151130676, + "mean_feedback_forward_cosine": 0.9989687345244668, + "mean_feedback_forward_relative_error": 0.044654008221219885, + "min_feedback_forward_cosine": 0.9975050091743469 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008190901937503409, + "instruction_rms": 0.0008190901936549919, + "raw_apical_rms": 0.15661127665436178, + "teaching_rms": 0.0008190901937503409, + "traffic_rms": 0.15660783013726773 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488563477993011, + "maximum_absolute_post_projection_soma_slope": 7.179834682347064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539510267280735e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005910763090338856, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.0088103628065851 + }, + { + "epoch": 124, + "eval_accuracy": 0.9312, + "eval_loss": 0.33396995239257815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07085384428501129, + "mean_feedback_forward_cosine": 0.9989688656546852, + "mean_feedback_forward_relative_error": 0.044650416922840205, + "min_feedback_forward_cosine": 0.9975050687789917 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008038917385117785, + "instruction_rms": 0.0008038917382657697, + "raw_apical_rms": 0.15659737569674384, + "teaching_rms": 0.0008038917385117785, + "traffic_rms": 0.1565940664957959 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.250633788340565e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5383427378892843e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005881761370236408, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.00854498468455341 + }, + { + "epoch": 125, + "eval_accuracy": 0.9296, + "eval_loss": 0.33593333740234377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07084941118955612, + "mean_feedback_forward_cosine": 0.9989690412174571, + "mean_feedback_forward_relative_error": 0.044646709530868314, + "min_feedback_forward_cosine": 0.9975054264068604 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000820239429463051, + "instruction_rms": 0.0008202394295456283, + "raw_apical_rms": 0.1565658226530157, + "teaching_rms": 0.000820239429463051, + "traffic_rms": 0.15656242031847997 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.694543313618851e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5364542856645387e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005836558479366115, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.008565741331958109 + }, + { + "epoch": 126, + "eval_accuracy": 0.93, + "eval_loss": 0.3363638122558594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708455815911293, + "mean_feedback_forward_cosine": 0.9989691755988381, + "mean_feedback_forward_relative_error": 0.044643311642787675, + "min_feedback_forward_cosine": 0.9975056648254395 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007721182285074628, + "instruction_rms": 0.0007721182286492539, + "raw_apical_rms": 0.15655390561368063, + "teaching_rms": 0.0007721182285074628, + "traffic_rms": 0.1565506457901433 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.784327410969127e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5363067740951922e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005898069685332526, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.007548671312464608 + }, + { + "epoch": 127, + "eval_accuracy": 0.9322, + "eval_loss": 0.33970142822265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07084105908870697, + "mean_feedback_forward_cosine": 0.998969309980219, + "mean_feedback_forward_relative_error": 0.04463963938707655, + "min_feedback_forward_cosine": 0.9975060224533081 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007952847310662166, + "instruction_rms": 0.000795284731119907, + "raw_apical_rms": 0.15649639667063278, + "teaching_rms": 0.0007952847310662166, + "traffic_rms": 0.1564932482741599 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.252860712014808e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5363136251838993e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005954777612223128, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.007928385191245212 + }, + { + "epoch": 128, + "eval_accuracy": 0.93, + "eval_loss": 0.34044767456054686, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07083731144666672, + "mean_feedback_forward_cosine": 0.998969432440671, + "mean_feedback_forward_relative_error": 0.044636607339436356, + "min_feedback_forward_cosine": 0.9975062608718872 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006701769134514322, + "instruction_rms": 0.0006701769134532176, + "raw_apical_rms": 0.1564775455340859, + "teaching_rms": 0.0006701769134514322, + "traffic_rms": 0.15647496268213462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.703119481699105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5411468270352428e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005862443182673322, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.006844878549252947 + }, + { + "epoch": 129, + "eval_accuracy": 0.9308, + "eval_loss": 0.3454027954101562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07083291560411453, + "mean_feedback_forward_cosine": 0.9989695819941434, + "mean_feedback_forward_relative_error": 0.04463285764848644, + "min_feedback_forward_cosine": 0.9975066781044006 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008105760127383864, + "instruction_rms": 0.0008105760127973803, + "raw_apical_rms": 0.15649128965338405, + "teaching_rms": 0.0008105760127383864, + "traffic_rms": 0.15648792415711557 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.919364636814862e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5357132180343736e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005847970109920921, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.00751388033653299 + }, + { + "epoch": 130, + "eval_accuracy": 0.9312, + "eval_loss": 0.3445435241699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07082948088645935, + "mean_feedback_forward_cosine": 0.9989696947011081, + "mean_feedback_forward_relative_error": 0.044630019857802176, + "min_feedback_forward_cosine": 0.997506856918335 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006418838499680199, + "instruction_rms": 0.0006418838497189099, + "raw_apical_rms": 0.15642447055607922, + "teaching_rms": 0.0006418838499680199, + "traffic_rms": 0.15642219504731347 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.848455586487944e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5415882295120003e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00581332910731474, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.005915861031826999 + }, + { + "epoch": 131, + "eval_accuracy": 0.9312, + "eval_loss": 0.3421359436035156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07082522660493851, + "mean_feedback_forward_cosine": 0.9989698323336514, + "mean_feedback_forward_relative_error": 0.04462645375593142, + "min_feedback_forward_cosine": 0.9975071549415588 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000737686763895328, + "instruction_rms": 0.000737686764102491, + "raw_apical_rms": 0.15641385282290396, + "teaching_rms": 0.000737686763895328, + "traffic_rms": 0.15641073588207244 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 8.277032037540266e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5418303191121762e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005896190408225254, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.006755238181166351 + }, + { + "epoch": 132, + "eval_accuracy": 0.9308, + "eval_loss": 0.3458807006835937, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07082144916057587, + "mean_feedback_forward_cosine": 0.9989699743010781, + "mean_feedback_forward_relative_error": 0.0446231294084679, + "min_feedback_forward_cosine": 0.9975073933601379 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007122230452864486, + "instruction_rms": 0.000712223045463341, + "raw_apical_rms": 0.15639152371594156, + "teaching_rms": 0.0007122230452864486, + "traffic_rms": 0.1563886586829158 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.187219352999819e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5373814648263543e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005826939346422026, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.006470164734808107 + }, + { + "epoch": 133, + "eval_accuracy": 0.9326, + "eval_loss": 0.34408472900390624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07081776857376099, + "mean_feedback_forward_cosine": 0.9989700924266468, + "mean_feedback_forward_relative_error": 0.04462002458220178, + "min_feedback_forward_cosine": 0.997507631778717 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006537943374323771, + "instruction_rms": 0.0006537943376771356, + "raw_apical_rms": 0.1563871952729514, + "teaching_rms": 0.0006537943374323771, + "traffic_rms": 0.15638462368018344 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.090667336091428e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536691766808993e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005813838483222319, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.005801812709122896 + }, + { + "epoch": 134, + "eval_accuracy": 0.9306, + "eval_loss": 0.3515958251953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07081308215856552, + "mean_feedback_forward_cosine": 0.9989702679894187, + "mean_feedback_forward_relative_error": 0.044616280546919865, + "min_feedback_forward_cosine": 0.9975079298019409 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007491625438862237, + "instruction_rms": 0.0007491625436995141, + "raw_apical_rms": 0.15636806297910094, + "teaching_rms": 0.0007491625438862237, + "traffic_rms": 0.15636475892267998 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.767201992692208e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5358320999069396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005863827401303301, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.006816023340572914 + }, + { + "epoch": 135, + "eval_accuracy": 0.9322, + "eval_loss": 0.3474887512207031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708087608218193, + "mean_feedback_forward_cosine": 0.9989703969521956, + "mean_feedback_forward_relative_error": 0.04461274790492925, + "min_feedback_forward_cosine": 0.9975084066390991 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007182916366688837, + "instruction_rms": 0.000718291636966421, + "raw_apical_rms": 0.15639258387918176, + "teaching_rms": 0.0007182916366688837, + "traffic_rms": 0.15638971540673072 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.585887746268781e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5351030459360198e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005775043328720412, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.00620322556015518 + }, + { + "epoch": 136, + "eval_accuracy": 0.93, + "eval_loss": 0.3605854431152344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0708053857088089, + "mean_feedback_forward_cosine": 0.9989705400033431, + "mean_feedback_forward_relative_error": 0.044609861262142655, + "min_feedback_forward_cosine": 0.9975085258483887 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006322296166297042, + "instruction_rms": 0.0006322296167166218, + "raw_apical_rms": 0.1564355575673448, + "teaching_rms": 0.0006322296166297042, + "traffic_rms": 0.15643309830365237 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.498969267771827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537036167272147e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005937657389980397, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.00543699383230673 + }, + { + "epoch": 137, + "eval_accuracy": 0.9328, + "eval_loss": 0.35638165893554685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07080122083425522, + "mean_feedback_forward_cosine": 0.9989706635475158, + "mean_feedback_forward_relative_error": 0.0446064278991385, + "min_feedback_forward_cosine": 0.9975087642669678 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006742951027451404, + "instruction_rms": 0.0006742951026956253, + "raw_apical_rms": 0.156422869061531, + "teaching_rms": 0.0006742951027451404, + "traffic_rms": 0.15642013032309593 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.080841868400967e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5443088115768734e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0059150798739307595, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.005787847943024503 + }, + { + "epoch": 138, + "eval_accuracy": 0.93, + "eval_loss": 0.36132265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707966685295105, + "mean_feedback_forward_cosine": 0.9989708206870339, + "mean_feedback_forward_relative_error": 0.04460295835001902, + "min_feedback_forward_cosine": 0.9975091218948364 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006623774581377063, + "instruction_rms": 0.0006623774580451413, + "raw_apical_rms": 0.15636598707233249, + "teaching_rms": 0.0006623774581377063, + "traffic_rms": 0.1563632726246208 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.556190612627688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536713059570718e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005861925669562594, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.005622351417938868 + }, + { + "epoch": 139, + "eval_accuracy": 0.932, + "eval_loss": 0.35857147216796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07079344242811203, + "mean_feedback_forward_cosine": 0.9989709236405112, + "mean_feedback_forward_relative_error": 0.04460024108940905, + "min_feedback_forward_cosine": 0.9975094199180603 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005859143428559092, + "instruction_rms": 0.0005859143429779436, + "raw_apical_rms": 0.15636178605641754, + "teaching_rms": 0.0005859143428559092, + "traffic_rms": 0.15635984807257855 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.712134302868435e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534403999828429e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005829575657965549, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.004863424326479435 + }, + { + "epoch": 140, + "eval_accuracy": 0.931, + "eval_loss": 0.36219781494140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07078968733549118, + "mean_feedback_forward_cosine": 0.9989710688591004, + "mean_feedback_forward_relative_error": 0.044596916471015324, + "min_feedback_forward_cosine": 0.9975095987319946 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006475315319334628, + "instruction_rms": 0.0006475315319258127, + "raw_apical_rms": 0.15632147733190216, + "teaching_rms": 0.0006475315319334628, + "traffic_rms": 0.15631878209154426 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.692059173474263e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5363117901857483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005813491649850502, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.005354508005910449 + }, + { + "epoch": 141, + "eval_accuracy": 0.9308, + "eval_loss": 0.35549102783203124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07078536599874496, + "mean_feedback_forward_cosine": 0.998971240086989, + "mean_feedback_forward_relative_error": 0.04459290558641607, + "min_feedback_forward_cosine": 0.9975098967552185 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0007365572518808323, + "instruction_rms": 0.0007365572518513475, + "raw_apical_rms": 0.15628183672241677, + "teaching_rms": 0.0007365572518808323, + "traffic_rms": 0.15627873155211572 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.233802534756251e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5344348896844735e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005786564292966944, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.005712501531182271 + }, + { + "epoch": 142, + "eval_accuracy": 0.9328, + "eval_loss": 0.3628958251953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07078168541193008, + "mean_feedback_forward_cosine": 0.9989713582125577, + "mean_feedback_forward_relative_error": 0.044589934091676366, + "min_feedback_forward_cosine": 0.9975100755691528 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005875167950028665, + "instruction_rms": 0.0005875167950690892, + "raw_apical_rms": 0.15631119822140535, + "teaching_rms": 0.0005875167950028665, + "traffic_rms": 0.1563089432548003 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.872849584420692e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5394041854233672e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058716752505594254, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.004544817179813981 + }, + { + "epoch": 143, + "eval_accuracy": 0.9324, + "eval_loss": 0.35895855712890623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07077816128730774, + "mean_feedback_forward_cosine": 0.9989714720032432, + "mean_feedback_forward_relative_error": 0.04458676227791743, + "min_feedback_forward_cosine": 0.9975103735923767 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006086106565886836, + "instruction_rms": 0.0006086106570433067, + "raw_apical_rms": 0.15632151175090112, + "teaching_rms": 0.0006086106565886836, + "traffic_rms": 0.15631922770924686 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.099928550109325e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5343036354917024e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005783009342741666, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.004699092442211178 + }, + { + "epoch": 144, + "eval_accuracy": 0.9296, + "eval_loss": 0.36100722045898437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07077439874410629, + "mean_feedback_forward_cosine": 0.9989716204729947, + "mean_feedback_forward_relative_error": 0.044583823006938805, + "min_feedback_forward_cosine": 0.9975106716156006 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005788136725466557, + "instruction_rms": 0.0005788136728108593, + "raw_apical_rms": 0.15631840573988948, + "teaching_rms": 0.0005788136725466557, + "traffic_rms": 0.15631622774328052 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.939645740862034e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5343369875683058e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058867044920564516, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.004655390022281144 + }, + { + "epoch": 145, + "eval_accuracy": 0.9302, + "eval_loss": 0.3650362976074219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07077116519212723, + "mean_feedback_forward_cosine": 0.9989717201753097, + "mean_feedback_forward_relative_error": 0.04458092269911008, + "min_feedback_forward_cosine": 0.9975109100341797 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005656258353966133, + "instruction_rms": 0.0005656258354606748, + "raw_apical_rms": 0.15630051021225427, + "teaching_rms": 0.0005656258353966133, + "traffic_rms": 0.15629844015168792 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.635451654801727e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5341038056365583e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0057506056176664455, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.004451047736489111 + }, + { + "epoch": 146, + "eval_accuracy": 0.9322, + "eval_loss": 0.3686236083984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07076733559370041, + "mean_feedback_forward_cosine": 0.9989718556404114, + "mean_feedback_forward_relative_error": 0.04457757064903324, + "min_feedback_forward_cosine": 0.9975112676620483 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0006118507959450226, + "instruction_rms": 0.0006118507956350694, + "raw_apical_rms": 0.15629460654171135, + "teaching_rms": 0.0006118507959450226, + "traffic_rms": 0.15629206355538705 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.179304617466187e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355107701321243e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005792362550473921, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.004486601050115294 + }, + { + "epoch": 147, + "eval_accuracy": 0.9312, + "eval_loss": 0.3708784484863281, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07076365500688553, + "mean_feedback_forward_cosine": 0.9989719791845841, + "mean_feedback_forward_relative_error": 0.044574448466300966, + "min_feedback_forward_cosine": 0.9975113868713379 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005917439001439649, + "instruction_rms": 0.0005917439003842913, + "raw_apical_rms": 0.1562591996962405, + "teaching_rms": 0.0005917439001439649, + "traffic_rms": 0.15625692916408343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 1.0164442443283406e-07, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5339368077170327e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005743765676402482, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.004553148381991519 + }, + { + "epoch": 148, + "eval_accuracy": 0.933, + "eval_loss": 0.35687979736328124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075996696949005, + "mean_feedback_forward_cosine": 0.9989721233194525, + "mean_feedback_forward_relative_error": 0.04457118883728981, + "min_feedback_forward_cosine": 0.9975116848945618 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000569190896028838, + "instruction_rms": 0.0005691908960499179, + "raw_apical_rms": 0.15621313399809966, + "teaching_rms": 0.000569190896028838, + "traffic_rms": 0.15621085011549987 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.872203073271521e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5366421053637076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811858511124198, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.004216510014779245 + }, + { + "epoch": 149, + "eval_accuracy": 0.9332, + "eval_loss": 0.36613702392578124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707562193274498, + "mean_feedback_forward_cosine": 0.9989722728729248, + "mean_feedback_forward_relative_error": 0.04456792431460185, + "min_feedback_forward_cosine": 0.9975119233131409 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005833205318357344, + "instruction_rms": 0.0005833205316495131, + "raw_apical_rms": 0.15622415389041572, + "teaching_rms": 0.0005833205318357344, + "traffic_rms": 0.15622189222459812 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.04262541009848e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5417710321320986e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811621761220833, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.004337725640171104 + }, + { + "epoch": 150, + "eval_accuracy": 0.9306, + "eval_loss": 0.36401038818359377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075214385986328, + "mean_feedback_forward_cosine": 0.998972417007793, + "mean_feedback_forward_relative_error": 0.04456454075195573, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005859148486319701, + "instruction_rms": 0.0005859148484117649, + "raw_apical_rms": 0.15617846671077237, + "teaching_rms": 0.0005859148486319701, + "traffic_rms": 0.15617611209861504 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.601874247280648e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355904813225256e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005827398066955642, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.004388301985752251 + }, + { + "epoch": 151, + "eval_accuracy": 0.9318, + "eval_loss": 0.36422869873046876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707520842552185, + "mean_feedback_forward_cosine": 0.9989724159240723, + "mean_feedback_forward_relative_error": 0.044564471613954416, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005565280756125213, + "instruction_rms": 0.0005565280757680321, + "raw_apical_rms": 0.15618927699685, + "teaching_rms": 0.0005565280756125213, + "traffic_rms": 0.156187117889629 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.69225741489754e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5346876037915427e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005892246531822571, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.003987497380210293 + }, + { + "epoch": 152, + "eval_accuracy": 0.9332, + "eval_loss": 0.36020493774414064, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075206190347672, + "mean_feedback_forward_cosine": 0.9989724224263972, + "mean_feedback_forward_relative_error": 0.044564413160763004, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005095751335896143, + "instruction_rms": 0.0005095751334227084, + "raw_apical_rms": 0.15618482505259446, + "teaching_rms": 0.0005095751335896143, + "traffic_rms": 0.15618299150455436 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.206531111274671e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.539423829837073e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005756206280104596, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.0036039010546687574 + }, + { + "epoch": 153, + "eval_accuracy": 0.9316, + "eval_loss": 0.3657336242675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075202465057373, + "mean_feedback_forward_cosine": 0.9989724180915139, + "mean_feedback_forward_relative_error": 0.044564356282353404, + "min_feedback_forward_cosine": 0.99751216173172 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0005144573056545982, + "instruction_rms": 0.0005144573057573081, + "raw_apical_rms": 0.15618785228516027, + "teaching_rms": 0.0005144573056545982, + "traffic_rms": 0.15618585031669577 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.958780846526679e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5397136800678003e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005782406628060175, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.003619852273642189 + }, + { + "epoch": 154, + "eval_accuracy": 0.9312, + "eval_loss": 0.361766259765625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075200229883194, + "mean_feedback_forward_cosine": 0.9989724180915139, + "mean_feedback_forward_relative_error": 0.04456430270590565, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00047462126330647184, + "instruction_rms": 0.00047462126314686654, + "raw_apical_rms": 0.1561874935552836, + "teaching_rms": 0.00047462126330647184, + "traffic_rms": 0.15618592888199648 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.252059930351606e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5359469216651334e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005833945475461095, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.003418941561298238 + }, + { + "epoch": 155, + "eval_accuracy": 0.9322, + "eval_loss": 0.3601679626464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075197249650955, + "mean_feedback_forward_cosine": 0.9989724332636053, + "mean_feedback_forward_relative_error": 0.044564251889559356, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004348900655850833, + "instruction_rms": 0.00043489006562027013, + "raw_apical_rms": 0.15618763321450302, + "teaching_rms": 0.0004348900655850833, + "traffic_rms": 0.15618636070868136 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.102023857896711e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360289173569167e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005871366295406924, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.0030920101912381747 + }, + { + "epoch": 156, + "eval_accuracy": 0.9334, + "eval_loss": 0.36134964599609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075195014476776, + "mean_feedback_forward_cosine": 0.9989724256775596, + "mean_feedback_forward_relative_error": 0.04456419878723947, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004496863362942985, + "instruction_rms": 0.000449686335963409, + "raw_apical_rms": 0.15618638196317103, + "teaching_rms": 0.0004496863362942985, + "traffic_rms": 0.15618490617362066 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.387681139585766e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5328539620243115e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005807898817690802, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.0033026031192392113 + }, + { + "epoch": 157, + "eval_accuracy": 0.933, + "eval_loss": 0.36484820556640624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075192034244537, + "mean_feedback_forward_cosine": 0.9989724191752347, + "mean_feedback_forward_relative_error": 0.04456414290788499, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00048817587151015504, + "instruction_rms": 0.0004881758713967517, + "raw_apical_rms": 0.15619299370344406, + "teaching_rms": 0.00048817587151015504, + "traffic_rms": 0.1561914335804525 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.844036448683255e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5339947055639765e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058057200265357834, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.0035868339443372354 + }, + { + "epoch": 158, + "eval_accuracy": 0.9332, + "eval_loss": 0.3618963806152344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075189054012299, + "mean_feedback_forward_cosine": 0.9989724245938387, + "mean_feedback_forward_relative_error": 0.044564094834707, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042458779175761664, + "instruction_rms": 0.00042458779165763, + "raw_apical_rms": 0.15619627830489188, + "teaching_rms": 0.00042458779175761664, + "traffic_rms": 0.1561949235736182 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 6.991372458742262e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536304876525867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058730891637487695, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0031154596315541615 + }, + { + "epoch": 159, + "eval_accuracy": 0.933, + "eval_loss": 0.3608845458984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707518681883812, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.04456403861668977, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004903175973574765, + "instruction_rms": 0.0004903175975856101, + "raw_apical_rms": 0.15619783319341446, + "teaching_rms": 0.0004903175973574765, + "traffic_rms": 0.15619598374444096 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.862629385930632e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5349964498701008e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005771215163845155, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.003495788243899329 + }, + { + "epoch": 160, + "eval_accuracy": 0.9338, + "eval_loss": 0.36179160766601565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075183838605881, + "mean_feedback_forward_cosine": 0.9989724289287221, + "mean_feedback_forward_relative_error": 0.04456398204307665, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004912375217397902, + "instruction_rms": 0.000491237521712185, + "raw_apical_rms": 0.15619367231437742, + "teaching_rms": 0.0004912375217397902, + "traffic_rms": 0.15619195051556656 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.83362699760437e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5367519236755603e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005842472803512416, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.003658046741928491 + }, + { + "epoch": 161, + "eval_accuracy": 0.9332, + "eval_loss": 0.36055188598632815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075180858373642, + "mean_feedback_forward_cosine": 0.9989724256775596, + "mean_feedback_forward_relative_error": 0.04456392541866411, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004898777905411831, + "instruction_rms": 0.0004898777904168807, + "raw_apical_rms": 0.15619185592653934, + "teaching_rms": 0.0004898777905411831, + "traffic_rms": 0.15619014678894416 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.207485370168797e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5398887761782835e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811061620240841, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.003570969367016935 + }, + { + "epoch": 162, + "eval_accuracy": 0.933, + "eval_loss": 0.35955615844726563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075179368257523, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.04456387668509375, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00039760659487526136, + "instruction_rms": 0.0003976065949684868, + "raw_apical_rms": 0.1561934914480541, + "teaching_rms": 0.00039760659487526136, + "traffic_rms": 0.15619229634519607 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.611171616237698e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5381311905987076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005755785324461161, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.0027860320918675924 + }, + { + "epoch": 163, + "eval_accuracy": 0.9334, + "eval_loss": 0.3592337158203125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075176388025284, + "mean_feedback_forward_cosine": 0.9989724386822093, + "mean_feedback_forward_relative_error": 0.044563823074779726, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004863620929113679, + "instruction_rms": 0.00048636209252903706, + "raw_apical_rms": 0.1561965802615731, + "teaching_rms": 0.0004863620929113679, + "traffic_rms": 0.1561948537767751 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.055869133409942e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5380353651053233e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00577728433299945, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.0035860527916759667 + }, + { + "epoch": 164, + "eval_accuracy": 0.9328, + "eval_loss": 0.3621196960449219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075174152851105, + "mean_feedback_forward_cosine": 0.9989724430170926, + "mean_feedback_forward_relative_error": 0.044563774375075645, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004124481990639766, + "instruction_rms": 0.0004124481989203935, + "raw_apical_rms": 0.1561922838465203, + "teaching_rms": 0.0004124481990639766, + "traffic_rms": 0.1561910573164225 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.283202473397978e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536488139497506e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00579818694022861, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.0028164971362799407 + }, + { + "epoch": 165, + "eval_accuracy": 0.932, + "eval_loss": 0.3607469970703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075171172618866, + "mean_feedback_forward_cosine": 0.9989724549380216, + "mean_feedback_forward_relative_error": 0.04456372271207246, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00044451223726630213, + "instruction_rms": 0.0004445122374880904, + "raw_apical_rms": 0.1561968047277056, + "teaching_rms": 0.00044451223726630213, + "traffic_rms": 0.15619524641360857 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.525871836833176e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5335534097653396e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005788216628066704, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.0030991528306156396 + }, + { + "epoch": 166, + "eval_accuracy": 0.9334, + "eval_loss": 0.3609993591308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075169682502747, + "mean_feedback_forward_cosine": 0.9989724267612804, + "mean_feedback_forward_relative_error": 0.04456367130306634, + "min_feedback_forward_cosine": 0.9975121021270752 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003992931518194371, + "instruction_rms": 0.00039929315184112865, + "raw_apical_rms": 0.1561993358125229, + "teaching_rms": 0.0003992931518194371, + "traffic_rms": 0.15619810784919824 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.569396214852532e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535884391979886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005841754779233307, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.0029006872836707365 + }, + { + "epoch": 167, + "eval_accuracy": 0.9334, + "eval_loss": 0.3611225158691406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075167447328568, + "mean_feedback_forward_cosine": 0.9989724278450012, + "mean_feedback_forward_relative_error": 0.044563623636283654, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00040215078600330475, + "instruction_rms": 0.00040215078585350195, + "raw_apical_rms": 0.15619606264845814, + "teaching_rms": 0.00040215078600330475, + "traffic_rms": 0.15619477717082902 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.738526702245508e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536509695498972e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005817911546870289, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.0028576889017182918 + }, + { + "epoch": 168, + "eval_accuracy": 0.9326, + "eval_loss": 0.3629449462890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075164467096329, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.04456357481804761, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003998887652516597, + "instruction_rms": 0.0003998887651836683, + "raw_apical_rms": 0.15619790996399205, + "teaching_rms": 0.0003998887652516597, + "traffic_rms": 0.1561966645374295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.799268265604951e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5362143182145867e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005799121166191996, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.002924299320164654 + }, + { + "epoch": 169, + "eval_accuracy": 0.9324, + "eval_loss": 0.36294425659179685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707516148686409, + "mean_feedback_forward_cosine": 0.9989724300124428, + "mean_feedback_forward_relative_error": 0.044563523358242076, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00044473610910030826, + "instruction_rms": 0.00044473610878450993, + "raw_apical_rms": 0.1562011833622537, + "teaching_rms": 0.00044473610910030826, + "traffic_rms": 0.15619974547398885 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.176765220468951e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538507105205204e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00591410992160838, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.0030470485738996003 + }, + { + "epoch": 170, + "eval_accuracy": 0.933, + "eval_loss": 0.3620413635253906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707515999674797, + "mean_feedback_forward_cosine": 0.998972435431047, + "mean_feedback_forward_relative_error": 0.044563474506139754, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042758043344175986, + "instruction_rms": 0.000427580433318418, + "raw_apical_rms": 0.15619711451121107, + "teaching_rms": 0.00042758043344175986, + "traffic_rms": 0.15619572164224343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.376089593615688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5369646367400142e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005832750105369693, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.0029799130838985242 + }, + { + "epoch": 171, + "eval_accuracy": 0.9328, + "eval_loss": 0.36404056396484374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075157761573792, + "mean_feedback_forward_cosine": 0.9989724484356967, + "mean_feedback_forward_relative_error": 0.04456342599270019, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004256224274186385, + "instruction_rms": 0.00042562242739830307, + "raw_apical_rms": 0.1562026264795551, + "teaching_rms": 0.0004256224274186385, + "traffic_rms": 0.1562012385918873 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.67029391113283e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535564534471448e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005755436385765845, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.003123174267510573 + }, + { + "epoch": 172, + "eval_accuracy": 0.9336, + "eval_loss": 0.3628458374023438, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075154781341553, + "mean_feedback_forward_cosine": 0.998972440849651, + "mean_feedback_forward_relative_error": 0.044563368250700564, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004919505852206358, + "instruction_rms": 0.0004919505851019495, + "raw_apical_rms": 0.15619272672224851, + "teaching_rms": 0.0004919505852206358, + "traffic_rms": 0.15619057426172062 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.396567897581917e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5318478637538845e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005892030299470198, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.003438941920403805 + }, + { + "epoch": 173, + "eval_accuracy": 0.934, + "eval_loss": 0.361624853515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075151801109314, + "mean_feedback_forward_cosine": 0.9989724375984885, + "mean_feedback_forward_relative_error": 0.044563313929194756, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004732160200341671, + "instruction_rms": 0.0004732160199897636, + "raw_apical_rms": 0.15619125526362268, + "teaching_rms": 0.0004732160200341671, + "traffic_rms": 0.15618951902258618 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.914952876968528e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534674797093549e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005832087390802645, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.0033569188530039457 + }, + { + "epoch": 174, + "eval_accuracy": 0.9334, + "eval_loss": 0.3606834411621094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075148820877075, + "mean_feedback_forward_cosine": 0.99897245277058, + "mean_feedback_forward_relative_error": 0.04456325716931712, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000504879789185604, + "instruction_rms": 0.0005048797891479873, + "raw_apical_rms": 0.15618947941493014, + "teaching_rms": 0.000504879789185604, + "traffic_rms": 0.15618761728279687 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.534652723961699e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534835066469787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005824511633885487, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.003320137251706587 + }, + { + "epoch": 175, + "eval_accuracy": 0.9328, + "eval_loss": 0.3632712341308594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075145095586777, + "mean_feedback_forward_cosine": 0.9989724516868591, + "mean_feedback_forward_relative_error": 0.04456320562484589, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004231173779611415, + "instruction_rms": 0.00042311737763738125, + "raw_apical_rms": 0.15618914205308448, + "teaching_rms": 0.0004231173779611415, + "traffic_rms": 0.1561877039875195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.140283031503714e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538623372146383e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005876878379789779, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.002933721011566619 + }, + { + "epoch": 176, + "eval_accuracy": 0.9332, + "eval_loss": 0.36194794311523437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075142860412598, + "mean_feedback_forward_cosine": 0.9989724419333718, + "mean_feedback_forward_relative_error": 0.04456315531649373, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00045991324012191817, + "instruction_rms": 0.0004599132399693753, + "raw_apical_rms": 0.15619017173639713, + "teaching_rms": 0.00045991324012191817, + "traffic_rms": 0.15618847591366844 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.06208717563095e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5391214217003264e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005789111506779168, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.00342260849043313 + }, + { + "epoch": 177, + "eval_accuracy": 0.9334, + "eval_loss": 0.3670292785644531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075142115354538, + "mean_feedback_forward_cosine": 0.9989724592729048, + "mean_feedback_forward_relative_error": 0.04456310263750228, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00040664537186046166, + "instruction_rms": 0.0004066453716585905, + "raw_apical_rms": 0.15619706280682688, + "teaching_rms": 0.00040664537186046166, + "traffic_rms": 0.1561957847530296 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.209095625488771e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534770102298202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005823902930677908, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.002844909175568157 + }, + { + "epoch": 178, + "eval_accuracy": 0.9332, + "eval_loss": 0.3643357971191406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707513839006424, + "mean_feedback_forward_cosine": 0.9989724495194175, + "mean_feedback_forward_relative_error": 0.044563054530458016, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004049486367224079, + "instruction_rms": 0.000404948636576789, + "raw_apical_rms": 0.15619552088355443, + "teaching_rms": 0.0004049486367224079, + "traffic_rms": 0.15619416135052885 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.023897552471681e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537750164351138e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0058023818346295665, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.0029311306476593017 + }, + { + "epoch": 179, + "eval_accuracy": 0.9342, + "eval_loss": 0.36326700439453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707513689994812, + "mean_feedback_forward_cosine": 0.9989724386822093, + "mean_feedback_forward_relative_error": 0.04456300999630581, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003612708095619896, + "instruction_rms": 0.00036127080952046164, + "raw_apical_rms": 0.15619400731217237, + "teaching_rms": 0.0003612708095619896, + "traffic_rms": 0.15619300828769123 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.921884931898603e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360207882004023e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005841793443784828, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.0024818976380758815 + }, + { + "epoch": 180, + "eval_accuracy": 0.9336, + "eval_loss": 0.36097264404296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075133919715881, + "mean_feedback_forward_cosine": 0.9989724603566257, + "mean_feedback_forward_relative_error": 0.04456295865503224, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004035473568848594, + "instruction_rms": 0.00040354735642829235, + "raw_apical_rms": 0.15619308976432028, + "teaching_rms": 0.0004035473568848594, + "traffic_rms": 0.15619173579574833 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.915981742849908e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.534677185287896e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005847928120514314, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.002834540325920615 + }, + { + "epoch": 181, + "eval_accuracy": 0.9336, + "eval_loss": 0.36409304809570314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075133174657822, + "mean_feedback_forward_cosine": 0.9989724506031383, + "mean_feedback_forward_relative_error": 0.04456291464580731, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003687424852903932, + "instruction_rms": 0.0003687424851625734, + "raw_apical_rms": 0.15620027226349922, + "teaching_rms": 0.0003687424852903932, + "traffic_rms": 0.1561991959946112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.412849273829124e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5348776588620148e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005791613943938917, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.0028244014726537797 + }, + { + "epoch": 182, + "eval_accuracy": 0.933, + "eval_loss": 0.36588348388671876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075129449367523, + "mean_feedback_forward_cosine": 0.9989724571054632, + "mean_feedback_forward_relative_error": 0.04456286396492611, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004362118070622091, + "instruction_rms": 0.0004362118068422204, + "raw_apical_rms": 0.1561978659340271, + "teaching_rms": 0.0004362118070622091, + "traffic_rms": 0.1561963286981839 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.406312908462496e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5351237634552045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00577301282666261, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.003047733818511996 + }, + { + "epoch": 183, + "eval_accuracy": 0.9336, + "eval_loss": 0.3611108093261719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075125724077225, + "mean_feedback_forward_cosine": 0.9989724571054632, + "mean_feedback_forward_relative_error": 0.04456281140446663, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004148569442338212, + "instruction_rms": 0.0004148569440265609, + "raw_apical_rms": 0.15619684428622288, + "teaching_rms": 0.0004148569442338212, + "traffic_rms": 0.15619553491803845 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.257590001723656e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5350059929103585e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005837293103899293, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.0029103908619326022 + }, + { + "epoch": 184, + "eval_accuracy": 0.9332, + "eval_loss": 0.36385486450195315, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075124233961105, + "mean_feedback_forward_cosine": 0.9989724668589506, + "mean_feedback_forward_relative_error": 0.04456276637925343, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00039054727424933097, + "instruction_rms": 0.00039054727427044125, + "raw_apical_rms": 0.15619733470302585, + "teaching_rms": 0.00039054727424933097, + "traffic_rms": 0.15619622904411004 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 9.030456027403488e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535347866608352e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0057423141327328725, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.002705052987134291 + }, + { + "epoch": 185, + "eval_accuracy": 0.9336, + "eval_loss": 0.3632457275390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075121253728867, + "mean_feedback_forward_cosine": 0.9989724516868591, + "mean_feedback_forward_relative_error": 0.04456271618943323, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003811323365473934, + "instruction_rms": 0.0003811323366254667, + "raw_apical_rms": 0.15619683555499747, + "teaching_rms": 0.0003811323365473934, + "traffic_rms": 0.15619564626965682 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.183086836448638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5369952899846847e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005860040617343112, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.002638033893249101 + }, + { + "epoch": 186, + "eval_accuracy": 0.9334, + "eval_loss": 0.36154000244140627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075117528438568, + "mean_feedback_forward_cosine": 0.9989724451845342, + "mean_feedback_forward_relative_error": 0.044562668624249374, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042452841534813666, + "instruction_rms": 0.00042452841530739896, + "raw_apical_rms": 0.1561949806606766, + "teaching_rms": 0.00042452841534813666, + "traffic_rms": 0.15619353826659707 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.10155134697743e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355028719230432e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005811353048002794, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.0030590131243069967 + }, + { + "epoch": 187, + "eval_accuracy": 0.935, + "eval_loss": 0.36507200927734373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075115293264389, + "mean_feedback_forward_cosine": 0.9989724625240672, + "mean_feedback_forward_relative_error": 0.04456262127919631, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00042261649333655434, + "instruction_rms": 0.0004226164933721387, + "raw_apical_rms": 0.156190922916446, + "teaching_rms": 0.00042261649333655434, + "traffic_rms": 0.15618943955163558 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.915867618000448e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5360549223900153e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005852771622581931, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.0030017898343917397 + }, + { + "epoch": 188, + "eval_accuracy": 0.9338, + "eval_loss": 0.36510807495117187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707511380314827, + "mean_feedback_forward_cosine": 0.9989724625240672, + "mean_feedback_forward_relative_error": 0.04456257711757313, + "min_feedback_forward_cosine": 0.9975124001502991 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003916084172831918, + "instruction_rms": 0.0003916084172182954, + "raw_apical_rms": 0.15619046241567305, + "teaching_rms": 0.0003916084172831918, + "traffic_rms": 0.15618924894105277 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.255608380773992e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5390736932889114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005833163123987753, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.0026285000529968077 + }, + { + "epoch": 189, + "eval_accuracy": 0.9338, + "eval_loss": 0.365090625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0707511231303215, + "mean_feedback_forward_cosine": 0.9989724560217423, + "mean_feedback_forward_relative_error": 0.04456252743574706, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00040755755962995714, + "instruction_rms": 0.00040755755963706907, + "raw_apical_rms": 0.15619771865988386, + "teaching_rms": 0.00040755755962995714, + "traffic_rms": 0.15619639373621746 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.970876936269633e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.538003063450597e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005825029032007757, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.002841304802708328 + }, + { + "epoch": 190, + "eval_accuracy": 0.9326, + "eval_loss": 0.36245419311523436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075109332799911, + "mean_feedback_forward_cosine": 0.9989724560217423, + "mean_feedback_forward_relative_error": 0.044562478295781395, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00038014948555528357, + "instruction_rms": 0.00038014948527374315, + "raw_apical_rms": 0.15619197940003895, + "teaching_rms": 0.00038014948555528357, + "traffic_rms": 0.15619094840030273 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.455015094137707e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5355496912772462e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005919079384556868, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.0027392928045656944 + }, + { + "epoch": 191, + "eval_accuracy": 0.9334, + "eval_loss": 0.36492646484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075107097625732, + "mean_feedback_forward_cosine": 0.9989724495194175, + "mean_feedback_forward_relative_error": 0.044562429782341824, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004018901788420274, + "instruction_rms": 0.00040189017884162333, + "raw_apical_rms": 0.15619137479569864, + "teaching_rms": 0.0004018901788420274, + "traffic_rms": 0.15619018727307954 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.211887265157202e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.535846275555213e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005815387216343496, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.0027275119118599428 + }, + { + "epoch": 192, + "eval_accuracy": 0.9324, + "eval_loss": 0.3650273254394531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075105607509613, + "mean_feedback_forward_cosine": 0.9989724679426714, + "mean_feedback_forward_relative_error": 0.04456238462166353, + "min_feedback_forward_cosine": 0.9975124001502991 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00038408126969583303, + "instruction_rms": 0.00038408126960202184, + "raw_apical_rms": 0.15618522479586572, + "teaching_rms": 0.00038408126969583303, + "traffic_rms": 0.1561841321759944 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.835953314521248e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5386043751217692e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005837757667829218, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.002694618460794704 + }, + { + "epoch": 193, + "eval_accuracy": 0.9328, + "eval_loss": 0.36529390869140627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075103372335434, + "mean_feedback_forward_cosine": 0.9989724722775546, + "mean_feedback_forward_relative_error": 0.0445623412897641, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003657145414101729, + "instruction_rms": 0.0003657145415227043, + "raw_apical_rms": 0.15618295268645463, + "teaching_rms": 0.0003657145414101729, + "traffic_rms": 0.1561818427703464 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.366949716924864e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5362354929076727e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00578793646396616, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.002424141924393674 + }, + { + "epoch": 194, + "eval_accuracy": 0.9334, + "eval_loss": 0.36309793090820314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075101137161255, + "mean_feedback_forward_cosine": 0.9989724668589506, + "mean_feedback_forward_relative_error": 0.04456229038875211, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004149085355985523, + "instruction_rms": 0.00041490853532579494, + "raw_apical_rms": 0.15619277628176287, + "teaching_rms": 0.0004149085355985523, + "traffic_rms": 0.15619127224170495 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.803163271091762e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5327822541004354e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005807142875516983, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.0029466939909590616 + }, + { + "epoch": 195, + "eval_accuracy": 0.933, + "eval_loss": 0.36418885498046877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075099647045135, + "mean_feedback_forward_cosine": 0.9989724841984835, + "mean_feedback_forward_relative_error": 0.044562242552638055, + "min_feedback_forward_cosine": 0.9975123405456543 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003596243837953494, + "instruction_rms": 0.000359624383627786, + "raw_apical_rms": 0.15619140879149387, + "teaching_rms": 0.0003596243837953494, + "traffic_rms": 0.15619041368152756 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.270116159612371e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.532461871659178e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005761611191290388, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.0023758877529866168 + }, + { + "epoch": 196, + "eval_accuracy": 0.933, + "eval_loss": 0.3622114013671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075097411870956, + "mean_feedback_forward_cosine": 0.9989724657752297, + "mean_feedback_forward_relative_error": 0.04456219828941605, + "min_feedback_forward_cosine": 0.9975124001502991 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0003788401002222492, + "instruction_rms": 0.00037884010010704724, + "raw_apical_rms": 0.15619391125265422, + "teaching_rms": 0.0003788401002222492, + "traffic_rms": 0.15619275072949035 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 5.44583045325453e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.536320428224063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005971374289712673, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.002593078167364001 + }, + { + "epoch": 197, + "eval_accuracy": 0.9332, + "eval_loss": 0.3630668518066406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075095176696777, + "mean_feedback_forward_cosine": 0.9989724690263921, + "mean_feedback_forward_relative_error": 0.04456215097822926, + "min_feedback_forward_cosine": 0.9975122809410095 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004019384723443367, + "instruction_rms": 0.00040193847233358605, + "raw_apical_rms": 0.1561901152440423, + "teaching_rms": 0.0004019384723443367, + "traffic_rms": 0.15618888084324176 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 6.272781405414207e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5356222845259446e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005798918022251434, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.003016370440295173 + }, + { + "epoch": 198, + "eval_accuracy": 0.934, + "eval_loss": 0.36500486450195313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075092196464539, + "mean_feedback_forward_cosine": 0.9989724636077881, + "mean_feedback_forward_relative_error": 0.044562104344367984, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004087684824610619, + "instruction_rms": 0.0004087684824830167, + "raw_apical_rms": 0.15618832771598362, + "teaching_rms": 0.0004087684824610619, + "traffic_rms": 0.15618686884398414 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 7.957885372888995e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.537445699149463e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005813321412645424, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.002747606110189938 + }, + { + "epoch": 199, + "eval_accuracy": 0.9334, + "eval_loss": 0.36417661743164065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.070750892162323, + "mean_feedback_forward_cosine": 0.9989724657752297, + "mean_feedback_forward_relative_error": 0.04456205044619062, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004871525299883906, + "instruction_rms": 0.00048715252976229, + "raw_apical_rms": 0.15619343033867647, + "teaching_rms": 0.0004871525299883906, + "traffic_rms": 0.15619158905469818 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488560497760773, + "maximum_absolute_post_projection_soma_slope": 8.859503708436023e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5383053457659095e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005776670791695986, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.003350807800433702 + }, + { + "epoch": 200, + "eval_accuracy": 0.9322, + "eval_loss": 0.3675199523925781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07075086236000061, + "mean_feedback_forward_cosine": 0.9989724636077881, + "mean_feedback_forward_relative_error": 0.044561998918652534, + "min_feedback_forward_cosine": 0.9975122213363647 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0004177410640374182, + "instruction_rms": 0.0004177410639151842, + "raw_apical_rms": 0.15618223981341522, + "teaching_rms": 0.0004177410640374182, + "traffic_rms": 0.15618101233638496 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.15488561987876892, + "maximum_absolute_post_projection_soma_slope": 8.024044007015618e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.5338302460993663e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.005836083964773954, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.002966017501304547 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9322, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.3675199523925781, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 3262279168, + "peak_memory_reserved_bytes": 5276434432, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 9.810454315584138e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 9.810454315584138e-08, + "min_residual_soma_slope": -8.382607319390445e-08, + "mse": 1.5715674817957412e-17, + "observations": 64, + "residual_traffic_rms_ratio": 7.224132997837907e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.224132997837907e-08, + "reuses_traffic_calibration_forward": true, + "steps": 1, + "task_loader_state_restored": true + }, + "protocol_family": "oral_a_cifar_local_resnet_development", + "provenance": { + "git_commit": "f55872beaee658ff76ac15fd421b73017820c320", + "git_tracked_dirty": false + }, + "schema_version": 1, + "split": { + "cifar_source_files": [ + { + "bytes": 31035704, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_1", + "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd" + }, + { + "bytes": 31035320, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_2", + "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8" + }, + { + "bytes": 31035999, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_3", + "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9" + }, + { + "bytes": 31035696, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_4", + "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b" + }, + { + "bytes": 31035623, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/data_batch_5", + "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154" + }, + { + "bytes": 31035526, + "path": "/orion/u/oscarwan/scratch/sdil-data/cifar-10-batches-py/test_batch", + "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831" + } + ], + "dataset": "cifar10", + "evaluation_split": "validation", + "input_layout": "NCHW", + "input_shape": [ + 3, + 32, + 32 + ], + "loader_seed": 0, + "normalization_mean": [ + 0.49140000343322754, + 0.4821999967098236, + 0.4465000033378601 + ], + "normalization_std": [ + 0.24699999392032623, + 0.2434999942779541, + 0.26159998774528503 + ], + "split_from_training_only": true, + "split_seed": 2027, + "test_examples": 10000, + "train_examples": 45000, + "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5", + "validation_class_counts": { + "0": 500, + "1": 500, + "2": 500, + "3": 500, + "4": 500, + "5": 500, + "6": 500, + "7": 500, + "8": 500, + "9": 500 + }, + "validation_examples": 5000, + "validation_index_sha256": "8328b206a97c420e49e54e3eca4abe3274c4756b084355784ea3fb8059e4515b" + }, + "timing": { + "apical_warmup_wall_s": 0.0, + "evaluation_wall_s": 60.00945520401001, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.3673684597015381, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 8344.301742315292, + "train_wall_s": 8280.477409601212 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.046645715832710266, + 0.036221180111169815, + 0.0219095591455698, + 0.024090727791190147, + 0.014862747862935066, + 0.017354732379317284, + 0.010909760370850563, + 0.012971695512533188, + 0.00854403991252184, + 0.010942773893475533, + 0.006431682035326958, + 0.007946319878101349, + 0.005390886217355728, + 0.007816178724169731, + 0.004525847267359495, + 0.0066139050759375095, + 0.003956272266805172, + 0.005918589886277914, + 0.0034256556537002325, + 0.00842861458659172, + 0.005142349284142256, + 0.006969384849071503, + 0.004458904732018709, + 0.005985064432024956, + 0.003961027599871159, + 0.005532111972570419, + 0.0035666259936988354, + 0.004960889462381601, + 0.0032047750428318977, + 0.004740030970424414, + 0.0028284969739615917, + 0.004207335412502289, + 0.00256044021807611, + 0.0038634713273495436, + 0.0023208747152239084, + 0.0035063333343714476, + 0.0021188128739595413, + 0.005200310610234737, + 0.003467726521193981, + 0.004882926121354103, + 0.0031611102167516947, + 0.00439144903793931, + 0.00288514606654644, + 0.003956384491175413, + 0.002680722391232848, + 0.003665727796033025, + 0.002519759815186262, + 0.003549956250935793, + 0.0023743261117488146, + 0.003312949789687991, + 0.0022739002015441656, + 0.0031900741159915924, + 0.002209193306043744, + 0.00286328187212348, + 0.0022419888991862535 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 3.999999523162842, + 4.0, + 4.000000476837158, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 3.999999523162842, + 3.999999761581421, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.000000476837158, + 4.0, + 3.999999523162842, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.24335139989852905, + 0.19567525386810303, + 0.07590705901384354, + 0.12499095499515533, + 0.039654169231653214, + 0.09477696567773819, + 0.02513667196035385, + 0.0682290568947792, + 0.017793802544474602, + 0.059991367161273956, + 0.012211106717586517, + 0.04119913652539253, + 0.009280920960009098, + 0.04255608469247818, + 0.0071248686872422695, + 0.03416771814227104, + 0.005649160128086805, + 0.03203295171260834, + 0.00472416402772069, + 0.046523354947566986, + 0.009258695878088474, + 0.03640424460172653, + 0.0073813083581626415, + 0.03129300847649574, + 0.006084667518734932, + 0.02907230518758297, + 0.005167320836335421, + 0.026048656553030014, + 0.004463230259716511, + 0.02523532323539257, + 0.0037100326735526323, + 0.022375741973519325, + 0.0031846852507442236, + 0.020636072382330894, + 0.002770467661321163, + 0.01875767856836319, + 0.0024420199915766716, + 0.027892472222447395, + 0.005269602406769991, + 0.026072902604937553, + 0.0044991569593548775, + 0.023685075342655182, + 0.0038742832839488983, + 0.020799512043595314, + 0.003461548825725913, + 0.019222838804125786, + 0.0030808867886662483, + 0.018855275586247444, + 0.0027911863289773464, + 0.017468731850385666, + 0.0025647676084190607, + 0.01688026450574398, + 0.002401146572083235, + 0.015071986243128777, + 0.0023585029412060976 + ], + "unscaled_traffic_rms": [ + 0.7667219638824463, + 0.7404345870018005, + 1.1545466184616089, + 0.7709590792655945, + 1.499236822128296, + 0.7324451804161072, + 1.7360707521438599, + 0.760479211807251, + 1.9206777811050415, + 0.7296232581138611, + 2.10683012008667, + 0.7715035080909729, + 2.323427200317383, + 0.7346708178520203, + 2.5408732891082764, + 0.7742869853973389, + 2.8013172149658203, + 0.7390626668930054, + 2.9005391597747803, + 0.7246781587600708, + 2.221630096435547, + 0.7657771706581116, + 2.4163222312927246, + 0.7650353908538818, + 2.60394024848938, + 0.7611521482467651, + 2.76090931892395, + 0.7617881298065186, + 2.872157573699951, + 0.751332700252533, + 3.049565553665161, + 0.7521244287490845, + 3.2159414291381836, + 0.748877227306366, + 3.3508780002593994, + 0.7477115988731384, + 3.470590591430664, + 0.7457654476165771, + 2.632249116897583, + 0.7491189241409302, + 2.8104023933410645, + 0.7416398525238037, + 2.9787662029266357, + 0.7608610391616821, + 3.097714424133301, + 0.7627859115600586, + 3.2714734077453613, + 0.7530955672264099, + 3.4026050567626953, + 0.7586011290550232, + 3.5463645458221436, + 0.7559298872947693, + 3.6802306175231934, + 0.7598950266838074, + 3.8023931980133057 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 125043328, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 1125397954772992, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 1125389952000000, + "local_weight_correlation_macs": 1129371264000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 1129371264000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 8031084544 + }, + "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs", + "elementwise_operations_estimate": 115016770519040, + "forward_macs_per_example": 125485696, + "logical_batch_loss_queries": 0, + "mirror_probe_examples": 0, + "mirror_readout_probe_examples": 0, + "neutral_projection_observations": 9000000, + "per_example_cross_entropy_terms": 0, + "total_clean_forward_examples": 9000064, + "total_forward_equivalent_examples": 9000064, + "total_macs_estimate": 4509538465857536 + } +} |
