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-d32-lr0p1.json | |
| parent | f55872beaee658ff76ac15fd421b73017820c320 (diff) | |
experiment: add completed A6000 crossover resultsa6000-x1-results
Diffstat (limited to 'results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json')
| -rw-r--r-- | results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json | 7472 |
1 files changed, 7472 insertions, 0 deletions
diff --git a/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json b/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json new file mode 100644 index 0000000..83cf766 --- /dev/null +++ b/results/resnet_crossover/r2/resnet-r2-sdil-d32-lr0p1.json @@ -0,0 +1,7472 @@ +{ + "architecture": { + "adaptive_apical_parameters": 606208, + "adaptive_feedback_parameters": 461440, + "base_width": 16, + "blocks_per_stage": 5, + "bn_eps": 1e-05, + "bn_momentum": 0.1, + "depth": 32, + "family": "CIFAR 6n+2 ResNet, option-A shortcuts", + "fixed_feedback_parameters": 0, + "fixed_traffic_coefficients": 303104, + "forward_parameters": 464154, + "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 + ], + [ + 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 + ] + ], + "normalization": "batchnorm", + "predictor_parameters": 606208, + "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": 32, + "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-d32-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": 0.9990641891956329, + "feedback_forward_cosine": [ + 0.9995248913764954, + 0.999525785446167, + 0.9993621110916138, + 0.9992882013320923, + 0.9992794990539551, + 0.9991533756256104, + 0.9995105266571045, + 0.9992536306381226, + 0.9993960857391357, + 0.9991365671157837, + 0.9993268847465515, + 0.9995407462120056, + 0.9993659257888794, + 0.9992746114730835, + 0.9994126558303833, + 0.9992644786834717, + 0.9995049834251404, + 0.9993177652359009, + 0.9994486570358276, + 0.9992416501045227, + 0.999461829662323, + 0.9996631145477295, + 0.9995964169502258, + 0.9995673298835754, + 0.9996181726455688, + 0.9994949102401733, + 0.9995560646057129, + 0.9993234872817993, + 0.9994652271270752, + 0.9987451434135437, + 0.9999646544456482 + ], + "feedback_forward_norm_ratio": [ + 0.9988867044448853, + 1.0008950233459473, + 0.9997606873512268, + 1.0032583475112915, + 0.9985292553901672, + 1.0024914741516113, + 0.998984158039093, + 1.0015926361083984, + 0.9991874098777771, + 1.0027505159378052, + 0.9982722401618958, + 1.0011152029037476, + 0.9988278746604919, + 1.0015062093734741, + 0.9987968802452087, + 1.0027403831481934, + 0.9992345571517944, + 1.0019012689590454, + 0.9989014267921448, + 1.0029679536819458, + 0.9991740584373474, + 1.0008102655410767, + 0.9995593428611755, + 1.001561164855957, + 1.0000790357589722, + 1.0014222860336304, + 0.9998015761375427, + 1.0020161867141724, + 0.999867856502533, + 1.0024110078811646, + 1.0053294897079468 + ], + "feedback_forward_relative_error": [ + 0.030829928815364838, + 0.03082478605210781, + 0.03571289777755737, + 0.037930626422166824, + 0.037959691137075424, + 0.041274625808000565, + 0.03128883242607117, + 0.03869929164648056, + 0.03474787622690201, + 0.04170121252536774, + 0.03669923543930054, + 0.030344000086188316, + 0.0356106236577034, + 0.038146961480379105, + 0.034275274723768234, + 0.03850400820374489, + 0.031461749225854874, + 0.0370202399790287, + 0.03320717811584473, + 0.039115216583013535, + 0.03280482068657875, + 0.02598089911043644, + 0.028407160192728043, + 0.029481027275323868, + 0.027640528976917267, + 0.03183675557374954, + 0.02979353256523609, + 0.036875028163194656, + 0.032700978219509125, + 0.05021461844444275, + 0.009983262047171593 + ], + "innovation_negative_gradient_cosine": [ + 0.9991377592086792, + 0.9990687966346741, + 0.9991105794906616, + 0.9988575577735901, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993926882743835, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.9999379515647888, + 0.7521367073059082 + ], + "instruction_negative_gradient_cosine": [ + 0.999137818813324, + 0.9990687966346741, + 0.9991105794906616, + 0.9988574981689453, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993927478790283, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.999937891960144, + 0.755734920501709 + ], + "matched_negative_gradient_cosine": [ + 0.004778770729899406, + 0.005657309666275978, + 0.0010874532163143158, + 0.018822133541107178, + 0.00964898057281971, + 0.016325540840625763, + 0.017202919349074364, + 0.017850112169981003, + 0.023066185414791107, + 0.041207410395145416, + 0.020595937967300415, + -0.04097568243741989, + -0.003883502446115017, + 0.03719305247068405, + 0.01840835250914097, + 0.05979178845882416, + 0.025737924501299858, + 0.10348624736070633, + 0.022145891562104225, + 0.05590894818305969, + 0.056562818586826324, + 0.0805809497833252, + 0.044539716094732285, + 0.16223689913749695, + 0.06719361245632172, + 0.2673830986022949, + 0.09676877409219742, + 0.24156257510185242, + 0.09677475690841675, + 0.1181703507900238, + 0.12388422340154648 + ], + "max_norm_match_direction_error": 1.7881393432617188e-07, + "max_norm_match_relative_error": 1.6217698828313587e-07, + "mean_feedback_forward_cosine": 0.9994059801101685, + "mean_feedback_forward_relative_error": 0.03390557637378093, + "neutral_projection": { + "instruction_observations": 0, + "max_absolute_correction_slope": 0.06770927459001541, + "max_absolute_post_projection_soma_slope": 1.9101433323953643e-08, + "max_absolute_pre_projection_soma_slope": 0.06770927459001541, + "max_positive_post_projection_soma_slope": 8.284928298962768e-09, + "min_post_projection_soma_slope": -1.9101433323953643e-08, + "observations": 32, + "post_projection_traffic_rms_ratio": 2.5151598565298195e-08, + "pre_projection_traffic_rms_ratio": 0.011312205665592186 + }, + "normalization_state": "training_batch_stats_without_running_update", + "predictor_traffic_residual_rms_ratio": 0.011312205665559983, + "raw_negative_gradient_cosine": [ + 0.004778770264238119, + 0.005657309666275978, + 0.0010874529834836721, + 0.018822133541107178, + 0.00964897871017456, + 0.016325540840625763, + 0.017202919349074364, + 0.0178501158952713, + 0.023066185414791107, + 0.041207410395145416, + 0.020595937967300415, + -0.04097568243741989, + -0.0038835033774375916, + 0.03719305619597435, + 0.01840835250914097, + 0.05979178845882416, + 0.025737926363945007, + 0.10348625481128693, + 0.022145891562104225, + 0.05590894818305969, + 0.056562818586826324, + 0.0805809497833252, + 0.044539712369441986, + 0.16223689913749695, + 0.06719361245632172, + 0.2673830986022949, + 0.09676877409219742, + 0.24156257510185242, + 0.09677475690841675, + 0.1181703582406044, + 0.12388423085212708 + ], + "teaching_negative_gradient_cosine": [ + 0.9991377592086792, + 0.9990687966346741, + 0.9991105794906616, + 0.9988575577735901, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993926882743835, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.9999379515647888, + 0.7521367073059082 + ], + "traffic_instruction_rms_ratio": 23.54690779022645, + "used_negative_gradient_cosine": [ + 0.9991377592086792, + 0.9990687966346741, + 0.9991105794906616, + 0.9988575577735901, + 0.9991105794906616, + 0.9988821148872375, + 0.9991482496261597, + 0.9991245269775391, + 0.9991716146469116, + 0.9990301132202148, + 0.9992015361785889, + 0.9991713762283325, + 0.9992712140083313, + 0.9990156292915344, + 0.9993025660514832, + 0.9990812540054321, + 0.999332070350647, + 0.9991285800933838, + 0.9993730783462524, + 0.9991375207901001, + 0.99940425157547, + 0.9995187520980835, + 0.9994936585426331, + 0.9993926882743835, + 0.9995537996292114, + 0.9995598793029785, + 0.9996328353881836, + 0.999687671661377, + 0.9998021125793457, + 0.9999379515647888, + 0.7521367073059082 + ], + "wall_s": 0.1316971778869629 + }, + "epochs": [ + { + "epoch": 1, + "eval_accuracy": 0.388, + "eval_loss": 1.6842045654296876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3977231979370117, + "mean_feedback_forward_cosine": 0.10176758456134027, + "mean_feedback_forward_relative_error": 1.3543651142428, + "min_feedback_forward_cosine": 0.03091779351234436 + }, + "lr": 0.02, + "mixed_apical": { + "innovation_rms": 0.004567544741317272, + "instruction_rms": 0.004567544740593721, + "raw_apical_rms": 0.03513459748022013, + "teaching_rms": 0.004567544741317272, + "traffic_rms": 0.03482145024585586 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677093043923378, + "maximum_absolute_post_projection_soma_slope": 3.303899731577076e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 3.048501115545045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007256994196668291, + "minimum_observations": 72 + }, + "output_lr": 0.02, + "step": 352, + "train_examples": 45000, + "train_loss": 2.0257865279939438 + }, + { + "epoch": 2, + "eval_accuracy": 0.474, + "eval_loss": 1.59408642578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3580535650253296, + "mean_feedback_forward_cosine": 0.24024918771559192, + "mean_feedback_forward_relative_error": 1.2470388316339063, + "min_feedback_forward_cosine": 0.09387169778347015 + }, + "lr": 0.04, + "mixed_apical": { + "innovation_rms": 0.005307031705626174, + "instruction_rms": 0.005307031705694782, + "raw_apical_rms": 0.040016573325931976, + "teaching_rms": 0.005307031705626174, + "traffic_rms": 0.039657659576885225 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.864828895279061e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.9509758007218805e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009549755634508534, + "minimum_observations": 72 + }, + "output_lr": 0.04, + "step": 704, + "train_examples": 45000, + "train_loss": 1.5833565533955891 + }, + { + "epoch": 3, + "eval_accuracy": 0.5938, + "eval_loss": 1.21504912109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.3161413669586182, + "mean_feedback_forward_cosine": 0.3675699810827932, + "mean_feedback_forward_relative_error": 1.1383580238588396, + "min_feedback_forward_cosine": 0.15631060302257538 + }, + "lr": 0.06000000000000001, + "mixed_apical": { + "innovation_rms": 0.005956704153025323, + "instruction_rms": 0.005956704150186274, + "raw_apical_rms": 0.04542083796175327, + "teaching_rms": 0.005956704153025323, + "traffic_rms": 0.04502470954108027 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4744238658058748e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8605195850272107e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00897240334648494, + "minimum_observations": 72 + }, + "output_lr": 0.06000000000000001, + "step": 1056, + "train_examples": 45000, + "train_loss": 1.1766535493850707 + }, + { + "epoch": 4, + "eval_accuracy": 0.6, + "eval_loss": 1.2461006103515626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2697638273239136, + "mean_feedback_forward_cosine": 0.46007657051086426, + "mean_feedback_forward_relative_error": 1.0509521999666769, + "min_feedback_forward_cosine": 0.2210283875465393 + }, + "lr": 0.08, + "mixed_apical": { + "innovation_rms": 0.005509644266008199, + "instruction_rms": 0.005509644266709996, + "raw_apical_rms": 0.04825561232693312, + "teaching_rms": 0.005509644266008199, + "traffic_rms": 0.047937556891173266 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.671688020290276e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.8255515906846483e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007464922171899973, + "minimum_observations": 72 + }, + "output_lr": 0.08, + "step": 1408, + "train_examples": 45000, + "train_loss": 0.9523900893847148 + }, + { + "epoch": 5, + "eval_accuracy": 0.7276, + "eval_loss": 0.7826447875976562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.2237781286239624, + "mean_feedback_forward_cosine": 0.5366203938761065, + "mean_feedback_forward_relative_error": 0.973043107217358, + "min_feedback_forward_cosine": 0.27918457984924316 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005091983841681295, + "instruction_rms": 0.00509198384066812, + "raw_apical_rms": 0.05179744654933709, + "teaching_rms": 0.005091983841681295, + "traffic_rms": 0.05154422627249688 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.839830059997439e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7884721745800226e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.007986453419867427, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 1760, + "train_examples": 45000, + "train_loss": 0.815070916832818 + }, + { + "epoch": 6, + "eval_accuracy": 0.7534, + "eval_loss": 0.7137961181640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1856058835983276, + "mean_feedback_forward_cosine": 0.5884564268973566, + "mean_feedback_forward_relative_error": 0.9165485866608158, + "min_feedback_forward_cosine": 0.32551515102386475 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.005027530190887641, + "instruction_rms": 0.005027530189482462, + "raw_apical_rms": 0.05350177493955319, + "teaching_rms": 0.005027530190887641, + "traffic_rms": 0.05326255006739937 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9742937002197323e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7749626564124105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008138300633405591, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2112, + "train_examples": 45000, + "train_loss": 0.7121988859388564 + }, + { + "epoch": 7, + "eval_accuracy": 0.719, + "eval_loss": 0.8742424682617187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1527107954025269, + "mean_feedback_forward_cosine": 0.6276193309214807, + "mean_feedback_forward_relative_error": 0.8711237195999392, + "min_feedback_forward_cosine": 0.3639024794101715 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004874586884066469, + "instruction_rms": 0.004874586884370733, + "raw_apical_rms": 0.05472647430139628, + "teaching_rms": 0.004874586884066469, + "traffic_rms": 0.054506565328151145 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6285386311997172e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.760684749344245e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008401987143937221, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2464, + "train_examples": 45000, + "train_loss": 0.6420152118682861 + }, + { + "epoch": 8, + "eval_accuracy": 0.7544, + "eval_loss": 0.7437487548828124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.1242905855178833, + "mean_feedback_forward_cosine": 0.6590101055560573, + "mean_feedback_forward_relative_error": 0.8331726616428744, + "min_feedback_forward_cosine": 0.3955609202384949 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0046604658948121765, + "instruction_rms": 0.00466046589273467, + "raw_apical_rms": 0.05602660090017935, + "teaching_rms": 0.0046604658948121765, + "traffic_rms": 0.05583040206733472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9735030437905152e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7574253798858607e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00839715642009344, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 2816, + "train_examples": 45000, + "train_loss": 0.596587984000312 + }, + { + "epoch": 9, + "eval_accuracy": 0.7946, + "eval_loss": 0.60154697265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0955640077590942, + "mean_feedback_forward_cosine": 0.6862208708640067, + "mean_feedback_forward_relative_error": 0.7990370306276506, + "min_feedback_forward_cosine": 0.42728692293167114 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004562242773027457, + "instruction_rms": 0.004562242772240408, + "raw_apical_rms": 0.05656990566652518, + "teaching_rms": 0.004562242773027457, + "traffic_rms": 0.05638386271176832 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.863355952342772e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7548746334258725e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008591035332460245, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3168, + "train_examples": 45000, + "train_loss": 0.5534716961648729 + }, + { + "epoch": 10, + "eval_accuracy": 0.7944, + "eval_loss": 0.6164186157226562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0699032545089722, + "mean_feedback_forward_cosine": 0.7099020461882314, + "mean_feedback_forward_relative_error": 0.7679428881214511, + "min_feedback_forward_cosine": 0.4542800188064575 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004563739370769258, + "instruction_rms": 0.004563739371583582, + "raw_apical_rms": 0.05763044514406948, + "teaching_rms": 0.004563739370769258, + "traffic_rms": 0.05744746317299652 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6442590339570415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.747594984543883e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008666427106516958, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3520, + "train_examples": 45000, + "train_loss": 0.5181713239722782 + }, + { + "epoch": 11, + "eval_accuracy": 0.814, + "eval_loss": 0.5557191284179688, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0450198650360107, + "mean_feedback_forward_cosine": 0.7315100171873646, + "mean_feedback_forward_relative_error": 0.7385791319031869, + "min_feedback_forward_cosine": 0.4804466664791107 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004469426111132943, + "instruction_rms": 0.004469426108247147, + "raw_apical_rms": 0.05807760817231134, + "teaching_rms": 0.004469426111132943, + "traffic_rms": 0.057902990804586524 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.267847598602657e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7500813606099063e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00868155962782433, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 3872, + "train_examples": 45000, + "train_loss": 0.49058791041904026 + }, + { + "epoch": 12, + "eval_accuracy": 0.8144, + "eval_loss": 0.5658951782226562, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 1.0195896625518799, + "mean_feedback_forward_cosine": 0.7507525624767426, + "mean_feedback_forward_relative_error": 0.7111556318498427, + "min_feedback_forward_cosine": 0.5057989954948425 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004474259591132489, + "instruction_rms": 0.004474259590376193, + "raw_apical_rms": 0.058553984622773635, + "teaching_rms": 0.004474259591132489, + "traffic_rms": 0.05838041382212191 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4798591624630717e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.736381041438012e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008935329311463529, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4224, + "train_examples": 45000, + "train_loss": 0.4756378570795059 + }, + { + "epoch": 13, + "eval_accuracy": 0.783, + "eval_loss": 0.6674646728515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9945835471153259, + "mean_feedback_forward_cosine": 0.7681244919376988, + "mean_feedback_forward_relative_error": 0.6856772361263153, + "min_feedback_forward_cosine": 0.5301342010498047 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004440338589064446, + "instruction_rms": 0.004440338590856765, + "raw_apical_rms": 0.05932498650991804, + "teaching_rms": 0.004440338589064446, + "traffic_rms": 0.05915643318048489 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.289915184723213e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7339047457168407e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009083964661700545, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4576, + "train_examples": 45000, + "train_loss": 0.4465404761367374 + }, + { + "epoch": 14, + "eval_accuracy": 0.8244, + "eval_loss": 0.5664008117675782, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9696456789970398, + "mean_feedback_forward_cosine": 0.7843829989433289, + "mean_feedback_forward_relative_error": 0.6608943179730447, + "min_feedback_forward_cosine": 0.5536136031150818 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004406709406016659, + "instruction_rms": 0.0044067094081843795, + "raw_apical_rms": 0.06019784191137481, + "teaching_rms": 0.004406709406016659, + "traffic_rms": 0.0600341441766403 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.981530400347765e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.726925638732207e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009200338237390608, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 4928, + "train_examples": 45000, + "train_loss": 0.43072997092141047 + }, + { + "epoch": 15, + "eval_accuracy": 0.819, + "eval_loss": 0.5506152587890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9452754855155945, + "mean_feedback_forward_cosine": 0.7992827103983972, + "mean_feedback_forward_relative_error": 0.6372344061251609, + "min_feedback_forward_cosine": 0.5759081840515137 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004326748108247472, + "instruction_rms": 0.004326748109206437, + "raw_apical_rms": 0.06030161495920084, + "teaching_rms": 0.004326748108247472, + "traffic_rms": 0.060144057022412205 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.354581774000053e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7276153912306198e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009178980940521502, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5280, + "train_examples": 45000, + "train_loss": 0.4152876753939523 + }, + { + "epoch": 16, + "eval_accuracy": 0.8252, + "eval_loss": 0.5355547790527344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.9220921993255615, + "mean_feedback_forward_cosine": 0.8126598596572876, + "mean_feedback_forward_relative_error": 0.6151499200251794, + "min_feedback_forward_cosine": 0.5968871116638184 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004373927017103245, + "instruction_rms": 0.00437392701692843, + "raw_apical_rms": 0.060641845687035054, + "teaching_rms": 0.004373927017103245, + "traffic_rms": 0.06048179286830912 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.218938360305401e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7335190174540215e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009351303294551121, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5632, + "train_examples": 45000, + "train_loss": 0.396997609117296 + }, + { + "epoch": 17, + "eval_accuracy": 0.8344, + "eval_loss": 0.49461121215820314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8985036015510559, + "mean_feedback_forward_cosine": 0.8253427820820962, + "mean_feedback_forward_relative_error": 0.5936438220162545, + "min_feedback_forward_cosine": 0.6173129081726074 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004339730517574678, + "instruction_rms": 0.0043397305169046016, + "raw_apical_rms": 0.06086485465183153, + "teaching_rms": 0.004339730517574678, + "traffic_rms": 0.06070770022252158 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7485743459010337e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.727211790915874e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009320647707649931, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 5984, + "train_examples": 45000, + "train_loss": 0.381810572581821 + }, + { + "epoch": 18, + "eval_accuracy": 0.8494, + "eval_loss": 0.44682206420898435, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8745174407958984, + "mean_feedback_forward_cosine": 0.8368679458095182, + "mean_feedback_forward_relative_error": 0.5733353779200585, + "min_feedback_forward_cosine": 0.6376420259475708 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004295681396640564, + "instruction_rms": 0.004295681397735242, + "raw_apical_rms": 0.061066769983406814, + "teaching_rms": 0.004295681396640564, + "traffic_rms": 0.06091296846652354 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.872836901701703e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7253816849063617e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009109225358336104, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6336, + "train_examples": 45000, + "train_loss": 0.372761104800966 + }, + { + "epoch": 19, + "eval_accuracy": 0.8438, + "eval_loss": 0.5001562072753907, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8507093787193298, + "mean_feedback_forward_cosine": 0.8478189783711587, + "mean_feedback_forward_relative_error": 0.5536001593835892, + "min_feedback_forward_cosine": 0.6569257974624634 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00430706482340492, + "instruction_rms": 0.004307064824701717, + "raw_apical_rms": 0.06156562532931305, + "teaching_rms": 0.00430706482340492, + "traffic_rms": 0.06141233044377472 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.177505820100123e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7231557651690612e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009345401061877862, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 6688, + "train_examples": 45000, + "train_loss": 0.36028081623183356 + }, + { + "epoch": 20, + "eval_accuracy": 0.8364, + "eval_loss": 0.5102594970703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8273930549621582, + "mean_feedback_forward_cosine": 0.8576913264489943, + "mean_feedback_forward_relative_error": 0.5349951569111117, + "min_feedback_forward_cosine": 0.6755627393722534 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004224719036629221, + "instruction_rms": 0.004224719038728415, + "raw_apical_rms": 0.061968769954962054, + "teaching_rms": 0.004224719036629221, + "traffic_rms": 0.061822003201752664 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.2150183909275256e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7203478299842856e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009124211101281124, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7040, + "train_examples": 45000, + "train_loss": 0.3493091468864017 + }, + { + "epoch": 21, + "eval_accuracy": 0.8274, + "eval_loss": 0.5551596557617188, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.8041688203811646, + "mean_feedback_forward_cosine": 0.8668903343139156, + "mean_feedback_forward_relative_error": 0.517221026843594, + "min_feedback_forward_cosine": 0.6933361291885376 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004196544092540644, + "instruction_rms": 0.0041965440922780905, + "raw_apical_rms": 0.0622468100375289, + "teaching_rms": 0.004196544092540644, + "traffic_rms": 0.06210289419973713 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 1.9702342868299638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7178518780542035e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00913035517495945, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7392, + "train_examples": 45000, + "train_loss": 0.33633084367116295 + }, + { + "epoch": 22, + "eval_accuracy": 0.8174, + "eval_loss": 0.5925237060546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7810643315315247, + "mean_feedback_forward_cosine": 0.8756693716972105, + "mean_feedback_forward_relative_error": 0.49964048016455864, + "min_feedback_forward_cosine": 0.7106687426567078 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004198794541783616, + "instruction_rms": 0.0041987945443736205, + "raw_apical_rms": 0.062473956208242275, + "teaching_rms": 0.004198794541783616, + "traffic_rms": 0.06233022832452929 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4426169531466257e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.715997397441632e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009176835462509384, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 7744, + "train_examples": 45000, + "train_loss": 0.3273309647401174 + }, + { + "epoch": 23, + "eval_accuracy": 0.8468, + "eval_loss": 0.4822109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7584139704704285, + "mean_feedback_forward_cosine": 0.8838707900816395, + "mean_feedback_forward_relative_error": 0.48255138868285763, + "min_feedback_forward_cosine": 0.7267588376998901 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004134608060019288, + "instruction_rms": 0.004134608059179785, + "raw_apical_rms": 0.06299561503763626, + "teaching_rms": 0.004134608060019288, + "traffic_rms": 0.06285719573841968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.3491308454026694e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7131197218623048e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009187891024563598, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8096, + "train_examples": 45000, + "train_loss": 0.3197494336181217 + }, + { + "epoch": 24, + "eval_accuracy": 0.8548, + "eval_loss": 0.47645836791992185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.7370368242263794, + "mean_feedback_forward_cosine": 0.8913365890902858, + "mean_feedback_forward_relative_error": 0.4664817803329037, + "min_feedback_forward_cosine": 0.7420064806938171 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004068535321030107, + "instruction_rms": 0.004068535321475998, + "raw_apical_rms": 0.06336379104187902, + "teaching_rms": 0.004068535321030107, + "traffic_rms": 0.06323072750014327 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.353047889869231e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7124145812856903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009404880109961884, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8448, + "train_examples": 45000, + "train_loss": 0.30924332915941877 + }, + { + "epoch": 25, + "eval_accuracy": 0.8522, + "eval_loss": 0.4688670715332031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.716600239276886, + "mean_feedback_forward_cosine": 0.8982402047803325, + "mean_feedback_forward_relative_error": 0.4512202475340136, + "min_feedback_forward_cosine": 0.7559291124343872 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004049261991271629, + "instruction_rms": 0.0040492619898989425, + "raw_apical_rms": 0.06351800839271533, + "teaching_rms": 0.004049261991271629, + "traffic_rms": 0.06338627554501733 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 2.3370080981521824e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7090493041667627e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00951750746944136, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 8800, + "train_examples": 45000, + "train_loss": 0.3014814792844984 + }, + { + "epoch": 26, + "eval_accuracy": 0.8586, + "eval_loss": 0.4448318176269531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6953040957450867, + "mean_feedback_forward_cosine": 0.9047312967238887, + "mean_feedback_forward_relative_error": 0.4362514446819982, + "min_feedback_forward_cosine": 0.7700546383857727 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004038551876452253, + "instruction_rms": 0.004038551875714417, + "raw_apical_rms": 0.06293602933749305, + "teaching_rms": 0.004038551876452253, + "traffic_rms": 0.06280328592312978 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.783110808479705e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7167502013725602e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009235098366898327, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9152, + "train_examples": 45000, + "train_loss": 0.2990670070648193 + }, + { + "epoch": 27, + "eval_accuracy": 0.8444, + "eval_loss": 0.5124790283203124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6745215058326721, + "mean_feedback_forward_cosine": 0.9107907472118255, + "mean_feedback_forward_relative_error": 0.42197790549647424, + "min_feedback_forward_cosine": 0.7833300828933716 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004002224396894647, + "instruction_rms": 0.004002224397603151, + "raw_apical_rms": 0.06332909380438531, + "teaching_rms": 0.004002224396894647, + "traffic_rms": 0.06319963985975724 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7412404790538858e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7122489179406515e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009382553349983636, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9504, + "train_examples": 45000, + "train_loss": 0.2910355213324229 + }, + { + "epoch": 28, + "eval_accuracy": 0.855, + "eval_loss": 0.45888658447265623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6546169519424438, + "mean_feedback_forward_cosine": 0.9165496057079684, + "mean_feedback_forward_relative_error": 0.40787488510531766, + "min_feedback_forward_cosine": 0.7957733869552612 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00405124829155215, + "instruction_rms": 0.004051248292788438, + "raw_apical_rms": 0.06371395587046495, + "teaching_rms": 0.00405124829155215, + "traffic_rms": 0.06358227999593934 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4874308834910153e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7035052282601972e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009621776229063744, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 9856, + "train_examples": 45000, + "train_loss": 0.27963798101743065 + }, + { + "epoch": 29, + "eval_accuracy": 0.8446, + "eval_loss": 0.5147614685058594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6332876682281494, + "mean_feedback_forward_cosine": 0.9220803841467826, + "mean_feedback_forward_relative_error": 0.3939231518776186, + "min_feedback_forward_cosine": 0.8088099956512451 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0040142515849250875, + "instruction_rms": 0.0040142515845335396, + "raw_apical_rms": 0.06378851239153105, + "teaching_rms": 0.0040142515849250875, + "traffic_rms": 0.0636594051496206 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.671583487767748e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7058148057442953e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00963041799819651, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10208, + "train_examples": 45000, + "train_loss": 0.28177171817620594 + }, + { + "epoch": 30, + "eval_accuracy": 0.8634, + "eval_loss": 0.42816656494140626, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.6144068837165833, + "mean_feedback_forward_cosine": 0.9271933878621748, + "mean_feedback_forward_relative_error": 0.38059873734751054, + "min_feedback_forward_cosine": 0.8199130296707153 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004047472878011346, + "instruction_rms": 0.00404747287724283, + "raw_apical_rms": 0.0639902815971012, + "teaching_rms": 0.004047472878011346, + "traffic_rms": 0.06385950306625889 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.661687759086817e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7010070054018895e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009610093831761419, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10560, + "train_examples": 45000, + "train_loss": 0.27227373503314123 + }, + { + "epoch": 31, + "eval_accuracy": 0.8554, + "eval_loss": 0.47649244384765627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5947118401527405, + "mean_feedback_forward_cosine": 0.9319607653925496, + "mean_feedback_forward_relative_error": 0.3677683886020414, + "min_feedback_forward_cosine": 0.8311227560043335 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003970332986572767, + "instruction_rms": 0.0039703329867750966, + "raw_apical_rms": 0.06403122964700471, + "teaching_rms": 0.003970332986572767, + "traffic_rms": 0.06390497861215359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6646426337938465e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.7014060207564076e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009607793921135612, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 10912, + "train_examples": 45000, + "train_loss": 0.26816212072902257 + }, + { + "epoch": 32, + "eval_accuracy": 0.821, + "eval_loss": 0.5862628173828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.575197160243988, + "mean_feedback_forward_cosine": 0.9365121702994069, + "mean_feedback_forward_relative_error": 0.35506819621209174, + "min_feedback_forward_cosine": 0.841862678527832 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.004041092568796793, + "instruction_rms": 0.004041092568857933, + "raw_apical_rms": 0.06434880632694744, + "teaching_rms": 0.004041092568796793, + "traffic_rms": 0.06421914223623654 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.928237918808918e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.700544054890139e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00970343761641436, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11264, + "train_examples": 45000, + "train_loss": 0.2629036833021376 + }, + { + "epoch": 33, + "eval_accuracy": 0.8614, + "eval_loss": 0.4444266784667969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5564904808998108, + "mean_feedback_forward_cosine": 0.9406849799617645, + "mean_feedback_forward_relative_error": 0.3430360435478149, + "min_feedback_forward_cosine": 0.8517687320709229 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00403412739904424, + "instruction_rms": 0.0040341274000684395, + "raw_apical_rms": 0.06480477899317917, + "teaching_rms": 0.00403412739904424, + "traffic_rms": 0.06467634344777613 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.132125314664336e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6950566314144252e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009332616035410659, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11616, + "train_examples": 45000, + "train_loss": 0.2601968579000897 + }, + { + "epoch": 34, + "eval_accuracy": 0.8706, + "eval_loss": 0.4194605712890625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.538070559501648, + "mean_feedback_forward_cosine": 0.9445796762743304, + "mean_feedback_forward_relative_error": 0.33144712976871, + "min_feedback_forward_cosine": 0.8612483143806458 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003983203975616594, + "instruction_rms": 0.00398320397674507, + "raw_apical_rms": 0.06473500908896433, + "teaching_rms": 0.003983203975616594, + "traffic_rms": 0.0646094019501837 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.1507609915261128e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6970872730951366e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00974665736367693, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 11968, + "train_examples": 45000, + "train_loss": 0.2567433127456241 + }, + { + "epoch": 35, + "eval_accuracy": 0.859, + "eval_loss": 0.4610693786621094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5206394195556641, + "mean_feedback_forward_cosine": 0.9482109585115986, + "mean_feedback_forward_relative_error": 0.3202533092229597, + "min_feedback_forward_cosine": 0.8700209856033325 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003933592381475546, + "instruction_rms": 0.003933592382589914, + "raw_apical_rms": 0.06499513738354835, + "teaching_rms": 0.003933592381475546, + "traffic_rms": 0.06487294588293592 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.1977518471771873e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.700391664730098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009886096025631225, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12320, + "train_examples": 45000, + "train_loss": 0.2506347870879703 + }, + { + "epoch": 36, + "eval_accuracy": 0.8618, + "eval_loss": 0.4532576416015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.5033196806907654, + "mean_feedback_forward_cosine": 0.9514969010506907, + "mean_feedback_forward_relative_error": 0.30979928566563514, + "min_feedback_forward_cosine": 0.878451943397522 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003896671065599783, + "instruction_rms": 0.003896671065423597, + "raw_apical_rms": 0.06514360563307033, + "teaching_rms": 0.003896671065599783, + "traffic_rms": 0.0650237827795661 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.4190093483866804e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.690560100655861e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009983557577443883, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 12672, + "train_examples": 45000, + "train_loss": 0.24602477144135368 + }, + { + "epoch": 37, + "eval_accuracy": 0.8672, + "eval_loss": 0.41887584838867187, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.486396461725235, + "mean_feedback_forward_cosine": 0.9546678143162881, + "mean_feedback_forward_relative_error": 0.29938822239637375, + "min_feedback_forward_cosine": 0.8864339590072632 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003893000874761815, + "instruction_rms": 0.0038930008759012007, + "raw_apical_rms": 0.06537661699246185, + "teaching_rms": 0.003893000874761815, + "traffic_rms": 0.06525766422221822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 1.950840022857392e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.69036997062304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009698828726604075, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13024, + "train_examples": 45000, + "train_loss": 0.2446405156109068 + }, + { + "epoch": 38, + "eval_accuracy": 0.8566, + "eval_loss": 0.4433000061035156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.471229612827301, + "mean_feedback_forward_cosine": 0.9574937397433866, + "mean_feedback_forward_relative_error": 0.289746877406874, + "min_feedback_forward_cosine": 0.8933587074279785 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003888784879666175, + "instruction_rms": 0.0038887848787239744, + "raw_apical_rms": 0.06561847299448788, + "teaching_rms": 0.003888784879666175, + "traffic_rms": 0.06550056080262137 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.945887445093831e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6862998767583053e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009867522792933625, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13376, + "train_examples": 45000, + "train_loss": 0.23308202556769053 + }, + { + "epoch": 39, + "eval_accuracy": 0.852, + "eval_loss": 0.5213807067871093, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4555465579032898, + "mean_feedback_forward_cosine": 0.9602498469814178, + "mean_feedback_forward_relative_error": 0.28014876885760215, + "min_feedback_forward_cosine": 0.9002827405929565 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003943040376982913, + "instruction_rms": 0.00394304037560111, + "raw_apical_rms": 0.06544258303738437, + "teaching_rms": 0.003943040376982913, + "traffic_rms": 0.06532078750332516 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.1349510603840827e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.691445734257437e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009442867302361439, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 13728, + "train_examples": 45000, + "train_loss": 0.22730018768575455 + }, + { + "epoch": 40, + "eval_accuracy": 0.8622, + "eval_loss": 0.4563710205078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4407300651073456, + "mean_feedback_forward_cosine": 0.9627599639277304, + "mean_feedback_forward_relative_error": 0.27106243035485666, + "min_feedback_forward_cosine": 0.9065840840339661 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038601983349361778, + "instruction_rms": 0.0038601983341153136, + "raw_apical_rms": 0.06575174846595079, + "teaching_rms": 0.0038601983349361778, + "traffic_rms": 0.06563548604744925 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.8701164112021615e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6795436523467688e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009744524816008222, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14080, + "train_examples": 45000, + "train_loss": 0.22868775055673388 + }, + { + "epoch": 41, + "eval_accuracy": 0.8586, + "eval_loss": 0.4705298645019531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.4253489375114441, + "mean_feedback_forward_cosine": 0.9651982476634364, + "mean_feedback_forward_relative_error": 0.26196353954653584, + "min_feedback_forward_cosine": 0.912874162197113 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038398989149705247, + "instruction_rms": 0.0038398989148078623, + "raw_apical_rms": 0.06570998625869473, + "teaching_rms": 0.0038398989149705247, + "traffic_rms": 0.06559499804204008 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.507201912749224e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6877465057365064e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00982549091408656, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14432, + "train_examples": 45000, + "train_loss": 0.2281675344573127 + }, + { + "epoch": 42, + "eval_accuracy": 0.8692, + "eval_loss": 0.4207826354980469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.41090819239616394, + "mean_feedback_forward_cosine": 0.9674562023532006, + "mean_feedback_forward_relative_error": 0.25319438043140596, + "min_feedback_forward_cosine": 0.9185685515403748 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003900329559278322, + "instruction_rms": 0.003900329559352517, + "raw_apical_rms": 0.06577654280179057, + "teaching_rms": 0.003900329559278322, + "traffic_rms": 0.06565717687693617 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7220789178272753e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.684044156398938e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009841091216831013, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 14784, + "train_examples": 45000, + "train_loss": 0.22557188702424366 + }, + { + "epoch": 43, + "eval_accuracy": 0.854, + "eval_loss": 0.4854722900390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3975459933280945, + "mean_feedback_forward_cosine": 0.9695534186978494, + "mean_feedback_forward_relative_error": 0.24481760037522163, + "min_feedback_forward_cosine": 0.9236574172973633 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038587651591665, + "instruction_rms": 0.003858765160739692, + "raw_apical_rms": 0.06620083047706758, + "teaching_rms": 0.0038587651591665, + "traffic_rms": 0.06608502767678665 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9172804616450776e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6827225905703166e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009815370464705665, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15136, + "train_examples": 45000, + "train_loss": 0.22222819827132756 + }, + { + "epoch": 44, + "eval_accuracy": 0.8568, + "eval_loss": 0.477352685546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3839803636074066, + "mean_feedback_forward_cosine": 0.9715552522290137, + "mean_feedback_forward_relative_error": 0.23654738549263246, + "min_feedback_forward_cosine": 0.9287176132202148 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038450055363336328, + "instruction_rms": 0.003845005535372914, + "raw_apical_rms": 0.06648666215531655, + "teaching_rms": 0.0038450055363336328, + "traffic_rms": 0.06637190449648765 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.916029551158772e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6823816055350486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009656137922083871, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15488, + "train_examples": 45000, + "train_loss": 0.22109255310694376 + }, + { + "epoch": 45, + "eval_accuracy": 0.8624, + "eval_loss": 0.4692766296386719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3707510530948639, + "mean_feedback_forward_cosine": 0.9734137385122238, + "mean_feedback_forward_relative_error": 0.2286106754214533, + "min_feedback_forward_cosine": 0.9334725141525269 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038689133874628477, + "instruction_rms": 0.0038689133876962404, + "raw_apical_rms": 0.06663124662534974, + "teaching_rms": 0.0038689133874628477, + "traffic_rms": 0.06651578034124844 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6223659688184853e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6800576791592795e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009691724331843043, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 15840, + "train_examples": 45000, + "train_loss": 0.21285866431660122 + }, + { + "epoch": 46, + "eval_accuracy": 0.858, + "eval_loss": 0.4939498779296875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3577093183994293, + "mean_feedback_forward_cosine": 0.975148985462804, + "mean_feedback_forward_relative_error": 0.22097072293681483, + "min_feedback_forward_cosine": 0.9380070567131042 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003831199979431922, + "instruction_rms": 0.003831199978927395, + "raw_apical_rms": 0.06692136029826536, + "teaching_rms": 0.003831199979431922, + "traffic_rms": 0.06680845661552556 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.1969630498451806e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.681096194116952e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009524549023077611, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16192, + "train_examples": 45000, + "train_loss": 0.21487696191999647 + }, + { + "epoch": 47, + "eval_accuracy": 0.872, + "eval_loss": 0.472917919921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3452940285205841, + "mean_feedback_forward_cosine": 0.9767827987670898, + "mean_feedback_forward_relative_error": 0.21350583530241443, + "min_feedback_forward_cosine": 0.9421908855438232 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0038345523695841512, + "instruction_rms": 0.0038345523689545485, + "raw_apical_rms": 0.0670646653799901, + "teaching_rms": 0.0038345523695841512, + "traffic_rms": 0.06695165768518628 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.331855324972821e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6728564451423404e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009493565128231124, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16544, + "train_examples": 45000, + "train_loss": 0.21163174434767829 + }, + { + "epoch": 48, + "eval_accuracy": 0.8688, + "eval_loss": 0.44108818359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.33314257860183716, + "mean_feedback_forward_cosine": 0.9782731494595928, + "mean_feedback_forward_relative_error": 0.20647436476522876, + "min_feedback_forward_cosine": 0.9461215734481812 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003764190321564789, + "instruction_rms": 0.0037641903210092367, + "raw_apical_rms": 0.0670691289351605, + "teaching_rms": 0.003764190321564789, + "traffic_rms": 0.06696056626715648 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.51406370668883e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.674099971128485e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009744211307131654, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 16896, + "train_examples": 45000, + "train_loss": 0.20839293567339578 + }, + { + "epoch": 49, + "eval_accuracy": 0.8862, + "eval_loss": 0.37858214111328126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.32079362869262695, + "mean_feedback_forward_cosine": 0.9797187486002522, + "mean_feedback_forward_relative_error": 0.19946648492928473, + "min_feedback_forward_cosine": 0.9499779939651489 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003796721602114192, + "instruction_rms": 0.0037967216032537697, + "raw_apical_rms": 0.06727567276809461, + "teaching_rms": 0.003796721602114192, + "traffic_rms": 0.06716515229479343 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.709113454737235e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6801163922097045e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009502915382779762, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17248, + "train_examples": 45000, + "train_loss": 0.20797124909559886 + }, + { + "epoch": 50, + "eval_accuracy": 0.8714, + "eval_loss": 0.4533976684570313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.3097163140773773, + "mean_feedback_forward_cosine": 0.9810209947247659, + "mean_feedback_forward_relative_error": 0.19289126487508898, + "min_feedback_forward_cosine": 0.9533424377441406 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003790985533641146, + "instruction_rms": 0.0037909855329868116, + "raw_apical_rms": 0.06751110619547468, + "teaching_rms": 0.003790985533641146, + "traffic_rms": 0.06740139136510716 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.339059600511064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6697007933493875e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009328096615369125, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17600, + "train_examples": 45000, + "train_loss": 0.20432197641796537 + }, + { + "epoch": 51, + "eval_accuracy": 0.8674, + "eval_loss": 0.442272802734375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.29837894439697266, + "mean_feedback_forward_cosine": 0.9822761243389498, + "mean_feedback_forward_relative_error": 0.18638599760109378, + "min_feedback_forward_cosine": 0.9566454887390137 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003752230948103122, + "instruction_rms": 0.003752230948219423, + "raw_apical_rms": 0.06714921863417124, + "teaching_rms": 0.003752230948103122, + "traffic_rms": 0.06704126659320363 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1463550815024064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6738598084129747e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009195574078491352, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 17952, + "train_examples": 45000, + "train_loss": 0.20099864941967857 + }, + { + "epoch": 52, + "eval_accuracy": 0.8576, + "eval_loss": 0.49777291259765627, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.28850314021110535, + "mean_feedback_forward_cosine": 0.9834090625086138, + "mean_feedback_forward_relative_error": 0.18029032479370793, + "min_feedback_forward_cosine": 0.9594382047653198 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037215057599723268, + "instruction_rms": 0.003721505758804505, + "raw_apical_rms": 0.06721710055288083, + "teaching_rms": 0.0037215057599723268, + "traffic_rms": 0.06711074629889201 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9702910353535117e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.673041938791294e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009731579987993283, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18304, + "train_examples": 45000, + "train_loss": 0.19583829987843832 + }, + { + "epoch": 53, + "eval_accuracy": 0.8728, + "eval_loss": 0.4199692565917969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.27802181243896484, + "mean_feedback_forward_cosine": 0.9845092200463818, + "mean_feedback_forward_relative_error": 0.17417119635689643, + "min_feedback_forward_cosine": 0.962307333946228 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037648703348918803, + "instruction_rms": 0.0037648703360029915, + "raw_apical_rms": 0.06741783690888394, + "teaching_rms": 0.0037648703348918803, + "traffic_rms": 0.06730939466965924 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1614129031586344e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.674570193260333e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009629303913869847, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 18656, + "train_examples": 45000, + "train_loss": 0.19910230238967472 + }, + { + "epoch": 54, + "eval_accuracy": 0.8592, + "eval_loss": 0.48668612670898437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.26857489347457886, + "mean_feedback_forward_cosine": 0.9855129680325908, + "mean_feedback_forward_relative_error": 0.16836787543950543, + "min_feedback_forward_cosine": 0.9647953510284424 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037347801412208263, + "instruction_rms": 0.0037347801417342602, + "raw_apical_rms": 0.06763126728019535, + "teaching_rms": 0.0037347801412208263, + "traffic_rms": 0.06752490857710151 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.12550163528158e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6735748017196808e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009732354467041557, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19008, + "train_examples": 45000, + "train_loss": 0.19133018584781222 + }, + { + "epoch": 55, + "eval_accuracy": 0.879, + "eval_loss": 0.4195354675292969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2584693133831024, + "mean_feedback_forward_cosine": 0.9865040856022989, + "mean_feedback_forward_relative_error": 0.1624837938335634, + "min_feedback_forward_cosine": 0.967373251914978 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003791996760680481, + "instruction_rms": 0.003791996760556632, + "raw_apical_rms": 0.06805450134213765, + "teaching_rms": 0.003791996760680481, + "traffic_rms": 0.06794524829129649 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.24055697831227e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.678159205079645e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009456198256669738, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19360, + "train_examples": 45000, + "train_loss": 0.2014359640068478 + }, + { + "epoch": 56, + "eval_accuracy": 0.8786, + "eval_loss": 0.3782403259277344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24904784560203552, + "mean_feedback_forward_cosine": 0.9874036119830224, + "mean_feedback_forward_relative_error": 0.1569397098595096, + "min_feedback_forward_cosine": 0.9696934819221497 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037843500449065964, + "instruction_rms": 0.0037843500451697504, + "raw_apical_rms": 0.06776119506531587, + "teaching_rms": 0.0037843500449065964, + "traffic_rms": 0.06765212949389462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.656533920093352e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.676397019751039e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009796610734344686, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 19712, + "train_examples": 45000, + "train_loss": 0.19607376774284574 + }, + { + "epoch": 57, + "eval_accuracy": 0.8538, + "eval_loss": 0.5347755859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.24004074931144714, + "mean_feedback_forward_cosine": 0.9882267367455267, + "mean_feedback_forward_relative_error": 0.1517001005911058, + "min_feedback_forward_cosine": 0.9718272089958191 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037077486941967386, + "instruction_rms": 0.0037077486959924476, + "raw_apical_rms": 0.06816540951085935, + "teaching_rms": 0.0037077486941967386, + "traffic_rms": 0.0680611902470229 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 2.8713960986692655e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6686866317146235e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009484137712073895, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20064, + "train_examples": 45000, + "train_loss": 0.1932436572737164 + }, + { + "epoch": 58, + "eval_accuracy": 0.867, + "eval_loss": 0.4802153991699219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.231326624751091, + "mean_feedback_forward_cosine": 0.9890082401614035, + "mean_feedback_forward_relative_error": 0.1465402961738648, + "min_feedback_forward_cosine": 0.9738050699234009 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003742871989211473, + "instruction_rms": 0.0037428719918443055, + "raw_apical_rms": 0.06845554943573484, + "teaching_rms": 0.003742871989211473, + "traffic_rms": 0.06835025965135379 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.002559623155321e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6692705455518372e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009609183502983362, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20416, + "train_examples": 45000, + "train_loss": 0.1894873697307375 + }, + { + "epoch": 59, + "eval_accuracy": 0.869, + "eval_loss": 0.46880623779296876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.22357633709907532, + "mean_feedback_forward_cosine": 0.9897216020091888, + "mean_feedback_forward_relative_error": 0.1416760240831683, + "min_feedback_forward_cosine": 0.9755195379257202 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037117681191604895, + "instruction_rms": 0.0037117681183885614, + "raw_apical_rms": 0.06885039561989607, + "teaching_rms": 0.0037117681191604895, + "traffic_rms": 0.0687473068281067 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.011009042097612e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6794532005972392e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009946817394249772, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 20768, + "train_examples": 45000, + "train_loss": 0.18652594333754646 + }, + { + "epoch": 60, + "eval_accuracy": 0.8708, + "eval_loss": 0.45631016845703126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.2157677412033081, + "mean_feedback_forward_cosine": 0.9903988684377363, + "mean_feedback_forward_relative_error": 0.13691440188596327, + "min_feedback_forward_cosine": 0.9771890044212341 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0037127934702630514, + "instruction_rms": 0.00371279346821834, + "raw_apical_rms": 0.0686732189890834, + "teaching_rms": 0.0037127934702630514, + "traffic_rms": 0.06856912437735939 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 2.8024947695826086e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6693847376850912e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009393395066284187, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21120, + "train_examples": 45000, + "train_loss": 0.18327303275797102 + }, + { + "epoch": 61, + "eval_accuracy": 0.8718, + "eval_loss": 0.4464225769042969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20803159475326538, + "mean_feedback_forward_cosine": 0.991038772367662, + "mean_feedback_forward_relative_error": 0.13224729607182165, + "min_feedback_forward_cosine": 0.9787874221801758 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036716750526718453, + "instruction_rms": 0.0036716750520768868, + "raw_apical_rms": 0.0687325340227867, + "teaching_rms": 0.0036716750526718453, + "traffic_rms": 0.068631049144822 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.627700451147575e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.679051637414625e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009369368561217686, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21472, + "train_examples": 45000, + "train_loss": 0.18633612112204234 + }, + { + "epoch": 62, + "eval_accuracy": 0.876, + "eval_loss": 0.4284907897949219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.20074361562728882, + "mean_feedback_forward_cosine": 0.9916233131962437, + "mean_feedback_forward_relative_error": 0.1278414705828313, + "min_feedback_forward_cosine": 0.9802231192588806 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036913750583233446, + "instruction_rms": 0.0036913750564064947, + "raw_apical_rms": 0.06897525226004436, + "teaching_rms": 0.0036913750583233446, + "traffic_rms": 0.0688732098451882 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.788874991528246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6674829261820502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009119546716608301, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 21824, + "train_examples": 45000, + "train_loss": 0.1868861578941345 + }, + { + "epoch": 63, + "eval_accuracy": 0.8764, + "eval_loss": 0.4421176025390625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.19378747045993805, + "mean_feedback_forward_cosine": 0.9921654770451207, + "mean_feedback_forward_relative_error": 0.12360262750617919, + "min_feedback_forward_cosine": 0.9815561771392822 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003675538245747924, + "instruction_rms": 0.0036755382454474795, + "raw_apical_rms": 0.06919303102746925, + "teaching_rms": 0.003675538245747924, + "traffic_rms": 0.06909157947821462 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5002013731855186e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6631604351137743e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009591347883657127, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22176, + "train_examples": 45000, + "train_loss": 0.18030287986596424 + }, + { + "epoch": 64, + "eval_accuracy": 0.8606, + "eval_loss": 0.4947023864746094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18710586428642273, + "mean_feedback_forward_cosine": 0.9926713493562513, + "mean_feedback_forward_relative_error": 0.11952662311734692, + "min_feedback_forward_cosine": 0.9827964305877686 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036685665645455005, + "instruction_rms": 0.003668566564958399, + "raw_apical_rms": 0.06899592224688407, + "teaching_rms": 0.0036685665645455005, + "traffic_rms": 0.06889483733996477 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.01061895413568e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6715958232016822e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009469844400082687, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22528, + "train_examples": 45000, + "train_loss": 0.179588240453932 + }, + { + "epoch": 65, + "eval_accuracy": 0.8856, + "eval_loss": 0.38915611572265624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.18032169342041016, + "mean_feedback_forward_cosine": 0.9931720533678609, + "mean_feedback_forward_relative_error": 0.11536907224405196, + "min_feedback_forward_cosine": 0.9840102791786194 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003749223339813597, + "instruction_rms": 0.003749223340732459, + "raw_apical_rms": 0.06937663825925827, + "teaching_rms": 0.003749223339813597, + "traffic_rms": 0.069272070622568 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.875665655073135e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.670059678456853e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009408578479519461, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 22880, + "train_examples": 45000, + "train_loss": 0.1879544606897566 + }, + { + "epoch": 66, + "eval_accuracy": 0.877, + "eval_loss": 0.414552197265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.17396798729896545, + "mean_feedback_forward_cosine": 0.9936253255413424, + "mean_feedback_forward_relative_error": 0.11146722529684344, + "min_feedback_forward_cosine": 0.9851115345954895 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003655129286387944, + "instruction_rms": 0.0036551292857377946, + "raw_apical_rms": 0.06990854412409622, + "teaching_rms": 0.003655129286387944, + "traffic_rms": 0.0698102541672528 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.541091331271673e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.674265402304113e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00986880363510469, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23232, + "train_examples": 45000, + "train_loss": 0.17746702675554488 + }, + { + "epoch": 67, + "eval_accuracy": 0.8752, + "eval_loss": 0.45304346923828126, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1677214503288269, + "mean_feedback_forward_cosine": 0.9940506239091197, + "mean_feedback_forward_relative_error": 0.10766576723225656, + "min_feedback_forward_cosine": 0.9861553311347961 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00364378350989667, + "instruction_rms": 0.003643783510117977, + "raw_apical_rms": 0.06982380075625995, + "teaching_rms": 0.00364378350989667, + "traffic_rms": 0.0697253093055836 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6275363157756146e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6698265352807232e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009722678805803367, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23584, + "train_examples": 45000, + "train_loss": 0.17719422046078576 + }, + { + "epoch": 68, + "eval_accuracy": 0.8842, + "eval_loss": 0.3809119873046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.16186578571796417, + "mean_feedback_forward_cosine": 0.9944415861560453, + "mean_feedback_forward_relative_error": 0.10406146186494059, + "min_feedback_forward_cosine": 0.9870941042900085 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036261218173337774, + "instruction_rms": 0.0036261218168315516, + "raw_apical_rms": 0.06978934825449166, + "teaching_rms": 0.0036261218173337774, + "traffic_rms": 0.069691834882275 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.079424715224377e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6676291423852776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.01003843050606504, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 23936, + "train_examples": 45000, + "train_loss": 0.171932643699646 + }, + { + "epoch": 69, + "eval_accuracy": 0.8854, + "eval_loss": 0.3779376342773437, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.1558031588792801, + "mean_feedback_forward_cosine": 0.9948173030730216, + "mean_feedback_forward_relative_error": 0.10046967719831774, + "min_feedback_forward_cosine": 0.9880361557006836 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003679628838903734, + "instruction_rms": 0.0036796288372781265, + "raw_apical_rms": 0.0700862147160062, + "teaching_rms": 0.003679628838903734, + "traffic_rms": 0.06998683022391407 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.7146118242171724e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6697609063299328e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009998567842615923, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24288, + "train_examples": 45000, + "train_loss": 0.17809613135523267 + }, + { + "epoch": 70, + "eval_accuracy": 0.8754, + "eval_loss": 0.42756370849609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.15017694234848022, + "mean_feedback_forward_cosine": 0.9951669727602312, + "mean_feedback_forward_relative_error": 0.09701778066735114, + "min_feedback_forward_cosine": 0.9888768792152405 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036259993143688937, + "instruction_rms": 0.003625999314513519, + "raw_apical_rms": 0.07033731329615805, + "teaching_rms": 0.0036259993143688937, + "traffic_rms": 0.07024063978982102 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0599758222015225e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6606407015584944e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009573684923473932, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24640, + "train_examples": 45000, + "train_loss": 0.17603131600353453 + }, + { + "epoch": 71, + "eval_accuracy": 0.8722, + "eval_loss": 0.46291195068359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.14495930075645447, + "mean_feedback_forward_cosine": 0.9954846212940831, + "mean_feedback_forward_relative_error": 0.09377915147812135, + "min_feedback_forward_cosine": 0.9896330833435059 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003627911470636652, + "instruction_rms": 0.003627911471955973, + "raw_apical_rms": 0.07046372460091366, + "teaching_rms": 0.003627911470636652, + "traffic_rms": 0.07036733456818349 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.187840391660757e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.655666551295883e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009569554454389147, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 24992, + "train_examples": 45000, + "train_loss": 0.17087456733385722 + }, + { + "epoch": 72, + "eval_accuracy": 0.8742, + "eval_loss": 0.41198401489257813, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13980896770954132, + "mean_feedback_forward_cosine": 0.9957855786046674, + "mean_feedback_forward_relative_error": 0.09058924752377695, + "min_feedback_forward_cosine": 0.9903514385223389 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036313193832344533, + "instruction_rms": 0.003631319381984086, + "raw_apical_rms": 0.07061298827791872, + "teaching_rms": 0.0036313193832344533, + "traffic_rms": 0.0705164096689381 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.685270399955698e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6516195090030695e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0093422294375728, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25344, + "train_examples": 45000, + "train_loss": 0.16949140020608902 + }, + { + "epoch": 73, + "eval_accuracy": 0.8708, + "eval_loss": 0.4712065979003906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.13476146757602692, + "mean_feedback_forward_cosine": 0.9960707118434291, + "mean_feedback_forward_relative_error": 0.08746329466662099, + "min_feedback_forward_cosine": 0.9910346269607544 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.00362695184190896, + "instruction_rms": 0.0036269518412897135, + "raw_apical_rms": 0.07047801939763346, + "teaching_rms": 0.00362695184190896, + "traffic_rms": 0.07038163708152971 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.089782817435662e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.660622801091213e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00924721079102012, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 25696, + "train_examples": 45000, + "train_loss": 0.1752525559955173 + }, + { + "epoch": 74, + "eval_accuracy": 0.8716, + "eval_loss": 0.4177692199707031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12982839345932007, + "mean_feedback_forward_cosine": 0.9963331895489846, + "mean_feedback_forward_relative_error": 0.08447716481262638, + "min_feedback_forward_cosine": 0.9916701316833496 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003608410703062105, + "instruction_rms": 0.003608410702433176, + "raw_apical_rms": 0.07073821852932612, + "teaching_rms": 0.003608410703062105, + "traffic_rms": 0.07064285059448229 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.683637572748921e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.666976742797306e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009209060033436514, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26048, + "train_examples": 45000, + "train_loss": 0.17139004131423102 + }, + { + "epoch": 75, + "eval_accuracy": 0.8912, + "eval_loss": 0.37933065185546877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12515747547149658, + "mean_feedback_forward_cosine": 0.9965746345058564, + "mean_feedback_forward_relative_error": 0.08164186985983003, + "min_feedback_forward_cosine": 0.9922549724578857 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036249178952964113, + "instruction_rms": 0.0036249178952551097, + "raw_apical_rms": 0.07086841209431215, + "teaching_rms": 0.0036249178952964113, + "traffic_rms": 0.07077238220220579 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.203561587701188e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.666084813064403e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009161638659102488, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26400, + "train_examples": 45000, + "train_loss": 0.16958862608273825 + }, + { + "epoch": 76, + "eval_accuracy": 0.8812, + "eval_loss": 0.42446873779296873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.12056756764650345, + "mean_feedback_forward_cosine": 0.9968070810840975, + "mean_feedback_forward_relative_error": 0.07882179145611101, + "min_feedback_forward_cosine": 0.9928070306777954 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036848084598481933, + "instruction_rms": 0.0036848084591509693, + "raw_apical_rms": 0.07092062240412486, + "teaching_rms": 0.0036848084598481933, + "traffic_rms": 0.07082153183190511 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5348755034192436e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6594730304805588e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009131354313175306, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 26752, + "train_examples": 45000, + "train_loss": 0.1727997622595893 + }, + { + "epoch": 77, + "eval_accuracy": 0.8844, + "eval_loss": 0.3910231018066406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11632562428712845, + "mean_feedback_forward_cosine": 0.997019596638218, + "mean_feedback_forward_relative_error": 0.07614792000141836, + "min_feedback_forward_cosine": 0.9933022260665894 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003596840719711829, + "instruction_rms": 0.0035968407196002488, + "raw_apical_rms": 0.07113584166134894, + "teaching_rms": 0.003596840719711829, + "traffic_rms": 0.0710414819832535 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6839654882214745e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.662309649777286e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00984470253604775, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27104, + "train_examples": 45000, + "train_loss": 0.16724212563302782 + }, + { + "epoch": 78, + "eval_accuracy": 0.882, + "eval_loss": 0.3831909851074219, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.11226841062307358, + "mean_feedback_forward_cosine": 0.9972206527186979, + "mean_feedback_forward_relative_error": 0.07352818993310775, + "min_feedback_forward_cosine": 0.9937601685523987 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036420830789038183, + "instruction_rms": 0.0036420830789345047, + "raw_apical_rms": 0.07121224261868106, + "teaching_rms": 0.0036420830789038183, + "traffic_rms": 0.07111619993692685 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.624504429922126e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.66994563362321e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009543623209031843, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27456, + "train_examples": 45000, + "train_loss": 0.17047343085077074 + }, + { + "epoch": 79, + "eval_accuracy": 0.8678, + "eval_loss": 0.44401912841796876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10821803659200668, + "mean_feedback_forward_cosine": 0.9974058635773198, + "mean_feedback_forward_relative_error": 0.07102661531779074, + "min_feedback_forward_cosine": 0.9942016005516052 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036054255838827444, + "instruction_rms": 0.0036054255846796443, + "raw_apical_rms": 0.07135991644014633, + "teaching_rms": 0.0036054255838827444, + "traffic_rms": 0.07126583954021509 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.244164403781724e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6555204898444926e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009561302727504852, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 27808, + "train_examples": 45000, + "train_loss": 0.1651164001862208 + }, + { + "epoch": 80, + "eval_accuracy": 0.8776, + "eval_loss": 0.4436854675292969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10441537946462631, + "mean_feedback_forward_cosine": 0.9975783498056473, + "mean_feedback_forward_relative_error": 0.06861883546075513, + "min_feedback_forward_cosine": 0.9946001172065735 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035964310372020012, + "instruction_rms": 0.003596431037086029, + "raw_apical_rms": 0.07132739490731534, + "teaching_rms": 0.0035964310372020012, + "traffic_rms": 0.07123316787910322 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.702899320501274e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6588788726412646e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009810967358417407, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28160, + "train_examples": 45000, + "train_loss": 0.16397848246362473 + }, + { + "epoch": 81, + "eval_accuracy": 0.8776, + "eval_loss": 0.4233134704589844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.10063564032316208, + "mean_feedback_forward_cosine": 0.9977411646996776, + "mean_feedback_forward_relative_error": 0.06626483101037241, + "min_feedback_forward_cosine": 0.9949819445610046 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003571335612018009, + "instruction_rms": 0.0035713356125407895, + "raw_apical_rms": 0.07163790382306386, + "teaching_rms": 0.003571335612018009, + "traffic_rms": 0.07154558551446401 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.6408196518445948e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.656869259830858e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009733629286904164, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28512, + "train_examples": 45000, + "train_loss": 0.1652719701078203 + }, + { + "epoch": 82, + "eval_accuracy": 0.8726, + "eval_loss": 0.44266607055664065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.09689801186323166, + "mean_feedback_forward_cosine": 0.9979009993614689, + "mean_feedback_forward_relative_error": 0.06387322474150889, + "min_feedback_forward_cosine": 0.9953445196151733 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036833393936125253, + "instruction_rms": 0.003683339394709965, + "raw_apical_rms": 0.07160597339924625, + "teaching_rms": 0.0036833393936125253, + "traffic_rms": 0.07150795155961165 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.135566473133622e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6618241543435608e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009807831831993425, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 28864, + "train_examples": 45000, + "train_loss": 0.1679715323501163 + }, + { + "epoch": 83, + "eval_accuracy": 0.8666, + "eval_loss": 0.5030519104003907, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0934315100312233, + "mean_feedback_forward_cosine": 0.9980408107080767, + "mean_feedback_forward_relative_error": 0.06170915283503071, + "min_feedback_forward_cosine": 0.995671272277832 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003554035555607399, + "instruction_rms": 0.0035540355554000607, + "raw_apical_rms": 0.07172303359046345, + "teaching_rms": 0.003554035555607399, + "traffic_rms": 0.07163143160456198 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.2331570309906965e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6646891743525186e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009595793136706894, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29216, + "train_examples": 45000, + "train_loss": 0.16389296588367885 + }, + { + "epoch": 84, + "eval_accuracy": 0.8506, + "eval_loss": 0.5674269653320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0899653360247612, + "mean_feedback_forward_cosine": 0.998176436270437, + "mean_feedback_forward_relative_error": 0.05953777473299734, + "min_feedback_forward_cosine": 0.9959843754768372 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036174064463937107, + "instruction_rms": 0.0036174064458526912, + "raw_apical_rms": 0.07178923803194671, + "teaching_rms": 0.0036174064463937107, + "traffic_rms": 0.0716947802706199 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.38670602673119e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6597809357246895e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009311851541260572, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29568, + "train_examples": 45000, + "train_loss": 0.1703937453561359 + }, + { + "epoch": 85, + "eval_accuracy": 0.8916, + "eval_loss": 0.3849796752929687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08688816428184509, + "mean_feedback_forward_cosine": 0.9982977009588673, + "mean_feedback_forward_relative_error": 0.05752050305806821, + "min_feedback_forward_cosine": 0.9962527751922607 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003597530378386697, + "instruction_rms": 0.003597530377097671, + "raw_apical_rms": 0.0720722297302695, + "teaching_rms": 0.003597530378386697, + "traffic_rms": 0.07197911854139254 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.321809754515016e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6607927820744984e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00932571139363934, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 29920, + "train_examples": 45000, + "train_loss": 0.16465434232287937 + }, + { + "epoch": 86, + "eval_accuracy": 0.893, + "eval_loss": 0.37870723266601564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08391399681568146, + "mean_feedback_forward_cosine": 0.9984068870544434, + "mean_feedback_forward_relative_error": 0.05564795470526142, + "min_feedback_forward_cosine": 0.9965047240257263 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035002124601188843, + "instruction_rms": 0.0035002124596508924, + "raw_apical_rms": 0.07226275588302505, + "teaching_rms": 0.0035002124601188843, + "traffic_rms": 0.07217460705782518 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.464167264155549e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.656678765377475e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00945063351076668, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30272, + "train_examples": 45000, + "train_loss": 0.1550357510195838 + }, + { + "epoch": 87, + "eval_accuracy": 0.878, + "eval_loss": 0.42328782958984373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.08091253787279129, + "mean_feedback_forward_cosine": 0.998512170007152, + "mean_feedback_forward_relative_error": 0.0537769113697352, + "min_feedback_forward_cosine": 0.9967488050460815 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035730514073078613, + "instruction_rms": 0.0035730514078760292, + "raw_apical_rms": 0.07224889820710166, + "teaching_rms": 0.0035730514073078613, + "traffic_rms": 0.0721564351530741 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.410326954167431e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.668245969136442e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009184684011630855, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30624, + "train_examples": 45000, + "train_loss": 0.16146544399261475 + }, + { + "epoch": 88, + "eval_accuracy": 0.8832, + "eval_loss": 0.42677001953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07806982100009918, + "mean_feedback_forward_cosine": 0.9986101600431627, + "mean_feedback_forward_relative_error": 0.0519687618699766, + "min_feedback_forward_cosine": 0.9969735741615295 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003487752935243331, + "instruction_rms": 0.003487752934749015, + "raw_apical_rms": 0.07215588748529213, + "teaching_rms": 0.003487752935243331, + "traffic_rms": 0.07206799095668322 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.830646377878111e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.660071052193981e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009321216980913925, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 30976, + "train_examples": 45000, + "train_loss": 0.16241874133480919 + }, + { + "epoch": 89, + "eval_accuracy": 0.872, + "eval_loss": 0.46930485229492186, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07518605142831802, + "mean_feedback_forward_cosine": 0.9987055313202643, + "mean_feedback_forward_relative_error": 0.0501535901980054, + "min_feedback_forward_cosine": 0.9971925616264343 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035720601461635072, + "instruction_rms": 0.0035720601469758596, + "raw_apical_rms": 0.07244703868875312, + "teaching_rms": 0.0035720601461635072, + "traffic_rms": 0.07235575863016837 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.132954600777339e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6593580799944038e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009185582520955396, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31328, + "train_examples": 45000, + "train_loss": 0.163161230580012 + }, + { + "epoch": 90, + "eval_accuracy": 0.8746, + "eval_loss": 0.4608991455078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.07252176105976105, + "mean_feedback_forward_cosine": 0.9987933501120536, + "mean_feedback_forward_relative_error": 0.04842322115455904, + "min_feedback_forward_cosine": 0.9973869919776917 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035652092843790066, + "instruction_rms": 0.003565209284478349, + "raw_apical_rms": 0.07258737420948207, + "teaching_rms": 0.0035652092843790066, + "traffic_rms": 0.07249694187569879 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.627100042535858e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.66188932376684e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009473667757679357, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 31680, + "train_examples": 45000, + "train_loss": 0.1598022187034289 + }, + { + "epoch": 91, + "eval_accuracy": 0.8812, + "eval_loss": 0.41463336181640625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06982004642486572, + "mean_feedback_forward_cosine": 0.9988757660312038, + "mean_feedback_forward_relative_error": 0.04674123125451227, + "min_feedback_forward_cosine": 0.9975777268409729 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035763155716032986, + "instruction_rms": 0.0035763155708816653, + "raw_apical_rms": 0.07281488449730049, + "teaching_rms": 0.0035763155716032986, + "traffic_rms": 0.072723704848616 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.574161056008052e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6492086060836144e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009355969896967818, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32032, + "train_examples": 45000, + "train_loss": 0.16621972343126934 + }, + { + "epoch": 92, + "eval_accuracy": 0.88, + "eval_loss": 0.45427987060546876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06721863150596619, + "mean_feedback_forward_cosine": 0.9989541172981262, + "mean_feedback_forward_relative_error": 0.04508476139556977, + "min_feedback_forward_cosine": 0.9977545142173767 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0036187418742760565, + "instruction_rms": 0.003618741871613057, + "raw_apical_rms": 0.07319300477937919, + "teaching_rms": 0.0036187418742760565, + "traffic_rms": 0.07309999209704568 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.3080969902575816e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6510856252466656e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009249153718346741, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32384, + "train_examples": 45000, + "train_loss": 0.16049772697289785 + }, + { + "epoch": 93, + "eval_accuracy": 0.8842, + "eval_loss": 0.40804675903320314, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.06487451493740082, + "mean_feedback_forward_cosine": 0.9990244584698831, + "mean_feedback_forward_relative_error": 0.04354102500984745, + "min_feedback_forward_cosine": 0.9979076981544495 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003525129912335197, + "instruction_rms": 0.0035251299124563706, + "raw_apical_rms": 0.07322521898968991, + "teaching_rms": 0.003525129912335197, + "traffic_rms": 0.07313673749452419 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6598152064470924e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.650618967878567e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008946322908574451, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 32736, + "train_examples": 45000, + "train_loss": 0.15889199346436395 + }, + { + "epoch": 94, + "eval_accuracy": 0.8774, + "eval_loss": 0.4351710021972656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0624665692448616, + "mean_feedback_forward_cosine": 0.9990899043698465, + "mean_feedback_forward_relative_error": 0.04205331461684358, + "min_feedback_forward_cosine": 0.9980596899986267 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035437455909674825, + "instruction_rms": 0.003543745589771452, + "raw_apical_rms": 0.07332060943510744, + "teaching_rms": 0.0035437455909674825, + "traffic_rms": 0.07323156261518245 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.717973839911792e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6439794761812034e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008958942905872078, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33088, + "train_examples": 45000, + "train_loss": 0.16275668821732203 + }, + { + "epoch": 95, + "eval_accuracy": 0.8834, + "eval_loss": 0.40972940673828123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.060325946658849716, + "mean_feedback_forward_cosine": 0.9991485514948445, + "mean_feedback_forward_relative_error": 0.04066988294042887, + "min_feedback_forward_cosine": 0.9981906414031982 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034620313225154366, + "instruction_rms": 0.003462031322280486, + "raw_apical_rms": 0.07341350454202532, + "teaching_rms": 0.0034620313225154366, + "traffic_rms": 0.07332870689643071 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5315704138838555e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.665924058018557e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009248571732534813, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33440, + "train_examples": 45000, + "train_loss": 0.1549742729028066 + }, + { + "epoch": 96, + "eval_accuracy": 0.8874, + "eval_loss": 0.3855793701171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.058349937200546265, + "mean_feedback_forward_cosine": 0.9992040837964704, + "mean_feedback_forward_relative_error": 0.03932008099171423, + "min_feedback_forward_cosine": 0.9983067512512207 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0034875497038143145, + "instruction_rms": 0.003487549704974714, + "raw_apical_rms": 0.07365003879253579, + "teaching_rms": 0.0034875497038143145, + "traffic_rms": 0.07356420552182218 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 4.434439659917189e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6458171200670373e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00937028715119181, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 33792, + "train_examples": 45000, + "train_loss": 0.14878102054595949 + }, + { + "epoch": 97, + "eval_accuracy": 0.8886, + "eval_loss": 0.3953185546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.056168243288993835, + "mean_feedback_forward_cosine": 0.9992594199795877, + "mean_feedback_forward_relative_error": 0.0379281306879655, + "min_feedback_forward_cosine": 0.9984309077262878 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.0035957013318513083, + "instruction_rms": 0.0035957013326367734, + "raw_apical_rms": 0.0733056118526088, + "teaching_rms": 0.0035957013318513083, + "traffic_rms": 0.07321380774767293 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9192386225295195e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.650805863676102e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009429207017063875, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34144, + "train_examples": 45000, + "train_loss": 0.15785606362554763 + }, + { + "epoch": 98, + "eval_accuracy": 0.8842, + "eval_loss": 0.41677572631835935, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05430087074637413, + "mean_feedback_forward_cosine": 0.9993068614313679, + "mean_feedback_forward_relative_error": 0.03669139430407555, + "min_feedback_forward_cosine": 0.9985330700874329 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003495759529560324, + "instruction_rms": 0.003495759529936718, + "raw_apical_rms": 0.07341240825660733, + "teaching_rms": 0.003495759529560324, + "traffic_rms": 0.07332585321366156 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.8842856042720086e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.650165960022111e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009514835634155968, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34496, + "train_examples": 45000, + "train_loss": 0.1526362185160319 + }, + { + "epoch": 99, + "eval_accuracy": 0.8944, + "eval_loss": 0.3704293640136719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05233167111873627, + "mean_feedback_forward_cosine": 0.9993537126048919, + "mean_feedback_forward_relative_error": 0.03543054300450509, + "min_feedback_forward_cosine": 0.9986371397972107 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003504387026545859, + "instruction_rms": 0.0035043870275930696, + "raw_apical_rms": 0.07377413682739672, + "teaching_rms": 0.003504387026545859, + "traffic_rms": 0.07368727359965827 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.901835654573915e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6477510326228728e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009302213669263211, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 34848, + "train_examples": 45000, + "train_loss": 0.15529265949726104 + }, + { + "epoch": 100, + "eval_accuracy": 0.882, + "eval_loss": 0.4139551330566406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050475358963012695, + "mean_feedback_forward_cosine": 0.9993980373105695, + "mean_feedback_forward_relative_error": 0.03419552521119194, + "min_feedback_forward_cosine": 0.998731791973114 + }, + "lr": 0.1, + "mixed_apical": { + "innovation_rms": 0.003581658465411083, + "instruction_rms": 0.0035816584649890552, + "raw_apical_rms": 0.07379287180973258, + "teaching_rms": 0.003581658465411083, + "traffic_rms": 0.0737022879116428 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.6121676316724916e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6539305751499018e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009126019775016257, + "minimum_observations": 72 + }, + "output_lr": 0.1, + "step": 35200, + "train_examples": 45000, + "train_loss": 0.15373161282274458 + }, + { + "epoch": 101, + "eval_accuracy": 0.9166, + "eval_loss": 0.296131787109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05046016722917557, + "mean_feedback_forward_cosine": 0.999398410320282, + "mean_feedback_forward_relative_error": 0.034182172779354356, + "min_feedback_forward_cosine": 0.9987326264381409 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002503702058118298, + "instruction_rms": 0.0025037020573505588, + "raw_apical_rms": 0.07391774718212694, + "teaching_rms": 0.002503702058118298, + "traffic_rms": 0.07387247257723574 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 4.1269547779165805e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.636368132771414e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008899219379347055, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35552, + "train_examples": 45000, + "train_loss": 0.08968734825385942 + }, + { + "epoch": 102, + "eval_accuracy": 0.9174, + "eval_loss": 0.29645615234375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05045269802212715, + "mean_feedback_forward_cosine": 0.9993986218206344, + "mean_feedback_forward_relative_error": 0.034173633602838364, + "min_feedback_forward_cosine": 0.9987329244613647 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0021944999453406113, + "instruction_rms": 0.002194499946388587, + "raw_apical_rms": 0.07379360535339856, + "teaching_rms": 0.0021944999453406113, + "traffic_rms": 0.07375858013327992 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.687942040642156e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.640653593188287e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008763187863656765, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 35904, + "train_examples": 45000, + "train_loss": 0.06357012153069179 + }, + { + "epoch": 103, + "eval_accuracy": 0.9176, + "eval_loss": 0.29464952392578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05044633895158768, + "mean_feedback_forward_cosine": 0.9993987890981859, + "mean_feedback_forward_relative_error": 0.034165723489657525, + "min_feedback_forward_cosine": 0.9987332820892334 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002103950848009906, + "instruction_rms": 0.0021039508485826415, + "raw_apical_rms": 0.07372151276783091, + "teaching_rms": 0.002103950848009906, + "traffic_rms": 0.07368904657404154 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5674055709478125e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6338468449839717e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008848141077239939, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36256, + "train_examples": 45000, + "train_loss": 0.053366583776805136 + }, + { + "epoch": 104, + "eval_accuracy": 0.9194, + "eval_loss": 0.29925628662109377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05044025182723999, + "mean_feedback_forward_cosine": 0.9993989756030421, + "mean_feedback_forward_relative_error": 0.03415835798988419, + "min_feedback_forward_cosine": 0.9987336993217468 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.002059360329649582, + "instruction_rms": 0.002059360328937544, + "raw_apical_rms": 0.07365805133806275, + "teaching_rms": 0.002059360329649582, + "traffic_rms": 0.07362615429563846 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.2976984698507295e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6291663740232836e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008897021931968223, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36608, + "train_examples": 45000, + "train_loss": 0.047539353975984786 + }, + { + "epoch": 105, + "eval_accuracy": 0.9178, + "eval_loss": 0.3023534362792969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05043430253863335, + "mean_feedback_forward_cosine": 0.9993991640306288, + "mean_feedback_forward_relative_error": 0.03415105467842471, + "min_feedback_forward_cosine": 0.9987341165542603 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0020590809279698, + "instruction_rms": 0.002059080928048657, + "raw_apical_rms": 0.07356907459457078, + "teaching_rms": 0.0020590809279698, + "traffic_rms": 0.07353715436783345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.57055824906638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6422143636760872e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009101876718829222, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 36960, + "train_examples": 45000, + "train_loss": 0.043767406027846865 + }, + { + "epoch": 106, + "eval_accuracy": 0.9214, + "eval_loss": 0.3021748046875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05042882636189461, + "mean_feedback_forward_cosine": 0.9993993466900241, + "mean_feedback_forward_relative_error": 0.03414431998446103, + "min_feedback_forward_cosine": 0.9987342953681946 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001980464299154211, + "instruction_rms": 0.0019804642994078595, + "raw_apical_rms": 0.07350965809847608, + "teaching_rms": 0.001980464299154211, + "traffic_rms": 0.07348000631026358 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.873495302286756e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.637047972949355e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008936199094665764, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37312, + "train_examples": 45000, + "train_loss": 0.039899225935671065 + }, + { + "epoch": 107, + "eval_accuracy": 0.9212, + "eval_loss": 0.3047294006347656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050423093140125275, + "mean_feedback_forward_cosine": 0.9993994985857317, + "mean_feedback_forward_relative_error": 0.03413757090006144, + "min_feedback_forward_cosine": 0.9987345933914185 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019584899976528005, + "instruction_rms": 0.00195848999787617, + "raw_apical_rms": 0.07348220034259192, + "teaching_rms": 0.0019584899976528005, + "traffic_rms": 0.07345267258990162 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.621288513500076e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6403314786546443e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.008975976658212542, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 37664, + "train_examples": 45000, + "train_loss": 0.0374519247604741 + }, + { + "epoch": 108, + "eval_accuracy": 0.9196, + "eval_loss": 0.3181728271484375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05041760951280594, + "mean_feedback_forward_cosine": 0.9993996524041698, + "mean_feedback_forward_relative_error": 0.034131177580885345, + "min_feedback_forward_cosine": 0.9987348914146423 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0019020930719547322, + "instruction_rms": 0.0019020930718985536, + "raw_apical_rms": 0.07347984602626077, + "teaching_rms": 0.0019020930719547322, + "traffic_rms": 0.0734517919918475 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.617163812918989e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6394616077975364e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009199146910600758, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38016, + "train_examples": 45000, + "train_loss": 0.0350735355509652 + }, + { + "epoch": 109, + "eval_accuracy": 0.9182, + "eval_loss": 0.3142703857421875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05041196942329407, + "mean_feedback_forward_cosine": 0.9993998177589909, + "mean_feedback_forward_relative_error": 0.034124843535884734, + "min_feedback_forward_cosine": 0.9987351298332214 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018793636401583705, + "instruction_rms": 0.0018793636398705615, + "raw_apical_rms": 0.07344885593576457, + "teaching_rms": 0.0018793636401583705, + "traffic_rms": 0.07342106875997677 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.9828502334794393e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6351668861964216e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009067827007928238, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38368, + "train_examples": 45000, + "train_loss": 0.031972534075048235 + }, + { + "epoch": 110, + "eval_accuracy": 0.9198, + "eval_loss": 0.3206904541015625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05040634423494339, + "mean_feedback_forward_cosine": 0.9993999773456205, + "mean_feedback_forward_relative_error": 0.034118656760021564, + "min_feedback_forward_cosine": 0.9987354874610901 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018419724681878135, + "instruction_rms": 0.0018419724681582822, + "raw_apical_rms": 0.07341183097499239, + "teaching_rms": 0.0018419724681878135, + "traffic_rms": 0.07338468595792112 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.3905721724968316e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.632848519724072e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00905213726479151, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 38720, + "train_examples": 45000, + "train_loss": 0.03032042328119278 + }, + { + "epoch": 111, + "eval_accuracy": 0.9198, + "eval_loss": 0.32301548461914065, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050401121377944946, + "mean_feedback_forward_cosine": 0.9994001542368243, + "mean_feedback_forward_relative_error": 0.03411253298362417, + "min_feedback_forward_cosine": 0.9987357258796692 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018661703224029228, + "instruction_rms": 0.0018661703226740712, + "raw_apical_rms": 0.07339671282876449, + "teaching_rms": 0.0018661703224029228, + "traffic_rms": 0.07336889522340968 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.526874081671849e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6362773445108326e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009158179741872831, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39072, + "train_examples": 45000, + "train_loss": 0.029052887268861134 + }, + { + "epoch": 112, + "eval_accuracy": 0.9198, + "eval_loss": 0.3287255310058594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05039596930146217, + "mean_feedback_forward_cosine": 0.9994002945961491, + "mean_feedback_forward_relative_error": 0.034106574922559724, + "min_feedback_forward_cosine": 0.9987360239028931 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0018375925809945362, + "instruction_rms": 0.001837592581293839, + "raw_apical_rms": 0.07334994477907034, + "teaching_rms": 0.0018375925809945362, + "traffic_rms": 0.07332291949214792 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5672488962745774e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6330009403873294e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009117030862201607, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39424, + "train_examples": 45000, + "train_loss": 0.028106970258553823 + }, + { + "epoch": 113, + "eval_accuracy": 0.9196, + "eval_loss": 0.3258778381347656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050390880554914474, + "mean_feedback_forward_cosine": 0.9994004368782043, + "mean_feedback_forward_relative_error": 0.03410087453742181, + "min_feedback_forward_cosine": 0.9987362027168274 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001753882422299409, + "instruction_rms": 0.001753882421137723, + "raw_apical_rms": 0.0733300083731065, + "teaching_rms": 0.001753882422299409, + "traffic_rms": 0.07330515633365575 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.279366467268119e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383620473785716e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009157197080994243, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 39776, + "train_examples": 45000, + "train_loss": 0.025644417638580004 + }, + { + "epoch": 114, + "eval_accuracy": 0.9194, + "eval_loss": 0.3260003479003906, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05038650706410408, + "mean_feedback_forward_cosine": 0.9994005695466073, + "mean_feedback_forward_relative_error": 0.034095589251768206, + "min_feedback_forward_cosine": 0.9987364411354065 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016507215790715448, + "instruction_rms": 0.0016507215784339788, + "raw_apical_rms": 0.07326140117464916, + "teaching_rms": 0.0016507215790715448, + "traffic_rms": 0.0732391549592118 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.903891254708469e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.640263418771098e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009264603588974272, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40128, + "train_examples": 45000, + "train_loss": 0.022590324265095923 + }, + { + "epoch": 115, + "eval_accuracy": 0.9228, + "eval_loss": 0.3252786376953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0503816194832325, + "mean_feedback_forward_cosine": 0.9994007272105063, + "mean_feedback_forward_relative_error": 0.03409001198146613, + "min_feedback_forward_cosine": 0.9987367391586304 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001714554079436268, + "instruction_rms": 0.0017145540793167373, + "raw_apical_rms": 0.07326508180322594, + "teaching_rms": 0.001714554079436268, + "traffic_rms": 0.07324081471506289 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0334526829610695e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6416379274780682e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009108295278790897, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40480, + "train_examples": 45000, + "train_loss": 0.023377065229415895 + }, + { + "epoch": 116, + "eval_accuracy": 0.9218, + "eval_loss": 0.3257878662109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05037691816687584, + "mean_feedback_forward_cosine": 0.9994008925653273, + "mean_feedback_forward_relative_error": 0.03408442137222136, + "min_feedback_forward_cosine": 0.9987369179725647 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001714413491093942, + "instruction_rms": 0.0017144134921156249, + "raw_apical_rms": 0.07323759252924394, + "teaching_rms": 0.001714413491093942, + "traffic_rms": 0.07321339172730634 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.17054195622768e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6411490192345613e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009176613455319031, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 40832, + "train_examples": 45000, + "train_loss": 0.022235628300905228 + }, + { + "epoch": 117, + "eval_accuracy": 0.9212, + "eval_loss": 0.3269914794921875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050372641533613205, + "mean_feedback_forward_cosine": 0.9994010060064255, + "mean_feedback_forward_relative_error": 0.034079157176517674, + "min_feedback_forward_cosine": 0.998737096786499 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016238884290508307, + "instruction_rms": 0.0016238884287559203, + "raw_apical_rms": 0.07328271932961998, + "teaching_rms": 0.0016238884290508307, + "traffic_rms": 0.07326048242890883 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.549012106418559e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6376774560338555e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0092378843029572, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41184, + "train_examples": 45000, + "train_loss": 0.020025251053439247 + }, + { + "epoch": 118, + "eval_accuracy": 0.9224, + "eval_loss": 0.33299217529296876, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05036759003996849, + "mean_feedback_forward_cosine": 0.999401169438516, + "mean_feedback_forward_relative_error": 0.03407376680162645, + "min_feedback_forward_cosine": 0.9987374544143677 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016813589612660286, + "instruction_rms": 0.0016813589623253867, + "raw_apical_rms": 0.07328028315887815, + "teaching_rms": 0.0016813589612660286, + "traffic_rms": 0.07325650206488526 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.0073060486301983e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6392372447890742e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009127665566403929, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41536, + "train_examples": 45000, + "train_loss": 0.021251448687952427 + }, + { + "epoch": 119, + "eval_accuracy": 0.9218, + "eval_loss": 0.33599387817382814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05036266893148422, + "mean_feedback_forward_cosine": 0.9994013251796845, + "mean_feedback_forward_relative_error": 0.034068433087198965, + "min_feedback_forward_cosine": 0.9987376928329468 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0016793044916914072, + "instruction_rms": 0.0016793044920357545, + "raw_apical_rms": 0.07320856823210507, + "teaching_rms": 0.0016793044916914072, + "traffic_rms": 0.0731843035091355 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.590515262885674e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6402947751876797e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009107391532663028, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 41888, + "train_examples": 45000, + "train_loss": 0.0202204686784082 + }, + { + "epoch": 120, + "eval_accuracy": 0.922, + "eval_loss": 0.3391268798828125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050358157604932785, + "mean_feedback_forward_cosine": 0.999401459770818, + "mean_feedback_forward_relative_error": 0.03406337510433889, + "min_feedback_forward_cosine": 0.9987379312515259 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015648042851772904, + "instruction_rms": 0.0015648042850207287, + "raw_apical_rms": 0.07318844722301304, + "teaching_rms": 0.0015648042851772904, + "traffic_rms": 0.07316722756940557 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.5761722472216206e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.649314614754579e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00920271237294771, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42240, + "train_examples": 45000, + "train_loss": 0.018327527635296187 + }, + { + "epoch": 121, + "eval_accuracy": 0.9204, + "eval_loss": 0.34658875732421873, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05035347491502762, + "mean_feedback_forward_cosine": 0.9994015885937598, + "mean_feedback_forward_relative_error": 0.03405820316965542, + "min_feedback_forward_cosine": 0.998738169670105 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001574837191435334, + "instruction_rms": 0.0015748371918382507, + "raw_apical_rms": 0.07316225684028811, + "teaching_rms": 0.001574837191435334, + "traffic_rms": 0.07314047159639461 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.4941205484528837e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.645228259004105e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009332900625260264, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42592, + "train_examples": 45000, + "train_loss": 0.01816672069641451 + }, + { + "epoch": 122, + "eval_accuracy": 0.9202, + "eval_loss": 0.34789984130859375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05034828931093216, + "mean_feedback_forward_cosine": 0.9994017520258504, + "mean_feedback_forward_relative_error": 0.03405260583085398, + "min_feedback_forward_cosine": 0.9987384676933289 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001662376227696435, + "instruction_rms": 0.0016623762277789668, + "raw_apical_rms": 0.07318676003390975, + "teaching_rms": 0.001662376227696435, + "traffic_rms": 0.0731627205265084 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.041743650873286e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.635776606278276e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009076971193582126, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 42944, + "train_examples": 45000, + "train_loss": 0.019140352391192897 + }, + { + "epoch": 123, + "eval_accuracy": 0.922, + "eval_loss": 0.35231456298828123, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05034376308321953, + "mean_feedback_forward_cosine": 0.9994018885397142, + "mean_feedback_forward_relative_error": 0.03404748508886945, + "min_feedback_forward_cosine": 0.998738706111908 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015854376293903563, + "instruction_rms": 0.001585437629724215, + "raw_apical_rms": 0.0731196388448265, + "teaching_rms": 0.0015854376293903563, + "traffic_rms": 0.0730979823387364 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.055430918332604e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6426968928782518e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009258208004321764, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43296, + "train_examples": 45000, + "train_loss": 0.017558129587107233 + }, + { + "epoch": 124, + "eval_accuracy": 0.9232, + "eval_loss": 0.34616173706054687, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05033883452415466, + "mean_feedback_forward_cosine": 0.9994020442808828, + "mean_feedback_forward_relative_error": 0.03404202627679033, + "min_feedback_forward_cosine": 0.9987388253211975 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015925764422237579, + "instruction_rms": 0.0015925764416803519, + "raw_apical_rms": 0.07313878644961624, + "teaching_rms": 0.0015925764422237579, + "traffic_rms": 0.073115404526152 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.379950269139954e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6441229389360017e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009274810200918145, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 43648, + "train_examples": 45000, + "train_loss": 0.016589551832692492 + }, + { + "epoch": 125, + "eval_accuracy": 0.9216, + "eval_loss": 0.3495522033691406, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05033367872238159, + "mean_feedback_forward_cosine": 0.9994022038675123, + "mean_feedback_forward_relative_error": 0.034036547636553165, + "min_feedback_forward_cosine": 0.9987391829490662 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015926960648308965, + "instruction_rms": 0.001592696064703634, + "raw_apical_rms": 0.07310383161859343, + "teaching_rms": 0.0015926960648308965, + "traffic_rms": 0.07308115086094133 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.832180439644617e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642661790056256e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00918368779003918, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44000, + "train_examples": 45000, + "train_loss": 0.017056103925241366 + }, + { + "epoch": 126, + "eval_accuracy": 0.9216, + "eval_loss": 0.35535875244140624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05032917112112045, + "mean_feedback_forward_cosine": 0.99940235576322, + "mean_feedback_forward_relative_error": 0.03403143389450927, + "min_feedback_forward_cosine": 0.9987393021583557 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001532232881593061, + "instruction_rms": 0.0015322328819102713, + "raw_apical_rms": 0.07310194505776765, + "teaching_rms": 0.001532232881593061, + "traffic_rms": 0.07308083408976394 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.162943494951833e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641713835608175e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009221676830231031, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44352, + "train_examples": 45000, + "train_loss": 0.015211372663577397 + }, + { + "epoch": 127, + "eval_accuracy": 0.9216, + "eval_loss": 0.35379696044921877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05032459646463394, + "mean_feedback_forward_cosine": 0.9994024980452753, + "mean_feedback_forward_relative_error": 0.034026270762326255, + "min_feedback_forward_cosine": 0.9987396001815796 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015258038387716882, + "instruction_rms": 0.0015258038388503378, + "raw_apical_rms": 0.07307827300679864, + "teaching_rms": 0.0015258038387716882, + "traffic_rms": 0.07305702099121773 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.6272851389185234e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6435527168634897e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009130628426376854, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 44704, + "train_examples": 45000, + "train_loss": 0.015641128700640466 + }, + { + "epoch": 128, + "eval_accuracy": 0.9222, + "eval_loss": 0.35718463745117185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05031966418027878, + "mean_feedback_forward_cosine": 0.9994026287909477, + "mean_feedback_forward_relative_error": 0.03402112616646674, + "min_feedback_forward_cosine": 0.9987397193908691 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0015096897160306836, + "instruction_rms": 0.0015096897163843694, + "raw_apical_rms": 0.07305015673281999, + "teaching_rms": 0.0015096897160306836, + "traffic_rms": 0.07302892372884263 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.438694745090288e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6452464620442568e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00925408248877178, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45056, + "train_examples": 45000, + "train_loss": 0.015048105148474376 + }, + { + "epoch": 129, + "eval_accuracy": 0.9222, + "eval_loss": 0.35528975830078124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05031528323888779, + "mean_feedback_forward_cosine": 0.9994027787639249, + "mean_feedback_forward_relative_error": 0.03401619577479939, + "min_feedback_forward_cosine": 0.9987400770187378 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001467657122464643, + "instruction_rms": 0.0014676571223915448, + "raw_apical_rms": 0.07306057313662985, + "teaching_rms": 0.001467657122464643, + "traffic_rms": 0.07304073399270669 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.058058339888703e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6418627325526776e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009300725272727057, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45408, + "train_examples": 45000, + "train_loss": 0.01413145841260751 + }, + { + "epoch": 130, + "eval_accuracy": 0.9212, + "eval_loss": 0.35853994750976564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05031029134988785, + "mean_feedback_forward_cosine": 0.9994029498869373, + "mean_feedback_forward_relative_error": 0.03401098742840752, + "min_feedback_forward_cosine": 0.9987403154373169 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014523157466972074, + "instruction_rms": 0.0014523157468996032, + "raw_apical_rms": 0.0730532038867201, + "teaching_rms": 0.0014523157466972074, + "traffic_rms": 0.07303325276072449 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.5567165213024055e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6394125717665666e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009217598344823634, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 45760, + "train_examples": 45000, + "train_loss": 0.014199002430588007 + }, + { + "epoch": 131, + "eval_accuracy": 0.9226, + "eval_loss": 0.35520433349609376, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05030573904514313, + "mean_feedback_forward_cosine": 0.9994030998599145, + "mean_feedback_forward_relative_error": 0.03400599851363128, + "min_feedback_forward_cosine": 0.9987406134605408 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014316561964262543, + "instruction_rms": 0.0014316561966733975, + "raw_apical_rms": 0.07306319828164792, + "teaching_rms": 0.0014316561964262543, + "traffic_rms": 0.07304411608409099 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9560333675581205e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6430615926998932e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009323446125707572, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46112, + "train_examples": 45000, + "train_loss": 0.013526438065618276 + }, + { + "epoch": 132, + "eval_accuracy": 0.9236, + "eval_loss": 0.36005784912109373, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05030115693807602, + "mean_feedback_forward_cosine": 0.9994032306055869, + "mean_feedback_forward_relative_error": 0.03400100389074895, + "min_feedback_forward_cosine": 0.9987407922744751 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014382695782894992, + "instruction_rms": 0.0014382695778631363, + "raw_apical_rms": 0.07307802445809876, + "teaching_rms": 0.0014382695782894992, + "traffic_rms": 0.07305819674230177 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.118337315617282e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642000280238486e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009196017944382502, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46464, + "train_examples": 45000, + "train_loss": 0.013548072084453371 + }, + { + "epoch": 133, + "eval_accuracy": 0.9214, + "eval_loss": 0.36589906005859374, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05029694363474846, + "mean_feedback_forward_cosine": 0.9994033517376069, + "mean_feedback_forward_relative_error": 0.03399634926069167, + "min_feedback_forward_cosine": 0.998741090297699 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013199388175307158, + "instruction_rms": 0.0013199388177408222, + "raw_apical_rms": 0.0730877369466105, + "teaching_rms": 0.0013199388175307158, + "traffic_rms": 0.07307120245075292 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.3117224518264265e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641820543158784e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009270692331866448, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 46816, + "train_examples": 45000, + "train_loss": 0.011600939117082291 + }, + { + "epoch": 134, + "eval_accuracy": 0.922, + "eval_loss": 0.3665016418457031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05029251053929329, + "mean_feedback_forward_cosine": 0.9994034997878536, + "mean_feedback_forward_relative_error": 0.033991458885852366, + "min_feedback_forward_cosine": 0.998741090297699 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001424050863529644, + "instruction_rms": 0.0014240508637996918, + "raw_apical_rms": 0.07304906485194272, + "teaching_rms": 0.001424050863529644, + "traffic_rms": 0.07302919407436295 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.679631532804706e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6435836091159347e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009450031870753627, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47168, + "train_examples": 45000, + "train_loss": 0.01283611283360256 + }, + { + "epoch": 135, + "eval_accuracy": 0.9224, + "eval_loss": 0.3726066833496094, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05028746649622917, + "mean_feedback_forward_cosine": 0.9994036516835613, + "mean_feedback_forward_relative_error": 0.033986137368746346, + "min_feedback_forward_cosine": 0.9987415075302124 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001479838315273072, + "instruction_rms": 0.0014798383162909754, + "raw_apical_rms": 0.07298577671049526, + "teaching_rms": 0.001479838315273072, + "traffic_rms": 0.07296484025762337 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.291577499453524e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6416886489302135e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009370776115057027, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47520, + "train_examples": 45000, + "train_loss": 0.013333194206986162 + }, + { + "epoch": 136, + "eval_accuracy": 0.923, + "eval_loss": 0.3642227478027344, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05028301104903221, + "mean_feedback_forward_cosine": 0.9994038228065737, + "mean_feedback_forward_relative_error": 0.033981250418770696, + "min_feedback_forward_cosine": 0.9987417459487915 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013616196736991172, + "instruction_rms": 0.0013616196737312946, + "raw_apical_rms": 0.07297882030325077, + "teaching_rms": 0.0013616196736991172, + "traffic_rms": 0.07296049545588662 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9056946121718283e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642310615525088e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009484388583996937, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 47872, + "train_examples": 45000, + "train_loss": 0.011724352257781559 + }, + { + "epoch": 137, + "eval_accuracy": 0.9236, + "eval_loss": 0.36472017822265623, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05027836188673973, + "mean_feedback_forward_cosine": 0.999403957397707, + "mean_feedback_forward_relative_error": 0.03397634877793251, + "min_feedback_forward_cosine": 0.9987421035766602 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001342822403989234, + "instruction_rms": 0.0013428224040910272, + "raw_apical_rms": 0.0729822857742375, + "teaching_rms": 0.001342822403989234, + "traffic_rms": 0.07296335682268686 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.8794461804391176e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6351492112505394e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009328691442835048, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48224, + "train_examples": 45000, + "train_loss": 0.011808820550143718 + }, + { + "epoch": 138, + "eval_accuracy": 0.921, + "eval_loss": 0.3714885986328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05027312412858009, + "mean_feedback_forward_cosine": 0.9994041208297976, + "mean_feedback_forward_relative_error": 0.03397086391886396, + "min_feedback_forward_cosine": 0.9987423419952393 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014630978232266752, + "instruction_rms": 0.0014630978235427672, + "raw_apical_rms": 0.07301775381295596, + "teaching_rms": 0.0014630978232266752, + "traffic_rms": 0.0729966487923406 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.932416430847297e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.637262266643452e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009256529178537804, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48576, + "train_examples": 45000, + "train_loss": 0.012508721241851649 + }, + { + "epoch": 139, + "eval_accuracy": 0.9248, + "eval_loss": 0.36586863403320313, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05026783421635628, + "mean_feedback_forward_cosine": 0.9994043227164976, + "mean_feedback_forward_relative_error": 0.03396528625800725, + "min_feedback_forward_cosine": 0.9987425804138184 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0014520432637409404, + "instruction_rms": 0.0014520432634355302, + "raw_apical_rms": 0.0730057830481586, + "teaching_rms": 0.0014520432637409404, + "traffic_rms": 0.07298413660330243 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1338825695902415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.635901133150022e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009179780753142939, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 48928, + "train_examples": 45000, + "train_loss": 0.012142394143508541 + }, + { + "epoch": 140, + "eval_accuracy": 0.9224, + "eval_loss": 0.376943359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05026276409626007, + "mean_feedback_forward_cosine": 0.9994044380803262, + "mean_feedback_forward_relative_error": 0.03396030199984389, + "min_feedback_forward_cosine": 0.9987427592277527 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013495141283864642, + "instruction_rms": 0.001349514128180668, + "raw_apical_rms": 0.07296538552645916, + "teaching_rms": 0.0013495141283864642, + "traffic_rms": 0.07294706516908853 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.5952443628038964e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6394641085446702e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009383143381762667, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49280, + "train_examples": 45000, + "train_loss": 0.011706718593173557 + }, + { + "epoch": 141, + "eval_accuracy": 0.9246, + "eval_loss": 0.36509554443359377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05025757476687431, + "mean_feedback_forward_cosine": 0.9994046015124167, + "mean_feedback_forward_relative_error": 0.03395504402297159, + "min_feedback_forward_cosine": 0.9987429976463318 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013513653636865344, + "instruction_rms": 0.0013513653627506327, + "raw_apical_rms": 0.07296948617265846, + "teaching_rms": 0.0013513653636865344, + "traffic_rms": 0.07295083631529936 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.399305370521688e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6372433844617886e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009366212179531961, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49632, + "train_examples": 45000, + "train_loss": 0.011151374283226 + }, + { + "epoch": 142, + "eval_accuracy": 0.9242, + "eval_loss": 0.37224436645507814, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050253115594387054, + "mean_feedback_forward_cosine": 0.999404739949011, + "mean_feedback_forward_relative_error": 0.03395028944097219, + "min_feedback_forward_cosine": 0.9987432360649109 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012526717766469747, + "instruction_rms": 0.001252671775948499, + "raw_apical_rms": 0.07294409291308425, + "teaching_rms": 0.0012526717766469747, + "traffic_rms": 0.07292772282963204 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1604881317880427e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642280184927957e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009348331796996548, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 49984, + "train_examples": 45000, + "train_loss": 0.009966925987145967 + }, + { + "epoch": 143, + "eval_accuracy": 0.9234, + "eval_loss": 0.38368392333984375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050248317420482635, + "mean_feedback_forward_cosine": 0.9994048918447187, + "mean_feedback_forward_relative_error": 0.033945409641150504, + "min_feedback_forward_cosine": 0.99874347448349 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001286189656071483, + "instruction_rms": 0.0012861896555978786, + "raw_apical_rms": 0.07294821075933687, + "teaching_rms": 0.001286189656071483, + "traffic_rms": 0.07293130182317659 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.6473304255177936e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6375652749029974e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009365291032438687, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50336, + "train_examples": 45000, + "train_loss": 0.010365842316134109 + }, + { + "epoch": 144, + "eval_accuracy": 0.9232, + "eval_loss": 0.383241064453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050243910402059555, + "mean_feedback_forward_cosine": 0.9994050168221996, + "mean_feedback_forward_relative_error": 0.03394059752745013, + "min_feedback_forward_cosine": 0.9987436532974243 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001276116556537042, + "instruction_rms": 0.0012761165568696384, + "raw_apical_rms": 0.07292939398541631, + "teaching_rms": 0.001276116556537042, + "traffic_rms": 0.07291261704185605 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.370652379430794e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.642175233349516e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.0091762770400409, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 50688, + "train_examples": 45000, + "train_loss": 0.00985723092638784 + }, + { + "epoch": 145, + "eval_accuracy": 0.9242, + "eval_loss": 0.37986845703125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05023950710892677, + "mean_feedback_forward_cosine": 0.9994051514133331, + "mean_feedback_forward_relative_error": 0.03393592354991744, + "min_feedback_forward_cosine": 0.9987438321113586 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012351150970882408, + "instruction_rms": 0.0012351150978570596, + "raw_apical_rms": 0.07293325650420523, + "teaching_rms": 0.0012351150970882408, + "traffic_rms": 0.07291708632823415 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.577432477503862e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6395470741251502e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009204737415403922, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51040, + "train_examples": 45000, + "train_loss": 0.009665863710972998 + }, + { + "epoch": 146, + "eval_accuracy": 0.923, + "eval_loss": 0.38360340576171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05023485794663429, + "mean_feedback_forward_cosine": 0.9994053302272674, + "mean_feedback_forward_relative_error": 0.03393060140191547, + "min_feedback_forward_cosine": 0.9987441301345825 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012814315542508852, + "instruction_rms": 0.001281431554315269, + "raw_apical_rms": 0.07287779382947894, + "teaching_rms": 0.0012814315542508852, + "traffic_rms": 0.07286031956870602 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.3912446184558576e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6364850692780833e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00932224725261283, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51392, + "train_examples": 45000, + "train_loss": 0.009358042332612806 + }, + { + "epoch": 147, + "eval_accuracy": 0.9232, + "eval_loss": 0.3824809326171875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05023019015789032, + "mean_feedback_forward_cosine": 0.9994054609729398, + "mean_feedback_forward_relative_error": 0.033925831738498904, + "min_feedback_forward_cosine": 0.9987443089485168 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0012637335666448634, + "instruction_rms": 0.0012637335668080345, + "raw_apical_rms": 0.07287240860193166, + "teaching_rms": 0.0012637335666448634, + "traffic_rms": 0.07285584999262126 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.68501895056761e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.641440568636575e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009306209725214905, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 51744, + "train_examples": 45000, + "train_loss": 0.009624289382663038 + }, + { + "epoch": 148, + "eval_accuracy": 0.9222, + "eval_loss": 0.38150341796875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05022507533431053, + "mean_feedback_forward_cosine": 0.9994056224822998, + "mean_feedback_forward_relative_error": 0.03392021559299, + "min_feedback_forward_cosine": 0.998744547367096 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0013185545309442087, + "instruction_rms": 0.0013185545306520783, + "raw_apical_rms": 0.07286871263122031, + "teaching_rms": 0.0013185545309442087, + "traffic_rms": 0.07285093943959148 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.156742505756483e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6395440969528884e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009326790193415034, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52096, + "train_examples": 45000, + "train_loss": 0.01013564128163788 + }, + { + "epoch": 149, + "eval_accuracy": 0.9204, + "eval_loss": 0.38396596069335936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050220828503370285, + "mean_feedback_forward_cosine": 0.9994057378461284, + "mean_feedback_forward_relative_error": 0.03391571517192548, + "min_feedback_forward_cosine": 0.9987448453903198 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011543139337207263, + "instruction_rms": 0.0011543139338205891, + "raw_apical_rms": 0.07282128451272232, + "teaching_rms": 0.0011543139337207263, + "traffic_rms": 0.0728072899533545 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0628336250847497e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63922328067474e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00925633488855441, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52448, + "train_examples": 45000, + "train_loss": 0.008327673875954416 + }, + { + "epoch": 150, + "eval_accuracy": 0.9214, + "eval_loss": 0.38366978759765624, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502161830663681, + "mean_feedback_forward_cosine": 0.9994059032009494, + "mean_feedback_forward_relative_error": 0.03391084285272706, + "min_feedback_forward_cosine": 0.9987449645996094 + }, + "lr": 0.010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001262952555487919, + "instruction_rms": 0.001262952555335136, + "raw_apical_rms": 0.07285522659217394, + "teaching_rms": 0.001262952555487919, + "traffic_rms": 0.07283838657013365 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.671799930771158e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6390587803751088e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009380069419239363, + "minimum_observations": 72 + }, + "output_lr": 0.010000000000000002, + "step": 52800, + "train_examples": 45000, + "train_loss": 0.009336604690717325 + }, + { + "epoch": 151, + "eval_accuracy": 0.922, + "eval_loss": 0.3788788818359375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021611601114273, + "mean_feedback_forward_cosine": 0.9994059051236799, + "mean_feedback_forward_relative_error": 0.03391070210284764, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011623118793956482, + "instruction_rms": 0.0011623118799001153, + "raw_apical_rms": 0.07284443897297617, + "teaching_rms": 0.0011623118793956482, + "traffic_rms": 0.0728301011721793 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.582945851121622e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6423221738816907e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009589571673323487, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53152, + "train_examples": 45000, + "train_loss": 0.008326395595901542 + }, + { + "epoch": 152, + "eval_accuracy": 0.9234, + "eval_loss": 0.3746349792480469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021607503294945, + "mean_feedback_forward_cosine": 0.9994059262737152, + "mean_feedback_forward_relative_error": 0.03391058439569127, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0011402023765820348, + "instruction_rms": 0.0011402023772544708, + "raw_apical_rms": 0.07284256320758146, + "teaching_rms": 0.0011402023765820348, + "traffic_rms": 0.07282858728433825 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9096363479984575e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6379234733280538e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009477752188391394, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53504, + "train_examples": 45000, + "train_loss": 0.00844092088109917 + }, + { + "epoch": 153, + "eval_accuracy": 0.9216, + "eval_loss": 0.37851375732421877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021604523062706, + "mean_feedback_forward_cosine": 0.9994059224282542, + "mean_feedback_forward_relative_error": 0.03391046849109473, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010433452263418365, + "instruction_rms": 0.0010433452267703878, + "raw_apical_rms": 0.07283875120750984, + "teaching_rms": 0.0010433452263418365, + "traffic_rms": 0.07282653566874077 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.2417408252504174e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6392866256194406e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009271696744672719, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 53856, + "train_examples": 45000, + "train_loss": 0.007334845918748114 + }, + { + "epoch": 154, + "eval_accuracy": 0.9218, + "eval_loss": 0.3769091857910156, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021600425243378, + "mean_feedback_forward_cosine": 0.999405941655559, + "mean_feedback_forward_relative_error": 0.03391035447918599, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010592078552617693, + "instruction_rms": 0.0010592078555815578, + "raw_apical_rms": 0.0728372357261761, + "teaching_rms": 0.0010592078552617693, + "traffic_rms": 0.0728243414193174 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.9540491769685104e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6361725490104595e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009313667526166636, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54208, + "train_examples": 45000, + "train_loss": 0.007543874210533169 + }, + { + "epoch": 155, + "eval_accuracy": 0.9214, + "eval_loss": 0.37496671142578125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021597072482109, + "mean_feedback_forward_cosine": 0.9994059051236799, + "mean_feedback_forward_relative_error": 0.033910241699026476, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010345856088621523, + "instruction_rms": 0.0010345856090650898, + "raw_apical_rms": 0.0728365529366768, + "teaching_rms": 0.0010345856088621523, + "traffic_rms": 0.07282455131023481 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.8732434976273e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638016824703517e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009354873480270223, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54560, + "train_examples": 45000, + "train_loss": 0.007022323927697209 + }, + { + "epoch": 156, + "eval_accuracy": 0.9212, + "eval_loss": 0.375509814453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021592974662781, + "mean_feedback_forward_cosine": 0.9994059262737152, + "mean_feedback_forward_relative_error": 0.03391013267419992, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001016232065481463, + "instruction_rms": 0.0010162320656357457, + "raw_apical_rms": 0.07282690160517119, + "teaching_rms": 0.001016232065481463, + "traffic_rms": 0.07281513292762037 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1301920106207035e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6382168537115793e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009296229783639766, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 54912, + "train_examples": 45000, + "train_loss": 0.006862791797270377 + }, + { + "epoch": 157, + "eval_accuracy": 0.922, + "eval_loss": 0.3778203308105469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021589994430542, + "mean_feedback_forward_cosine": 0.9994059301191761, + "mean_feedback_forward_relative_error": 0.03391001493700089, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010073372393457304, + "instruction_rms": 0.0010073372392150456, + "raw_apical_rms": 0.07282213680545249, + "teaching_rms": 0.0010073372393457304, + "traffic_rms": 0.07281038702447838 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.664934311586876e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6348023378875587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009261372796807976, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55264, + "train_examples": 45000, + "train_loss": 0.007121293383919531 + }, + { + "epoch": 158, + "eval_accuracy": 0.9224, + "eval_loss": 0.3786832702636719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215862691402435, + "mean_feedback_forward_cosine": 0.9994059281964456, + "mean_feedback_forward_relative_error": 0.03390990684349691, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009815531392525657, + "instruction_rms": 0.0009815531391464748, + "raw_apical_rms": 0.07282119378546283, + "teaching_rms": 0.0009815531392525657, + "traffic_rms": 0.072810389042785 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.7329989766021754e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6373333040449887e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009549073501831705, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55616, + "train_examples": 45000, + "train_loss": 0.0069797852007879154 + }, + { + "epoch": 159, + "eval_accuracy": 0.9238, + "eval_loss": 0.374742138671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215814262628555, + "mean_feedback_forward_cosine": 0.9994059281964456, + "mean_feedback_forward_relative_error": 0.033909794604105335, + "min_feedback_forward_cosine": 0.9987450242042542 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010119271228425945, + "instruction_rms": 0.0010119271229596044, + "raw_apical_rms": 0.07282141540928444, + "teaching_rms": 0.0010119271228425945, + "traffic_rms": 0.07280968992056051 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1533684935757265e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63807837813095e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009365390021928427, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 55968, + "train_examples": 45000, + "train_loss": 0.007130757482639617 + }, + { + "epoch": 160, + "eval_accuracy": 0.9234, + "eval_loss": 0.37453284912109375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215788185596466, + "mean_feedback_forward_cosine": 0.999405941655559, + "mean_feedback_forward_relative_error": 0.033909681403348525, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0010204216858368256, + "instruction_rms": 0.0010204216861331291, + "raw_apical_rms": 0.07281743260799332, + "teaching_rms": 0.0010204216858368256, + "traffic_rms": 0.0728055896562762 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.4760297751290636e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.636717054432205e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009435143997698714, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56320, + "train_examples": 45000, + "train_loss": 0.007176053299734162 + }, + { + "epoch": 161, + "eval_accuracy": 0.9244, + "eval_loss": 0.3757555358886719, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021575093269348, + "mean_feedback_forward_cosine": 0.9994059397328284, + "mean_feedback_forward_relative_error": 0.03390957120685808, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009790675754742858, + "instruction_rms": 0.0009790675752193514, + "raw_apical_rms": 0.07281026428724338, + "teaching_rms": 0.0009790675754742858, + "traffic_rms": 0.07279962064713329 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.136117499025204e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639591172495471e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009261054781229678, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 56672, + "train_examples": 45000, + "train_loss": 0.0067126722757187155 + }, + { + "epoch": 162, + "eval_accuracy": 0.9214, + "eval_loss": 0.37511260375976563, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021572858095169, + "mean_feedback_forward_cosine": 0.9994059320419065, + "mean_feedback_forward_relative_error": 0.03390947230640919, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008729864156213054, + "instruction_rms": 0.0008729864154180971, + "raw_apical_rms": 0.07280216728271632, + "teaching_rms": 0.0008729864156213054, + "traffic_rms": 0.07279293527425085 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.601536846531417e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6374553336649922e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009376385483806822, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57024, + "train_examples": 45000, + "train_loss": 0.006105931394402351 + }, + { + "epoch": 163, + "eval_accuracy": 0.9228, + "eval_loss": 0.3784532287597656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502156987786293, + "mean_feedback_forward_cosine": 0.9994059147373322, + "mean_feedback_forward_relative_error": 0.03390936520431311, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.001044296922700967, + "instruction_rms": 0.0010442969225778944, + "raw_apical_rms": 0.07280095812172736, + "teaching_rms": 0.001044296922700967, + "traffic_rms": 0.07278821216121345 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092969417572, + "maximum_absolute_post_projection_soma_slope": 3.5597199854464634e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6371352304211653e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009319832459995793, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57376, + "train_examples": 45000, + "train_loss": 0.007222668740732802 + }, + { + "epoch": 164, + "eval_accuracy": 0.9208, + "eval_loss": 0.3755808349609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021568015217781, + "mean_feedback_forward_cosine": 0.9994059339646371, + "mean_feedback_forward_relative_error": 0.03390925957430755, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009426020207152024, + "instruction_rms": 0.0009426020209865341, + "raw_apical_rms": 0.07279991349257399, + "teaching_rms": 0.0009426020207152024, + "traffic_rms": 0.07278962840524515 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.144392834997234e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639210997898199e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009353989720007325, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 57728, + "train_examples": 45000, + "train_loss": 0.006089698071612252 + }, + { + "epoch": 165, + "eval_accuracy": 0.923, + "eval_loss": 0.37624190063476565, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215646624565125, + "mean_feedback_forward_cosine": 0.9994059397328284, + "mean_feedback_forward_relative_error": 0.03390915436489928, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009311302803166897, + "instruction_rms": 0.000931130280125532, + "raw_apical_rms": 0.07280005213689716, + "teaching_rms": 0.0009311302803166897, + "traffic_rms": 0.07278956904321929 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.6460868929898425e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638568239501903e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009272727348378633, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58080, + "train_examples": 45000, + "train_loss": 0.0064649609093864755 + }, + { + "epoch": 166, + "eval_accuracy": 0.922, + "eval_loss": 0.3788002136230469, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021560937166214, + "mean_feedback_forward_cosine": 0.9994059339646371, + "mean_feedback_forward_relative_error": 0.03390904711258988, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000951799096096911, + "instruction_rms": 0.0009517990966686603, + "raw_apical_rms": 0.07279990962474571, + "teaching_rms": 0.000951799096096911, + "traffic_rms": 0.07278931803572794 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.383095759090793e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6378296091705208e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009335598022536896, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58432, + "train_examples": 45000, + "train_loss": 0.006698690681656201 + }, + { + "epoch": 167, + "eval_accuracy": 0.9224, + "eval_loss": 0.3780606689453125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021558329463005, + "mean_feedback_forward_cosine": 0.9994059608828637, + "mean_feedback_forward_relative_error": 0.033908947491117065, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008896655708649884, + "instruction_rms": 0.0008896655708132983, + "raw_apical_rms": 0.07279797216228476, + "teaching_rms": 0.0008896655708649884, + "traffic_rms": 0.07278825528259195 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.671396342497246e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6407614709913652e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009282221743622334, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 58784, + "train_examples": 45000, + "train_loss": 0.006275268090681897 + }, + { + "epoch": 168, + "eval_accuracy": 0.9224, + "eval_loss": 0.378640185546875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021555349230766, + "mean_feedback_forward_cosine": 0.9994059320419065, + "mean_feedback_forward_relative_error": 0.033908843483415343, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009209827422351977, + "instruction_rms": 0.0009209827423907628, + "raw_apical_rms": 0.0727964476001508, + "teaching_rms": 0.0009209827422351977, + "traffic_rms": 0.07278594311247243 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.299055961747399e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638953556226787e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00936715458701123, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59136, + "train_examples": 45000, + "train_loss": 0.006200297907160388 + }, + { + "epoch": 169, + "eval_accuracy": 0.923, + "eval_loss": 0.37887355346679685, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021551996469498, + "mean_feedback_forward_cosine": 0.9994059551146722, + "mean_feedback_forward_relative_error": 0.033908742149510694, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009711995502875797, + "instruction_rms": 0.0009711995498512051, + "raw_apical_rms": 0.07279644517307739, + "teaching_rms": 0.0009711995502875797, + "traffic_rms": 0.07278561888103524 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1259067273813343e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6411877288308163e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009288771153875197, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59488, + "train_examples": 45000, + "train_loss": 0.0066708320299784345 + }, + { + "epoch": 170, + "eval_accuracy": 0.9224, + "eval_loss": 0.37823710327148435, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021549388766289, + "mean_feedback_forward_cosine": 0.9994059281964456, + "mean_feedback_forward_relative_error": 0.033908638021638314, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009068830681940569, + "instruction_rms": 0.0009068830681429819, + "raw_apical_rms": 0.07279731809941654, + "teaching_rms": 0.0009068830681940569, + "traffic_rms": 0.07278748630276233 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1985181553627626e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6417179504213202e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00934180178498991, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 59840, + "train_examples": 45000, + "train_loss": 0.005992806352302432 + }, + { + "epoch": 171, + "eval_accuracy": 0.9222, + "eval_loss": 0.3777138671875, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502154640853405, + "mean_feedback_forward_cosine": 0.9994059685737856, + "mean_feedback_forward_relative_error": 0.033908533683467294, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009218364303518459, + "instruction_rms": 0.0009218364301352782, + "raw_apical_rms": 0.07279907283003745, + "teaching_rms": 0.0009218364303518459, + "traffic_rms": 0.07278936574241918 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.051080537692542e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63948518168127e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009399914198352297, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60192, + "train_examples": 45000, + "train_loss": 0.006182804728216595 + }, + { + "epoch": 172, + "eval_accuracy": 0.9234, + "eval_loss": 0.37877182006835936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021543800830841, + "mean_feedback_forward_cosine": 0.9994059781874379, + "mean_feedback_forward_relative_error": 0.033908430216533524, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008757880830134921, + "instruction_rms": 0.0008757880829026351, + "raw_apical_rms": 0.0727909295614263, + "teaching_rms": 0.0008757880830134921, + "traffic_rms": 0.07278188463544438 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.394451735834082e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639762290664436e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00939934921304198, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60544, + "train_examples": 45000, + "train_loss": 0.005911937013144295 + }, + { + "epoch": 173, + "eval_accuracy": 0.923, + "eval_loss": 0.37629825439453124, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215404480695724, + "mean_feedback_forward_cosine": 0.9994059551146722, + "mean_feedback_forward_relative_error": 0.033908325397679885, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000973114015998277, + "instruction_rms": 0.0009731140163191162, + "raw_apical_rms": 0.07279270316450596, + "teaching_rms": 0.000973114015998277, + "traffic_rms": 0.0727814818875947 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.15684695553864e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6398265330088304e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009303746652720676, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 60896, + "train_examples": 45000, + "train_loss": 0.006478916634453668 + }, + { + "epoch": 174, + "eval_accuracy": 0.9222, + "eval_loss": 0.3773844543457031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215382128953934, + "mean_feedback_forward_cosine": 0.9994059493464809, + "mean_feedback_forward_relative_error": 0.033908218565967777, + "min_feedback_forward_cosine": 0.9987452626228333 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009448674316323791, + "instruction_rms": 0.0009448674318010515, + "raw_apical_rms": 0.07279174859924621, + "teaching_rms": 0.0009448674316323791, + "traffic_rms": 0.07278141079728034 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.272727949228283e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6418756013933296e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009380181118971624, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61248, + "train_examples": 45000, + "train_loss": 0.006255613459543221 + }, + { + "epoch": 175, + "eval_accuracy": 0.923, + "eval_loss": 0.3780188659667969, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021534487605095, + "mean_feedback_forward_cosine": 0.9994059551146722, + "mean_feedback_forward_relative_error": 0.033908120416585474, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008671276321794992, + "instruction_rms": 0.0008671276314825126, + "raw_apical_rms": 0.07279193205743949, + "teaching_rms": 0.0008671276321794992, + "traffic_rms": 0.07278263841080512 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.534815462558072e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6410912552650877e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009327561037759096, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61600, + "train_examples": 45000, + "train_loss": 0.006052574612283044 + }, + { + "epoch": 176, + "eval_accuracy": 0.9226, + "eval_loss": 0.3774455078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021531879901886, + "mean_feedback_forward_cosine": 0.999405974341977, + "mean_feedback_forward_relative_error": 0.03390802196677654, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009107835894467998, + "instruction_rms": 0.000910783589541975, + "raw_apical_rms": 0.07279263183190296, + "teaching_rms": 0.0009107835894467998, + "traffic_rms": 0.07278268993508986 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.600662523695064e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6396375303343103e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009284183196265975, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 61952, + "train_examples": 45000, + "train_loss": 0.006236939753923151 + }, + { + "epoch": 177, + "eval_accuracy": 0.9246, + "eval_loss": 0.3832741027832031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021528899669647, + "mean_feedback_forward_cosine": 0.9994059531919418, + "mean_feedback_forward_relative_error": 0.03390791567583238, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009633222918402966, + "instruction_rms": 0.0009633222919751328, + "raw_apical_rms": 0.07280173134169926, + "teaching_rms": 0.0009633222918402966, + "traffic_rms": 0.07279096102970496 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.57055824906638e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638033545376644e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009280104216846495, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62304, + "train_examples": 45000, + "train_loss": 0.006763021509597699 + }, + { + "epoch": 178, + "eval_accuracy": 0.9244, + "eval_loss": 0.3776351867675781, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215255469083786, + "mean_feedback_forward_cosine": 0.9994059647283247, + "mean_feedback_forward_relative_error": 0.033907806140280536, + "min_feedback_forward_cosine": 0.998745322227478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009586253475755312, + "instruction_rms": 0.000958625347693359, + "raw_apical_rms": 0.07280250700127605, + "teaching_rms": 0.0009586253475755312, + "traffic_rms": 0.07279176941753737 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.719749841479825e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6398562880900587e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00944796904913204, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 62656, + "train_examples": 45000, + "train_loss": 0.006566029266516367 + }, + { + "epoch": 179, + "eval_accuracy": 0.9234, + "eval_loss": 0.37907015380859377, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502152293920517, + "mean_feedback_forward_cosine": 0.9994059666510551, + "mean_feedback_forward_relative_error": 0.03390770447590659, + "min_feedback_forward_cosine": 0.9987449645996094 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008790105723924426, + "instruction_rms": 0.0008790105720771522, + "raw_apical_rms": 0.07279674862877619, + "teaching_rms": 0.0008790105723924426, + "traffic_rms": 0.07278790064682795 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 2.993412806517881e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6405543416716802e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009364554261458803, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63008, + "train_examples": 45000, + "train_loss": 0.005639267408682241 + }, + { + "epoch": 180, + "eval_accuracy": 0.9234, + "eval_loss": 0.3765585388183594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021520331501961, + "mean_feedback_forward_cosine": 0.9994059724192466, + "mean_feedback_forward_relative_error": 0.03390759593176265, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008834507133467337, + "instruction_rms": 0.0008834507134877638, + "raw_apical_rms": 0.07279706868424211, + "teaching_rms": 0.0008834507133467337, + "traffic_rms": 0.07278753699624642 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.1553420260243e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6402966158823036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009301459478879287, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63360, + "train_examples": 45000, + "train_loss": 0.005865725637806786 + }, + { + "epoch": 181, + "eval_accuracy": 0.9238, + "eval_loss": 0.3767367126464844, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021516606211662, + "mean_feedback_forward_cosine": 0.9994059397328284, + "mean_feedback_forward_relative_error": 0.03390748928030653, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009059746277008423, + "instruction_rms": 0.0009059746276981573, + "raw_apical_rms": 0.07280177074087833, + "teaching_rms": 0.0009059746277008423, + "traffic_rms": 0.07279164452071894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.174826173652946e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.639320196533143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00936177934788029, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 63712, + "train_examples": 45000, + "train_loss": 0.006068159006536007 + }, + { + "epoch": 182, + "eval_accuracy": 0.9246, + "eval_loss": 0.3834036926269531, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215139985084534, + "mean_feedback_forward_cosine": 0.9994059647283247, + "mean_feedback_forward_relative_error": 0.03390739034981497, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009031406778541334, + "instruction_rms": 0.0009031406777414056, + "raw_apical_rms": 0.07279577581150613, + "teaching_rms": 0.0009031406778541334, + "traffic_rms": 0.07278685137317638 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.092711591852094e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6351176539766795e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009371290056919238, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64064, + "train_examples": 45000, + "train_loss": 0.006130151518848207 + }, + { + "epoch": 183, + "eval_accuracy": 0.9226, + "eval_loss": 0.38016424560546874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021510645747185, + "mean_feedback_forward_cosine": 0.9994059724192466, + "mean_feedback_forward_relative_error": 0.033907286882881194, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009198295107836894, + "instruction_rms": 0.0009198295107451514, + "raw_apical_rms": 0.07279343644432036, + "teaching_rms": 0.0009198295107836894, + "traffic_rms": 0.07278314696048392 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 3.4009541849400193e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383791361941714e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009477513198940484, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64416, + "train_examples": 45000, + "train_loss": 0.006037287366555797 + }, + { + "epoch": 184, + "eval_accuracy": 0.9222, + "eval_loss": 0.37987239379882815, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021507665514946, + "mean_feedback_forward_cosine": 0.9994059724192466, + "mean_feedback_forward_relative_error": 0.03390718251466751, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008858086400521034, + "instruction_rms": 0.0008858086399444548, + "raw_apical_rms": 0.07279044923353264, + "teaching_rms": 0.0008858086400521034, + "traffic_rms": 0.07278121669807068 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.3412252520292895e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6363775237892308e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009410989802460348, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 64768, + "train_examples": 45000, + "train_loss": 0.0060372411655883 + }, + { + "epoch": 185, + "eval_accuracy": 0.9238, + "eval_loss": 0.3833714599609375, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050215043127536774, + "mean_feedback_forward_cosine": 0.9994059685737856, + "mean_feedback_forward_relative_error": 0.03390708415498657, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000824911630066183, + "instruction_rms": 0.0008249116295215703, + "raw_apical_rms": 0.07279042653889208, + "teaching_rms": 0.000824911630066183, + "traffic_rms": 0.07278195475173309 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.077775355744052e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6367446567474114e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009301936742821348, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65120, + "train_examples": 45000, + "train_loss": 0.005888968814992243 + }, + { + "epoch": 186, + "eval_accuracy": 0.924, + "eval_loss": 0.3802086730957031, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021502077579498, + "mean_feedback_forward_cosine": 0.9994059666510551, + "mean_feedback_forward_relative_error": 0.03390698176958869, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.00089744708681856, + "instruction_rms": 0.0008974470863714073, + "raw_apical_rms": 0.07278875310093366, + "teaching_rms": 0.00089744708681856, + "traffic_rms": 0.0727795669934635 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.340492765209092e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6354787432597972e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009585534170728247, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65472, + "train_examples": 45000, + "train_loss": 0.005864890639566713 + }, + { + "epoch": 187, + "eval_accuracy": 0.9234, + "eval_loss": 0.3811006103515625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.0502149872481823, + "mean_feedback_forward_cosine": 0.999405982032899, + "mean_feedback_forward_relative_error": 0.03390688521246756, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000916919782043793, + "instruction_rms": 0.0009169197822677663, + "raw_apical_rms": 0.0727924594210436, + "teaching_rms": 0.000916919782043793, + "traffic_rms": 0.07278210274929854 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 4.196382263899068e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383436759923482e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009338556265023695, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 65824, + "train_examples": 45000, + "train_loss": 0.006068551631271839 + }, + { + "epoch": 188, + "eval_accuracy": 0.9244, + "eval_loss": 0.38284039916992185, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021495372056961, + "mean_feedback_forward_cosine": 0.9994059839556294, + "mean_feedback_forward_relative_error": 0.03390678682274396, + "min_feedback_forward_cosine": 0.9987450838088989 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008812204341248592, + "instruction_rms": 0.0008812204343298186, + "raw_apical_rms": 0.07278655818563996, + "teaching_rms": 0.0008812204341248592, + "traffic_rms": 0.07277747691378347 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 4.115873153409666e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6383831814105143e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009336102800834519, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66176, + "train_examples": 45000, + "train_loss": 0.00590759240090847 + }, + { + "epoch": 189, + "eval_accuracy": 0.924, + "eval_loss": 0.38746590576171874, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021492764353752, + "mean_feedback_forward_cosine": 0.9994059608828637, + "mean_feedback_forward_relative_error": 0.03390668215410363, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000944013221245689, + "instruction_rms": 0.000944013221230113, + "raw_apical_rms": 0.07279046141483732, + "teaching_rms": 0.000944013221245689, + "traffic_rms": 0.07277950835120359 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 5.1209863016765667e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6368435021535036e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009340577283093121, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66528, + "train_examples": 45000, + "train_loss": 0.00627474368541605 + }, + { + "epoch": 190, + "eval_accuracy": 0.9242, + "eval_loss": 0.3782501953125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021490156650543, + "mean_feedback_forward_cosine": 0.9994059781874379, + "mean_feedback_forward_relative_error": 0.033906579858833746, + "min_feedback_forward_cosine": 0.998745322227478 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009228944355058448, + "instruction_rms": 0.0009228944359649598, + "raw_apical_rms": 0.07278233118548318, + "teaching_rms": 0.0009228944355058448, + "traffic_rms": 0.07277222320074045 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.167023703871564e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6409180082698417e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00929254526828421, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 66880, + "train_examples": 45000, + "train_loss": 0.006109715344508489 + }, + { + "epoch": 191, + "eval_accuracy": 0.9236, + "eval_loss": 0.381222705078125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214868038892746, + "mean_feedback_forward_cosine": 0.999405974341977, + "mean_feedback_forward_relative_error": 0.033906479366123676, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0009125403572763838, + "instruction_rms": 0.0009125403574922392, + "raw_apical_rms": 0.07277732131233654, + "teaching_rms": 0.0009125403572763838, + "traffic_rms": 0.07276739159796018 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.160996169844111e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6368375345992205e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009306256766792505, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67232, + "train_examples": 45000, + "train_loss": 0.005790863518623842 + }, + { + "epoch": 192, + "eval_accuracy": 0.924, + "eval_loss": 0.37977269897460936, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021483451128006, + "mean_feedback_forward_cosine": 0.9994059897238209, + "mean_feedback_forward_relative_error": 0.03390637695068313, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008634372559339853, + "instruction_rms": 0.0008634372557363993, + "raw_apical_rms": 0.07277595362353002, + "teaching_rms": 0.0008634372559339853, + "traffic_rms": 0.07276696431622127 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.573541818013837e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6418940704509305e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009294865945736475, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67584, + "train_examples": 45000, + "train_loss": 0.0055582207631423245 + }, + { + "epoch": 193, + "eval_accuracy": 0.9232, + "eval_loss": 0.38124073486328125, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021481215953827, + "mean_feedback_forward_cosine": 0.9994060051056647, + "mean_feedback_forward_relative_error": 0.03390627768972228, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000852372129311672, + "instruction_rms": 0.0008523721291413863, + "raw_apical_rms": 0.07277602368993115, + "teaching_rms": 0.000852372129311672, + "traffic_rms": 0.07276717651199026 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.0884806651320105e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.63812014918529e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009340981492915228, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 67936, + "train_examples": 45000, + "train_loss": 0.005360517854657438 + }, + { + "epoch": 194, + "eval_accuracy": 0.9228, + "eval_loss": 0.3823780700683594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214774906635284, + "mean_feedback_forward_cosine": 0.9994059878010904, + "mean_feedback_forward_relative_error": 0.033906179780681285, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008369135264463215, + "instruction_rms": 0.0008369135264636173, + "raw_apical_rms": 0.07277888262188072, + "teaching_rms": 0.0008369135264463215, + "traffic_rms": 0.07277062523838189 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.558799477332286e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6423136984203254e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009384849108658035, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68288, + "train_examples": 45000, + "train_loss": 0.005631521541501085 + }, + { + "epoch": 195, + "eval_accuracy": 0.9236, + "eval_loss": 0.3792059997558594, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214748829603195, + "mean_feedback_forward_cosine": 0.9994059858783599, + "mean_feedback_forward_relative_error": 0.033906079918867155, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008770085033274025, + "instruction_rms": 0.0008770085031596293, + "raw_apical_rms": 0.07277805580198711, + "teaching_rms": 0.0008770085033274025, + "traffic_rms": 0.0727688635778701 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.804241982014901e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6397546781794007e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009314376633056905, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68640, + "train_examples": 45000, + "train_loss": 0.00563206836965142 + }, + { + "epoch": 196, + "eval_accuracy": 0.9254, + "eval_loss": 0.37827137451171877, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.050214722752571106, + "mean_feedback_forward_cosine": 0.9994059916465513, + "mean_feedback_forward_relative_error": 0.033905980087095694, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000863613228595432, + "instruction_rms": 0.0008636132283742729, + "raw_apical_rms": 0.07277717260645288, + "teaching_rms": 0.000863613228595432, + "traffic_rms": 0.07276802659885094 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.043533863296034e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6419506779648067e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00930628926678014, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 68992, + "train_examples": 45000, + "train_loss": 0.005555464403993553 + }, + { + "epoch": 197, + "eval_accuracy": 0.9222, + "eval_loss": 0.37736356811523436, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021468549966812, + "mean_feedback_forward_cosine": 0.9994059878010904, + "mean_feedback_forward_relative_error": 0.033905881607244094, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000900053276765676, + "instruction_rms": 0.0009000532764372408, + "raw_apical_rms": 0.07277405393904711, + "teaching_rms": 0.000900053276765676, + "traffic_rms": 0.07276462906558215 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.7921108858872685e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6407543926550647e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009467302561984978, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69344, + "train_examples": 45000, + "train_loss": 0.0058654394092659155 + }, + { + "epoch": 198, + "eval_accuracy": 0.9244, + "eval_loss": 0.37802588500976564, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021467059850693, + "mean_feedback_forward_cosine": 0.9994059839556294, + "mean_feedback_forward_relative_error": 0.033905782015813936, + "min_feedback_forward_cosine": 0.9987452030181885 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.000811120721353702, + "instruction_rms": 0.0008111207213972554, + "raw_apical_rms": 0.07277230863245111, + "teaching_rms": 0.000811120721353702, + "traffic_rms": 0.07276464107952894 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 4.067615577696415e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6391286366009206e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009343913485228653, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 69696, + "train_examples": 45000, + "train_loss": 0.005283751911007696 + }, + { + "epoch": 199, + "eval_accuracy": 0.9224, + "eval_loss": 0.3791741271972656, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021464079618454, + "mean_feedback_forward_cosine": 0.9994059839556294, + "mean_feedback_forward_relative_error": 0.033905680531695966, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008688715170549379, + "instruction_rms": 0.0008688715163338572, + "raw_apical_rms": 0.07277670697438565, + "teaching_rms": 0.0008688715170549379, + "traffic_rms": 0.07276745220179037 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.0677092894911766, + "maximum_absolute_post_projection_soma_slope": 3.409411419852404e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.638858979983995e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.009275071927212605, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70048, + "train_examples": 45000, + "train_loss": 0.005599399071849055 + }, + { + "epoch": 200, + "eval_accuracy": 0.9236, + "eval_loss": 0.3822314697265625, + "eval_split": "validation", + "feedback_tracking": { + "max_feedback_forward_relative_error": 0.05021461844444275, + "mean_feedback_forward_cosine": 0.9994059801101685, + "mean_feedback_forward_relative_error": 0.03390557637378093, + "min_feedback_forward_cosine": 0.9987451434135437 + }, + "lr": 0.0010000000000000002, + "mixed_apical": { + "innovation_rms": 0.0008638414231774739, + "instruction_rms": 0.0008638414230827563, + "raw_apical_rms": 0.0727772008884368, + "teaching_rms": 0.0008638414231774739, + "traffic_rms": 0.07276843471150797 + }, + "neutral_projection": { + "instruction_observations": 0, + "maximum_absolute_correction_slope": 0.06770928204059601, + "maximum_absolute_post_projection_soma_slope": 2.943882115857832e-08, + "maximum_observations": 128, + "maximum_post_projection_traffic_rms_ratio": 2.6374803630092046e-08, + "maximum_pre_projection_traffic_rms_ratio": 0.00941933233090907, + "minimum_observations": 72 + }, + "output_lr": 0.0010000000000000002, + "step": 70400, + "train_examples": 45000, + "train_loss": 0.005543160841696792 + } + ], + "evaluation_protocol": { + "test_evaluations": 0, + "test_used_for_selection": false, + "validation_evaluations": 201 + }, + "final": { + "accuracy": 0.9236, + "epoch": 200, + "evaluation_split": "validation", + "finite": true, + "loss": 0.3822314697265625, + "step": 70400 + }, + "hardware": { + "cuda_device_name": "NVIDIA RTX A6000", + "cuda_visible_devices": "0", + "device": "cuda", + "device_total_memory_bytes": 50896568320, + "peak_memory_allocated_bytes": 2329526272, + "peak_memory_reserved_bytes": 3575644160, + "torch_version": "2.3.1+cu118" + }, + "predictor_warmup": { + "closed_form_fit": { + "max_absolute_residual_soma_slope": 4.4873448956650464e-08, + "max_applied_stability_margin": 0.0, + "max_positive_residual_soma_slope": 4.4873448956650464e-08, + "min_residual_soma_slope": -4.4004142552012127e-08, + "mse": 5.220318620602023e-18, + "observations": 64, + "residual_traffic_rms_ratio": 7.213750884517225e-08, + "stability_margin": 0.0 + }, + "examples": 64, + "instruction_present": false, + "mode": "closed_form", + "post_warmup_traffic_residual_rms_ratio": 7.213750884517225e-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": 34.19122052192688, + "mirror_warmup_wall_s": 0.0, + "predictor_warmup_wall_s": 0.6957113742828369, + "timing_excludes_data_loading_hashing_and_model_construction": true, + "total_timed_wall_s": 4736.380233287811, + "train_wall_s": 4699.746609687805 + }, + "traffic_calibration": { + "data_source": "first unaugmented development-training examples", + "examples": 64, + "instruction_rms": [ + 0.020279373973608017, + 0.01588572934269905, + 0.009585018269717693, + 0.01053142361342907, + 0.006489923689514399, + 0.007580762729048729, + 0.004748271778225899, + 0.0056568942964077, + 0.0037147034890949726, + 0.004711497575044632, + 0.0028121231589466333, + 0.005797920748591423, + 0.004072000738233328, + 0.004937782883644104, + 0.0032058130018413067, + 0.004144369624555111, + 0.002652793889865279, + 0.0034981879871338606, + 0.0022634966298937798, + 0.003057616064324975, + 0.0020107743330299854, + 0.004512992687523365, + 0.003177475184202194, + 0.004072428680956364, + 0.002793447347357869, + 0.0033600262831896544, + 0.0025724933948367834, + 0.0031278016977012157, + 0.0024165462236851454, + 0.0028540485072880983, + 0.0024009225890040398 + ], + "realized_traffic_instruction_rms_ratio": [ + 4.000000476837158, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.000000476837158, + 4.000000476837158, + 4.000000476837158, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 4.0, + 3.999999761581421, + 4.0, + 4.0, + 4.0 + ], + "target_ratio": 4.0, + "traffic_gain": [ + 0.10579779744148254, + 0.08581840991973877, + 0.03320790454745293, + 0.05464063584804535, + 0.017315272241830826, + 0.0413997545838356, + 0.010940272361040115, + 0.029754366725683212, + 0.007736234460026026, + 0.025829754769802094, + 0.005339059978723526, + 0.03059656172990799, + 0.009305471554398537, + 0.026019593700766563, + 0.006473024841398001, + 0.02154967002570629, + 0.004868118092417717, + 0.01855073869228363, + 0.003812844166532159, + 0.01592211239039898, + 0.0030986128840595484, + 0.024262895807623863, + 0.006371741183102131, + 0.021971462294459343, + 0.005072675179690123, + 0.01743730716407299, + 0.004265559371560812, + 0.01702703908085823, + 0.0037584376987069845, + 0.01519029587507248, + 0.0034847366623580456 + ], + "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, + 0.7579833269119263, + 1.750368356704712, + 0.7590868473052979, + 1.9810292720794678, + 0.7692683339118958, + 2.1797285079956055, + 0.7542961835861206, + 2.3746018409729004, + 0.768143355846405, + 2.5957090854644775, + 0.7440155148506165, + 1.9947296380996704, + 0.7414032816886902, + 2.2027409076690674, + 0.7707672119140625, + 2.4123384952545166, + 0.7347846031188965, + 2.5718624591827393, + 0.7515451908111572, + 2.755929946899414 + ], + "uses_validation_endpoint": false + }, + "work": { + "apical_macs_per_example": 68420224, + "causal_scalar_observations": 0, + "components": { + "apical_projection_macs": 615786394894336, + "apical_regression_macs": 0, + "bp_reverse_macs_estimate": 0, + "kp_reciprocal_correlation_macs": 615782016000000, + "local_weight_correlation_macs": 619763328000000, + "mirror_feedback_prediction_macs": 0, + "mirror_local_correlation_macs": 0, + "mirror_response_macs": 0, + "ordinary_forward_macs": 619763328000000, + "perturbation_forward_macs": 0, + "warmup_clean_forward_macs": 4407205888 + }, + "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": 65471084756992, + "forward_macs_per_example": 68862592, + "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": 2471099474100224 + } +} |
