{
  "architecture": {
    "adaptive_apical_parameters": 0,
    "adaptive_feedback_parameters": 848512,
    "base_width": 16,
    "blocks_per_stage": 9,
    "bn_eps": 1e-05,
    "bn_momentum": 0.1,
    "depth": 56,
    "family": "CIFAR 6n+2 ResNet, option-A shortcuts",
    "fixed_feedback_parameters": 0,
    "fixed_traffic_coefficients": 0,
    "forward_parameters": 853018,
    "hidden_shapes": [
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ]
    ],
    "normalization": "batchnorm",
    "predictor_parameters": 0,
    "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": "/home/yurenh2/sdrn/data",
    "depth": 56,
    "device": "cuda",
    "epochs": 200,
    "eta_A": 0.01,
    "eta_P": 0.01,
    "eval_every": 0,
    "eval_split": "test",
    "learn_P": 0,
    "loader_seed": 11,
    "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",
    "momentum": 0.9,
    "neutral_projection": 0,
    "normalization": "batchnorm",
    "nuisance_scale": 0.0,
    "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d56_s11.json",
    "output_lr": 0.1,
    "pert_directions": 1,
    "pert_every": 4,
    "pert_sigma": 0.01,
    "perturb_seed": 1000,
    "predictor_every": 0,
    "predictor_mode": "nlms",
    "predictor_warmup_steps": 0,
    "residual_scale": null,
    "seed": 11,
    "split_seed": 2027,
    "traffic_calibration_examples": 64,
    "traffic_ratio": 4.0,
    "traffic_rule": "innovation",
    "traffic_seed": 4000,
    "train_limit": 0,
    "use_residual": 1,
    "val_examples": 0,
    "vectorizer_mode": "spatial_template",
    "warmup_epochs": 0,
    "weight_decay": 0.0001,
    "weight_scale": 1.0,
    "width": 16
  },
  "calibration_metric_space": "reciprocal_local_activity_products",
  "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": 0,
    "ordinary_examples": 10000000,
    "per_example_loss_terms": 0,
    "perturbation_events": 0,
    "perturbation_forward_examples": 0,
    "predictor_update_examples": 0,
    "predictor_warmup_examples": 0,
    "traffic_audit_examples": 0,
    "traffic_calibration_examples": 0
  },
  "diagnostics": {
    "early_third_mean": 0.9994188480907016,
    "feedback_forward_cosine": [
      0.9996993541717529,
      0.9996830821037292,
      0.9995632767677307,
      0.9994942545890808,
      0.9996981620788574,
      0.999686598777771,
      0.9995149374008179,
      0.9995197057723999,
      0.9996012449264526,
      0.999535083770752,
      0.9995328187942505,
      0.9994897246360779,
      0.9996421933174133,
      0.9995424151420593,
      0.9996671676635742,
      0.999509334564209,
      0.9997736215591431,
      0.9996488690376282,
      0.9997099041938782,
      0.9997872114181519,
      0.9996638298034668,
      0.9996258020401001,
      0.9996758103370667,
      0.9996204376220703,
      0.9996803402900696,
      0.9995815753936768,
      0.9996604323387146,
      0.9995313882827759,
      0.9996680617332458,
      0.9995308518409729,
      0.9996833801269531,
      0.9995642304420471,
      0.9996212720870972,
      0.9994816780090332,
      0.999650776386261,
      0.9994839429855347,
      0.9997327923774719,
      0.9998137354850769,
      0.9997699856758118,
      0.9997382760047913,
      0.9997545480728149,
      0.9996718764305115,
      0.999742865562439,
      0.9995942115783691,
      0.9997011423110962,
      0.9995160102844238,
      0.9994035959243774,
      0.9991016983985901,
      0.9994409084320068,
      0.999116063117981,
      0.9989480972290039,
      0.9984638690948486,
      0.9990649223327637,
      0.998519241809845,
      0.9999802708625793
    ],
    "feedback_forward_norm_ratio": [
      0.9994032382965088,
      1.0002179145812988,
      0.9992724061012268,
      1.0009822845458984,
      0.9994022846221924,
      1.0009336471557617,
      0.9990427494049072,
      1.0013375282287598,
      0.9988051056861877,
      1.0024958848953247,
      0.9994823336601257,
      1.002917766571045,
      1.0000933408737183,
      1.00282621383667,
      0.9992473721504211,
      1.0023692846298218,
      0.9999445676803589,
      1.0007997751235962,
      0.99896240234375,
      1.0006413459777832,
      0.9993771314620972,
      1.0014231204986572,
      0.9992647171020508,
      1.0011144876480103,
      0.999278724193573,
      1.0012441873550415,
      0.9986801743507385,
      1.0012577772140503,
      0.9993001222610474,
      1.0015628337860107,
      0.9993846416473389,
      1.0021334886550903,
      0.9988804459571838,
      1.0015361309051514,
      0.9989018440246582,
      1.0020201206207275,
      0.9992831349372864,
      1.0004626512527466,
      0.9996106624603271,
      1.0006550550460815,
      0.9993950128555298,
      1.0010838508605957,
      0.9993111491203308,
      1.001451849937439,
      0.9995183348655701,
      1.001678228378296,
      0.9984020590782166,
      1.0025371313095093,
      0.9987742304801941,
      1.0025242567062378,
      0.9982123970985413,
      1.0027648210525513,
      0.9986278414726257,
      1.0024317502975464,
      1.0019079446792603
    ],
    "feedback_forward_relative_error": [
      0.02452433668076992,
      0.025184625759720802,
      0.02955569326877594,
      0.03183719143271446,
      0.02456962876021862,
      0.02506248652935028,
      0.0311441570520401,
      0.031041115522384644,
      0.0282449834048748,
      0.030632462352514267,
      0.03056534379720688,
      0.032126203179359436,
      0.026748819276690483,
      0.030421754345297813,
      0.0258016437292099,
      0.031452398747205734,
      0.021281296387314796,
      0.026522092521190643,
      0.024094196036458015,
      0.020647048950195312,
      0.025927532464265823,
      0.027411917224526405,
      0.0254663098603487,
      0.027587715536355972,
      0.02528620883822441,
      0.028972430154681206,
      0.026078246533870697,
      0.030659740790724754,
      0.02576402947306633,
      0.03069719485938549,
      0.02516496740281582,
      0.02963043749332428,
      0.027530504390597343,
      0.03225620836019516,
      0.026434797793626785,
      0.03222276642918587,
      0.02312055230140686,
      0.019309254363179207,
      0.021448656916618347,
      0.022896774113178253,
      0.022158363834023476,
      0.0256564412266016,
      0.02268141880631447,
      0.028542323037981987,
      0.024445435032248497,
      0.031184131279587746,
      0.034544941037893295,
      0.042515937238931656,
      0.033438790589571,
      0.04217500612139702,
      0.04586270824074745,
      0.05557382106781006,
      0.04323612526059151,
      0.05453914776444435,
      0.00657545356079936
    ],
    "innovation_negative_gradient_cosine": [
      0.9994824528694153,
      0.9993388056755066,
      0.999483585357666,
      0.9993628263473511,
      0.9994879364967346,
      0.9994150996208191,
      0.9994694590568542,
      0.9991943836212158,
      0.9994771480560303,
      0.9993506669998169,
      0.9994744658470154,
      0.9993091821670532,
      0.9994770288467407,
      0.999363899230957,
      0.9994924068450928,
      0.9993765354156494,
      0.9995057582855225,
      0.9994776248931885,
      0.9995115995407104,
      0.9995065927505493,
      0.9995415210723877,
      0.9993911981582642,
      0.9995527267456055,
      0.9994387626647949,
      0.9995670318603516,
      0.9994246363639832,
      0.9995906352996826,
      0.9994567632675171,
      0.999603271484375,
      0.9995082020759583,
      0.9996105432510376,
      0.9995136260986328,
      0.9996271133422852,
      0.9995301365852356,
      0.9996365904808044,
      0.9995220899581909,
      0.9996331930160522,
      0.999711275100708,
      0.999653160572052,
      0.9996668696403503,
      0.9996533989906311,
      0.9996808767318726,
      0.9996672868728638,
      0.9997168779373169,
      0.9997159242630005,
      0.999731183052063,
      0.9996582865715027,
      0.9996111392974854,
      0.9997155666351318,
      0.9997341632843018,
      0.999699056148529,
      0.9996869564056396,
      0.9998605251312256,
      0.9998820424079895,
      0.8115973472595215
    ],
    "mean_feedback_forward_cosine": 0.9995472160252658,
    "mean_feedback_forward_relative_error": 0.02906279576603662,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.9994824528694153,
      0.9993388056755066,
      0.999483585357666,
      0.9993628263473511,
      0.9994879364967346,
      0.9994150996208191,
      0.9994694590568542,
      0.9991943836212158,
      0.9994771480560303,
      0.9993506669998169,
      0.9994744658470154,
      0.9993091821670532,
      0.9994770288467407,
      0.999363899230957,
      0.9994924068450928,
      0.9993765354156494,
      0.9995057582855225,
      0.9994776248931885,
      0.9995115995407104,
      0.9995065927505493,
      0.9995415210723877,
      0.9993911981582642,
      0.9995527267456055,
      0.9994387626647949,
      0.9995670318603516,
      0.9994246363639832,
      0.9995906352996826,
      0.9994567632675171,
      0.999603271484375,
      0.9995082020759583,
      0.9996105432510376,
      0.9995136260986328,
      0.9996271133422852,
      0.9995301365852356,
      0.9996365904808044,
      0.9995220899581909,
      0.9996331930160522,
      0.999711275100708,
      0.999653160572052,
      0.9996668696403503,
      0.9996533989906311,
      0.9996808767318726,
      0.9996672868728638,
      0.9997168779373169,
      0.9997159242630005,
      0.999731183052063,
      0.9996582865715027,
      0.9996111392974854,
      0.9997155666351318,
      0.9997341632843018,
      0.999699056148529,
      0.9996869564056396,
      0.9998605251312256,
      0.9998820424079895,
      0.8115973472595215
    ],
    "teaching_negative_gradient_cosine": [
      0.9994824528694153,
      0.9993388056755066,
      0.999483585357666,
      0.9993628263473511,
      0.9994879364967346,
      0.9994150996208191,
      0.9994694590568542,
      0.9991943836212158,
      0.9994771480560303,
      0.9993506669998169,
      0.9994744658470154,
      0.9993091821670532,
      0.9994770288467407,
      0.999363899230957,
      0.9994924068450928,
      0.9993765354156494,
      0.9995057582855225,
      0.9994776248931885,
      0.9995115995407104,
      0.9995065927505493,
      0.9995415210723877,
      0.9993911981582642,
      0.9995527267456055,
      0.9994387626647949,
      0.9995670318603516,
      0.9994246363639832,
      0.9995906352996826,
      0.9994567632675171,
      0.999603271484375,
      0.9995082020759583,
      0.9996105432510376,
      0.9995136260986328,
      0.9996271133422852,
      0.9995301365852356,
      0.9996365904808044,
      0.9995220899581909,
      0.9996331930160522,
      0.999711275100708,
      0.999653160572052,
      0.9996668696403503,
      0.9996533989906311,
      0.9996808767318726,
      0.9996672868728638,
      0.9997168779373169,
      0.9997159242630005,
      0.999731183052063,
      0.9996582865715027,
      0.9996111392974854,
      0.9997155666351318,
      0.9997341632843018,
      0.999699056148529,
      0.9996869564056396,
      0.9998605251312256,
      0.9998820424079895,
      0.8115973472595215
    ],
    "wall_s": 0.1958451271057129
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3753533363342285,
        "mean_feedback_forward_cosine": 0.23858112272891133,
        "mean_feedback_forward_relative_error": 1.227240677313371,
        "min_feedback_forward_cosine": 0.05976404994726181
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 2.3715767669677734
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3492971658706665,
        "mean_feedback_forward_cosine": 0.2779971172863787,
        "mean_feedback_forward_relative_error": 1.1959425774487582,
        "min_feedback_forward_cosine": 0.09428754448890686
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 1.7599987884521484
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3413547277450562,
        "mean_feedback_forward_cosine": 0.33180117837407375,
        "mean_feedback_forward_relative_error": 1.1510712428526444,
        "min_feedback_forward_cosine": 0.10544475167989731
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.3671581301879883
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3315541744232178,
        "mean_feedback_forward_cosine": 0.3783432330597531,
        "mean_feedback_forward_relative_error": 1.11073296232657,
        "min_feedback_forward_cosine": 0.12001358717679977
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 1.1184045793533326
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3182597160339355,
        "mean_feedback_forward_cosine": 0.4222381101413207,
        "mean_feedback_forward_relative_error": 1.0710616138848392,
        "min_feedback_forward_cosine": 0.13862502574920654
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 0.9559098788452148
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.302269458770752,
        "mean_feedback_forward_cosine": 0.46252228292551906,
        "mean_feedback_forward_relative_error": 1.0328783571720124,
        "min_feedback_forward_cosine": 0.16006067395210266
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 0.8277665467453003
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2851325273513794,
        "mean_feedback_forward_cosine": 0.49840348931876094,
        "mean_feedback_forward_relative_error": 0.9977190592072227,
        "min_feedback_forward_cosine": 0.18014489114284515
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.7322896691513061
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2729781866073608,
        "mean_feedback_forward_cosine": 0.5305414419282567,
        "mean_feedback_forward_relative_error": 0.9645393799651752,
        "min_feedback_forward_cosine": 0.19644924998283386
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.6643429850769043
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2598854303359985,
        "mean_feedback_forward_cosine": 0.5591073819182136,
        "mean_feedback_forward_relative_error": 0.9337524311109022,
        "min_feedback_forward_cosine": 0.21356301009655
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.6060961878585815
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2453932762145996,
        "mean_feedback_forward_cosine": 0.5866251238367775,
        "mean_feedback_forward_relative_error": 0.9031165372241627,
        "min_feedback_forward_cosine": 0.23207230865955353
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.56962231924057
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2283390760421753,
        "mean_feedback_forward_cosine": 0.6119551198049025,
        "mean_feedback_forward_relative_error": 0.8740448279814287,
        "min_feedback_forward_cosine": 0.25264260172843933
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.5373372936439514
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2145673036575317,
        "mean_feedback_forward_cosine": 0.6354925567453558,
        "mean_feedback_forward_relative_error": 0.846172617782246,
        "min_feedback_forward_cosine": 0.2757928967475891
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.5023022323799133
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.200329065322876,
        "mean_feedback_forward_cosine": 0.6576699701222506,
        "mean_feedback_forward_relative_error": 0.8189129428430038,
        "min_feedback_forward_cosine": 0.297935426235199
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.480546227684021
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1844172477722168,
        "mean_feedback_forward_cosine": 0.6783240502530878,
        "mean_feedback_forward_relative_error": 0.7928331624377858,
        "min_feedback_forward_cosine": 0.3203599452972412
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.4619817748260498
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1689317226409912,
        "mean_feedback_forward_cosine": 0.6979443685574965,
        "mean_feedback_forward_relative_error": 0.7671210226687518,
        "min_feedback_forward_cosine": 0.3392482101917267
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.43396106838226317
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.152108073234558,
        "mean_feedback_forward_cosine": 0.7162653863430023,
        "mean_feedback_forward_relative_error": 0.7422217447649349,
        "min_feedback_forward_cosine": 0.3589640259742737
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.4181311864089966
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1348786354064941,
        "mean_feedback_forward_cosine": 0.7338095366954803,
        "mean_feedback_forward_relative_error": 0.7178020859306509,
        "min_feedback_forward_cosine": 0.37888485193252563
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.40221583515167236
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1163133382797241,
        "mean_feedback_forward_cosine": 0.7505907714366913,
        "mean_feedback_forward_relative_error": 0.6936617954210802,
        "min_feedback_forward_cosine": 0.39955729246139526
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.38683955910682677
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.096354603767395,
        "mean_feedback_forward_cosine": 0.7664685032584451,
        "mean_feedback_forward_relative_error": 0.6701816315000707,
        "min_feedback_forward_cosine": 0.4216799736022949
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.37687294452667236
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.075522780418396,
        "mean_feedback_forward_cosine": 0.7813193516297774,
        "mean_feedback_forward_relative_error": 0.6474809026176279,
        "min_feedback_forward_cosine": 0.44336777925491333
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.35998005069732664
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0544183254241943,
        "mean_feedback_forward_cosine": 0.7955406438220631,
        "mean_feedback_forward_relative_error": 0.6251224211671136,
        "min_feedback_forward_cosine": 0.46595484018325806
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.35017641264915467
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.033109426498413,
        "mean_feedback_forward_cosine": 0.808514407006177,
        "mean_feedback_forward_relative_error": 0.6041197836399078,
        "min_feedback_forward_cosine": 0.4874805510044098
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.33736838609695435
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0119140148162842,
        "mean_feedback_forward_cosine": 0.8209045312621377,
        "mean_feedback_forward_relative_error": 0.583318842812018,
        "min_feedback_forward_cosine": 0.5093268156051636
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.3264688785266876
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9892695546150208,
        "mean_feedback_forward_cosine": 0.8328157869252292,
        "mean_feedback_forward_relative_error": 0.5627880638295953,
        "min_feedback_forward_cosine": 0.5313945412635803
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.32062696184158324
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9671341180801392,
        "mean_feedback_forward_cosine": 0.8439171400937167,
        "mean_feedback_forward_relative_error": 0.5429453088478609,
        "min_feedback_forward_cosine": 0.5521942973136902
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.31335874647140505
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9442291259765625,
        "mean_feedback_forward_cosine": 0.854020869731903,
        "mean_feedback_forward_relative_error": 0.5243836768648841,
        "min_feedback_forward_cosine": 0.5734392404556274
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.2967264782428741
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9206990003585815,
        "mean_feedback_forward_cosine": 0.8638439904559743,
        "mean_feedback_forward_relative_error": 0.5056991010904313,
        "min_feedback_forward_cosine": 0.5943516492843628
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.29112177519798277
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8963457942008972,
        "mean_feedback_forward_cosine": 0.873080080205744,
        "mean_feedback_forward_relative_error": 0.4876327503811229,
        "min_feedback_forward_cosine": 0.615612268447876
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.2879420889377594
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8727693557739258,
        "mean_feedback_forward_cosine": 0.8818722540682012,
        "mean_feedback_forward_relative_error": 0.46990829543633894,
        "min_feedback_forward_cosine": 0.6358089447021484
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.28146713822364805
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8501355648040771,
        "mean_feedback_forward_cosine": 0.8899538072672757,
        "mean_feedback_forward_relative_error": 0.45294053459709344,
        "min_feedback_forward_cosine": 0.654342770576477
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.2742295644712448
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8255464434623718,
        "mean_feedback_forward_cosine": 0.8975248510187323,
        "mean_feedback_forward_relative_error": 0.436677206104452,
        "min_feedback_forward_cosine": 0.6740566492080688
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.26588446420669554
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8018335700035095,
        "mean_feedback_forward_cosine": 0.9046640190211209,
        "mean_feedback_forward_relative_error": 0.42078411958434364,
        "min_feedback_forward_cosine": 0.6920709609985352
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.2695349159812927
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7783890962600708,
        "mean_feedback_forward_cosine": 0.911278574033217,
        "mean_feedback_forward_relative_error": 0.4056242719292641,
        "min_feedback_forward_cosine": 0.7097922563552856
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.25668339400291446
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7556774616241455,
        "mean_feedback_forward_cosine": 0.9173880251971158,
        "mean_feedback_forward_relative_error": 0.3910275107080286,
        "min_feedback_forward_cosine": 0.7264081239700317
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.2529852655506134
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7331120371818542,
        "mean_feedback_forward_cosine": 0.9231266433542425,
        "mean_feedback_forward_relative_error": 0.3769001170992851,
        "min_feedback_forward_cosine": 0.7425932288169861
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.24265406985282897
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7092484831809998,
        "mean_feedback_forward_cosine": 0.9287206346338446,
        "mean_feedback_forward_relative_error": 0.36265066442164506,
        "min_feedback_forward_cosine": 0.7588396072387695
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.24575304817199706
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.686872661113739,
        "mean_feedback_forward_cosine": 0.933739878914573,
        "mean_feedback_forward_relative_error": 0.34935173134912145,
        "min_feedback_forward_cosine": 0.7736266851425171
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.2325409859275818
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6640751361846924,
        "mean_feedback_forward_cosine": 0.9385467746041037,
        "mean_feedback_forward_relative_error": 0.3362755129283125,
        "min_feedback_forward_cosine": 0.7882757186889648
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.2368836960887909
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.642682671546936,
        "mean_feedback_forward_cosine": 0.9429273377765309,
        "mean_feedback_forward_relative_error": 0.3238413172689351,
        "min_feedback_forward_cosine": 0.8015169501304626
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.22704051488876342
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6210477352142334,
        "mean_feedback_forward_cosine": 0.9470651431517167,
        "mean_feedback_forward_relative_error": 0.31177527890963985,
        "min_feedback_forward_cosine": 0.8147600889205933
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.2250134210729599
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5990719199180603,
        "mean_feedback_forward_cosine": 0.9510523882779208,
        "mean_feedback_forward_relative_error": 0.2996232462200252,
        "min_feedback_forward_cosine": 0.8275330662727356
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.22723024928092955
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5792933106422424,
        "mean_feedback_forward_cosine": 0.9545747236772018,
        "mean_feedback_forward_relative_error": 0.2884721588004719,
        "min_feedback_forward_cosine": 0.8386154770851135
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.2188000938987732
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5588022470474243,
        "mean_feedback_forward_cosine": 0.9579399780793624,
        "mean_feedback_forward_relative_error": 0.27745696793903,
        "min_feedback_forward_cosine": 0.8496239185333252
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.21868580892562867
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.538779079914093,
        "mean_feedback_forward_cosine": 0.9610339793291959,
        "mean_feedback_forward_relative_error": 0.26698302951726044,
        "min_feedback_forward_cosine": 0.8600154519081116
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.21327681936264037
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5196689367294312,
        "mean_feedback_forward_cosine": 0.9639250018379905,
        "mean_feedback_forward_relative_error": 0.2568584924394434,
        "min_feedback_forward_cosine": 0.8696984052658081
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.2090343994283676
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5012470483779907,
        "mean_feedback_forward_cosine": 0.966585939580744,
        "mean_feedback_forward_relative_error": 0.24716081849553367,
        "min_feedback_forward_cosine": 0.8787448406219482
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.20594262621879578
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.48256897926330566,
        "mean_feedback_forward_cosine": 0.9691115661100907,
        "mean_feedback_forward_relative_error": 0.23761470968073065,
        "min_feedback_forward_cosine": 0.8873358964920044
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.2056799218273163
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.465013325214386,
        "mean_feedback_forward_cosine": 0.9714327422055331,
        "mean_feedback_forward_relative_error": 0.2285067179663615,
        "min_feedback_forward_cosine": 0.8953490853309631
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.20313076534509658
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4464685022830963,
        "mean_feedback_forward_cosine": 0.9736069094051014,
        "mean_feedback_forward_relative_error": 0.2196408591487191,
        "min_feedback_forward_cosine": 0.9034064412117004
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.20379007936239243
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.42973530292510986,
        "mean_feedback_forward_cosine": 0.975603437423706,
        "mean_feedback_forward_relative_error": 0.2111600256778977,
        "min_feedback_forward_cosine": 0.910393238067627
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.1951459578704834
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4128367006778717,
        "mean_feedback_forward_cosine": 0.9774770899252458,
        "mean_feedback_forward_relative_error": 0.2028905689716339,
        "min_feedback_forward_cosine": 0.9172727465629578
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.1930615305519104
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.39636051654815674,
        "mean_feedback_forward_cosine": 0.9791891726580533,
        "mean_feedback_forward_relative_error": 0.1950628221712329,
        "min_feedback_forward_cosine": 0.923702597618103
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.1915807810497284
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.38107752799987793,
        "mean_feedback_forward_cosine": 0.9807723077860746,
        "mean_feedback_forward_relative_error": 0.1875022035430778,
        "min_feedback_forward_cosine": 0.9294193983078003
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.1889551772403717
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3662196397781372,
        "mean_feedback_forward_cosine": 0.9822606726126237,
        "mean_feedback_forward_relative_error": 0.18007641305977648,
        "min_feedback_forward_cosine": 0.9347590208053589
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.19030168395996094
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.352567583322525,
        "mean_feedback_forward_cosine": 0.9835671229795976,
        "mean_feedback_forward_relative_error": 0.17330904772335834,
        "min_feedback_forward_cosine": 0.939473032951355
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.18390124263048171
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3388565480709076,
        "mean_feedback_forward_cosine": 0.9848230567845431,
        "mean_feedback_forward_relative_error": 0.1665922653268684,
        "min_feedback_forward_cosine": 0.9440066814422607
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.18452228686332703
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3252885341644287,
        "mean_feedback_forward_cosine": 0.9859845107251948,
        "mean_feedback_forward_relative_error": 0.1601314303549853,
        "min_feedback_forward_cosine": 0.9483773708343506
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.18178530210494995
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3123828172683716,
        "mean_feedback_forward_cosine": 0.9870757471431385,
        "mean_feedback_forward_relative_error": 0.15379290289499542,
        "min_feedback_forward_cosine": 0.9523506760597229
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.183211806974411
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.30033665895462036,
        "mean_feedback_forward_cosine": 0.988059955293482,
        "mean_feedback_forward_relative_error": 0.14786873784932222,
        "min_feedback_forward_cosine": 0.9559201002120972
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.1828134522819519
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2893706262111664,
        "mean_feedback_forward_cosine": 0.9889495286074552,
        "mean_feedback_forward_relative_error": 0.14224021170626988,
        "min_feedback_forward_cosine": 0.9590567350387573
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.17575358436584473
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2780655026435852,
        "mean_feedback_forward_cosine": 0.9897775075652383,
        "mean_feedback_forward_relative_error": 0.1368156373500824,
        "min_feedback_forward_cosine": 0.9621654748916626
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.17398421032905578
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2671937048435211,
        "mean_feedback_forward_cosine": 0.9905405618927695,
        "mean_feedback_forward_relative_error": 0.1316248351877386,
        "min_feedback_forward_cosine": 0.9650338292121887
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.1702138798713684
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2561606764793396,
        "mean_feedback_forward_cosine": 0.9912769718603655,
        "mean_feedback_forward_relative_error": 0.1264395376498049,
        "min_feedback_forward_cosine": 0.9678294062614441
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.17647948033809663
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.24617137014865875,
        "mean_feedback_forward_cosine": 0.9919285492463545,
        "mean_feedback_forward_relative_error": 0.12165466289628636,
        "min_feedback_forward_cosine": 0.970259428024292
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.17145303641319276
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.23637861013412476,
        "mean_feedback_forward_cosine": 0.9925379102880304,
        "mean_feedback_forward_relative_error": 0.11700458303093911,
        "min_feedback_forward_cosine": 0.9725724458694458
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.16560191806793212
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22676199674606323,
        "mean_feedback_forward_cosine": 0.9931233644485473,
        "mean_feedback_forward_relative_error": 0.11233989095146006,
        "min_feedback_forward_cosine": 0.9747302532196045
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.16914532986164094
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.21755750477313995,
        "mean_feedback_forward_cosine": 0.9936565973541953,
        "mean_feedback_forward_relative_error": 0.1079208682206544,
        "min_feedback_forward_cosine": 0.9767346382141113
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.16609831602573394
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20901095867156982,
        "mean_feedback_forward_cosine": 0.9941426092928106,
        "mean_feedback_forward_relative_error": 0.10371322702955116,
        "min_feedback_forward_cosine": 0.9785107374191284
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.1675348688173294
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20092988014221191,
        "mean_feedback_forward_cosine": 0.994590209830891,
        "mean_feedback_forward_relative_error": 0.09967515681954947,
        "min_feedback_forward_cosine": 0.9801209568977356
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.1662090193605423
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19294902682304382,
        "mean_feedback_forward_cosine": 0.9950010332194241,
        "mean_feedback_forward_relative_error": 0.09581900222057646,
        "min_feedback_forward_cosine": 0.981660008430481
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.16575380907058715
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18515849113464355,
        "mean_feedback_forward_cosine": 0.9953876462849703,
        "mean_feedback_forward_relative_error": 0.09206989001144063,
        "min_feedback_forward_cosine": 0.9830981492996216
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.16140517864227294
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17758528888225555,
        "mean_feedback_forward_cosine": 0.9957432486794212,
        "mean_feedback_forward_relative_error": 0.08848337564956057,
        "min_feedback_forward_cosine": 0.9844441413879395
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.1637605294418335
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1703767478466034,
        "mean_feedback_forward_cosine": 0.9960744359276511,
        "mean_feedback_forward_relative_error": 0.08500558852472088,
        "min_feedback_forward_cosine": 0.9856750965118408
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.16280171936035157
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16341739892959595,
        "mean_feedback_forward_cosine": 0.996375438300046,
        "mean_feedback_forward_relative_error": 0.08170108446343378,
        "min_feedback_forward_cosine": 0.9868167638778687
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.16216033221244813
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1568925380706787,
        "mean_feedback_forward_cosine": 0.9966534831307151,
        "mean_feedback_forward_relative_error": 0.07852334237911485,
        "min_feedback_forward_cosine": 0.9878414869308472
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.1593558922433853
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15079909563064575,
        "mean_feedback_forward_cosine": 0.99690823988481,
        "mean_feedback_forward_relative_error": 0.07549517093734308,
        "min_feedback_forward_cosine": 0.988762378692627
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.15664530474185945
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14507438242435455,
        "mean_feedback_forward_cosine": 0.9971384102647954,
        "mean_feedback_forward_relative_error": 0.07264490733769807,
        "min_feedback_forward_cosine": 0.9895980358123779
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.15697357017040253
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13921557366847992,
        "mean_feedback_forward_cosine": 0.9973604917526245,
        "mean_feedback_forward_relative_error": 0.06978270492770455,
        "min_feedback_forward_cosine": 0.990416944026947
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.1557736623430252
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1335224062204361,
        "mean_feedback_forward_cosine": 0.9975588289174167,
        "mean_feedback_forward_relative_error": 0.0671323896470395,
        "min_feedback_forward_cosine": 0.9911783337593079
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.15450497795581816
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12787528336048126,
        "mean_feedback_forward_cosine": 0.9977456851439043,
        "mean_feedback_forward_relative_error": 0.06453609778122468,
        "min_feedback_forward_cosine": 0.9919034242630005
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.15260362037658692
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12294699251651764,
        "mean_feedback_forward_cosine": 0.9979152842001482,
        "mean_feedback_forward_relative_error": 0.06206949288872155,
        "min_feedback_forward_cosine": 0.9925139546394348
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.15245496213912965
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11825887858867645,
        "mean_feedback_forward_cosine": 0.9980723045089028,
        "mean_feedback_forward_relative_error": 0.05969261093573137,
        "min_feedback_forward_cosine": 0.9930729866027832
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.14681518635749816
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11360993981361389,
        "mean_feedback_forward_cosine": 0.9982220443812284,
        "mean_feedback_forward_relative_error": 0.05734120180661028,
        "min_feedback_forward_cosine": 0.9936062097549438
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.15030074099302293
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10911266505718231,
        "mean_feedback_forward_cosine": 0.998359325799075,
        "mean_feedback_forward_relative_error": 0.055096853101118046,
        "min_feedback_forward_cosine": 0.9941010475158691
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.1492630126428604
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10453247278928757,
        "mean_feedback_forward_cosine": 0.9984864820133555,
        "mean_feedback_forward_relative_error": 0.05293607467954809,
        "min_feedback_forward_cosine": 0.9945822954177856
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.15211886331558228
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10039085149765015,
        "mean_feedback_forward_cosine": 0.9986011526801369,
        "mean_feedback_forward_relative_error": 0.05090133160014044,
        "min_feedback_forward_cosine": 0.995002031326294
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.14970635938167573
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09618359804153442,
        "mean_feedback_forward_cosine": 0.9987105098637667,
        "mean_feedback_forward_relative_error": 0.04889079356058077,
        "min_feedback_forward_cosine": 0.9954111576080322
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.15076603876113892
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09216465801000595,
        "mean_feedback_forward_cosine": 0.9988103324716742,
        "mean_feedback_forward_relative_error": 0.04697381913323294,
        "min_feedback_forward_cosine": 0.9957854747772217
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.14972564540863037
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08837541192770004,
        "mean_feedback_forward_cosine": 0.9989030296152288,
        "mean_feedback_forward_relative_error": 0.045120776749470015,
        "min_feedback_forward_cosine": 0.9961245059967041
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.15021497856140137
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08487039059400558,
        "mean_feedback_forward_cosine": 0.9989857966249639,
        "mean_feedback_forward_relative_error": 0.04339315025982532,
        "min_feedback_forward_cosine": 0.9964252710342407
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.14488318905830383
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08149995654821396,
        "mean_feedback_forward_cosine": 0.9990638689561324,
        "mean_feedback_forward_relative_error": 0.041694756072353234,
        "min_feedback_forward_cosine": 0.9967029094696045
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.14355681781768798
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07798758894205093,
        "mean_feedback_forward_cosine": 0.9991379531947049,
        "mean_feedback_forward_relative_error": 0.040026017481630496,
        "min_feedback_forward_cosine": 0.996979832649231
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.1520788672018051
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07485543191432953,
        "mean_feedback_forward_cosine": 0.9992034706202421,
        "mean_feedback_forward_relative_error": 0.03848231861537153,
        "min_feedback_forward_cosine": 0.997217059135437
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.1431914198923111
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07189079374074936,
        "mean_feedback_forward_cosine": 0.9992649891159752,
        "mean_feedback_forward_relative_error": 0.036976625448600815,
        "min_feedback_forward_cosine": 0.9974318146705627
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.14487737963199615
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06921626627445221,
        "mean_feedback_forward_cosine": 0.999318236654455,
        "mean_feedback_forward_relative_error": 0.035616513290865855,
        "min_feedback_forward_cosine": 0.9976187944412231
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.1393111728811264
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06634382903575897,
        "mean_feedback_forward_cosine": 0.9993712479417974,
        "mean_feedback_forward_relative_error": 0.03421463214538314,
        "min_feedback_forward_cosine": 0.9978121519088745
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.14071884412527083
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0636911615729332,
        "mean_feedback_forward_cosine": 0.9994187701832165,
        "mean_feedback_forward_relative_error": 0.03290359213609587,
        "min_feedback_forward_cosine": 0.9979828596115112
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.13622581929683686
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06113015115261078,
        "mean_feedback_forward_cosine": 0.9994621732018211,
        "mean_feedback_forward_relative_error": 0.03166139036078345,
        "min_feedback_forward_cosine": 0.9981414079666138
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.13988637959480285
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05856306850910187,
        "mean_feedback_forward_cosine": 0.9995021874254401,
        "mean_feedback_forward_relative_error": 0.030473380942236294,
        "min_feedback_forward_cosine": 0.9982942342758179
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.1389887262105942
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05612877383828163,
        "mean_feedback_forward_cosine": 0.9995406811887567,
        "mean_feedback_forward_relative_error": 0.029282100541009146,
        "min_feedback_forward_cosine": 0.9984330534934998
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.1357506630563736
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.056110598146915436,
        "mean_feedback_forward_cosine": 0.9995409900491888,
        "mean_feedback_forward_relative_error": 0.029271923581307584,
        "min_feedback_forward_cosine": 0.9984340071678162
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.07303508335351944
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05610087141394615,
        "mean_feedback_forward_cosine": 0.999541123346849,
        "mean_feedback_forward_relative_error": 0.029266326561231504,
        "min_feedback_forward_cosine": 0.9984345436096191
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.05084166835069656
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05609184876084328,
        "mean_feedback_forward_cosine": 0.9995412588119507,
        "mean_feedback_forward_relative_error": 0.029261213820427658,
        "min_feedback_forward_cosine": 0.9984351396560669
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.04009084811389446
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05608273670077324,
        "mean_feedback_forward_cosine": 0.999541396444494,
        "mean_feedback_forward_relative_error": 0.02925638123838739,
        "min_feedback_forward_cosine": 0.9984355568885803
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.034704581493735315
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0560733936727047,
        "mean_feedback_forward_cosine": 0.9995415373282,
        "mean_feedback_forward_relative_error": 0.02925169115716761,
        "min_feedback_forward_cosine": 0.9984361529350281
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.030660836898684502
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.056064337491989136,
        "mean_feedback_forward_cosine": 0.9995416316119108,
        "mean_feedback_forward_relative_error": 0.029247118439525367,
        "min_feedback_forward_cosine": 0.9984366297721863
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.027093489158153534
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05605465546250343,
        "mean_feedback_forward_cosine": 0.9995417800816623,
        "mean_feedback_forward_relative_error": 0.02924259211021391,
        "min_feedback_forward_cosine": 0.9984371662139893
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.025626796327233314
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05604396015405655,
        "mean_feedback_forward_cosine": 0.99954191012816,
        "mean_feedback_forward_relative_error": 0.029237881903959945,
        "min_feedback_forward_cosine": 0.9984377026557922
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.024102603558897973
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05603327602148056,
        "mean_feedback_forward_cosine": 0.9995420585979115,
        "mean_feedback_forward_relative_error": 0.029233268165791578,
        "min_feedback_forward_cosine": 0.9984383583068848
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.021994172270596028
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05602308362722397,
        "mean_feedback_forward_cosine": 0.9995421821420843,
        "mean_feedback_forward_relative_error": 0.029228808586909013,
        "min_feedback_forward_cosine": 0.9984388947486877
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.02064167999714613
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.056012753397226334,
        "mean_feedback_forward_cosine": 0.9995423143560236,
        "mean_feedback_forward_relative_error": 0.029224336858500132,
        "min_feedback_forward_cosine": 0.9984394907951355
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.020304233653247357
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05600287392735481,
        "mean_feedback_forward_cosine": 0.9995424368164756,
        "mean_feedback_forward_relative_error": 0.029220171382820063,
        "min_feedback_forward_cosine": 0.9984400868415833
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.01767860997915268
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055992644280195236,
        "mean_feedback_forward_cosine": 0.9995425560257651,
        "mean_feedback_forward_relative_error": 0.029216162453998218,
        "min_feedback_forward_cosine": 0.998440682888031
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.015958202982246875
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05598311126232147,
        "mean_feedback_forward_cosine": 0.9995426719838922,
        "mean_feedback_forward_relative_error": 0.029212310834025795,
        "min_feedback_forward_cosine": 0.998441219329834
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.014467060579061507
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055972620844841,
        "mean_feedback_forward_cosine": 0.9995427922769027,
        "mean_feedback_forward_relative_error": 0.02920824975113977,
        "min_feedback_forward_cosine": 0.998441755771637
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.01501357210099697
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055961620062589645,
        "mean_feedback_forward_cosine": 0.9995429212396795,
        "mean_feedback_forward_relative_error": 0.029203972245820543,
        "min_feedback_forward_cosine": 0.9984422922134399
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.01460032881140709
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05595085769891739,
        "mean_feedback_forward_cosine": 0.9995430317792026,
        "mean_feedback_forward_relative_error": 0.029199915938079357,
        "min_feedback_forward_cosine": 0.9984430074691772
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.013433075107038022
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055939629673957825,
        "mean_feedback_forward_cosine": 0.999543145569888,
        "mean_feedback_forward_relative_error": 0.02919578467580405,
        "min_feedback_forward_cosine": 0.998443603515625
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.013182361350655555
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05592821165919304,
        "mean_feedback_forward_cosine": 0.9995432810349898,
        "mean_feedback_forward_relative_error": 0.029191561229526996,
        "min_feedback_forward_cosine": 0.998444139957428
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.013328653957396746
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05591729283332825,
        "mean_feedback_forward_cosine": 0.999543407830325,
        "mean_feedback_forward_relative_error": 0.029187543444674124,
        "min_feedback_forward_cosine": 0.998444676399231
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.012116881666332483
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05590762570500374,
        "mean_feedback_forward_cosine": 0.9995435259558938,
        "mean_feedback_forward_relative_error": 0.029183953966606748,
        "min_feedback_forward_cosine": 0.9984453916549683
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.010798799199536442
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05589691922068596,
        "mean_feedback_forward_cosine": 0.9995436321605335,
        "mean_feedback_forward_relative_error": 0.029179944580590183,
        "min_feedback_forward_cosine": 0.9984459280967712
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.011075613817870616
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055886756628751755,
        "mean_feedback_forward_cosine": 0.9995437318628485,
        "mean_feedback_forward_relative_error": 0.029176139501346784,
        "min_feedback_forward_cosine": 0.9984466433525085
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.010073255243003369
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055876828730106354,
        "mean_feedback_forward_cosine": 0.9995438564907421,
        "mean_feedback_forward_relative_error": 0.029172447289932856,
        "min_feedback_forward_cosine": 0.9984471797943115
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.009847601467296482
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05586809292435646,
        "mean_feedback_forward_cosine": 0.9995439496907321,
        "mean_feedback_forward_relative_error": 0.029169122959402473,
        "min_feedback_forward_cosine": 0.9984475374221802
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.008988191752880811
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05585648864507675,
        "mean_feedback_forward_cosine": 0.9995440851558338,
        "mean_feedback_forward_relative_error": 0.029165107350457797,
        "min_feedback_forward_cosine": 0.9984482526779175
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.008952195992469788
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05584488436579704,
        "mean_feedback_forward_cosine": 0.9995442032814026,
        "mean_feedback_forward_relative_error": 0.029161272807554765,
        "min_feedback_forward_cosine": 0.9984488487243652
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.008961161298081279
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05583440884947777,
        "mean_feedback_forward_cosine": 0.9995443257418546,
        "mean_feedback_forward_relative_error": 0.029157466780055652,
        "min_feedback_forward_cosine": 0.998449444770813
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.008415414428785442
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05582287907600403,
        "mean_feedback_forward_cosine": 0.9995444547046315,
        "mean_feedback_forward_relative_error": 0.029153058406981555,
        "min_feedback_forward_cosine": 0.9984500408172607
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.009376009065210819
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05581072345376015,
        "mean_feedback_forward_cosine": 0.9995445782488043,
        "mean_feedback_forward_relative_error": 0.0291485387662595,
        "min_feedback_forward_cosine": 0.9984506964683533
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.00964955476462841
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05579955875873566,
        "mean_feedback_forward_cosine": 0.9995447158813476,
        "mean_feedback_forward_relative_error": 0.02914439800449393,
        "min_feedback_forward_cosine": 0.9984513521194458
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.00819142333328724
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05579020828008652,
        "mean_feedback_forward_cosine": 0.9995448177511042,
        "mean_feedback_forward_relative_error": 0.02914082907478918,
        "min_feedback_forward_cosine": 0.998451828956604
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.007724165812730789
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055780064314603806,
        "mean_feedback_forward_cosine": 0.9995449196208607,
        "mean_feedback_forward_relative_error": 0.029137031750922852,
        "min_feedback_forward_cosine": 0.9984524846076965
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.007425479439049959
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055770739912986755,
        "mean_feedback_forward_cosine": 0.9995450550859625,
        "mean_feedback_forward_relative_error": 0.029133377884599295,
        "min_feedback_forward_cosine": 0.9984529614448547
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.006916446788460017
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05575976148247719,
        "mean_feedback_forward_cosine": 0.9995451580394398,
        "mean_feedback_forward_relative_error": 0.029129705451090226,
        "min_feedback_forward_cosine": 0.9984534978866577
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.006984633325561881
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05575165897607803,
        "mean_feedback_forward_cosine": 0.9995452599091963,
        "mean_feedback_forward_relative_error": 0.029126468952745198,
        "min_feedback_forward_cosine": 0.9984540343284607
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.006382851370871067
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055739447474479675,
        "mean_feedback_forward_cosine": 0.9995453964580189,
        "mean_feedback_forward_relative_error": 0.02912215185774998,
        "min_feedback_forward_cosine": 0.9984546303749084
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.007196955539733172
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055727098137140274,
        "mean_feedback_forward_cosine": 0.999545524337075,
        "mean_feedback_forward_relative_error": 0.029117773837325247,
        "min_feedback_forward_cosine": 0.9984555244445801
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.007402901085168123
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055713534355163574,
        "mean_feedback_forward_cosine": 0.9995456760579889,
        "mean_feedback_forward_relative_error": 0.029112934812226078,
        "min_feedback_forward_cosine": 0.9984561800956726
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.007763749913834036
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05570363998413086,
        "mean_feedback_forward_cosine": 0.9995457898486745,
        "mean_feedback_forward_relative_error": 0.029109399545599112,
        "min_feedback_forward_cosine": 0.998456597328186
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.005878015903830528
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05569057911634445,
        "mean_feedback_forward_cosine": 0.9995459350672635,
        "mean_feedback_forward_relative_error": 0.029105050074444576,
        "min_feedback_forward_cosine": 0.9984574317932129
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.006828229729831219
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05567900836467743,
        "mean_feedback_forward_cosine": 0.999546060778878,
        "mean_feedback_forward_relative_error": 0.02910095618360422,
        "min_feedback_forward_cosine": 0.9984580278396606
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.006254217918552458
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05566876754164696,
        "mean_feedback_forward_cosine": 0.9995461604811928,
        "mean_feedback_forward_relative_error": 0.029097308506342498,
        "min_feedback_forward_cosine": 0.998458743095398
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.005695980285033583
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05565546452999115,
        "mean_feedback_forward_cosine": 0.999546328457919,
        "mean_feedback_forward_relative_error": 0.029092716916718265,
        "min_feedback_forward_cosine": 0.9984594583511353
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.006483047075793147
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05564580857753754,
        "mean_feedback_forward_cosine": 0.9995464227416299,
        "mean_feedback_forward_relative_error": 0.02908916412429376,
        "min_feedback_forward_cosine": 0.9984599351882935
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.005303248546496033
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055633921176195145,
        "mean_feedback_forward_cosine": 0.9995465430346402,
        "mean_feedback_forward_relative_error": 0.029084920036521824,
        "min_feedback_forward_cosine": 0.9984606504440308
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.00568860914632678
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05562036111950874,
        "mean_feedback_forward_cosine": 0.9995467110113664,
        "mean_feedback_forward_relative_error": 0.02908001181415536,
        "min_feedback_forward_cosine": 0.9984612464904785
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.005940030800290406
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05560300499200821,
        "mean_feedback_forward_cosine": 0.9995468909090216,
        "mean_feedback_forward_relative_error": 0.02907430969009345,
        "min_feedback_forward_cosine": 0.9984622001647949
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.006679464721567929
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05558954179286957,
        "mean_feedback_forward_cosine": 0.9995470307090065,
        "mean_feedback_forward_relative_error": 0.029069276746701112,
        "min_feedback_forward_cosine": 0.9984630346298218
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.005969574514850974
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055577341467142105,
        "mean_feedback_forward_cosine": 0.9995471824299206,
        "mean_feedback_forward_relative_error": 0.029064963478595017,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.005214287418797612
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557722598314285,
        "mean_feedback_forward_cosine": 0.9995471705089916,
        "mean_feedback_forward_relative_error": 0.02906490105458281,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.0047789658159390095
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557713657617569,
        "mean_feedback_forward_cosine": 0.9995471672578291,
        "mean_feedback_forward_relative_error": 0.029064845920286395,
        "min_feedback_forward_cosine": 0.9984636306762695
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.004308129639588296
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055577050894498825,
        "mean_feedback_forward_cosine": 0.9995471770113165,
        "mean_feedback_forward_relative_error": 0.029064794054085557,
        "min_feedback_forward_cosine": 0.9984636306762695
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.0042297825917229056
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557696521282196,
        "mean_feedback_forward_cosine": 0.9995471802624789,
        "mean_feedback_forward_relative_error": 0.029064744550057434,
        "min_feedback_forward_cosine": 0.9984638094902039
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.004123410360803828
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557689443230629,
        "mean_feedback_forward_cosine": 0.9995472008531744,
        "mean_feedback_forward_relative_error": 0.02906469644301317,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.004064238927587867
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557681620121002,
        "mean_feedback_forward_cosine": 0.9995471889322455,
        "mean_feedback_forward_relative_error": 0.029064647286114367,
        "min_feedback_forward_cosine": 0.9984638094902039
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.0038662421111017465
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557674169540405,
        "mean_feedback_forward_cosine": 0.9995471878485246,
        "mean_feedback_forward_relative_error": 0.029064595834775404,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.00411278917491436
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557667836546898,
        "mean_feedback_forward_cosine": 0.9995471900159663,
        "mean_feedback_forward_relative_error": 0.029064552579075097,
        "min_feedback_forward_cosine": 0.9984638094902039
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.0034468719584774227
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557659640908241,
        "mean_feedback_forward_cosine": 0.9995471921834078,
        "mean_feedback_forward_relative_error": 0.029064508315853096,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.0034755243319645524
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557649955153465,
        "mean_feedback_forward_cosine": 0.9995471802624789,
        "mean_feedback_forward_relative_error": 0.029064458371563393,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.004001413870276883
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557642877101898,
        "mean_feedback_forward_cosine": 0.9995471791787581,
        "mean_feedback_forward_relative_error": 0.02906441172076897,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.0034888430537469684
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055576350539922714,
        "mean_feedback_forward_cosine": 0.9995471867648038,
        "mean_feedback_forward_relative_error": 0.029064367863942276,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.003523206314407289
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557628720998764,
        "mean_feedback_forward_cosine": 0.9995471965182912,
        "mean_feedback_forward_relative_error": 0.02906432424417951,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.0032350585881620645
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557620897889137,
        "mean_feedback_forward_cosine": 0.9995472008531744,
        "mean_feedback_forward_relative_error": 0.02906427825377746,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.003561518343985081
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055576130747795105,
        "mean_feedback_forward_cosine": 0.9995472019368952,
        "mean_feedback_forward_relative_error": 0.02906423352489417,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.003654435398131609
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055576056241989136,
        "mean_feedback_forward_cosine": 0.9995471900159663,
        "mean_feedback_forward_relative_error": 0.029064187161963094,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.0037600355511903764
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055575959384441376,
        "mean_feedback_forward_cosine": 0.9995472084392201,
        "mean_feedback_forward_relative_error": 0.02906414189968597,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.003887376354970038
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557587742805481,
        "mean_feedback_forward_cosine": 0.9995471867648038,
        "mean_feedback_forward_relative_error": 0.02906409820372408,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.0031937511023879053
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557580664753914,
        "mean_feedback_forward_cosine": 0.999547185681083,
        "mean_feedback_forward_relative_error": 0.029064055159687997,
        "min_feedback_forward_cosine": 0.9984636306762695
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.003321297593843192
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557574704289436,
        "mean_feedback_forward_cosine": 0.9995471976020119,
        "mean_feedback_forward_relative_error": 0.02906401441855864,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.0030844264754094184
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557568371295929,
        "mean_feedback_forward_cosine": 0.9995471954345703,
        "mean_feedback_forward_relative_error": 0.029063970350067723,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.0032436885320395233
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557561665773392,
        "mean_feedback_forward_cosine": 0.9995471900159663,
        "mean_feedback_forward_relative_error": 0.029063926459374754,
        "min_feedback_forward_cosine": 0.9984636306762695
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.0031704610666260123
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557556077837944,
        "mean_feedback_forward_cosine": 0.9995471943508495,
        "mean_feedback_forward_relative_error": 0.029063886302438648,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.003225151681303978
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557548627257347,
        "mean_feedback_forward_cosine": 0.9995471954345703,
        "mean_feedback_forward_relative_error": 0.02906383995644071,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.003295274883285165
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0555754229426384,
        "mean_feedback_forward_cosine": 0.9995472041043368,
        "mean_feedback_forward_relative_error": 0.029063797403465617,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.003042548473458737
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557537451386452,
        "mean_feedback_forward_cosine": 0.9995472127741033,
        "mean_feedback_forward_relative_error": 0.02906376075168902,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.0027685807963088154
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557529255747795,
        "mean_feedback_forward_cosine": 0.9995471997694536,
        "mean_feedback_forward_relative_error": 0.02906371160325679,
        "min_feedback_forward_cosine": 0.9984639286994934
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.003699375984184444
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557522550225258,
        "mean_feedback_forward_cosine": 0.9995471976020119,
        "mean_feedback_forward_relative_error": 0.0290636691857468,
        "min_feedback_forward_cosine": 0.9984638690948486
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.003018076034467667
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0555751696228981,
        "mean_feedback_forward_cosine": 0.9995472116903825,
        "mean_feedback_forward_relative_error": 0.02906362803822214,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.002852506909370422
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055575091391801834,
        "mean_feedback_forward_cosine": 0.9995472008531744,
        "mean_feedback_forward_relative_error": 0.029063586543568155,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.003247974327132106
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557503178715706,
        "mean_feedback_forward_cosine": 0.9995471997694536,
        "mean_feedback_forward_relative_error": 0.029063544481654057,
        "min_feedback_forward_cosine": 0.9984638094902039
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.002867137948926538
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055574964731931686,
        "mean_feedback_forward_cosine": 0.9995472106066617,
        "mean_feedback_forward_relative_error": 0.029063502555205064,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.0031107386077940465
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557490140199661,
        "mean_feedback_forward_cosine": 0.9995472030206161,
        "mean_feedback_forward_relative_error": 0.029063460002229972,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.00310584289137274
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557483434677124,
        "mean_feedback_forward_cosine": 0.9995472095229408,
        "mean_feedback_forward_relative_error": 0.02906341812658039,
        "min_feedback_forward_cosine": 0.9984638094902039
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.0031988932495750485
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557479336857796,
        "mean_feedback_forward_cosine": 0.9995472073554993,
        "mean_feedback_forward_relative_error": 0.029063381754200566,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.0027497278521955012
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055574726313352585,
        "mean_feedback_forward_cosine": 0.9995472268624739,
        "mean_feedback_forward_relative_error": 0.029063340310345995,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.003436942494735122
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557466670870781,
        "mean_feedback_forward_cosine": 0.9995472171089866,
        "mean_feedback_forward_relative_error": 0.02906330324370753,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.002816575135476887
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557459220290184,
        "mean_feedback_forward_cosine": 0.9995472084392201,
        "mean_feedback_forward_relative_error": 0.02906325902281837,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.0032262862982973457
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557452142238617,
        "mean_feedback_forward_cosine": 0.9995472116903825,
        "mean_feedback_forward_relative_error": 0.029063219179145314,
        "min_feedback_forward_cosine": 0.9984638690948486
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.002962641342394054
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0555744469165802,
        "mean_feedback_forward_cosine": 0.9995472116903825,
        "mean_feedback_forward_relative_error": 0.02906317760829221,
        "min_feedback_forward_cosine": 0.9984638690948486
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.002968811162225902
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055574383586645126,
        "mean_feedback_forward_cosine": 0.9995472225275907,
        "mean_feedback_forward_relative_error": 0.02906313966959715,
        "min_feedback_forward_cosine": 0.9984638690948486
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.0027009287017956377
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557432025671005,
        "mean_feedback_forward_cosine": 0.9995472116903825,
        "mean_feedback_forward_relative_error": 0.029063107183372433,
        "min_feedback_forward_cosine": 0.9984636902809143
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.0027876313353702425
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557425320148468,
        "mean_feedback_forward_cosine": 0.9995472116903825,
        "mean_feedback_forward_relative_error": 0.029063069049946287,
        "min_feedback_forward_cosine": 0.9984639883041382
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.0029258847868442535
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557418614625931,
        "mean_feedback_forward_cosine": 0.9995471986857327,
        "mean_feedback_forward_relative_error": 0.0290630292147398,
        "min_feedback_forward_cosine": 0.9984638690948486
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.0028698310617730023
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557411536574364,
        "mean_feedback_forward_cosine": 0.9995472225275907,
        "mean_feedback_forward_relative_error": 0.029062989591197533,
        "min_feedback_forward_cosine": 0.9984638094902039
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.002892410690151155
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557403340935707,
        "mean_feedback_forward_cosine": 0.9995472290299156,
        "mean_feedback_forward_relative_error": 0.029062945395708084,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.0031911729769781233
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557398870587349,
        "mean_feedback_forward_cosine": 0.999547214941545,
        "mean_feedback_forward_relative_error": 0.02906290939585729,
        "min_feedback_forward_cosine": 0.9984637498855591
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.002494597954954952
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055573929101228714,
        "mean_feedback_forward_cosine": 0.9995472095229408,
        "mean_feedback_forward_relative_error": 0.029062868646261366,
        "min_feedback_forward_cosine": 0.9984638690948486
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.0028874194427765906
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557387322187424,
        "mean_feedback_forward_cosine": 0.9995472203601491,
        "mean_feedback_forward_relative_error": 0.029062832316214388,
        "min_feedback_forward_cosine": 0.9984639286994934
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.002676534458883107
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05557382106781006,
        "mean_feedback_forward_cosine": 0.9995472160252658,
        "mean_feedback_forward_relative_error": 0.02906279576603662,
        "min_feedback_forward_cosine": 0.9984638690948486
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.0026516037723701446
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9298,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.3818249603271484,
    "step": 78200
  },
  "hardware": {
    "cuda_device_name": "NVIDIA GeForce GTX 1080",
    "cuda_visible_devices": "5",
    "device": "cuda",
    "device_total_memory_bytes": 8507949056,
    "peak_memory_allocated_bytes": 3318545920,
    "peak_memory_reserved_bytes": 4754243584,
    "torch_version": "2.3.1+cu118"
  },
  "protocol_family": "oral_a_cifar_local_resnet_development",
  "provenance": {
    "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c",
    "git_tracked_dirty": false
  },
  "schema_version": 1,
  "split": {
    "cifar_source_files": [
      {
        "bytes": 31035704,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1",
        "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd"
      },
      {
        "bytes": 31035320,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2",
        "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8"
      },
      {
        "bytes": 31035999,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3",
        "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9"
      },
      {
        "bytes": 31035696,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4",
        "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b"
      },
      {
        "bytes": 31035623,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5",
        "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154"
      },
      {
        "bytes": 31035526,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch",
        "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831"
      }
    ],
    "dataset": "cifar10",
    "evaluation_split": "test",
    "input_layout": "NCHW",
    "input_shape": [
      3,
      32,
      32
    ],
    "loader_seed": 11,
    "normalization_mean": [
      0.49140000343322754,
      0.4821999967098236,
      0.4465000033378601
    ],
    "normalization_std": [
      0.24699999392032623,
      0.2434999942779541,
      0.26159998774528503
    ],
    "split_from_training_only": true,
    "split_seed": null,
    "test_examples": 10000,
    "train_examples": 50000,
    "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5",
    "validation_class_counts": {},
    "validation_examples": 0,
    "validation_index_sha256": null
  },
  "timing": {
    "apical_warmup_wall_s": 0.0,
    "evaluation_wall_s": 2.0549397468566895,
    "mirror_warmup_wall_s": 0.0,
    "predictor_warmup_wall_s": 0.0,
    "timing_excludes_data_loading_hashing_and_model_construction": true,
    "total_timed_wall_s": 18262.86109972,
    "train_wall_s": 18256.888276338577
  },
  "work": {
    "apical_macs_per_example": 125043328,
    "causal_scalar_observations": 0,
    "components": {
      "apical_projection_macs": 1250433280000000,
      "apical_regression_macs": 0,
      "bp_reverse_macs_estimate": 0,
      "kp_reciprocal_correlation_macs": 1250433280000000,
      "local_weight_correlation_macs": 1254856960000000,
      "mirror_feedback_prediction_macs": 0,
      "mirror_local_correlation_macs": 0,
      "mirror_response_macs": 0,
      "ordinary_forward_macs": 1254856960000000,
      "perturbation_forward_macs": 0,
      "warmup_clean_forward_macs": 0
    },
    "definition": "multiply-accumulates in conv/linear maps; one local weight correlation equals one forward-weight MAC count; BP reverse is estimated as one weight-gradient plus one activation-gradient convolution per forward convolution; mixed-traffic/predictor elementwise arithmetic is reported as a conservative operation estimate separately and is not folded into MACs",
    "elementwise_operations_estimate": 0,
    "forward_macs_per_example": 125485696,
    "logical_batch_loss_queries": 0,
    "mirror_probe_examples": 0,
    "mirror_readout_probe_examples": 0,
    "neutral_projection_observations": 0,
    "per_example_cross_entropy_terms": 0,
    "total_clean_forward_examples": 10000000,
    "total_forward_equivalent_examples": 10000000,
    "total_macs_estimate": 5010580480000000
  }
}