{
  "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": 14,
    "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_s14.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": 14,
    "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": 0.9993968506654104,
    "feedback_forward_cosine": [
      0.9995486736297607,
      0.9991993308067322,
      0.9991708993911743,
      0.9989316463470459,
      0.9989402294158936,
      0.998775839805603,
      0.9987432956695557,
      0.9984483122825623,
      0.999407172203064,
      0.9991537928581238,
      0.9992674589157104,
      0.9991320371627808,
      0.9986305236816406,
      0.9982662796974182,
      0.9992818236351013,
      0.9991592764854431,
      0.9995694756507874,
      0.9995255470275879,
      0.9996371865272522,
      0.9996630549430847,
      0.9987449645996094,
      0.9988316297531128,
      0.9991240501403809,
      0.999101996421814,
      0.9987897872924805,
      0.9986447691917419,
      0.9992947578430176,
      0.999209463596344,
      0.9993540048599243,
      0.9992989301681519,
      0.9994040131568909,
      0.999302089214325,
      0.9996514916419983,
      0.999592661857605,
      0.9998459219932556,
      0.9997779130935669,
      0.9997722506523132,
      0.9998602271080017,
      0.9997929334640503,
      0.9997841119766235,
      0.9997772574424744,
      0.9997502565383911,
      0.9997769594192505,
      0.9997308850288391,
      0.9997909069061279,
      0.999720573425293,
      0.9997861385345459,
      0.9996883869171143,
      0.9997413158416748,
      0.9996073246002197,
      0.9997631907463074,
      0.99956876039505,
      0.9996873140335083,
      0.9993549585342407,
      0.9999806880950928
    ],
    "feedback_forward_norm_ratio": [
      0.999349057674408,
      1.0004899501800537,
      0.9993137121200562,
      1.0019783973693848,
      0.9970383644104004,
      1.001090168952942,
      0.9982687830924988,
      1.0007696151733398,
      0.9992757439613342,
      1.0008498430252075,
      0.9987574219703674,
      1.00188410282135,
      1.002251148223877,
      1.0031952857971191,
      0.9994503259658813,
      0.9994155168533325,
      0.9998971819877625,
      1.001420497894287,
      0.9988576769828796,
      1.000478982925415,
      0.9980266690254211,
      1.001484990119934,
      0.9985701441764832,
      1.0008782148361206,
      0.9986444115638733,
      1.0008747577667236,
      0.9995606541633606,
      1.0017122030258179,
      0.9987946152687073,
      1.0015406608581543,
      0.9989224672317505,
      1.0014458894729614,
      0.9994480609893799,
      1.0008728504180908,
      0.9996688961982727,
      1.0007503032684326,
      0.9992460012435913,
      1.0002858638763428,
      0.9995903372764587,
      1.0005089044570923,
      0.9994708895683289,
      1.0008928775787354,
      0.9995566606521606,
      1.0010942220687866,
      0.9995842576026917,
      1.0011290311813354,
      0.9995908141136169,
      1.0012075901031494,
      0.9993408918380737,
      1.0018969774246216,
      0.9994351267814636,
      1.0011729001998901,
      0.9998877048492432,
      1.0017391443252563,
      1.0021624565124512
    ],
    "feedback_forward_relative_error": [
      0.03004281036555767,
      0.04002983868122101,
      0.04071551933884621,
      0.04631281644105911,
      0.04606624320149422,
      0.0495198555290699,
      0.05012083426117897,
      0.05573458969593048,
      0.034428082406520844,
      0.041166696697473526,
      0.03827217221260071,
      0.0417473204433918,
      0.05244080722332001,
      0.059065550565719604,
      0.03788788989186287,
      0.04099579527974129,
      0.029339220374822617,
      0.030859628692269325,
      0.02694813348352909,
      0.02597028762102127,
      0.050091758370399475,
      0.04839812591671944,
      0.04184954985976219,
      0.04240509867668152,
      0.04918261989951134,
      0.052094217389822006,
      0.03755047917366028,
      0.03983255475759506,
      0.03594253957271576,
      0.0375065915286541,
      0.03452565148472786,
      0.03741465136408806,
      0.026397136971354485,
      0.028567994013428688,
      0.017552809789776802,
      0.02109622210264206,
      0.021345283836126328,
      0.016726097092032433,
      0.020349428057670593,
      0.020791415125131607,
      0.021108200773596764,
      0.022380851209163666,
      0.02111971750855446,
      0.023237867280840874,
      0.02044793963432312,
      0.023680970072746277,
      0.02068115770816803,
      0.025010209530591965,
      0.022751644253730774,
      0.028116144239902496,
      0.021761473268270493,
      0.02940809354186058,
      0.02500557154417038,
      0.03599012270569801,
      0.006583130918443203
    ],
    "innovation_negative_gradient_cosine": [
      0.9995798468589783,
      0.9994189739227295,
      0.9995826482772827,
      0.999254584312439,
      0.9995844960212708,
      0.9990823268890381,
      0.9995860457420349,
      0.9990700483322144,
      0.9995884299278259,
      0.9992920160293579,
      0.9995938539505005,
      0.9992960691452026,
      0.9995992183685303,
      0.9988152384757996,
      0.9995999336242676,
      0.9992086887359619,
      0.9996014833450317,
      0.9993894100189209,
      0.9996145367622375,
      0.999541163444519,
      0.9996229410171509,
      0.9991058111190796,
      0.9996281862258911,
      0.9992707371711731,
      0.999628484249115,
      0.9991620779037476,
      0.9996294975280762,
      0.9992156624794006,
      0.9996325969696045,
      0.9993586540222168,
      0.9996317625045776,
      0.9994008541107178,
      0.9996525049209595,
      0.9994752407073975,
      0.9996588230133057,
      0.9996336102485657,
      0.9996620416641235,
      0.9996757507324219,
      0.9997068047523499,
      0.9996463656425476,
      0.999724805355072,
      0.9996492862701416,
      0.9997373819351196,
      0.9996373653411865,
      0.9997483491897583,
      0.9996845722198486,
      0.9997813701629639,
      0.999725878238678,
      0.9998012781143188,
      0.999775767326355,
      0.9998584985733032,
      0.9998941421508789,
      0.9999178647994995,
      0.9999518394470215,
      0.8171511888504028
    ],
    "mean_feedback_forward_cosine": 0.9993700861930848,
    "mean_feedback_forward_relative_error": 0.03353758930143985,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.9995798468589783,
      0.9994189739227295,
      0.9995826482772827,
      0.999254584312439,
      0.9995844960212708,
      0.9990823268890381,
      0.9995860457420349,
      0.9990700483322144,
      0.9995884299278259,
      0.9992920160293579,
      0.9995938539505005,
      0.9992960691452026,
      0.9995992183685303,
      0.9988152384757996,
      0.9995999336242676,
      0.9992086887359619,
      0.9996014833450317,
      0.9993894100189209,
      0.9996145367622375,
      0.999541163444519,
      0.9996229410171509,
      0.9991058111190796,
      0.9996281862258911,
      0.9992707371711731,
      0.999628484249115,
      0.9991620779037476,
      0.9996294975280762,
      0.9992156624794006,
      0.9996325969696045,
      0.9993586540222168,
      0.9996317625045776,
      0.9994008541107178,
      0.9996525049209595,
      0.9994752407073975,
      0.9996588230133057,
      0.9996336102485657,
      0.9996620416641235,
      0.9996757507324219,
      0.9997068047523499,
      0.9996463656425476,
      0.999724805355072,
      0.9996492862701416,
      0.9997373819351196,
      0.9996373653411865,
      0.9997483491897583,
      0.9996845722198486,
      0.9997813701629639,
      0.999725878238678,
      0.9998012781143188,
      0.999775767326355,
      0.9998584985733032,
      0.9998941421508789,
      0.9999178647994995,
      0.9999518394470215,
      0.8171511888504028
    ],
    "teaching_negative_gradient_cosine": [
      0.9995798468589783,
      0.9994189739227295,
      0.9995826482772827,
      0.999254584312439,
      0.9995844960212708,
      0.9990823268890381,
      0.9995860457420349,
      0.9990700483322144,
      0.9995884299278259,
      0.9992920160293579,
      0.9995938539505005,
      0.9992960691452026,
      0.9995992183685303,
      0.9988152384757996,
      0.9995999336242676,
      0.9992086887359619,
      0.9996014833450317,
      0.9993894100189209,
      0.9996145367622375,
      0.999541163444519,
      0.9996229410171509,
      0.9991058111190796,
      0.9996281862258911,
      0.9992707371711731,
      0.999628484249115,
      0.9991620779037476,
      0.9996294975280762,
      0.9992156624794006,
      0.9996325969696045,
      0.9993586540222168,
      0.9996317625045776,
      0.9994008541107178,
      0.9996525049209595,
      0.9994752407073975,
      0.9996588230133057,
      0.9996336102485657,
      0.9996620416641235,
      0.9996757507324219,
      0.9997068047523499,
      0.9996463656425476,
      0.999724805355072,
      0.9996492862701416,
      0.9997373819351196,
      0.9996373653411865,
      0.9997483491897583,
      0.9996845722198486,
      0.9997813701629639,
      0.999725878238678,
      0.9998012781143188,
      0.999775767326355,
      0.9998584985733032,
      0.9998941421508789,
      0.9999178647994995,
      0.9999518394470215,
      0.8171511888504028
    ],
    "wall_s": 0.17903757095336914
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3360395431518555,
        "mean_feedback_forward_cosine": 0.5398452989079735,
        "mean_feedback_forward_relative_error": 0.9057000404054468,
        "min_feedback_forward_cosine": 0.11729566752910614
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 2.261840078430176
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.301427960395813,
        "mean_feedback_forward_cosine": 0.5570171255956996,
        "mean_feedback_forward_relative_error": 0.8909265252676877,
        "min_feedback_forward_cosine": 0.1651471108198166
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 1.632542042617798
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2671263217926025,
        "mean_feedback_forward_cosine": 0.5770768756216222,
        "mean_feedback_forward_relative_error": 0.8736525118350983,
        "min_feedback_forward_cosine": 0.2004789412021637
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.43858590965271
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2385221719741821,
        "mean_feedback_forward_cosine": 0.5996637119488283,
        "mean_feedback_forward_relative_error": 0.8534463828260248,
        "min_feedback_forward_cosine": 0.23759151995182037
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 1.2779202951812745
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2044333219528198,
        "mean_feedback_forward_cosine": 0.6234834736043756,
        "mean_feedback_forward_relative_error": 0.8310065833005038,
        "min_feedback_forward_cosine": 0.28002744913101196
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 1.125606082649231
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.168691873550415,
        "mean_feedback_forward_cosine": 0.6454366548494859,
        "mean_feedback_forward_relative_error": 0.8094265899874947,
        "min_feedback_forward_cosine": 0.3218994140625
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 1.001765019683838
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1300506591796875,
        "mean_feedback_forward_cosine": 0.6663690940900282,
        "mean_feedback_forward_relative_error": 0.7878504964438352,
        "min_feedback_forward_cosine": 0.36677074432373047
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.8995718880844116
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0915017127990723,
        "mean_feedback_forward_cosine": 0.6862029666250402,
        "mean_feedback_forward_relative_error": 0.7663991510868072,
        "min_feedback_forward_cosine": 0.40944093465805054
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.8154410832023621
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.062315583229065,
        "mean_feedback_forward_cosine": 0.7044076865369623,
        "mean_feedback_forward_relative_error": 0.7458936360749331,
        "min_feedback_forward_cosine": 0.45194125175476074
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.7438993234443665
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0368866920471191,
        "mean_feedback_forward_cosine": 0.7206309345635501,
        "mean_feedback_forward_relative_error": 0.7267518672076139,
        "min_feedback_forward_cosine": 0.482383131980896
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.6833764509677887
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.01106858253479,
        "mean_feedback_forward_cosine": 0.7355885321443731,
        "mean_feedback_forward_relative_error": 0.7083800781856884,
        "min_feedback_forward_cosine": 0.5088209509849548
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.6408538368225097
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9844040274620056,
        "mean_feedback_forward_cosine": 0.7500098434361544,
        "mean_feedback_forward_relative_error": 0.690076062896035,
        "min_feedback_forward_cosine": 0.5345862507820129
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.6051172892189026
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9599705338478088,
        "mean_feedback_forward_cosine": 0.7629830067807978,
        "mean_feedback_forward_relative_error": 0.6729642987251282,
        "min_feedback_forward_cosine": 0.5579694509506226
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.5670238526153565
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9349731206893921,
        "mean_feedback_forward_cosine": 0.7752044851129706,
        "mean_feedback_forward_relative_error": 0.656314297968691,
        "min_feedback_forward_cosine": 0.5811662077903748
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.5346621171951293
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9106687307357788,
        "mean_feedback_forward_cosine": 0.7867725946686485,
        "mean_feedback_forward_relative_error": 0.6399037128145044,
        "min_feedback_forward_cosine": 0.6029323935508728
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.5101796304035187
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8859759569168091,
        "mean_feedback_forward_cosine": 0.7978461677377874,
        "mean_feedback_forward_relative_error": 0.6236773769963871,
        "min_feedback_forward_cosine": 0.6241481304168701
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.49010099882125857
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8601457476615906,
        "mean_feedback_forward_cosine": 0.8083188620480624,
        "mean_feedback_forward_relative_error": 0.6077677271582863,
        "min_feedback_forward_cosine": 0.6459009647369385
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.4727135664749145
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8357653617858887,
        "mean_feedback_forward_cosine": 0.8180293473330411,
        "mean_feedback_forward_relative_error": 0.592511474815282,
        "min_feedback_forward_cosine": 0.6656611561775208
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.45012226643562314
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8113493919372559,
        "mean_feedback_forward_cosine": 0.8274970444765958,
        "mean_feedback_forward_relative_error": 0.5771354071118615,
        "min_feedback_forward_cosine": 0.6852105855941772
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.4389621475124359
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7879949808120728,
        "mean_feedback_forward_cosine": 0.8362948872826316,
        "mean_feedback_forward_relative_error": 0.5622997183691372,
        "min_feedback_forward_cosine": 0.7030802369117737
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.4171404622268677
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7637096047401428,
        "mean_feedback_forward_cosine": 0.8449254285205494,
        "mean_feedback_forward_relative_error": 0.5472864966500889,
        "min_feedback_forward_cosine": 0.7209082841873169
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.4103449504566193
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7398356795310974,
        "mean_feedback_forward_cosine": 0.8530320265076378,
        "mean_feedback_forward_relative_error": 0.5328117982907729,
        "min_feedback_forward_cosine": 0.7381588220596313
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.3970993436813354
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.716636598110199,
        "mean_feedback_forward_cosine": 0.8606991063464772,
        "mean_feedback_forward_relative_error": 0.5185809793797407,
        "min_feedback_forward_cosine": 0.75416499376297
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.3831087645149231
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6941056251525879,
        "mean_feedback_forward_cosine": 0.8678555889563127,
        "mean_feedback_forward_relative_error": 0.5049192954193462,
        "min_feedback_forward_cosine": 0.7692728042602539
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.36635501386642455
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6716936230659485,
        "mean_feedback_forward_cosine": 0.8749791730533947,
        "mean_feedback_forward_relative_error": 0.49094241992993787,
        "min_feedback_forward_cosine": 0.7837886810302734
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.3610792467689514
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6493027210235596,
        "mean_feedback_forward_cosine": 0.8815253160216592,
        "mean_feedback_forward_relative_error": 0.47770702784711666,
        "min_feedback_forward_cosine": 0.7960513234138489
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.3519402063179016
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6338154673576355,
        "mean_feedback_forward_cosine": 0.8878368030894886,
        "mean_feedback_forward_relative_error": 0.4645320017229427,
        "min_feedback_forward_cosine": 0.8039942979812622
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.34612735505104064
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6198217272758484,
        "mean_feedback_forward_cosine": 0.8939330339431762,
        "mean_feedback_forward_relative_error": 0.4514224285429174,
        "min_feedback_forward_cosine": 0.8125402331352234
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.3347622646522522
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6074491739273071,
        "mean_feedback_forward_cosine": 0.8996492028236389,
        "mean_feedback_forward_relative_error": 0.43872043680060996,
        "min_feedback_forward_cosine": 0.8199578523635864
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.327846246509552
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5957520604133606,
        "mean_feedback_forward_cosine": 0.9050342408093539,
        "mean_feedback_forward_relative_error": 0.42643835246562956,
        "min_feedback_forward_cosine": 0.8265140652656555
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.31661326223373415
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5830815434455872,
        "mean_feedback_forward_cosine": 0.9101586894555526,
        "mean_feedback_forward_relative_error": 0.4143924686041745,
        "min_feedback_forward_cosine": 0.8340471386909485
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.303727802362442
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5698826313018799,
        "mean_feedback_forward_cosine": 0.9152525284073569,
        "mean_feedback_forward_relative_error": 0.40209951116280124,
        "min_feedback_forward_cosine": 0.8409469127655029
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.30891136761665344
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5572258830070496,
        "mean_feedback_forward_cosine": 0.9199291597713124,
        "mean_feedback_forward_relative_error": 0.3904243734749881,
        "min_feedback_forward_cosine": 0.8477466106414795
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.2947652544689178
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5451564788818359,
        "mean_feedback_forward_cosine": 0.9245218396186828,
        "mean_feedback_forward_relative_error": 0.37870549031279305,
        "min_feedback_forward_cosine": 0.8543558120727539
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.29327453715324403
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5320948958396912,
        "mean_feedback_forward_cosine": 0.928714085708965,
        "mean_feedback_forward_relative_error": 0.36763537986712025,
        "min_feedback_forward_cosine": 0.8603912591934204
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.28498666655540467
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5205531716346741,
        "mean_feedback_forward_cosine": 0.9328078356656161,
        "mean_feedback_forward_relative_error": 0.35648485327308826,
        "min_feedback_forward_cosine": 0.8662968873977661
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.27884571946144104
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5085384249687195,
        "mean_feedback_forward_cosine": 0.9367399844256314,
        "mean_feedback_forward_relative_error": 0.34547765525904567,
        "min_feedback_forward_cosine": 0.8721722960472107
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.2728859224414825
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4989911913871765,
        "mean_feedback_forward_cosine": 0.9404762939973311,
        "mean_feedback_forward_relative_error": 0.33477059453725816,
        "min_feedback_forward_cosine": 0.8767096996307373
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.2726372246932983
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4896278977394104,
        "mean_feedback_forward_cosine": 0.9438534920865839,
        "mean_feedback_forward_relative_error": 0.3247550377791578,
        "min_feedback_forward_cosine": 0.8813626766204834
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.2597659772014618
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4794587194919586,
        "mean_feedback_forward_cosine": 0.9471759416840293,
        "mean_feedback_forward_relative_error": 0.31466083838181064,
        "min_feedback_forward_cosine": 0.8863492608070374
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.2566597058677673
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.46974116563796997,
        "mean_feedback_forward_cosine": 0.9502717776732011,
        "mean_feedback_forward_relative_error": 0.3049257840622555,
        "min_feedback_forward_cosine": 0.8909269571304321
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.2535866453552246
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.45840635895729065,
        "mean_feedback_forward_cosine": 0.953336640921506,
        "mean_feedback_forward_relative_error": 0.29503738731145857,
        "min_feedback_forward_cosine": 0.8961610794067383
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.2539304351520538
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4476323127746582,
        "mean_feedback_forward_cosine": 0.956215010989796,
        "mean_feedback_forward_relative_error": 0.2854661605574868,
        "min_feedback_forward_cosine": 0.9009397625923157
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.24867324002742766
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4384644627571106,
        "mean_feedback_forward_cosine": 0.9589088060639122,
        "mean_feedback_forward_relative_error": 0.276235669715838,
        "min_feedback_forward_cosine": 0.9049193263053894
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.24410620104789735
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.42801737785339355,
        "mean_feedback_forward_cosine": 0.9613754695112054,
        "mean_feedback_forward_relative_error": 0.2675473301248117,
        "min_feedback_forward_cosine": 0.9095538854598999
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.23810331386089326
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.41825419664382935,
        "mean_feedback_forward_cosine": 0.9637145074931058,
        "mean_feedback_forward_relative_error": 0.2590205722234466,
        "min_feedback_forward_cosine": 0.9137883186340332
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.23533988403320313
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4078068435192108,
        "mean_feedback_forward_cosine": 0.9660646536133506,
        "mean_feedback_forward_relative_error": 0.2501983148130504,
        "min_feedback_forward_cosine": 0.9179442524909973
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.23375252881526948
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3972949683666229,
        "mean_feedback_forward_cosine": 0.9681946667757901,
        "mean_feedback_forward_relative_error": 0.24198557592251085,
        "min_feedback_forward_cosine": 0.9221304059028625
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.22916575040340423
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3864702880382538,
        "mean_feedback_forward_cosine": 0.9702675407583063,
        "mean_feedback_forward_relative_error": 0.23377114974639632,
        "min_feedback_forward_cosine": 0.9264482259750366
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.2279023641014099
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3768327236175537,
        "mean_feedback_forward_cosine": 0.9721953045238149,
        "mean_feedback_forward_relative_error": 0.22584167129614136,
        "min_feedback_forward_cosine": 0.9300549626350403
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.22254173853874207
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3676289916038513,
        "mean_feedback_forward_cosine": 0.9739825508811257,
        "mean_feedback_forward_relative_error": 0.21825949299064548,
        "min_feedback_forward_cosine": 0.9334088563919067
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.216740937871933
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3589218854904175,
        "mean_feedback_forward_cosine": 0.9756713390350342,
        "mean_feedback_forward_relative_error": 0.2108152225613594,
        "min_feedback_forward_cosine": 0.9364010095596313
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.21592123507976532
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.34930217266082764,
        "mean_feedback_forward_cosine": 0.97728990316391,
        "mean_feedback_forward_relative_error": 0.20350951477885246,
        "min_feedback_forward_cosine": 0.939773440361023
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.21098066742420196
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3383598029613495,
        "mean_feedback_forward_cosine": 0.9788350506262345,
        "mean_feedback_forward_relative_error": 0.19632480801506477,
        "min_feedback_forward_cosine": 0.9434282779693604
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.21545665407657624
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3280719518661499,
        "mean_feedback_forward_cosine": 0.9802916006608443,
        "mean_feedback_forward_relative_error": 0.18931772085753354,
        "min_feedback_forward_cosine": 0.9468652009963989
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.2137152345609665
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3173877000808716,
        "mean_feedback_forward_cosine": 0.9816158673980019,
        "mean_feedback_forward_relative_error": 0.182749734277075,
        "min_feedback_forward_cosine": 0.9502993226051331
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.2024301117658615
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3077414333820343,
        "mean_feedback_forward_cosine": 0.9828643029386347,
        "mean_feedback_forward_relative_error": 0.17626704580404542,
        "min_feedback_forward_cosine": 0.9531734585762024
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.2101277583503723
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.29863837361335754,
        "mean_feedback_forward_cosine": 0.9840318376367743,
        "mean_feedback_forward_relative_error": 0.1699988741766323,
        "min_feedback_forward_cosine": 0.9558677673339844
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.20592128714084626
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.28895822167396545,
        "mean_feedback_forward_cosine": 0.985127288644964,
        "mean_feedback_forward_relative_error": 0.16398801268501714,
        "min_feedback_forward_cosine": 0.9586999416351318
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.20172549364566802
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2780231237411499,
        "mean_feedback_forward_cosine": 0.9861920714378357,
        "mean_feedback_forward_relative_error": 0.15795523747801782,
        "min_feedback_forward_cosine": 0.9618793725967407
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.20062646031856537
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.26860901713371277,
        "mean_feedback_forward_cosine": 0.9871679197658192,
        "mean_feedback_forward_relative_error": 0.1521740548312664,
        "min_feedback_forward_cosine": 0.9643744826316833
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.1979875206375122
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.25864025950431824,
        "mean_feedback_forward_cosine": 0.9880671089345758,
        "mean_feedback_forward_relative_error": 0.14666806825182654,
        "min_feedback_forward_cosine": 0.9669016599655151
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.19849038482666015
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.24903854727745056,
        "mean_feedback_forward_cosine": 0.9888936259529807,
        "mean_feedback_forward_relative_error": 0.1414306450296532,
        "min_feedback_forward_cosine": 0.9693467617034912
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.19536169333457948
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2402242124080658,
        "mean_feedback_forward_cosine": 0.9896845037286932,
        "mean_feedback_forward_relative_error": 0.13624327758496457,
        "min_feedback_forward_cosine": 0.9714638590812683
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.19034876211166382
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.23262028396129608,
        "mean_feedback_forward_cosine": 0.9904228037053888,
        "mean_feedback_forward_relative_error": 0.13119625354355033,
        "min_feedback_forward_cosine": 0.9732394218444824
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.19610364086151122
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22418992221355438,
        "mean_feedback_forward_cosine": 0.9911059455438094,
        "mean_feedback_forward_relative_error": 0.12638776359910314,
        "min_feedback_forward_cosine": 0.9751158952713013
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.18961908777713776
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.21725112199783325,
        "mean_feedback_forward_cosine": 0.9917246786030856,
        "mean_feedback_forward_relative_error": 0.12187630513852293,
        "min_feedback_forward_cosine": 0.9766388535499573
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.18855149097442628
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.21030044555664062,
        "mean_feedback_forward_cosine": 0.9923040400851857,
        "mean_feedback_forward_relative_error": 0.11746241463856263,
        "min_feedback_forward_cosine": 0.9781090021133423
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.1827023038291931
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20252923667430878,
        "mean_feedback_forward_cosine": 0.9928736892613498,
        "mean_feedback_forward_relative_error": 0.11299660514024171,
        "min_feedback_forward_cosine": 0.9796693921089172
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.18777125893592833
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19475862383842468,
        "mean_feedback_forward_cosine": 0.9933944236148488,
        "mean_feedback_forward_relative_error": 0.1087731986221942,
        "min_feedback_forward_cosine": 0.9812123775482178
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.1827207085800171
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18700455129146576,
        "mean_feedback_forward_cosine": 0.993892765045166,
        "mean_feedback_forward_relative_error": 0.10458857572891496,
        "min_feedback_forward_cosine": 0.9826465845108032
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.18099756442070009
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18052984774112701,
        "mean_feedback_forward_cosine": 0.9943497278473594,
        "mean_feedback_forward_relative_error": 0.1005988916551525,
        "min_feedback_forward_cosine": 0.9838318228721619
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.180845529088974
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17324385046958923,
        "mean_feedback_forward_cosine": 0.9947746981273998,
        "mean_feedback_forward_relative_error": 0.09673330394381827,
        "min_feedback_forward_cosine": 0.9851095676422119
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.18006097470283508
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1666959673166275,
        "mean_feedback_forward_cosine": 0.9951619343324141,
        "mean_feedback_forward_relative_error": 0.09307177717035467,
        "min_feedback_forward_cosine": 0.9862115979194641
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.1777750191783905
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16139131784439087,
        "mean_feedback_forward_cosine": 0.9955085570161993,
        "mean_feedback_forward_relative_error": 0.08965976339849559,
        "min_feedback_forward_cosine": 0.987093448638916
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.1749332488489151
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15509504079818726,
        "mean_feedback_forward_cosine": 0.9958382075483149,
        "mean_feedback_forward_relative_error": 0.08627855002202771,
        "min_feedback_forward_cosine": 0.9880722165107727
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.17808644639015198
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1499006450176239,
        "mean_feedback_forward_cosine": 0.9961370966651223,
        "mean_feedback_forward_relative_error": 0.08310288553210822,
        "min_feedback_forward_cosine": 0.9888517260551453
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.17081560897827147
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1446233093738556,
        "mean_feedback_forward_cosine": 0.9964214520020919,
        "mean_feedback_forward_relative_error": 0.07997748160904104,
        "min_feedback_forward_cosine": 0.9896301627159119
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.1707915583229065
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13814088702201843,
        "mean_feedback_forward_cosine": 0.996705302325162,
        "mean_feedback_forward_relative_error": 0.07675201486457478,
        "min_feedback_forward_cosine": 0.990538477897644
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.17181168204307556
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13184763491153717,
        "mean_feedback_forward_cosine": 0.9969579989259894,
        "mean_feedback_forward_relative_error": 0.07376226691359823,
        "min_feedback_forward_cosine": 0.9913761615753174
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.16899568368911744
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12670230865478516,
        "mean_feedback_forward_cosine": 0.9971872037107294,
        "mean_feedback_forward_relative_error": 0.07091320035132495,
        "min_feedback_forward_cosine": 0.9920350313186646
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.17493566308021546
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12146414816379547,
        "mean_feedback_forward_cosine": 0.9973919738422741,
        "mean_feedback_forward_relative_error": 0.06827812296423046,
        "min_feedback_forward_cosine": 0.9926751255989075
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.1669038712310791
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11693733185529709,
        "mean_feedback_forward_cosine": 0.9975873134352944,
        "mean_feedback_forward_relative_error": 0.06566752669486133,
        "min_feedback_forward_cosine": 0.9932080507278442
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.1658772466468811
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11253160238265991,
        "mean_feedback_forward_cosine": 0.9977676835927096,
        "mean_feedback_forward_relative_error": 0.06316453343765302,
        "min_feedback_forward_cosine": 0.9937189817428589
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.1679341847229004
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10800867527723312,
        "mean_feedback_forward_cosine": 0.99793504151431,
        "mean_feedback_forward_relative_error": 0.06075672393834049,
        "min_feedback_forward_cosine": 0.9942063093185425
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.16437279180526734
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10382198542356491,
        "mean_feedback_forward_cosine": 0.9980900504372336,
        "mean_feedback_forward_relative_error": 0.058428227579729125,
        "min_feedback_forward_cosine": 0.9946441650390625
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.16653217509269713
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10005079954862595,
        "mean_feedback_forward_cosine": 0.9982323559847746,
        "mean_feedback_forward_relative_error": 0.05619608253579248,
        "min_feedback_forward_cosine": 0.9950335621833801
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.1683104187965393
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09619463235139847,
        "mean_feedback_forward_cosine": 0.9983689481561834,
        "mean_feedback_forward_relative_error": 0.053986180065707726,
        "min_feedback_forward_cosine": 0.9954037070274353
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.16667857217788695
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09265120327472687,
        "mean_feedback_forward_cosine": 0.998490385575728,
        "mean_feedback_forward_relative_error": 0.0519376022063873,
        "min_feedback_forward_cosine": 0.9957320690155029
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.1615715386390686
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08872567117214203,
        "mean_feedback_forward_cosine": 0.9986063913865523,
        "mean_feedback_forward_relative_error": 0.049906102246181534,
        "min_feedback_forward_cosine": 0.996086061000824
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.16552179415225982
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08490674197673798,
        "mean_feedback_forward_cosine": 0.99870783849196,
        "mean_feedback_forward_relative_error": 0.04805871864611452,
        "min_feedback_forward_cosine": 0.9964166879653931
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.157160287899971
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08148565143346786,
        "mean_feedback_forward_cosine": 0.9988041726025668,
        "mean_feedback_forward_relative_error": 0.04622622855346311,
        "min_feedback_forward_cosine": 0.9967026114463806
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.1610899450969696
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07781962305307388,
        "mean_feedback_forward_cosine": 0.9988956841555509,
        "mean_feedback_forward_relative_error": 0.04442393962632526,
        "min_feedback_forward_cosine": 0.9969951510429382
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.1601667630839348
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07474031299352646,
        "mean_feedback_forward_cosine": 0.9989781065420671,
        "mean_feedback_forward_relative_error": 0.04273093168369748,
        "min_feedback_forward_cosine": 0.997226357460022
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.15819521648168564
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07225430011749268,
        "mean_feedback_forward_cosine": 0.9990535009991038,
        "mean_feedback_forward_relative_error": 0.04112511758099903,
        "min_feedback_forward_cosine": 0.9974077343940735
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.15748157766342163
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06968426704406738,
        "mean_feedback_forward_cosine": 0.9991265687075528,
        "mean_feedback_forward_relative_error": 0.03950692435557192,
        "min_feedback_forward_cosine": 0.9975906610488892
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.16175335730075835
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06721122562885284,
        "mean_feedback_forward_cosine": 0.9991913286122409,
        "mean_feedback_forward_relative_error": 0.03802152044055137,
        "min_feedback_forward_cosine": 0.9977591037750244
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.1524793500804901
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06464006751775742,
        "mean_feedback_forward_cosine": 0.9992528611963446,
        "mean_feedback_forward_relative_error": 0.03654412842271003,
        "min_feedback_forward_cosine": 0.9979234933853149
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.1604527052783966
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06187574937939644,
        "mean_feedback_forward_cosine": 0.9993114037947222,
        "mean_feedback_forward_relative_error": 0.03509028789333322,
        "min_feedback_forward_cosine": 0.9980965852737427
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.15484731866598128
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05939973518252373,
        "mean_feedback_forward_cosine": 0.999362301826477,
        "mean_feedback_forward_relative_error": 0.03376985603774136,
        "min_feedback_forward_cosine": 0.998246967792511
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.15393349830150604
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059377092868089676,
        "mean_feedback_forward_cosine": 0.9993627006357366,
        "mean_feedback_forward_relative_error": 0.033757664347914135,
        "min_feedback_forward_cosine": 0.9982478618621826
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.08632886606931686
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05936736986041069,
        "mean_feedback_forward_cosine": 0.9993629076264121,
        "mean_feedback_forward_relative_error": 0.03375121339816939,
        "min_feedback_forward_cosine": 0.9982483386993408
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.05908495314359665
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05935855582356453,
        "mean_feedback_forward_cosine": 0.9993630907752297,
        "mean_feedback_forward_relative_error": 0.03374522678892721,
        "min_feedback_forward_cosine": 0.9982488751411438
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.048748821417093274
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05935034528374672,
        "mean_feedback_forward_cosine": 0.9993633042682301,
        "mean_feedback_forward_relative_error": 0.033739580772817135,
        "min_feedback_forward_cosine": 0.998249351978302
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.04319611127018928
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05934198945760727,
        "mean_feedback_forward_cosine": 0.9993634288961237,
        "mean_feedback_forward_relative_error": 0.033734094114466144,
        "min_feedback_forward_cosine": 0.998249888420105
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.038027093428373335
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05933482199907303,
        "mean_feedback_forward_cosine": 0.9993636033751748,
        "mean_feedback_forward_relative_error": 0.0337288758802143,
        "min_feedback_forward_cosine": 0.9982502460479736
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.036271356392502785
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05932800844311714,
        "mean_feedback_forward_cosine": 0.9993637550960888,
        "mean_feedback_forward_relative_error": 0.03372401965951378,
        "min_feedback_forward_cosine": 0.9982507228851318
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.030873922350406646
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05932128429412842,
        "mean_feedback_forward_cosine": 0.9993639165704901,
        "mean_feedback_forward_relative_error": 0.03371912017464638,
        "min_feedback_forward_cosine": 0.9982513189315796
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.02892630367606878
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05931437388062477,
        "mean_feedback_forward_cosine": 0.999364068291404,
        "mean_feedback_forward_relative_error": 0.033714231103658676,
        "min_feedback_forward_cosine": 0.9982515573501587
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.026974181639552116
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0593072883784771,
        "mean_feedback_forward_cosine": 0.9993642265146429,
        "mean_feedback_forward_relative_error": 0.03370935560458086,
        "min_feedback_forward_cosine": 0.9982519745826721
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.025269123996645212
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059300776571035385,
        "mean_feedback_forward_cosine": 0.9993643739006736,
        "mean_feedback_forward_relative_error": 0.03370474155314944,
        "min_feedback_forward_cosine": 0.9982523322105408
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.023698409345149996
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05929463356733322,
        "mean_feedback_forward_cosine": 0.9993645191192627,
        "mean_feedback_forward_relative_error": 0.033700264897197484,
        "min_feedback_forward_cosine": 0.9982526898384094
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.021694072423279285
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059288155287504196,
        "mean_feedback_forward_cosine": 0.999364686012268,
        "mean_feedback_forward_relative_error": 0.03369569815695286,
        "min_feedback_forward_cosine": 0.9982531070709229
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.020212705764472486
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05928143858909607,
        "mean_feedback_forward_cosine": 0.9993648214773698,
        "mean_feedback_forward_relative_error": 0.03369119951332157,
        "min_feedback_forward_cosine": 0.9982537031173706
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.019997652606964112
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059275150299072266,
        "mean_feedback_forward_cosine": 0.9993649797006087,
        "mean_feedback_forward_relative_error": 0.03368665474382314,
        "min_feedback_forward_cosine": 0.9982538819313049
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.018869017801880838
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059269119054079056,
        "mean_feedback_forward_cosine": 0.9993651238354769,
        "mean_feedback_forward_relative_error": 0.033682371489703655,
        "min_feedback_forward_cosine": 0.9982541799545288
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.017170048105716706
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05926356837153435,
        "mean_feedback_forward_cosine": 0.999365263635462,
        "mean_feedback_forward_relative_error": 0.03367843947119333,
        "min_feedback_forward_cosine": 0.9982545971870422
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.015152093956172466
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05925753712654114,
        "mean_feedback_forward_cosine": 0.9993653915145181,
        "mean_feedback_forward_relative_error": 0.03367414422841235,
        "min_feedback_forward_cosine": 0.9982547760009766
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.015982890694737434
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05925161391496658,
        "mean_feedback_forward_cosine": 0.9993655323982239,
        "mean_feedback_forward_relative_error": 0.033669975645501504,
        "min_feedback_forward_cosine": 0.99825519323349
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.015350015664100648
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059244535863399506,
        "mean_feedback_forward_cosine": 0.9993656971237876,
        "mean_feedback_forward_relative_error": 0.03366517693820325,
        "min_feedback_forward_cosine": 0.9982555508613586
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.01586255150437355
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05923856422305107,
        "mean_feedback_forward_cosine": 0.999365834756331,
        "mean_feedback_forward_relative_error": 0.0336608559909192,
        "min_feedback_forward_cosine": 0.9982557892799377
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.013963704481124878
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05923229455947876,
        "mean_feedback_forward_cosine": 0.9993660081516612,
        "mean_feedback_forward_relative_error": 0.033656191275539725,
        "min_feedback_forward_cosine": 0.9982560873031616
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.014500608921051025
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059226661920547485,
        "mean_feedback_forward_cosine": 0.9993661522865296,
        "mean_feedback_forward_relative_error": 0.0336520003662868,
        "min_feedback_forward_cosine": 0.9982565641403198
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.012911818793714046
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05922118201851845,
        "mean_feedback_forward_cosine": 0.9993662931702354,
        "mean_feedback_forward_relative_error": 0.03364786172116345,
        "min_feedback_forward_cosine": 0.9982568025588989
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.012699424402415753
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05921560898423195,
        "mean_feedback_forward_cosine": 0.9993664340539412,
        "mean_feedback_forward_relative_error": 0.033643715862523425,
        "min_feedback_forward_cosine": 0.998257040977478
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.01211710981324315
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05921022593975067,
        "mean_feedback_forward_cosine": 0.999366557598114,
        "mean_feedback_forward_relative_error": 0.03363977646963163,
        "min_feedback_forward_cosine": 0.9982572197914124
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.011623655118197202
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059204477816820145,
        "mean_feedback_forward_cosine": 0.9993666941469366,
        "mean_feedback_forward_relative_error": 0.0336357313462279,
        "min_feedback_forward_cosine": 0.998257577419281
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.011208783966600895
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059199266135692596,
        "mean_feedback_forward_cosine": 0.9993668328632008,
        "mean_feedback_forward_relative_error": 0.03363182963803411,
        "min_feedback_forward_cosine": 0.998258113861084
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.010452382577955723
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059193745255470276,
        "mean_feedback_forward_cosine": 0.9993669683283025,
        "mean_feedback_forward_relative_error": 0.03362774834544821,
        "min_feedback_forward_cosine": 0.9982582926750183
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.010859078953117132
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05918765813112259,
        "mean_feedback_forward_cosine": 0.9993671048771251,
        "mean_feedback_forward_relative_error": 0.033623597567731686,
        "min_feedback_forward_cosine": 0.9982587099075317
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.009921169189661742
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059181831777095795,
        "mean_feedback_forward_cosine": 0.9993672414259477,
        "mean_feedback_forward_relative_error": 0.03361950904469598,
        "min_feedback_forward_cosine": 0.9982592463493347
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.009995801807530224
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05917598307132721,
        "mean_feedback_forward_cosine": 0.9993673833933744,
        "mean_feedback_forward_relative_error": 0.033615434880961075,
        "min_feedback_forward_cosine": 0.9982595443725586
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.009514382094442844
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05917006731033325,
        "mean_feedback_forward_cosine": 0.999367525360801,
        "mean_feedback_forward_relative_error": 0.033611282359131356,
        "min_feedback_forward_cosine": 0.9982598423957825
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.009537497631013393
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05916416645050049,
        "mean_feedback_forward_cosine": 0.9993676684119485,
        "mean_feedback_forward_relative_error": 0.03360732892866839,
        "min_feedback_forward_cosine": 0.9982601404190063
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.009508014184907078
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05915813893079758,
        "mean_feedback_forward_cosine": 0.9993678255514665,
        "mean_feedback_forward_relative_error": 0.03360312380404635,
        "min_feedback_forward_cosine": 0.9982608556747437
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.009375937500596046
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05915252864360809,
        "mean_feedback_forward_cosine": 0.9993679523468018,
        "mean_feedback_forward_relative_error": 0.03359934392131188,
        "min_feedback_forward_cosine": 0.9982609748840332
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.008746569739431142
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059146810322999954,
        "mean_feedback_forward_cosine": 0.9993680899793451,
        "mean_feedback_forward_relative_error": 0.03359527747908776,
        "min_feedback_forward_cosine": 0.9982613325119019
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.008365975003316999
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05914038419723511,
        "mean_feedback_forward_cosine": 0.9993682514537464,
        "mean_feedback_forward_relative_error": 0.03359085150401701,
        "min_feedback_forward_cosine": 0.9982618093490601
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.008953102668970822
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05913488566875458,
        "mean_feedback_forward_cosine": 0.9993683934211731,
        "mean_feedback_forward_relative_error": 0.03358684871345759,
        "min_feedback_forward_cosine": 0.9982622265815735
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.008177163695096969
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05912889167666435,
        "mean_feedback_forward_cosine": 0.9993685267188332,
        "mean_feedback_forward_relative_error": 0.033582697419280355,
        "min_feedback_forward_cosine": 0.9982626438140869
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.008396954881548882
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05912240967154503,
        "mean_feedback_forward_cosine": 0.9993686773560264,
        "mean_feedback_forward_relative_error": 0.03357825619591908,
        "min_feedback_forward_cosine": 0.9982631206512451
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.008055837308987976
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059114810079336166,
        "mean_feedback_forward_cosine": 0.9993688615885647,
        "mean_feedback_forward_relative_error": 0.03357319518585097,
        "min_feedback_forward_cosine": 0.9982635974884033
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.009275788169205188
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05910937115550041,
        "mean_feedback_forward_cosine": 0.9993690013885498,
        "mean_feedback_forward_relative_error": 0.03356921449303627,
        "min_feedback_forward_cosine": 0.998263955116272
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.007554302980676293
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0591028667986393,
        "mean_feedback_forward_cosine": 0.9993691628629511,
        "mean_feedback_forward_relative_error": 0.03356482491574504,
        "min_feedback_forward_cosine": 0.9982640743255615
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.007218919408172369
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05909779667854309,
        "mean_feedback_forward_cosine": 0.9993693037466569,
        "mean_feedback_forward_relative_error": 0.03356129590591246,
        "min_feedback_forward_cosine": 0.998264491558075
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.006407227513724938
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05909143015742302,
        "mean_feedback_forward_cosine": 0.9993694240396673,
        "mean_feedback_forward_relative_error": 0.03355686375871301,
        "min_feedback_forward_cosine": 0.998264729976654
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.0071185772954672575
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0590851828455925,
        "mean_feedback_forward_cosine": 0.9993696093559266,
        "mean_feedback_forward_relative_error": 0.03355245374312455,
        "min_feedback_forward_cosine": 0.9982651472091675
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.007279688682090491
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05907873064279556,
        "mean_feedback_forward_cosine": 0.9993697556582364,
        "mean_feedback_forward_relative_error": 0.0335481343452226,
        "min_feedback_forward_cosine": 0.9982655048370361
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.006789383641332388
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05907309427857399,
        "mean_feedback_forward_cosine": 0.9993698889558965,
        "mean_feedback_forward_relative_error": 0.033544281125068666,
        "min_feedback_forward_cosine": 0.99826580286026
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.006515305521786213
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059067174792289734,
        "mean_feedback_forward_cosine": 0.9993700200861151,
        "mean_feedback_forward_relative_error": 0.0335402156480334,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.006334109747409821
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906711891293526,
        "mean_feedback_forward_cosine": 0.9993700222535566,
        "mean_feedback_forward_relative_error": 0.0335401299494234,
        "min_feedback_forward_cosine": 0.9982661008834839
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.005846326352842152
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906708538532257,
        "mean_feedback_forward_cosine": 0.9993700211698359,
        "mean_feedback_forward_relative_error": 0.03354006827046925,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.0057507517351955174
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0590670220553875,
        "mean_feedback_forward_cosine": 0.999370032007044,
        "mean_feedback_forward_relative_error": 0.033540005033666435,
        "min_feedback_forward_cosine": 0.9982661008834839
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.00579348510928452
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906699597835541,
        "mean_feedback_forward_cosine": 0.9993700417605313,
        "mean_feedback_forward_relative_error": 0.03353994405743751,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.005419869238585234
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906695872545242,
        "mean_feedback_forward_cosine": 0.9993700395930897,
        "mean_feedback_forward_relative_error": 0.03353988913480531,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.004929324226006866
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906691029667854,
        "mean_feedback_forward_cosine": 0.999370032007044,
        "mean_feedback_forward_relative_error": 0.033539826871657914,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.005693616818711162
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906688794493675,
        "mean_feedback_forward_cosine": 0.999370032007044,
        "mean_feedback_forward_relative_error": 0.03353976836766709,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.005635865669995546
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906684696674347,
        "mean_feedback_forward_cosine": 0.9993700385093689,
        "mean_feedback_forward_relative_error": 0.03353971331803636,
        "min_feedback_forward_cosine": 0.9982661008834839
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.004806031485870481
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066832065582275,
        "mean_feedback_forward_cosine": 0.9993700417605313,
        "mean_feedback_forward_relative_error": 0.03353965641422705,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.005307988227382303
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906680226325989,
        "mean_feedback_forward_cosine": 0.9993700471791354,
        "mean_feedback_forward_relative_error": 0.033539599027823316,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.00502156273201108
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066761285066605,
        "mean_feedback_forward_cosine": 0.9993700395930897,
        "mean_feedback_forward_relative_error": 0.033539544689384373,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.004966159328147769
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066712856292725,
        "mean_feedback_forward_cosine": 0.9993700395930897,
        "mean_feedback_forward_relative_error": 0.03353949117220261,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.004718669764697552
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906669422984123,
        "mean_feedback_forward_cosine": 0.9993700504302978,
        "mean_feedback_forward_relative_error": 0.03353944017805836,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.004746594812879339
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066660702228546,
        "mean_feedback_forward_cosine": 0.999370043927973,
        "mean_feedback_forward_relative_error": 0.03353939016603611,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.004119446426257491
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906660109758377,
        "mean_feedback_forward_cosine": 0.9993700591000644,
        "mean_feedback_forward_relative_error": 0.03353933249176903,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.0048942151668667794
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906656011939049,
        "mean_feedback_forward_cosine": 0.9993700601837852,
        "mean_feedback_forward_relative_error": 0.033539283995262605,
        "min_feedback_forward_cosine": 0.998266339302063
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.00413149460606277
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066515415906906,
        "mean_feedback_forward_cosine": 0.9993700482628562,
        "mean_feedback_forward_relative_error": 0.03353923006321896,
        "min_feedback_forward_cosine": 0.9982661008834839
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.004647674621641636
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0590665265917778,
        "mean_feedback_forward_cosine": 0.999370049346577,
        "mean_feedback_forward_relative_error": 0.03353917565704747,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.004325697136372328
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906650051474571,
        "mean_feedback_forward_cosine": 0.9993700482628562,
        "mean_feedback_forward_relative_error": 0.033539126661013476,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.0044166714534163475
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906650051474571,
        "mean_feedback_forward_cosine": 0.9993700580163436,
        "mean_feedback_forward_relative_error": 0.03353908182206479,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.0040626617965474725
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906646326184273,
        "mean_feedback_forward_cosine": 0.9993700374256481,
        "mean_feedback_forward_relative_error": 0.033539028863676566,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.004413442970365286
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906642600893974,
        "mean_feedback_forward_cosine": 0.9993700536814603,
        "mean_feedback_forward_relative_error": 0.033538979520513254,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.0041954827751964335
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906641110777855,
        "mean_feedback_forward_cosine": 0.9993700536814603,
        "mean_feedback_forward_relative_error": 0.03353893079540946,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.004174803726151585
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066370129585266,
        "mean_feedback_forward_cosine": 0.9993700580163436,
        "mean_feedback_forward_relative_error": 0.03353887637230483,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.004461090686693788
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906631425023079,
        "mean_feedback_forward_cosine": 0.9993700580163436,
        "mean_feedback_forward_relative_error": 0.03353881750425154,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.004873574165888131
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066273272037506,
        "mean_feedback_forward_cosine": 0.9993700591000644,
        "mean_feedback_forward_relative_error": 0.033538767661560664,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.004191665274426341
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906626209616661,
        "mean_feedback_forward_cosine": 0.9993700623512268,
        "mean_feedback_forward_relative_error": 0.033538717124611137,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.004248902084603906
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906623601913452,
        "mean_feedback_forward_cosine": 0.9993700525977395,
        "mean_feedback_forward_relative_error": 0.03353866261684082,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.004457712961398065
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066202491521835,
        "mean_feedback_forward_cosine": 0.9993700591000644,
        "mean_feedback_forward_relative_error": 0.033538611978292464,
        "min_feedback_forward_cosine": 0.998266339302063
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.0043210549486428496
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906614288687706,
        "mean_feedback_forward_cosine": 0.9993700623512268,
        "mean_feedback_forward_relative_error": 0.03353856156834147,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.0045159763356298205
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906612053513527,
        "mean_feedback_forward_cosine": 0.9993700699372725,
        "mean_feedback_forward_relative_error": 0.03353851396082477,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.00403653068780899
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059066079556941986,
        "mean_feedback_forward_cosine": 0.9993700656023893,
        "mean_feedback_forward_relative_error": 0.033538462966680525,
        "min_feedback_forward_cosine": 0.998266339302063
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.004224807922951877
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906602367758751,
        "mean_feedback_forward_cosine": 0.9993700677698308,
        "mean_feedback_forward_relative_error": 0.03353841178627177,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.004067436493560672
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906599387526512,
        "mean_feedback_forward_cosine": 0.9993700634349476,
        "mean_feedback_forward_relative_error": 0.033538361875848334,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.004148853787258267
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059065982699394226,
        "mean_feedback_forward_cosine": 0.9993700829419223,
        "mean_feedback_forward_relative_error": 0.03353831536898559,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.004028445797674358
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906595289707184,
        "mean_feedback_forward_cosine": 0.999370073188435,
        "mean_feedback_forward_relative_error": 0.03353826442564076,
        "min_feedback_forward_cosine": 0.9982661604881287
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.00399504036322236
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906594544649124,
        "mean_feedback_forward_cosine": 0.9993700666861101,
        "mean_feedback_forward_relative_error": 0.033538217800246045,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.004098718171343207
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906591936945915,
        "mean_feedback_forward_cosine": 0.9993700829419223,
        "mean_feedback_forward_relative_error": 0.03353816772049124,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.004304251219257712
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906587094068527,
        "mean_feedback_forward_cosine": 0.9993700699372725,
        "mean_feedback_forward_relative_error": 0.033538121289827603,
        "min_feedback_forward_cosine": 0.9982661008834839
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.004008610866479576
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906582996249199,
        "mean_feedback_forward_cosine": 0.9993700861930848,
        "mean_feedback_forward_relative_error": 0.0335380723530596,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.004255639990828931
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0590657964348793,
        "mean_feedback_forward_cosine": 0.9993700645186684,
        "mean_feedback_forward_relative_error": 0.03353802318769422,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.004154543425813317
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906577780842781,
        "mean_feedback_forward_cosine": 0.9993700721047142,
        "mean_feedback_forward_relative_error": 0.0335379774343561,
        "min_feedback_forward_cosine": 0.9982663989067078
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.0037911088526621462
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906573683023453,
        "mean_feedback_forward_cosine": 0.9993700840256431,
        "mean_feedback_forward_relative_error": 0.03353792656721039,
        "min_feedback_forward_cosine": 0.998266339302063
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.003857225678190589
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906572565436363,
        "mean_feedback_forward_cosine": 0.9993700796907599,
        "mean_feedback_forward_relative_error": 0.03353787823156877,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.003955270027332008
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906571447849274,
        "mean_feedback_forward_cosine": 0.9993700710209933,
        "mean_feedback_forward_relative_error": 0.03353783100504767,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.004000426731593907
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906567722558975,
        "mean_feedback_forward_cosine": 0.999370078607039,
        "mean_feedback_forward_relative_error": 0.033537784447385505,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.003992643709369003
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05906563252210617,
        "mean_feedback_forward_cosine": 0.9993700688535517,
        "mean_feedback_forward_relative_error": 0.033537734460762955,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.004213414700143039
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059065595269203186,
        "mean_feedback_forward_cosine": 0.9993700775233182,
        "mean_feedback_forward_relative_error": 0.0335376863791184,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.0037768240635842085
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0590655654668808,
        "mean_feedback_forward_cosine": 0.9993700807744806,
        "mean_feedback_forward_relative_error": 0.03353764149783687,
        "min_feedback_forward_cosine": 0.9982662200927734
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.0033817386390268805
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.059065550565719604,
        "mean_feedback_forward_cosine": 0.9993700861930848,
        "mean_feedback_forward_relative_error": 0.03353758930143985,
        "min_feedback_forward_cosine": 0.9982662796974182
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.004372435570731759
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9212,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.4281075653076172,
    "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": 14,
    "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": 2.059422492980957,
    "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": 18256.374680042267,
    "train_wall_s": 18250.738517045975
  },
  "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
  }
}