{
  "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": 13,
    "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_s13.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": 13,
    "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": 0.9995148744848039,
    "feedback_forward_cosine": [
      0.9997456073760986,
      0.9997330904006958,
      0.9995865821838379,
      0.9995840787887573,
      0.9996457099914551,
      0.9996064901351929,
      0.9997146129608154,
      0.9996328353881836,
      0.999509334564209,
      0.9994516968727112,
      0.9995710849761963,
      0.9995086193084717,
      0.9996844530105591,
      0.9996521472930908,
      0.9997467994689941,
      0.9996289014816284,
      0.9997332692146301,
      0.9996172785758972,
      0.9996885657310486,
      0.9997808337211609,
      0.9997159242630005,
      0.9996783137321472,
      0.999729335308075,
      0.999666154384613,
      0.9997511506080627,
      0.999657154083252,
      0.9996792078018188,
      0.9995862245559692,
      0.9996955394744873,
      0.9995726346969604,
      0.9995874166488647,
      0.9994576573371887,
      0.9996607303619385,
      0.9995304346084595,
      0.9995937943458557,
      0.9994197487831116,
      0.9997514486312866,
      0.9998229742050171,
      0.9997981190681458,
      0.9997464418411255,
      0.9997518062591553,
      0.9996287226676941,
      0.9996957778930664,
      0.9995288848876953,
      0.9996234774589539,
      0.999355673789978,
      0.9994724988937378,
      0.9991140961647034,
      0.9988570213317871,
      0.9983136057853699,
      0.9983664751052856,
      0.9977402687072754,
      0.9987434148788452,
      0.9979894161224365,
      0.9999787211418152
    ],
    "feedback_forward_norm_ratio": [
      0.9988730549812317,
      1.0006078481674194,
      0.9993394613265991,
      1.001127004623413,
      0.9990338683128357,
      1.0004154443740845,
      0.9995367527008057,
      1.0011768341064453,
      0.9985411167144775,
      1.0020065307617188,
      0.9981188178062439,
      1.0027170181274414,
      0.9990379214286804,
      1.0009950399398804,
      0.9998155832290649,
      1.0005804300308228,
      0.9991078972816467,
      1.0011104345321655,
      0.9990886449813843,
      1.0007953643798828,
      0.9995934963226318,
      1.000977635383606,
      0.9990978837013245,
      1.0008434057235718,
      0.9994692802429199,
      1.0015718936920166,
      0.9989431500434875,
      1.0015720129013062,
      0.9993394017219543,
      1.001583456993103,
      0.9989436864852905,
      1.0021345615386963,
      0.9989631175994873,
      1.001844048500061,
      0.9989639520645142,
      1.003008484840393,
      0.9993592500686646,
      1.0002747774124146,
      0.9998732805252075,
      1.00054931640625,
      0.9995817542076111,
      1.0010743141174316,
      0.999226450920105,
      1.0015519857406616,
      0.9992843270301819,
      1.0018341541290283,
      0.998745322227478,
      1.0020575523376465,
      0.9983550906181335,
      1.0026588439941406,
      0.9973151087760925,
      1.0030194520950317,
      0.9985994100570679,
      1.0021740198135376,
      1.001080870628357
    ],
    "feedback_forward_relative_error": [
      0.022569570690393448,
      0.023116396740078926,
      0.028753137215971947,
      0.02887820266187191,
      0.02662196196615696,
      0.028064891695976257,
      0.023886168375611305,
      0.027140865102410316,
      0.03134024888277054,
      0.033206596970558167,
      0.02931816317141056,
      0.031509820371866226,
      0.025130072608590126,
      0.026406964287161827,
      0.02250189147889614,
      0.02725551277399063,
      0.02310079149901867,
      0.027704205363988876,
      0.024963723495602608,
      0.02095712162554264,
      0.023833798244595528,
      0.025395900011062622,
      0.02327672205865383,
      0.0258642490953207,
      0.022308772429823875,
      0.02625497803092003,
      0.025336265563964844,
      0.028834355995059013,
      0.024675320833921432,
      0.0293009914457798,
      0.028729842975735664,
      0.033037349581718445,
      0.02605678141117096,
      0.030729036778211594,
      0.02850693091750145,
      0.03424990922212601,
      0.022296937182545662,
      0.01881887950003147,
      0.020092979073524475,
      0.02253291755914688,
      0.022278066724538803,
      0.027286378666758537,
      0.024672068655490875,
      0.03076157532632351,
      0.027442889288067818,
      0.035978950560092926,
      0.03248658403754234,
      0.04218292236328125,
      0.04779919609427452,
      0.058212172240018845,
      0.057143598794937134,
      0.06739622354507446,
      0.05011552944779396,
      0.06351963430643082,
      0.006617867853492498
    ],
    "innovation_negative_gradient_cosine": [
      0.9995577931404114,
      0.9995124340057373,
      0.9995529055595398,
      0.9994076490402222,
      0.999546468257904,
      0.9993767738342285,
      0.9995553493499756,
      0.9994248151779175,
      0.9995616674423218,
      0.9994311332702637,
      0.9995665550231934,
      0.9993959665298462,
      0.9995825886726379,
      0.9995402097702026,
      0.9995884895324707,
      0.9995632767677307,
      0.9995993375778198,
      0.9995043277740479,
      0.999602735042572,
      0.999591588973999,
      0.9996074438095093,
      0.9995112419128418,
      0.9996109008789062,
      0.9995533227920532,
      0.9996251463890076,
      0.9995191693305969,
      0.9996445178985596,
      0.9995243549346924,
      0.9996548891067505,
      0.9995096921920776,
      0.999665379524231,
      0.9994922876358032,
      0.9996774792671204,
      0.9995546340942383,
      0.9996901154518127,
      0.9994900226593018,
      0.9996975660324097,
      0.9997286200523376,
      0.9997000694274902,
      0.9996534585952759,
      0.9996994137763977,
      0.9996539354324341,
      0.9997164011001587,
      0.9996889233589172,
      0.9997352957725525,
      0.9996839761734009,
      0.9997507333755493,
      0.9997549057006836,
      0.9996919631958008,
      0.9995550513267517,
      0.9996441602706909,
      0.999585747718811,
      0.9998013973236084,
      0.9998618364334106,
      0.7703130841255188
    ],
    "mean_feedback_forward_cosine": 0.9994924046776511,
    "mean_feedback_forward_relative_error": 0.0299355069598691,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.9995577931404114,
      0.9995124340057373,
      0.9995529055595398,
      0.9994076490402222,
      0.999546468257904,
      0.9993767738342285,
      0.9995553493499756,
      0.9994248151779175,
      0.9995616674423218,
      0.9994311332702637,
      0.9995665550231934,
      0.9993959665298462,
      0.9995825886726379,
      0.9995402097702026,
      0.9995884895324707,
      0.9995632767677307,
      0.9995993375778198,
      0.9995043277740479,
      0.999602735042572,
      0.999591588973999,
      0.9996074438095093,
      0.9995112419128418,
      0.9996109008789062,
      0.9995533227920532,
      0.9996251463890076,
      0.9995191693305969,
      0.9996445178985596,
      0.9995243549346924,
      0.9996548891067505,
      0.9995096921920776,
      0.999665379524231,
      0.9994922876358032,
      0.9996774792671204,
      0.9995546340942383,
      0.9996901154518127,
      0.9994900226593018,
      0.9996975660324097,
      0.9997286200523376,
      0.9997000694274902,
      0.9996534585952759,
      0.9996994137763977,
      0.9996539354324341,
      0.9997164011001587,
      0.9996889233589172,
      0.9997352957725525,
      0.9996839761734009,
      0.9997507333755493,
      0.9997549057006836,
      0.9996919631958008,
      0.9995550513267517,
      0.9996441602706909,
      0.999585747718811,
      0.9998013973236084,
      0.9998618364334106,
      0.7703130841255188
    ],
    "teaching_negative_gradient_cosine": [
      0.9995577931404114,
      0.9995124340057373,
      0.9995529055595398,
      0.9994076490402222,
      0.999546468257904,
      0.9993767738342285,
      0.9995553493499756,
      0.9994248151779175,
      0.9995616674423218,
      0.9994311332702637,
      0.9995665550231934,
      0.9993959665298462,
      0.9995825886726379,
      0.9995402097702026,
      0.9995884895324707,
      0.9995632767677307,
      0.9995993375778198,
      0.9995043277740479,
      0.999602735042572,
      0.999591588973999,
      0.9996074438095093,
      0.9995112419128418,
      0.9996109008789062,
      0.9995533227920532,
      0.9996251463890076,
      0.9995191693305969,
      0.9996445178985596,
      0.9995243549346924,
      0.9996548891067505,
      0.9995096921920776,
      0.999665379524231,
      0.9994922876358032,
      0.9996774792671204,
      0.9995546340942383,
      0.9996901154518127,
      0.9994900226593018,
      0.9996975660324097,
      0.9997286200523376,
      0.9997000694274902,
      0.9996534585952759,
      0.9996994137763977,
      0.9996539354324341,
      0.9997164011001587,
      0.9996889233589172,
      0.9997352957725525,
      0.9996839761734009,
      0.9997507333755493,
      0.9997549057006836,
      0.9996919631958008,
      0.9995550513267517,
      0.9996441602706909,
      0.999585747718811,
      0.9998013973236084,
      0.9998618364334106,
      0.7703130841255188
    ],
    "wall_s": 0.1903843879699707
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3721873760223389,
        "mean_feedback_forward_cosine": 0.22906434779817408,
        "mean_feedback_forward_relative_error": 1.23561264926737,
        "min_feedback_forward_cosine": 0.06822853535413742
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 2.618657088623047
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3618451356887817,
        "mean_feedback_forward_cosine": 0.2514899350025437,
        "mean_feedback_forward_relative_error": 1.216604634848508,
        "min_feedback_forward_cosine": 0.0825369656085968
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 2.0443328186035155
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3392257690429688,
        "mean_feedback_forward_cosine": 0.27437169619581914,
        "mean_feedback_forward_relative_error": 1.1974607321349058,
        "min_feedback_forward_cosine": 0.1043594554066658
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.8029051454925538
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3336762189865112,
        "mean_feedback_forward_cosine": 0.3102063487876545,
        "mean_feedback_forward_relative_error": 1.1679667462002148,
        "min_feedback_forward_cosine": 0.11178690195083618
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 1.5733120203781128
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3269165754318237,
        "mean_feedback_forward_cosine": 0.3568539035591212,
        "mean_feedback_forward_relative_error": 1.1278487942435524,
        "min_feedback_forward_cosine": 0.11940619349479675
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 1.3117623218536376
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3227629661560059,
        "mean_feedback_forward_cosine": 0.4043669891628352,
        "mean_feedback_forward_relative_error": 1.085144471580332,
        "min_feedback_forward_cosine": 0.12460734695196152
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 1.0928271724700929
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3171675205230713,
        "mean_feedback_forward_cosine": 0.4497559509494088,
        "mean_feedback_forward_relative_error": 1.0418285944245078,
        "min_feedback_forward_cosine": 0.13213735818862915
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.9260222872543334
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3109973669052124,
        "mean_feedback_forward_cosine": 0.4906862039457668,
        "mean_feedback_forward_relative_error": 1.0011875304308804,
        "min_feedback_forward_cosine": 0.14014112949371338
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.8086126829910278
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3044464588165283,
        "mean_feedback_forward_cosine": 0.5249264020811427,
        "mean_feedback_forward_relative_error": 0.9653921685435555,
        "min_feedback_forward_cosine": 0.14885416626930237
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.7224795008468627
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2969162464141846,
        "mean_feedback_forward_cosine": 0.555767253854058,
        "mean_feedback_forward_relative_error": 0.9317991077899933,
        "min_feedback_forward_cosine": 0.1582750529050827
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.6612757705783844
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2884743213653564,
        "mean_feedback_forward_cosine": 0.5836254799907857,
        "mean_feedback_forward_relative_error": 0.9001913368701935,
        "min_feedback_forward_cosine": 0.1689441055059433
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.6117791198348999
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2787386178970337,
        "mean_feedback_forward_cosine": 0.609176770936359,
        "mean_feedback_forward_relative_error": 0.8700118612159382,
        "min_feedback_forward_cosine": 0.18143343925476074
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.574167943572998
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.268519401550293,
        "mean_feedback_forward_cosine": 0.6319928404959765,
        "mean_feedback_forward_relative_error": 0.8422089395197955,
        "min_feedback_forward_cosine": 0.19407938420772552
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.5307482202529907
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2560867071151733,
        "mean_feedback_forward_cosine": 0.6534666595133868,
        "mean_feedback_forward_relative_error": 0.8151899307966233,
        "min_feedback_forward_cosine": 0.20934352278709412
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.5125123575782776
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2439846992492676,
        "mean_feedback_forward_cosine": 0.6732584809715098,
        "mean_feedback_forward_relative_error": 0.7894332414323634,
        "min_feedback_forward_cosine": 0.22344081103801727
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.4776908352661133
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2297369241714478,
        "mean_feedback_forward_cosine": 0.6926673897288063,
        "mean_feedback_forward_relative_error": 0.7637290185148066,
        "min_feedback_forward_cosine": 0.2404581755399704
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.46589656856536865
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2135229110717773,
        "mean_feedback_forward_cosine": 0.7107760201800953,
        "mean_feedback_forward_relative_error": 0.7388164438984611,
        "min_feedback_forward_cosine": 0.2596164643764496
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.44467153213500976
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.197867512702942,
        "mean_feedback_forward_cosine": 0.7274938036095012,
        "mean_feedback_forward_relative_error": 0.7151795102791353,
        "min_feedback_forward_cosine": 0.27816063165664673
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.42247454542160034
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1803089380264282,
        "mean_feedback_forward_cosine": 0.7433888825503263,
        "mean_feedback_forward_relative_error": 0.6923050625757737,
        "min_feedback_forward_cosine": 0.2987276315689087
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.40903840388298035
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.164432168006897,
        "mean_feedback_forward_cosine": 0.7582452020861886,
        "mean_feedback_forward_relative_error": 0.6701002207669344,
        "min_feedback_forward_cosine": 0.31991180777549744
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.3893443667793274
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1477042436599731,
        "mean_feedback_forward_cosine": 0.7729029227386821,
        "mean_feedback_forward_relative_error": 0.6478099530393427,
        "min_feedback_forward_cosine": 0.34410470724105835
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.37688736102104187
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1307538747787476,
        "mean_feedback_forward_cosine": 0.7863774174993688,
        "mean_feedback_forward_relative_error": 0.6267310367389158,
        "min_feedback_forward_cosine": 0.36850911378860474
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.36979484375
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1123244762420654,
        "mean_feedback_forward_cosine": 0.7996666431427002,
        "mean_feedback_forward_relative_error": 0.6054776159199802,
        "min_feedback_forward_cosine": 0.39363932609558105
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.3512759153175354
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0937360525131226,
        "mean_feedback_forward_cosine": 0.8120680473067544,
        "mean_feedback_forward_relative_error": 0.584982961958105,
        "min_feedback_forward_cosine": 0.41800665855407715
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.3465504659366608
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0731472969055176,
        "mean_feedback_forward_cosine": 0.823938865553249,
        "mean_feedback_forward_relative_error": 0.5647742089900103,
        "min_feedback_forward_cosine": 0.44163089990615845
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.3362824455356598
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.054181694984436,
        "mean_feedback_forward_cosine": 0.8349253486503254,
        "mean_feedback_forward_relative_error": 0.5455743025649678,
        "min_feedback_forward_cosine": 0.46197107434272766
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.3249240168380737
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0323097705841064,
        "mean_feedback_forward_cosine": 0.8456882639364762,
        "mean_feedback_forward_relative_error": 0.5262904961000789,
        "min_feedback_forward_cosine": 0.4847239851951599
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.31498544347763063
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0116260051727295,
        "mean_feedback_forward_cosine": 0.8558149587024342,
        "mean_feedback_forward_relative_error": 0.507576184110208,
        "min_feedback_forward_cosine": 0.5057708024978638
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.3097911694049835
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9895654916763306,
        "mean_feedback_forward_cosine": 0.8650211746042425,
        "mean_feedback_forward_relative_error": 0.49017953926866703,
        "min_feedback_forward_cosine": 0.5277377367019653
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.29969036979675295
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9655799269676208,
        "mean_feedback_forward_cosine": 0.8739804907278581,
        "mean_feedback_forward_relative_error": 0.47264781160788105,
        "min_feedback_forward_cosine": 0.5506198406219482
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.29819255435943604
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9415902495384216,
        "mean_feedback_forward_cosine": 0.8824259974739769,
        "mean_feedback_forward_relative_error": 0.45579604953527453,
        "min_feedback_forward_cosine": 0.5728276371955872
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.2815729883956909
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9179940223693848,
        "mean_feedback_forward_cosine": 0.8905050201849504,
        "mean_feedback_forward_relative_error": 0.43920987912199716,
        "min_feedback_forward_cosine": 0.5942811369895935
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.28759058656692504
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.894770085811615,
        "mean_feedback_forward_cosine": 0.8977142171426253,
        "mean_feedback_forward_relative_error": 0.423809283294461,
        "min_feedback_forward_cosine": 0.6146277785301208
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.2714368555927277
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8704290390014648,
        "mean_feedback_forward_cosine": 0.9047920725562356,
        "mean_feedback_forward_relative_error": 0.40825533162463795,
        "min_feedback_forward_cosine": 0.6355072259902954
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.2704858200263977
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8463582992553711,
        "mean_feedback_forward_cosine": 0.9114495082335039,
        "mean_feedback_forward_relative_error": 0.393126321922649,
        "min_feedback_forward_cosine": 0.6554253101348877
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.2692690270805359
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8230829834938049,
        "mean_feedback_forward_cosine": 0.9175222591920332,
        "mean_feedback_forward_relative_error": 0.3787562095306136,
        "min_feedback_forward_cosine": 0.67411869764328
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.25900745127677915
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7997261881828308,
        "mean_feedback_forward_cosine": 0.9232426578348333,
        "mean_feedback_forward_relative_error": 0.36491667099974373,
        "min_feedback_forward_cosine": 0.6921552419662476
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.2505539571809769
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7771807909011841,
        "mean_feedback_forward_cosine": 0.928544446555051,
        "mean_feedback_forward_relative_error": 0.35170431380922146,
        "min_feedback_forward_cosine": 0.709445059299469
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.2463171188735962
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7533586621284485,
        "mean_feedback_forward_cosine": 0.933710492740978,
        "mean_feedback_forward_relative_error": 0.3383773974396966,
        "min_feedback_forward_cosine": 0.726951539516449
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.2437849562740326
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7309722304344177,
        "mean_feedback_forward_cosine": 0.9382276155731895,
        "mean_feedback_forward_relative_error": 0.3263433360240676,
        "min_feedback_forward_cosine": 0.7429503798484802
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.23817058839797975
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7088060975074768,
        "mean_feedback_forward_cosine": 0.9425449761477384,
        "mean_feedback_forward_relative_error": 0.31445055292411284,
        "min_feedback_forward_cosine": 0.7582069635391235
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.23429553953170776
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6824594140052795,
        "mean_feedback_forward_cosine": 0.9470177336172624,
        "mean_feedback_forward_relative_error": 0.3017793464389714,
        "min_feedback_forward_cosine": 0.7757521867752075
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.23840826580524443
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6601223349571228,
        "mean_feedback_forward_cosine": 0.9508190794424577,
        "mean_feedback_forward_relative_error": 0.2904875550757755,
        "min_feedback_forward_cosine": 0.7900207042694092
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.22959980630874635
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6384415030479431,
        "mean_feedback_forward_cosine": 0.9544151338663969,
        "mean_feedback_forward_relative_error": 0.27948091301051053,
        "min_feedback_forward_cosine": 0.8036364316940308
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.23084515382766724
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6156907677650452,
        "mean_feedback_forward_cosine": 0.9577961065552452,
        "mean_feedback_forward_relative_error": 0.26881950958208606,
        "min_feedback_forward_cosine": 0.8170660138130188
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.22511683757781983
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5947313904762268,
        "mean_feedback_forward_cosine": 0.9608323324810375,
        "mean_feedback_forward_relative_error": 0.25886820771477437,
        "min_feedback_forward_cosine": 0.8291815519332886
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.21716432525157928
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5736299753189087,
        "mean_feedback_forward_cosine": 0.9637927716428584,
        "mean_feedback_forward_relative_error": 0.24875448203899644,
        "min_feedback_forward_cosine": 0.8410825133323669
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.2182045086479187
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5530214905738831,
        "mean_feedback_forward_cosine": 0.9664847081357782,
        "mean_feedback_forward_relative_error": 0.23927537900480358,
        "min_feedback_forward_cosine": 0.8522039651870728
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.21188023831367492
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5337027907371521,
        "mean_feedback_forward_cosine": 0.968951736796986,
        "mean_feedback_forward_relative_error": 0.2302019331942905,
        "min_feedback_forward_cosine": 0.8622238636016846
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.20758117275238036
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5153356790542603,
        "mean_feedback_forward_cosine": 0.9713034532286904,
        "mean_feedback_forward_relative_error": 0.22121331122788515,
        "min_feedback_forward_cosine": 0.8714470863342285
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.2128644209384918
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4960263669490814,
        "mean_feedback_forward_cosine": 0.9734993869608098,
        "mean_feedback_forward_relative_error": 0.21256887032227081,
        "min_feedback_forward_cosine": 0.8808215260505676
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.20747516375541686
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4773900806903839,
        "mean_feedback_forward_cosine": 0.9755356550216675,
        "mean_feedback_forward_relative_error": 0.2042575067417188,
        "min_feedback_forward_cosine": 0.8895806074142456
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.20645665825366974
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4604026973247528,
        "mean_feedback_forward_cosine": 0.9773342045870694,
        "mean_feedback_forward_relative_error": 0.19654388956048272,
        "min_feedback_forward_cosine": 0.8970670700073242
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.20073466979026794
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4430951476097107,
        "mean_feedback_forward_cosine": 0.9790165261788801,
        "mean_feedback_forward_relative_error": 0.1891219674186273,
        "min_feedback_forward_cosine": 0.904595136642456
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.19337763548851014
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.42659634351730347,
        "mean_feedback_forward_cosine": 0.9805894006382335,
        "mean_feedback_forward_relative_error": 0.18193596174771134,
        "min_feedback_forward_cosine": 0.9115451574325562
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.193649558801651
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4102657735347748,
        "mean_feedback_forward_cosine": 0.9820752284743569,
        "mean_feedback_forward_relative_error": 0.1748588041825728,
        "min_feedback_forward_cosine": 0.91810142993927
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.1931214778614044
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3945562541484833,
        "mean_feedback_forward_cosine": 0.9834433187137951,
        "mean_feedback_forward_relative_error": 0.16806023865938186,
        "min_feedback_forward_cosine": 0.9241763353347778
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.19203194622039796
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.380015105009079,
        "mean_feedback_forward_cosine": 0.9846876285292886,
        "mean_feedback_forward_relative_error": 0.16161761378700082,
        "min_feedback_forward_cosine": 0.9296014904975891
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.18920780365467071
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.36605533957481384,
        "mean_feedback_forward_cosine": 0.9858385129408402,
        "mean_feedback_forward_relative_error": 0.1554398418150165,
        "min_feedback_forward_cosine": 0.9346195459365845
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.1875186634349823
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.35228097438812256,
        "mean_feedback_forward_cosine": 0.9869261893359098,
        "mean_feedback_forward_relative_error": 0.1493348742750558,
        "min_feedback_forward_cosine": 0.939432680606842
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.18835279341697692
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.33810895681381226,
        "mean_feedback_forward_cosine": 0.9879302880980751,
        "mean_feedback_forward_relative_error": 0.14352331920103592,
        "min_feedback_forward_cosine": 0.9441512823104858
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.18333342970848082
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3247024714946747,
        "mean_feedback_forward_cosine": 0.9888717066157948,
        "mean_feedback_forward_relative_error": 0.1378449701450088,
        "min_feedback_forward_cosine": 0.9484421014785767
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.1856263377380371
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.31059858202934265,
        "mean_feedback_forward_cosine": 0.9897600022229281,
        "mean_feedback_forward_relative_error": 0.13233231245116753,
        "min_feedback_forward_cosine": 0.9527953267097473
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.18473754248142243
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2986745834350586,
        "mean_feedback_forward_cosine": 0.9905232527039268,
        "mean_feedback_forward_relative_error": 0.1273461725224148,
        "min_feedback_forward_cosine": 0.9563075304031372
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.17438865018367766
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2868397831916809,
        "mean_feedback_forward_cosine": 0.9912605014714327,
        "mean_feedback_forward_relative_error": 0.12232954075390642,
        "min_feedback_forward_cosine": 0.9596667289733887
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.180476394944191
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.27480974793434143,
        "mean_feedback_forward_cosine": 0.9919500871138139,
        "mean_feedback_forward_relative_error": 0.11744141053747047,
        "min_feedback_forward_cosine": 0.9629510641098022
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.18252969101667404
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.26383450627326965,
        "mean_feedback_forward_cosine": 0.9925798947160894,
        "mean_feedback_forward_relative_error": 0.11280005262656645,
        "min_feedback_forward_cosine": 0.9658384323120117
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.17417837971687317
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.25343167781829834,
        "mean_feedback_forward_cosine": 0.9931712020527232,
        "mean_feedback_forward_relative_error": 0.10825635316697034,
        "min_feedback_forward_cosine": 0.9684686064720154
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.18076635927200319
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.24370114505290985,
        "mean_feedback_forward_cosine": 0.9936928879130971,
        "mean_feedback_forward_relative_error": 0.10406356636773456,
        "min_feedback_forward_cosine": 0.9708285927772522
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.17136417557716369
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2343776673078537,
        "mean_feedback_forward_cosine": 0.9941819310188293,
        "mean_feedback_forward_relative_error": 0.09996809884905815,
        "min_feedback_forward_cosine": 0.9730112552642822
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.17325837555885315
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22468312084674835,
        "mean_feedback_forward_cosine": 0.9946417169137435,
        "mean_feedback_forward_relative_error": 0.09599588062952866,
        "min_feedback_forward_cosine": 0.9751811027526855
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.17303499470472336
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2157231569290161,
        "mean_feedback_forward_cosine": 0.9950508767908269,
        "mean_feedback_forward_relative_error": 0.09230448867787014,
        "min_feedback_forward_cosine": 0.977104663848877
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.16835947505950927
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20719926059246063,
        "mean_feedback_forward_cosine": 0.9954314134337685,
        "mean_feedback_forward_relative_error": 0.0887195079841397,
        "min_feedback_forward_cosine": 0.978866696357727
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.16926348195552826
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19898490607738495,
        "mean_feedback_forward_cosine": 0.995786114172502,
        "mean_feedback_forward_relative_error": 0.08523530509661545,
        "min_feedback_forward_cosine": 0.9804971218109131
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.16732301572322844
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1906488537788391,
        "mean_feedback_forward_cosine": 0.9961130423979325,
        "mean_feedback_forward_relative_error": 0.0818956772712144,
        "min_feedback_forward_cosine": 0.9820825457572937
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.1667278288602829
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18334421515464783,
        "mean_feedback_forward_cosine": 0.9964060230688615,
        "mean_feedback_forward_relative_error": 0.07877858199856498,
        "min_feedback_forward_cosine": 0.9834163188934326
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.16291458018779756
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17592990398406982,
        "mean_feedback_forward_cosine": 0.9966853965412487,
        "mean_feedback_forward_relative_error": 0.07569103329019113,
        "min_feedback_forward_cosine": 0.9847314357757568
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.1661173464488983
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16859185695648193,
        "mean_feedback_forward_cosine": 0.9969464638016441,
        "mean_feedback_forward_relative_error": 0.07268435887315057,
        "min_feedback_forward_cosine": 0.9859713315963745
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.1632753385400772
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16158096492290497,
        "mean_feedback_forward_cosine": 0.9971798214045438,
        "mean_feedback_forward_relative_error": 0.06990636316212741,
        "min_feedback_forward_cosine": 0.9871110916137695
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.15748083257198334
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15530893206596375,
        "mean_feedback_forward_cosine": 0.9974013209342957,
        "mean_feedback_forward_relative_error": 0.06712234782224352,
        "min_feedback_forward_cosine": 0.9880821704864502
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.1633152895450592
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14913444221019745,
        "mean_feedback_forward_cosine": 0.9976030414754694,
        "mean_feedback_forward_relative_error": 0.06448946229436181,
        "min_feedback_forward_cosine": 0.9890108108520508
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.16058050225019455
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14306572079658508,
        "mean_feedback_forward_cosine": 0.9977879004044966,
        "mean_feedback_forward_relative_error": 0.06198189661584117,
        "min_feedback_forward_cosine": 0.9898815155029297
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.1555505397939682
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13722650706768036,
        "mean_feedback_forward_cosine": 0.9979615753347223,
        "mean_feedback_forward_relative_error": 0.05952306567606601,
        "min_feedback_forward_cosine": 0.990685760974884
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.15881990154266357
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13209199905395508,
        "mean_feedback_forward_cosine": 0.9981146498159928,
        "mean_feedback_forward_relative_error": 0.0572621196677739,
        "min_feedback_forward_cosine": 0.9913676381111145
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.1541928218650818
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1266280859708786,
        "mean_feedback_forward_cosine": 0.9982635487209667,
        "mean_feedback_forward_relative_error": 0.054980828579176556,
        "min_feedback_forward_cosine": 0.9920614957809448
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.15678585725784303
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12115006148815155,
        "mean_feedback_forward_cosine": 0.9984057946638627,
        "mean_feedback_forward_relative_error": 0.05271399278532375,
        "min_feedback_forward_cosine": 0.9927307367324829
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.1581881642627716
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11628136783838272,
        "mean_feedback_forward_cosine": 0.9985274737531489,
        "mean_feedback_forward_relative_error": 0.05068704489279877,
        "min_feedback_forward_cosine": 0.9933011531829834
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.15509069224357605
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1117628961801529,
        "mean_feedback_forward_cosine": 0.9986399585550482,
        "mean_feedback_forward_relative_error": 0.04872960036789829,
        "min_feedback_forward_cosine": 0.9938088655471802
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.15393281336784362
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10720320045948029,
        "mean_feedback_forward_cosine": 0.998745450106534,
        "mean_feedback_forward_relative_error": 0.04682655361565677,
        "min_feedback_forward_cosine": 0.9943026900291443
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.1532562497806549
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10307206958532333,
        "mean_feedback_forward_cosine": 0.9988405346870423,
        "mean_feedback_forward_relative_error": 0.04503027690066533,
        "min_feedback_forward_cosine": 0.9947317838668823
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.15164671294689178
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0990542471408844,
        "mean_feedback_forward_cosine": 0.9989296945658597,
        "mean_feedback_forward_relative_error": 0.043277259001677684,
        "min_feedback_forward_cosine": 0.9951339960098267
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.14897760011672972
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09500788897275925,
        "mean_feedback_forward_cosine": 0.9990142898126082,
        "mean_feedback_forward_relative_error": 0.04155274192717942,
        "min_feedback_forward_cosine": 0.995521605014801
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.15326515437602997
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09107612818479538,
        "mean_feedback_forward_cosine": 0.9990905924276872,
        "mean_feedback_forward_relative_error": 0.03993578402494842,
        "min_feedback_forward_cosine": 0.9958833456039429
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.14831679260253905
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0873541310429573,
        "mean_feedback_forward_cosine": 0.9991615978154269,
        "mean_feedback_forward_relative_error": 0.03835834117436951,
        "min_feedback_forward_cosine": 0.9962108731269836
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.15207230830192567
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08383745700120926,
        "mean_feedback_forward_cosine": 0.9992264563387091,
        "mean_feedback_forward_relative_error": 0.036856697042557326,
        "min_feedback_forward_cosine": 0.9965096712112427
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.15156425470352172
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08054554462432861,
        "mean_feedback_forward_cosine": 0.9992848027836193,
        "mean_feedback_forward_relative_error": 0.03545508719980717,
        "min_feedback_forward_cosine": 0.9967774748802185
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.14365550803661348
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0773683413863182,
        "mean_feedback_forward_cosine": 0.9993400530381636,
        "mean_feedback_forward_relative_error": 0.034076379087160935,
        "min_feedback_forward_cosine": 0.997025728225708
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.14493907814264298
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07412941008806229,
        "mean_feedback_forward_cosine": 0.9993923219767484,
        "mean_feedback_forward_relative_error": 0.03272376771677624,
        "min_feedback_forward_cosine": 0.9972688555717468
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.1477852155828476
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07104866951704025,
        "mean_feedback_forward_cosine": 0.999440254948356,
        "mean_feedback_forward_relative_error": 0.031422443501651284,
        "min_feedback_forward_cosine": 0.997490644454956
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.14734023728847503
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06826168298721313,
        "mean_feedback_forward_cosine": 0.999483335018158,
        "mean_feedback_forward_relative_error": 0.030201662348752673,
        "min_feedback_forward_cosine": 0.9976828694343567
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.146364845328331
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0682401955127716,
        "mean_feedback_forward_cosine": 0.9994836807250976,
        "mean_feedback_forward_relative_error": 0.030190771013836968,
        "min_feedback_forward_cosine": 0.9976843595504761
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.0802191784787178
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0682283490896225,
        "mean_feedback_forward_cosine": 0.9994838562878695,
        "mean_feedback_forward_relative_error": 0.0301847822540863,
        "min_feedback_forward_cosine": 0.9976852536201477
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.05366209157705307
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06821569800376892,
        "mean_feedback_forward_cosine": 0.9994840231808749,
        "mean_feedback_forward_relative_error": 0.030179313011467458,
        "min_feedback_forward_cosine": 0.9976860880851746
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.043359219493865966
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06820192188024521,
        "mean_feedback_forward_cosine": 0.9994842019948093,
        "mean_feedback_forward_relative_error": 0.03017386038025672,
        "min_feedback_forward_cosine": 0.9976872205734253
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.03950208593964577
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06818819791078568,
        "mean_feedback_forward_cosine": 0.9994843537157232,
        "mean_feedback_forward_relative_error": 0.030168596663596954,
        "min_feedback_forward_cosine": 0.9976880550384521
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.034691714227199556
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06817445158958435,
        "mean_feedback_forward_cosine": 0.9994845184412869,
        "mean_feedback_forward_relative_error": 0.030163414945656605,
        "min_feedback_forward_cosine": 0.997688889503479
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.030728123164772986
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06816031783819199,
        "mean_feedback_forward_cosine": 0.9994846723296426,
        "mean_feedback_forward_relative_error": 0.030158490137281745,
        "min_feedback_forward_cosine": 0.9976897835731506
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.02796993443250656
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06814504414796829,
        "mean_feedback_forward_cosine": 0.9994848413900895,
        "mean_feedback_forward_relative_error": 0.030153311797502365,
        "min_feedback_forward_cosine": 0.9976907968521118
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.026594881633855403
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06812984496355057,
        "mean_feedback_forward_cosine": 0.9994850147854198,
        "mean_feedback_forward_relative_error": 0.030148253691467373,
        "min_feedback_forward_cosine": 0.9976917505264282
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.024129524217247962
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0681147351861,
        "mean_feedback_forward_cosine": 0.9994851632551713,
        "mean_feedback_forward_relative_error": 0.030143321855840358,
        "min_feedback_forward_cosine": 0.9976927638053894
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.02287360542178154
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06809647381305695,
        "mean_feedback_forward_cosine": 0.9994853464039889,
        "mean_feedback_forward_relative_error": 0.030138056649064476,
        "min_feedback_forward_cosine": 0.9976940751075745
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.022134124634861946
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06808172911405563,
        "mean_feedback_forward_cosine": 0.9994855057109486,
        "mean_feedback_forward_relative_error": 0.030133291105316443,
        "min_feedback_forward_cosine": 0.9976949095726013
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.019623418084681033
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0680660679936409,
        "mean_feedback_forward_cosine": 0.9994856552644209,
        "mean_feedback_forward_relative_error": 0.030128194112330677,
        "min_feedback_forward_cosine": 0.997696042060852
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.019880822781324387
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06805063039064407,
        "mean_feedback_forward_cosine": 0.9994858264923095,
        "mean_feedback_forward_relative_error": 0.030123510711233724,
        "min_feedback_forward_cosine": 0.997697114944458
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.018001913035809995
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06803379207849503,
        "mean_feedback_forward_cosine": 0.9994859890504317,
        "mean_feedback_forward_relative_error": 0.030118542156097564,
        "min_feedback_forward_cosine": 0.997698187828064
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.016911001521945
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06801775097846985,
        "mean_feedback_forward_cosine": 0.9994861570271578,
        "mean_feedback_forward_relative_error": 0.030113677976822312,
        "min_feedback_forward_cosine": 0.9976992607116699
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.015953069033920766
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06800206005573273,
        "mean_feedback_forward_cosine": 0.9994863217527216,
        "mean_feedback_forward_relative_error": 0.030109035147523337,
        "min_feedback_forward_cosine": 0.9977003931999207
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.01511427465081215
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06798448413610458,
        "mean_feedback_forward_cosine": 0.9994864918968894,
        "mean_feedback_forward_relative_error": 0.03010388956489888,
        "min_feedback_forward_cosine": 0.9977014064788818
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.015267606116458774
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06796781718730927,
        "mean_feedback_forward_cosine": 0.9994866620410573,
        "mean_feedback_forward_relative_error": 0.030099049760875376,
        "min_feedback_forward_cosine": 0.9977025985717773
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.014295138305872679
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0679512470960617,
        "mean_feedback_forward_cosine": 0.9994868137619712,
        "mean_feedback_forward_relative_error": 0.030094342331656,
        "min_feedback_forward_cosine": 0.9977036714553833
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.013587301034033298
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06793487071990967,
        "mean_feedback_forward_cosine": 0.9994869860735807,
        "mean_feedback_forward_relative_error": 0.03008950800550255,
        "min_feedback_forward_cosine": 0.9977046847343445
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.012850923472344875
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0679175928235054,
        "mean_feedback_forward_cosine": 0.9994871638037942,
        "mean_feedback_forward_relative_error": 0.03008446695113724,
        "min_feedback_forward_cosine": 0.9977058172225952
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.012713236034810543
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06789865344762802,
        "mean_feedback_forward_cosine": 0.9994873437014493,
        "mean_feedback_forward_relative_error": 0.030079286494715647,
        "min_feedback_forward_cosine": 0.9977070689201355
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.012308287309929729
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06788161396980286,
        "mean_feedback_forward_cosine": 0.9994875030084089,
        "mean_feedback_forward_relative_error": 0.030074561658230695,
        "min_feedback_forward_cosine": 0.9977082014083862
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.011468153971470892
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06786490976810455,
        "mean_feedback_forward_cosine": 0.9994876883246682,
        "mean_feedback_forward_relative_error": 0.03006964431736957,
        "min_feedback_forward_cosine": 0.9977092742919922
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.011168898286223411
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06784682720899582,
        "mean_feedback_forward_cosine": 0.9994878628037193,
        "mean_feedback_forward_relative_error": 0.030064505321735685,
        "min_feedback_forward_cosine": 0.9977104663848877
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.010675356038808823
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06783173233270645,
        "mean_feedback_forward_cosine": 0.9994880145246332,
        "mean_feedback_forward_relative_error": 0.03006000839681788,
        "min_feedback_forward_cosine": 0.9977114796638489
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.009513134524971247
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06781570613384247,
        "mean_feedback_forward_cosine": 0.9994881749153137,
        "mean_feedback_forward_relative_error": 0.030055251615968618,
        "min_feedback_forward_cosine": 0.9977124929428101
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.009519220564886927
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06779937446117401,
        "mean_feedback_forward_cosine": 0.9994883407245982,
        "mean_feedback_forward_relative_error": 0.030050520903684878,
        "min_feedback_forward_cosine": 0.997713565826416
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.009649479085206986
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0677826851606369,
        "mean_feedback_forward_cosine": 0.9994885217059742,
        "mean_feedback_forward_relative_error": 0.030045833684165368,
        "min_feedback_forward_cosine": 0.9977146983146667
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.008831491970270873
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06776418536901474,
        "mean_feedback_forward_cosine": 0.9994887016036293,
        "mean_feedback_forward_relative_error": 0.03004076196388765,
        "min_feedback_forward_cosine": 0.997715950012207
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.009418278399966657
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06774482131004333,
        "mean_feedback_forward_cosine": 0.9994888923384927,
        "mean_feedback_forward_relative_error": 0.03003532240505923,
        "min_feedback_forward_cosine": 0.9977171421051025
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.009592618333250284
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06772809475660324,
        "mean_feedback_forward_cosine": 0.9994890505617315,
        "mean_feedback_forward_relative_error": 0.030030591946772555,
        "min_feedback_forward_cosine": 0.997718334197998
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.007983197383247317
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06770776212215424,
        "mean_feedback_forward_cosine": 0.9994892597198486,
        "mean_feedback_forward_relative_error": 0.030025187134742736,
        "min_feedback_forward_cosine": 0.9977196455001831
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.009227880974523723
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06768783181905746,
        "mean_feedback_forward_cosine": 0.9994894396175038,
        "mean_feedback_forward_relative_error": 0.030019957292824983,
        "min_feedback_forward_cosine": 0.9977210760116577
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.008448275340795517
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06767150014638901,
        "mean_feedback_forward_cosine": 0.9994896032593467,
        "mean_feedback_forward_relative_error": 0.030015156358819118,
        "min_feedback_forward_cosine": 0.9977221488952637
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.008099065444543958
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06765541434288025,
        "mean_feedback_forward_cosine": 0.9994897788221185,
        "mean_feedback_forward_relative_error": 0.030010465922003442,
        "min_feedback_forward_cosine": 0.9977231025695801
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.007605297755300998
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06763419508934021,
        "mean_feedback_forward_cosine": 0.9994899728081443,
        "mean_feedback_forward_relative_error": 0.030004414077848197,
        "min_feedback_forward_cosine": 0.9977245330810547
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.008371457745283842
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06761518865823746,
        "mean_feedback_forward_cosine": 0.9994901700453325,
        "mean_feedback_forward_relative_error": 0.029999074331400068,
        "min_feedback_forward_cosine": 0.9977258443832397
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.007859050422012806
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06759700179100037,
        "mean_feedback_forward_cosine": 0.9994903640313582,
        "mean_feedback_forward_relative_error": 0.02999353918500922,
        "min_feedback_forward_cosine": 0.9977270364761353
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.0077766713855415585
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06757808476686478,
        "mean_feedback_forward_cosine": 0.9994905764406378,
        "mean_feedback_forward_relative_error": 0.02998782027174126,
        "min_feedback_forward_cosine": 0.9977282881736755
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.00798575831759721
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06755691766738892,
        "mean_feedback_forward_cosine": 0.9994907899336382,
        "mean_feedback_forward_relative_error": 0.029981972234831615,
        "min_feedback_forward_cosine": 0.9977295994758606
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.007603930989652872
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06753963232040405,
        "mean_feedback_forward_cosine": 0.9994909600778059,
        "mean_feedback_forward_relative_error": 0.02997714238749309,
        "min_feedback_forward_cosine": 0.9977308511734009
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.0069355460249260065
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06752147525548935,
        "mean_feedback_forward_cosine": 0.9994911237196489,
        "mean_feedback_forward_relative_error": 0.029971912274645133,
        "min_feedback_forward_cosine": 0.9977318048477173
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.006878562457635999
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06750191003084183,
        "mean_feedback_forward_cosine": 0.9994913252917204,
        "mean_feedback_forward_relative_error": 0.029966585913842374,
        "min_feedback_forward_cosine": 0.9977332949638367
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.006689532620459795
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06748202443122864,
        "mean_feedback_forward_cosine": 0.9994915268637917,
        "mean_feedback_forward_relative_error": 0.029960997360335154,
        "min_feedback_forward_cosine": 0.9977346658706665
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.006697340769506991
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06746382266283035,
        "mean_feedback_forward_cosine": 0.9994917284358632,
        "mean_feedback_forward_relative_error": 0.029955772632224993,
        "min_feedback_forward_cosine": 0.997735857963562
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.006545494780167937
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06744419038295746,
        "mean_feedback_forward_cosine": 0.9994919245893304,
        "mean_feedback_forward_relative_error": 0.029950090844861485,
        "min_feedback_forward_cosine": 0.9977370500564575
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.006664705861844122
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06742171198129654,
        "mean_feedback_forward_cosine": 0.999492143500935,
        "mean_feedback_forward_relative_error": 0.029943787331946872,
        "min_feedback_forward_cosine": 0.9977386593818665
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.00722001444298774
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740182638168335,
        "mean_feedback_forward_cosine": 0.9994923548264937,
        "mean_feedback_forward_relative_error": 0.02993803441693837,
        "min_feedback_forward_cosine": 0.9977399110794067
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.006274170103864744
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740158796310425,
        "mean_feedback_forward_cosine": 0.9994923656637018,
        "mean_feedback_forward_relative_error": 0.029937946440821345,
        "min_feedback_forward_cosine": 0.9977399110794067
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.005793132462501526
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0674014464020729,
        "mean_feedback_forward_cosine": 0.9994923548264937,
        "mean_feedback_forward_relative_error": 0.029937878962267527,
        "min_feedback_forward_cosine": 0.9977399110794067
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.005183473689854145
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740128248929977,
        "mean_feedback_forward_cosine": 0.999492369998585,
        "mean_feedback_forward_relative_error": 0.02993781239810315,
        "min_feedback_forward_cosine": 0.9977399110794067
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.0055938471915386615
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740112602710724,
        "mean_feedback_forward_cosine": 0.9994923613288186,
        "mean_feedback_forward_relative_error": 0.029937750126489183,
        "min_feedback_forward_cosine": 0.9977399706840515
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.004966788945272565
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740099936723709,
        "mean_feedback_forward_cosine": 0.9994923537427729,
        "mean_feedback_forward_relative_error": 0.029937695457854054,
        "min_feedback_forward_cosine": 0.9977399110794067
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.004283378521986306
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740088015794754,
        "mean_feedback_forward_cosine": 0.9994923580776561,
        "mean_feedback_forward_relative_error": 0.029937641305679625,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.004494202756518498
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740075349807739,
        "mean_feedback_forward_cosine": 0.9994923580776561,
        "mean_feedback_forward_relative_error": 0.029937585951252416,
        "min_feedback_forward_cosine": 0.9977399110794067
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.004588802584782243
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740062683820724,
        "mean_feedback_forward_cosine": 0.9994923765009099,
        "mean_feedback_forward_relative_error": 0.02993752999569882,
        "min_feedback_forward_cosine": 0.9977399706840515
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.004372061618436128
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740052253007889,
        "mean_feedback_forward_cosine": 0.9994923743334684,
        "mean_feedback_forward_relative_error": 0.029937479526481844,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.0039116406575217845
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740038096904755,
        "mean_feedback_forward_cosine": 0.9994923656637018,
        "mean_feedback_forward_relative_error": 0.02993742266500538,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.00434201705057174
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0674002394080162,
        "mean_feedback_forward_cosine": 0.9994923808357933,
        "mean_feedback_forward_relative_error": 0.029937365227802232,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.0043814171719551084
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06740012019872665,
        "mean_feedback_forward_cosine": 0.999492369998585,
        "mean_feedback_forward_relative_error": 0.02993730908598412,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.004154835649877787
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0674000158905983,
        "mean_feedback_forward_cosine": 0.9994923765009099,
        "mean_feedback_forward_relative_error": 0.029937259150160985,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.003897070373259485
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739989668130875,
        "mean_feedback_forward_cosine": 0.9994923569939353,
        "mean_feedback_forward_relative_error": 0.029937205649912358,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.004187560288645327
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0673997774720192,
        "mean_feedback_forward_cosine": 0.9994923754171892,
        "mean_feedback_forward_relative_error": 0.029937153445048766,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.0038901928127743303
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739966571331024,
        "mean_feedback_forward_cosine": 0.9994923830032348,
        "mean_feedback_forward_relative_error": 0.029937104101885447,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.003830749564897269
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0673995390534401,
        "mean_feedback_forward_cosine": 0.9994923786683516,
        "mean_feedback_forward_relative_error": 0.029937052735212175,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.0034973140881955625
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739944219589233,
        "mean_feedback_forward_cosine": 0.9994923797520724,
        "mean_feedback_forward_relative_error": 0.029937005034563216,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.00363929495383054
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739933788776398,
        "mean_feedback_forward_cosine": 0.9994924014264887,
        "mean_feedback_forward_relative_error": 0.029936956089328635,
        "min_feedback_forward_cosine": 0.9977400302886963
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.0038257140539586544
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739923357963562,
        "mean_feedback_forward_cosine": 0.9994923830032348,
        "mean_feedback_forward_relative_error": 0.02993690621277148,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.0037405504473671317
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739912927150726,
        "mean_feedback_forward_cosine": 0.9994923754171892,
        "mean_feedback_forward_relative_error": 0.029936855489557438,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.0036961491615790875
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0673990324139595,
        "mean_feedback_forward_cosine": 0.9994923743334684,
        "mean_feedback_forward_relative_error": 0.029936808796430178,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.003587280821725726
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739892810583115,
        "mean_feedback_forward_cosine": 0.9994923765009099,
        "mean_feedback_forward_relative_error": 0.029936761358244853,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.003504292621985078
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739881634712219,
        "mean_feedback_forward_cosine": 0.9994923851706765,
        "mean_feedback_forward_relative_error": 0.02993671512231231,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.003447736492715776
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739870458841324,
        "mean_feedback_forward_cosine": 0.9994923689148643,
        "mean_feedback_forward_relative_error": 0.029936667091467164,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.003741029561981559
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739861518144608,
        "mean_feedback_forward_cosine": 0.9994923895055597,
        "mean_feedback_forward_relative_error": 0.02993662009354342,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.003317566669508815
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739848852157593,
        "mean_feedback_forward_cosine": 0.9994923992590471,
        "mean_feedback_forward_relative_error": 0.029936568532139062,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.003686123422048986
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739836186170578,
        "mean_feedback_forward_cosine": 0.9994923819195141,
        "mean_feedback_forward_relative_error": 0.02993651358410716,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.003723716419087723
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739827245473862,
        "mean_feedback_forward_cosine": 0.9994923981753263,
        "mean_feedback_forward_relative_error": 0.029936465468596327,
        "min_feedback_forward_cosine": 0.9977400898933411
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.0032948575318977238
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739818304777145,
        "mean_feedback_forward_cosine": 0.9994923927567222,
        "mean_feedback_forward_relative_error": 0.029936423313549974,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.003050474014468491
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0673980712890625,
        "mean_feedback_forward_cosine": 0.9994923960078846,
        "mean_feedback_forward_relative_error": 0.029936372802000153,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.0038788184327445924
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739798188209534,
        "mean_feedback_forward_cosine": 0.9994923862543973,
        "mean_feedback_forward_relative_error": 0.029936325973407788,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.0031438035936653614
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739788502454758,
        "mean_feedback_forward_cosine": 0.9994923992590471,
        "mean_feedback_forward_relative_error": 0.029936281659386374,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.0033135824465192853
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739780306816101,
        "mean_feedback_forward_cosine": 0.9994923808357933,
        "mean_feedback_forward_relative_error": 0.029936238877813926,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.00299471766974777
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739768385887146,
        "mean_feedback_forward_cosine": 0.999492388421839,
        "mean_feedback_forward_relative_error": 0.029936187942935663,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.0033631774935498834
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0673975944519043,
        "mean_feedback_forward_cosine": 0.9994923851706765,
        "mean_feedback_forward_relative_error": 0.02993614422157407,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.0030226892620697616
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739750504493713,
        "mean_feedback_forward_cosine": 0.9994923949241639,
        "mean_feedback_forward_relative_error": 0.029936096859587865,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.003323344582170248
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739738583564758,
        "mean_feedback_forward_cosine": 0.9994924025102095,
        "mean_feedback_forward_relative_error": 0.029936048735610465,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.003273675970211625
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739726662635803,
        "mean_feedback_forward_cosine": 0.9994923992590471,
        "mean_feedback_forward_relative_error": 0.029936002347279677,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.003436235188674182
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739716976881027,
        "mean_feedback_forward_cosine": 0.9994924003427679,
        "mean_feedback_forward_relative_error": 0.029935957135801966,
        "min_feedback_forward_cosine": 0.9977402687072754
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.0031318929681554437
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739707291126251,
        "mean_feedback_forward_cosine": 0.9994923927567222,
        "mean_feedback_forward_relative_error": 0.029935914345762947,
        "min_feedback_forward_cosine": 0.9977401494979858
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.003197696532420814
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739697605371475,
        "mean_feedback_forward_cosine": 0.9994923960078846,
        "mean_feedback_forward_relative_error": 0.029935869252817198,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.0032609035214222967
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739687919616699,
        "mean_feedback_forward_cosine": 0.9994924176823009,
        "mean_feedback_forward_relative_error": 0.029935823677277023,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.0031454689562600106
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739677488803864,
        "mean_feedback_forward_cosine": 0.9994924133474177,
        "mean_feedback_forward_relative_error": 0.029935773572122508,
        "min_feedback_forward_cosine": 0.9977402687072754
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.003257605445832014
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739669293165207,
        "mean_feedback_forward_cosine": 0.9994923970916054,
        "mean_feedback_forward_relative_error": 0.02993572960523042,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.0028618731648474933
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0673966109752655,
        "mean_feedback_forward_cosine": 0.9994924133474177,
        "mean_feedback_forward_relative_error": 0.029935687831179662,
        "min_feedback_forward_cosine": 0.9977402091026306
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.002991606337893754
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739653646945953,
        "mean_feedback_forward_cosine": 0.9994923992590471,
        "mean_feedback_forward_relative_error": 0.029935647928240626,
        "min_feedback_forward_cosine": 0.9977402687072754
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.0027888119074143468
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739643961191177,
        "mean_feedback_forward_cosine": 0.9994924133474177,
        "mean_feedback_forward_relative_error": 0.029935601328245617,
        "min_feedback_forward_cosine": 0.9977402687072754
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.0034980247987434266
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739633530378342,
        "mean_feedback_forward_cosine": 0.9994924187660217,
        "mean_feedback_forward_relative_error": 0.029935555456375534,
        "min_feedback_forward_cosine": 0.9977402687072754
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.003266522671226412
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06739622354507446,
        "mean_feedback_forward_cosine": 0.9994924046776511,
        "mean_feedback_forward_relative_error": 0.0299355069598691,
        "min_feedback_forward_cosine": 0.9977402687072754
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.0032272437416017057
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9273,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.390718896484375,
    "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": 13,
    "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": 2.0538651943206787,
    "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": 18248.176662921906,
    "train_wall_s": 18242.354815006256
  },
  "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
  }
}