{
  "architecture": {
    "adaptive_apical_parameters": 0,
    "adaptive_feedback_parameters": 461440,
    "base_width": 16,
    "blocks_per_stage": 5,
    "bn_eps": 1e-05,
    "bn_momentum": 0.1,
    "depth": 32,
    "family": "CIFAR 6n+2 ResNet, option-A shortcuts",
    "fixed_feedback_parameters": 0,
    "fixed_traffic_coefficients": 0,
    "forward_parameters": 464154,
    "hidden_shapes": [
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        16,
        32,
        32
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        32,
        16,
        16
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ],
      [
        64,
        8,
        8
      ]
    ],
    "normalization": "batchnorm",
    "predictor_parameters": 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": 32,
    "device": "cuda",
    "epochs": 200,
    "eta_A": 0.01,
    "eta_P": 0.01,
    "eval_every": 0,
    "eval_split": "test",
    "learn_P": 0,
    "loader_seed": 14,
    "lr": 0.1,
    "lr_gamma": 0.1,
    "lr_milestones": "100,150",
    "lr_schedule": "step",
    "max_steps": 0,
    "mirror_batch_size": 1,
    "mirror_eta": 0.1,
    "mirror_every": 16,
    "mirror_noise_std": 1.0,
    "mirror_seed": 3000,
    "mirror_warmup_steps": 0,
    "mode": "kp",
    "momentum": 0.9,
    "neutral_projection": 0,
    "normalization": "batchnorm",
    "nuisance_scale": 0.0,
    "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s14.json",
    "output_lr": 0.1,
    "pert_directions": 1,
    "pert_every": 4,
    "pert_sigma": 0.01,
    "perturb_seed": 1000,
    "predictor_every": 0,
    "predictor_mode": "nlms",
    "predictor_warmup_steps": 0,
    "residual_scale": null,
    "seed": 14,
    "split_seed": 2027,
    "traffic_calibration_examples": 64,
    "traffic_ratio": 4.0,
    "traffic_rule": "innovation",
    "traffic_seed": 4000,
    "train_limit": 0,
    "use_residual": 1,
    "val_examples": 0,
    "vectorizer_mode": "spatial_template",
    "warmup_epochs": 0,
    "weight_decay": 0.0001,
    "weight_scale": 1.0,
    "width": 16
  },
  "calibration_metric_space": "reciprocal_local_activity_products",
  "counters": {
    "apical_warmup_examples": 0,
    "calibration_event_examples": 0,
    "causal_scalar_observations": 0,
    "logical_batch_loss_queries": 0,
    "mirror_conv_examples": 0,
    "mirror_events": 0,
    "mirror_readout_examples": 0,
    "neutral_projection_examples": 0,
    "ordinary_examples": 10000000,
    "per_example_loss_terms": 0,
    "perturbation_events": 0,
    "perturbation_forward_examples": 0,
    "predictor_update_examples": 0,
    "predictor_warmup_examples": 0,
    "traffic_audit_examples": 0,
    "traffic_calibration_examples": 0
  },
  "diagnostics": {
    "early_third_mean": 0.9996268689632416,
    "feedback_forward_cosine": [
      0.9997919797897339,
      0.9997849464416504,
      0.9997409582138062,
      0.9997344613075256,
      0.999788224697113,
      0.9997414350509644,
      0.9998248815536499,
      0.9997788667678833,
      0.9997860789299011,
      0.9996800422668457,
      0.9997262954711914,
      0.9998047351837158,
      0.9997510313987732,
      0.9997111558914185,
      0.9997680187225342,
      0.999701976776123,
      0.9997506737709045,
      0.9996946454048157,
      0.999782919883728,
      0.9996932744979858,
      0.9997825622558594,
      0.9998618364334106,
      0.99983149766922,
      0.9998146295547485,
      0.9998298287391663,
      0.9997783899307251,
      0.9998220205307007,
      0.9996968507766724,
      0.9997162222862244,
      0.9994392395019531,
      0.9999843835830688
    ],
    "feedback_forward_norm_ratio": [
      0.9989004731178284,
      1.0007294416427612,
      1.0002119541168213,
      1.001495122909546,
      1.0001250505447388,
      1.0014092922210693,
      0.9999539256095886,
      1.0007349252700806,
      0.9999049305915833,
      1.0011711120605469,
      0.9987809658050537,
      1.001015543937683,
      0.9990402460098267,
      1.0010043382644653,
      0.999497652053833,
      1.0015132427215576,
      0.9995678663253784,
      1.0010426044464111,
      0.9991808533668518,
      1.0010385513305664,
      0.999531626701355,
      1.0005073547363281,
      0.9997082352638245,
      1.000762939453125,
      0.9995765089988708,
      1.000929594039917,
      0.9997575879096985,
      1.0009090900421143,
      0.9997023940086365,
      1.0010747909545898,
      1.0025800466537476
    ],
    "feedback_forward_relative_error": [
      0.020418580621480942,
      0.020756781101226807,
      0.022764239460229874,
      0.023108825087547302,
      0.020585954189300537,
      0.022801287472248077,
      0.018710697069764137,
      0.021051138639450073,
      0.020687395706772804,
      0.025339433923363686,
      0.02341407909989357,
      0.019799573346972466,
      0.022325001657009125,
      0.02407108061015606,
      0.02154187485575676,
      0.024480070918798447,
      0.02232399955391884,
      0.02474418468773365,
      0.02083893120288849,
      0.024802938103675842,
      0.020856020972132683,
      0.016635606065392494,
      0.018359307199716568,
      0.019281823188066483,
      0.018446631729602814,
      0.021085331216454506,
      0.01886802539229393,
      0.024646619334816933,
      0.023822002112865448,
      0.03352376073598862,
      0.006160619668662548
    ],
    "innovation_negative_gradient_cosine": [
      0.999638557434082,
      0.9996102452278137,
      0.9996362924575806,
      0.9995453357696533,
      0.9996436834335327,
      0.9995719194412231,
      0.9996561408042908,
      0.999657154083252,
      0.999669075012207,
      0.9996402859687805,
      0.9996742010116577,
      0.9996456503868103,
      0.999691903591156,
      0.9995970129966736,
      0.9997073411941528,
      0.9996359348297119,
      0.9997234344482422,
      0.999672532081604,
      0.9997422695159912,
      0.9996891617774963,
      0.9997507333755493,
      0.9997844696044922,
      0.9997909069061279,
      0.9997697472572327,
      0.9998173713684082,
      0.9998141527175903,
      0.9998586177825928,
      0.9998947381973267,
      0.9999035596847534,
      0.9999581575393677,
      0.7821329832077026
    ],
    "mean_feedback_forward_cosine": 0.9997610988155488,
    "mean_feedback_forward_relative_error": 0.021491994029812275,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.999638557434082,
      0.9996102452278137,
      0.9996362924575806,
      0.9995453357696533,
      0.9996436834335327,
      0.9995719194412231,
      0.9996561408042908,
      0.999657154083252,
      0.999669075012207,
      0.9996402859687805,
      0.9996742010116577,
      0.9996456503868103,
      0.999691903591156,
      0.9995970129966736,
      0.9997073411941528,
      0.9996359348297119,
      0.9997234344482422,
      0.999672532081604,
      0.9997422695159912,
      0.9996891617774963,
      0.9997507333755493,
      0.9997844696044922,
      0.9997909069061279,
      0.9997697472572327,
      0.9998173713684082,
      0.9998141527175903,
      0.9998586177825928,
      0.9998947381973267,
      0.9999035596847534,
      0.9999581575393677,
      0.7821329832077026
    ],
    "teaching_negative_gradient_cosine": [
      0.999638557434082,
      0.9996102452278137,
      0.9996362924575806,
      0.9995453357696533,
      0.9996436834335327,
      0.9995719194412231,
      0.9996561408042908,
      0.999657154083252,
      0.999669075012207,
      0.9996402859687805,
      0.9996742010116577,
      0.9996456503868103,
      0.999691903591156,
      0.9995970129966736,
      0.9997073411941528,
      0.9996359348297119,
      0.9997234344482422,
      0.999672532081604,
      0.9997422695159912,
      0.9996891617774963,
      0.9997507333755493,
      0.9997844696044922,
      0.9997909069061279,
      0.9997697472572327,
      0.9998173713684082,
      0.9998141527175903,
      0.9998586177825928,
      0.9998947381973267,
      0.9999035596847534,
      0.9999581575393677,
      0.7821329832077026
    ],
    "wall_s": 0.12763762474060059
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.332453727722168,
        "mean_feedback_forward_cosine": 0.28569606547394105,
        "mean_feedback_forward_relative_error": 1.1923096929827044,
        "min_feedback_forward_cosine": 0.11065932363271713
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 1.8864616674423218
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2947325706481934,
        "mean_feedback_forward_cosine": 0.37175222558359944,
        "mean_feedback_forward_relative_error": 1.1189814344529183,
        "min_feedback_forward_cosine": 0.16187171638011932
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 1.3447055550384521
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2630192041397095,
        "mean_feedback_forward_cosine": 0.4468510333568819,
        "mean_feedback_forward_relative_error": 1.050396325126771,
        "min_feedback_forward_cosine": 0.20473386347293854
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.0783762699890136
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2285550832748413,
        "mean_feedback_forward_cosine": 0.506152376051872,
        "mean_feedback_forward_relative_error": 0.9932994640642597,
        "min_feedback_forward_cosine": 0.25025951862335205
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 0.8908661921310425
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1947357654571533,
        "mean_feedback_forward_cosine": 0.5549704615146883,
        "mean_feedback_forward_relative_error": 0.9433515715983606,
        "min_feedback_forward_cosine": 0.29240745306015015
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 0.7713095247745514
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1657177209854126,
        "mean_feedback_forward_cosine": 0.5965976099814138,
        "mean_feedback_forward_relative_error": 0.898323812792378,
        "min_feedback_forward_cosine": 0.333939790725708
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 0.6912768111038208
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1398651599884033,
        "mean_feedback_forward_cosine": 0.6339097003782949,
        "mean_feedback_forward_relative_error": 0.8561380467107219,
        "min_feedback_forward_cosine": 0.36467909812927246
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.6317887006759644
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1156481504440308,
        "mean_feedback_forward_cosine": 0.6650987209812287,
        "mean_feedback_forward_relative_error": 0.8189005140335329,
        "min_feedback_forward_cosine": 0.3936135768890381
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.5794082787704468
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0911922454833984,
        "mean_feedback_forward_cosine": 0.6936676425318564,
        "mean_feedback_forward_relative_error": 0.7829820350293191,
        "min_feedback_forward_cosine": 0.4210509657859802
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.5458818233299255
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0665953159332275,
        "mean_feedback_forward_cosine": 0.7185857988172962,
        "mean_feedback_forward_relative_error": 0.7501856107865611,
        "min_feedback_forward_cosine": 0.4483541250228882
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.5144352616310119
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0408952236175537,
        "mean_feedback_forward_cosine": 0.7419745210678347,
        "mean_feedback_forward_relative_error": 0.7179888842567321,
        "min_feedback_forward_cosine": 0.4757261276245117
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.48909983785629274
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0145304203033447,
        "mean_feedback_forward_cosine": 0.7628409727927177,
        "mean_feedback_forward_relative_error": 0.6880684966041196,
        "min_feedback_forward_cosine": 0.5025681853294373
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.46696600249290465
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9898214936256409,
        "mean_feedback_forward_cosine": 0.7811945292257494,
        "mean_feedback_forward_relative_error": 0.660570744545229,
        "min_feedback_forward_cosine": 0.5271745324134827
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.44200228766441346
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.965601921081543,
        "mean_feedback_forward_cosine": 0.7980669429225307,
        "mean_feedback_forward_relative_error": 0.6340958384736892,
        "min_feedback_forward_cosine": 0.550247311592102
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.41981414313316345
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9398868083953857,
        "mean_feedback_forward_cosine": 0.8137554430192516,
        "mean_feedback_forward_relative_error": 0.608631880052628,
        "min_feedback_forward_cosine": 0.5739184617996216
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.40447162753105165
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9141412973403931,
        "mean_feedback_forward_cosine": 0.828351866814398,
        "mean_feedback_forward_relative_error": 0.5839648977402718,
        "min_feedback_forward_cosine": 0.597303032875061
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.39567691297531127
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.889471173286438,
        "mean_feedback_forward_cosine": 0.8413230834468719,
        "mean_feedback_forward_relative_error": 0.5611338033791511,
        "min_feedback_forward_cosine": 0.6190862655639648
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.37986853382110597
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8651845455169678,
        "mean_feedback_forward_cosine": 0.8531636115043394,
        "mean_feedback_forward_relative_error": 0.539392416515658,
        "min_feedback_forward_cosine": 0.6397159099578857
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.36365509808063506
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.839250922203064,
        "mean_feedback_forward_cosine": 0.8643700384324596,
        "mean_feedback_forward_relative_error": 0.5179809389575836,
        "min_feedback_forward_cosine": 0.6609064340591431
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.3630038592338562
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8157009482383728,
        "mean_feedback_forward_cosine": 0.8745473238729662,
        "mean_feedback_forward_relative_error": 0.4977624238498749,
        "min_feedback_forward_cosine": 0.679825484752655
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.34615828795433046
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7918817400932312,
        "mean_feedback_forward_cosine": 0.8837778452903994,
        "mean_feedback_forward_relative_error": 0.4788322655423995,
        "min_feedback_forward_cosine": 0.6985763311386108
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.3352479345989227
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7668979167938232,
        "mean_feedback_forward_cosine": 0.8925440811341808,
        "mean_feedback_forward_relative_error": 0.4600896590179013,
        "min_feedback_forward_cosine": 0.7172738909721375
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.3282865807914734
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7425411939620972,
        "mean_feedback_forward_cosine": 0.900546173895559,
        "mean_feedback_forward_relative_error": 0.4423999296080682,
        "min_feedback_forward_cosine": 0.7347917556762695
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.3158574280834198
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.719512939453125,
        "mean_feedback_forward_cosine": 0.9078300268419327,
        "mean_feedback_forward_relative_error": 0.42559024835786513,
        "min_feedback_forward_cosine": 0.7508101463317871
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.30837035751342773
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6968992352485657,
        "mean_feedback_forward_cosine": 0.9147653137483904,
        "mean_feedback_forward_relative_error": 0.40906073008814164,
        "min_feedback_forward_cosine": 0.766478955745697
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.29827840731620786
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6743670701980591,
        "mean_feedback_forward_cosine": 0.9211131814987429,
        "mean_feedback_forward_relative_error": 0.39326262089513964,
        "min_feedback_forward_cosine": 0.781310498714447
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.2950888151741028
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.651057243347168,
        "mean_feedback_forward_cosine": 0.9269627544187731,
        "mean_feedback_forward_relative_error": 0.37825878493247495,
        "min_feedback_forward_cosine": 0.796197235584259
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.2906055842828751
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6293326616287231,
        "mean_feedback_forward_cosine": 0.9323475995371419,
        "mean_feedback_forward_relative_error": 0.36379277369668406,
        "min_feedback_forward_cosine": 0.8094980716705322
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.2810540374565125
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6074268817901611,
        "mean_feedback_forward_cosine": 0.9374242501874124,
        "mean_feedback_forward_relative_error": 0.34971984788294763,
        "min_feedback_forward_cosine": 0.8221464157104492
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.2781663143348694
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5866720080375671,
        "mean_feedback_forward_cosine": 0.9419786180219343,
        "mean_feedback_forward_relative_error": 0.33660862138194425,
        "min_feedback_forward_cosine": 0.8340702056884766
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.27273870217323304
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5648839473724365,
        "mean_feedback_forward_cosine": 0.9463978294403322,
        "mean_feedback_forward_relative_error": 0.3234559158163686,
        "min_feedback_forward_cosine": 0.8459969162940979
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.26634878540039064
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5445200800895691,
        "mean_feedback_forward_cosine": 0.9504142595875648,
        "mean_feedback_forward_relative_error": 0.31098151182936085,
        "min_feedback_forward_cosine": 0.856764554977417
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.26126067794322966
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5249777436256409,
        "mean_feedback_forward_cosine": 0.9541119606264176,
        "mean_feedback_forward_relative_error": 0.29899491441826664,
        "min_feedback_forward_cosine": 0.8667631149291992
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.2520839551353455
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5054250359535217,
        "mean_feedback_forward_cosine": 0.9575571475490448,
        "mean_feedback_forward_relative_error": 0.28742779647150346,
        "min_feedback_forward_cosine": 0.8762861490249634
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.25377710359573363
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.48607558012008667,
        "mean_feedback_forward_cosine": 0.9607740486821821,
        "mean_feedback_forward_relative_error": 0.2762442082166672,
        "min_feedback_forward_cosine": 0.8854256868362427
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.24896999197483063
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4677378535270691,
        "mean_feedback_forward_cosine": 0.963708339198943,
        "mean_feedback_forward_relative_error": 0.26564024220551213,
        "min_feedback_forward_cosine": 0.8937792181968689
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.2468858616542816
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4498147666454315,
        "mean_feedback_forward_cosine": 0.9664514295516475,
        "mean_feedback_forward_relative_error": 0.2553777086638635,
        "min_feedback_forward_cosine": 0.9017024040222168
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.23782768265247345
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.43192774057388306,
        "mean_feedback_forward_cosine": 0.9689808353301017,
        "mean_feedback_forward_relative_error": 0.24550913610766012,
        "min_feedback_forward_cosine": 0.9092162847518921
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.23959526492118835
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.41520217061042786,
        "mean_feedback_forward_cosine": 0.9713096599425038,
        "mean_feedback_forward_relative_error": 0.23603895787269838,
        "min_feedback_forward_cosine": 0.9160403609275818
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.23518191647052764
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.39894381165504456,
        "mean_feedback_forward_cosine": 0.9734609646181906,
        "mean_feedback_forward_relative_error": 0.2269706735687871,
        "min_feedback_forward_cosine": 0.9224241971969604
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.23041978515148162
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.38303983211517334,
        "mean_feedback_forward_cosine": 0.975447416305542,
        "mean_feedback_forward_relative_error": 0.2182956261019553,
        "min_feedback_forward_cosine": 0.92841637134552
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.23074689755916594
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.368954598903656,
        "mean_feedback_forward_cosine": 0.9772347577156559,
        "mean_feedback_forward_relative_error": 0.21015597807784234,
        "min_feedback_forward_cosine": 0.9335581064224243
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.22321409762859346
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3541412353515625,
        "mean_feedback_forward_cosine": 0.9790231431684187,
        "mean_feedback_forward_relative_error": 0.2016921742789207,
        "min_feedback_forward_cosine": 0.9387012124061584
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.22403297308921813
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.33939701318740845,
        "mean_feedback_forward_cosine": 0.9805971268684633,
        "mean_feedback_forward_relative_error": 0.19396936436814646,
        "min_feedback_forward_cosine": 0.9436639547348022
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.22272661297798158
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.32659563422203064,
        "mean_feedback_forward_cosine": 0.98200483475962,
        "mean_feedback_forward_relative_error": 0.18678255018688017,
        "min_feedback_forward_cosine": 0.9477999210357666
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.21533804458618164
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3134482502937317,
        "mean_feedback_forward_cosine": 0.9833629381272101,
        "mean_feedback_forward_relative_error": 0.17959226283334917,
        "min_feedback_forward_cosine": 0.9518828392028809
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.21674478680610657
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3012486398220062,
        "mean_feedback_forward_cosine": 0.9845931433862255,
        "mean_feedback_forward_relative_error": 0.17279463826167968,
        "min_feedback_forward_cosine": 0.9555051326751709
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.2155820286989212
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.28929030895233154,
        "mean_feedback_forward_cosine": 0.985737721766195,
        "mean_feedback_forward_relative_error": 0.1662241930682813,
        "min_feedback_forward_cosine": 0.9589306116104126
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.2111910402202606
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2772831618785858,
        "mean_feedback_forward_cosine": 0.986845201061618,
        "mean_feedback_forward_relative_error": 0.15963495522737503,
        "min_feedback_forward_cosine": 0.9622498154640198
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.21108394720077514
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.26609328389167786,
        "mean_feedback_forward_cosine": 0.9878137342391475,
        "mean_feedback_forward_relative_error": 0.1536394064464877,
        "min_feedback_forward_cosine": 0.9651902914047241
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.20221650484085082
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.25522199273109436,
        "mean_feedback_forward_cosine": 0.9887455471100346,
        "mean_feedback_forward_relative_error": 0.1476403996348381,
        "min_feedback_forward_cosine": 0.9679491519927979
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.20754753598213196
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2448800802230835,
        "mean_feedback_forward_cosine": 0.9895845670853892,
        "mean_feedback_forward_relative_error": 0.14203302297861345,
        "min_feedback_forward_cosine": 0.9704720973968506
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.2019150578212738
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.23494644463062286,
        "mean_feedback_forward_cosine": 0.9903780145029868,
        "mean_feedback_forward_relative_error": 0.13650183896383933,
        "min_feedback_forward_cosine": 0.9727993607521057
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.20128608113765717
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22498035430908203,
        "mean_feedback_forward_cosine": 0.9911072619499699,
        "mean_feedback_forward_relative_error": 0.1312497033467216,
        "min_feedback_forward_cosine": 0.9750555157661438
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.20113794761180878
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.21596264839172363,
        "mean_feedback_forward_cosine": 0.9917765778879966,
        "mean_feedback_forward_relative_error": 0.1262037168827749,
        "min_feedback_forward_cosine": 0.9769986867904663
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.19641307671546937
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20744924247264862,
        "mean_feedback_forward_cosine": 0.9923929975878808,
        "mean_feedback_forward_relative_error": 0.12137739264195965,
        "min_feedback_forward_cosine": 0.9787628054618835
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.1954195665884018
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19935357570648193,
        "mean_feedback_forward_cosine": 0.9929594359090251,
        "mean_feedback_forward_relative_error": 0.11676642827449306,
        "min_feedback_forward_cosine": 0.9803847670555115
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.19436430928230286
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19114062190055847,
        "mean_feedback_forward_cosine": 0.9934934050806107,
        "mean_feedback_forward_relative_error": 0.11225502865929757,
        "min_feedback_forward_cosine": 0.9819468259811401
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.19422180081844329
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.183742955327034,
        "mean_feedback_forward_cosine": 0.9939748471783053,
        "mean_feedback_forward_relative_error": 0.10800776967117863,
        "min_feedback_forward_cosine": 0.983309268951416
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.1890528978586197
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17630770802497864,
        "mean_feedback_forward_cosine": 0.9944276944283517,
        "mean_feedback_forward_relative_error": 0.1038708767342952,
        "min_feedback_forward_cosine": 0.9846312999725342
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.1900973216152191
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16908420622348785,
        "mean_feedback_forward_cosine": 0.9948449615509279,
        "mean_feedback_forward_relative_error": 0.09990192457072196,
        "min_feedback_forward_cosine": 0.985857367515564
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.19067359004735945
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16221991181373596,
        "mean_feedback_forward_cosine": 0.9952308247166295,
        "mean_feedback_forward_relative_error": 0.09608427531296207,
        "min_feedback_forward_cosine": 0.9869732856750488
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.18816538153648377
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1557830423116684,
        "mean_feedback_forward_cosine": 0.9955845071423438,
        "mean_feedback_forward_relative_error": 0.0924553124895019,
        "min_feedback_forward_cosine": 0.9879834651947021
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.1851732423210144
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14936737716197968,
        "mean_feedback_forward_cosine": 0.9959167568914352,
        "mean_feedback_forward_relative_error": 0.0889079782270616,
        "min_feedback_forward_cosine": 0.9889500141143799
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.18822476876735686
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14342321455478668,
        "mean_feedback_forward_cosine": 0.9962218461498138,
        "mean_feedback_forward_relative_error": 0.08552141439530157,
        "min_feedback_forward_cosine": 0.989808201789856
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.18450284621238708
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13741940259933472,
        "mean_feedback_forward_cosine": 0.9965088386689464,
        "mean_feedback_forward_relative_error": 0.0822157223498629,
        "min_feedback_forward_cosine": 0.9906416535377502
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.1846496206665039
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1318826675415039,
        "mean_feedback_forward_cosine": 0.9967711721697161,
        "mean_feedback_forward_relative_error": 0.07906677039159883,
        "min_feedback_forward_cosine": 0.991375207901001
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.18396508299827577
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12647992372512817,
        "mean_feedback_forward_cosine": 0.9970187756323046,
        "mean_feedback_forward_relative_error": 0.07597769808865362,
        "min_feedback_forward_cosine": 0.9920645952224731
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.18446759902000429
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12145280092954636,
        "mean_feedback_forward_cosine": 0.9972406164292367,
        "mean_feedback_forward_relative_error": 0.07309857100969361,
        "min_feedback_forward_cosine": 0.992681086063385
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.18181212815999984
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11659641563892365,
        "mean_feedback_forward_cosine": 0.9974452833975515,
        "mean_feedback_forward_relative_error": 0.07033463843887852,
        "min_feedback_forward_cosine": 0.9932525157928467
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.17683307625293732
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1117180585861206,
        "mean_feedback_forward_cosine": 0.9976399214037003,
        "mean_feedback_forward_relative_error": 0.06761032252782775,
        "min_feedback_forward_cosine": 0.993803858757019
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.1790683585214615
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10718255490064621,
        "mean_feedback_forward_cosine": 0.9978178970275386,
        "mean_feedback_forward_relative_error": 0.06500570896652437,
        "min_feedback_forward_cosine": 0.9942947626113892
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.1744948213863373
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10274895280599594,
        "mean_feedback_forward_cosine": 0.9979846592872373,
        "mean_feedback_forward_relative_error": 0.062479867930373835,
        "min_feedback_forward_cosine": 0.9947553873062134
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.18304942828178405
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09887866675853729,
        "mean_feedback_forward_cosine": 0.9981283007129547,
        "mean_feedback_forward_relative_error": 0.06020873045969394,
        "min_feedback_forward_cosine": 0.9951422810554504
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.17089463544845582
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09468246251344681,
        "mean_feedback_forward_cosine": 0.9982725804851901,
        "mean_feedback_forward_relative_error": 0.05784883536398411,
        "min_feedback_forward_cosine": 0.9955440759658813
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.17783002980709076
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09086678177118301,
        "mean_feedback_forward_cosine": 0.9984012207677287,
        "mean_feedback_forward_relative_error": 0.05565292400217826,
        "min_feedback_forward_cosine": 0.995894193649292
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.1737877887916565
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08716202527284622,
        "mean_feedback_forward_cosine": 0.998520556957491,
        "mean_feedback_forward_relative_error": 0.05353486916470912,
        "min_feedback_forward_cosine": 0.9962215423583984
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.17157120014190674
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08369570225477219,
        "mean_feedback_forward_cosine": 0.9986321349297801,
        "mean_feedback_forward_relative_error": 0.05147713903457888,
        "min_feedback_forward_cosine": 0.9965160489082336
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.17163206961631775
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0802949070930481,
        "mean_feedback_forward_cosine": 0.9987352817289291,
        "mean_feedback_forward_relative_error": 0.0495004809551662,
        "min_feedback_forward_cosine": 0.9967924952507019
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.1683024906373024
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07682712376117706,
        "mean_feedback_forward_cosine": 0.9988350791315879,
        "mean_feedback_forward_relative_error": 0.0475140098602541,
        "min_feedback_forward_cosine": 0.9970628023147583
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.1741444329929352
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07371344417333603,
        "mean_feedback_forward_cosine": 0.9989226691184505,
        "mean_feedback_forward_relative_error": 0.04569508850334152,
        "min_feedback_forward_cosine": 0.9972962141036987
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.1721878899717331
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07070252299308777,
        "mean_feedback_forward_cosine": 0.9990031642298545,
        "mean_feedback_forward_relative_error": 0.0439572018841582,
        "min_feedback_forward_cosine": 0.9975121021270752
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.16856893158912659
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06776734441518784,
        "mean_feedback_forward_cosine": 0.9990783141505334,
        "mean_feedback_forward_relative_error": 0.04227025079871378,
        "min_feedback_forward_cosine": 0.9977139234542847
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.16814912214279173
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06500975042581558,
        "mean_feedback_forward_cosine": 0.9991486053312978,
        "mean_feedback_forward_relative_error": 0.04062697895231747,
        "min_feedback_forward_cosine": 0.9978955984115601
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.16887235580444335
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06233934313058853,
        "mean_feedback_forward_cosine": 0.9992128937475143,
        "mean_feedback_forward_relative_error": 0.039068897134594376,
        "min_feedback_forward_cosine": 0.9980646371841431
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.16984825755119323
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05984418839216232,
        "mean_feedback_forward_cosine": 0.9992719600277562,
        "mean_feedback_forward_relative_error": 0.037575110522729736,
        "min_feedback_forward_cosine": 0.9982162714004517
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.16876170399665832
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0574023611843586,
        "mean_feedback_forward_cosine": 0.9993275230930697,
        "mean_feedback_forward_relative_error": 0.036111079306612094,
        "min_feedback_forward_cosine": 0.9983586072921753
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.16502491756916046
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05504690110683441,
        "mean_feedback_forward_cosine": 0.9993792676156567,
        "mean_feedback_forward_relative_error": 0.034695310157633594,
        "min_feedback_forward_cosine": 0.9984903335571289
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.1693146181845665
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.052905917167663574,
        "mean_feedback_forward_cosine": 0.9994260822573016,
        "mean_feedback_forward_relative_error": 0.03336298264442913,
        "min_feedback_forward_cosine": 0.998605489730835
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.16608294499397278
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05085453763604164,
        "mean_feedback_forward_cosine": 0.9994678977997072,
        "mean_feedback_forward_relative_error": 0.032121037133038044,
        "min_feedback_forward_cosine": 0.9987109899520874
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.16007911418914794
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04878352954983711,
        "mean_feedback_forward_cosine": 0.9995081136303563,
        "mean_feedback_forward_relative_error": 0.030884851701557636,
        "min_feedback_forward_cosine": 0.9988136887550354
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.1665627337384224
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04683680459856987,
        "mean_feedback_forward_cosine": 0.9995447685641627,
        "mean_feedback_forward_relative_error": 0.029712052264761542,
        "min_feedback_forward_cosine": 0.9989064335823059
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.16070930757522583
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.044917646795511246,
        "mean_feedback_forward_cosine": 0.9995796122858601,
        "mean_feedback_forward_relative_error": 0.028554636534423597,
        "min_feedback_forward_cosine": 0.9989941120147705
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.1618361934566498
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.043176449835300446,
        "mean_feedback_forward_cosine": 0.9996113450296463,
        "mean_feedback_forward_relative_error": 0.027455753827046965,
        "min_feedback_forward_cosine": 0.999070405960083
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.16187013481616974
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.041375935077667236,
        "mean_feedback_forward_cosine": 0.9996412434885579,
        "mean_feedback_forward_relative_error": 0.02638026164664376,
        "min_feedback_forward_cosine": 0.9991462826728821
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.1638966773366928
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039716508239507675,
        "mean_feedback_forward_cosine": 0.9996683982110792,
        "mean_feedback_forward_relative_error": 0.02536310447800544,
        "min_feedback_forward_cosine": 0.9992133378982544
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.16252423438310623
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038125619292259216,
        "mean_feedback_forward_cosine": 0.9996935590620963,
        "mean_feedback_forward_relative_error": 0.024381615101329742,
        "min_feedback_forward_cosine": 0.9992749691009521
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.16267103942394256
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.036539141088724136,
        "mean_feedback_forward_cosine": 0.9997172394106465,
        "mean_feedback_forward_relative_error": 0.023422856334476702,
        "min_feedback_forward_cosine": 0.9993340969085693
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.16701118450164795
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03513658419251442,
        "mean_feedback_forward_cosine": 0.9997374261579206,
        "mean_feedback_forward_relative_error": 0.022570521690912785,
        "min_feedback_forward_cosine": 0.9993841648101807
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.15273886563301087
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03375523164868355,
        "mean_feedback_forward_cosine": 0.9997568380448126,
        "mean_feedback_forward_relative_error": 0.021721016255117232,
        "min_feedback_forward_cosine": 0.9994314908981323
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.15564070889234544
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033744946122169495,
        "mean_feedback_forward_cosine": 0.9997570226269383,
        "mean_feedback_forward_relative_error": 0.021712409422522592,
        "min_feedback_forward_cosine": 0.9994319677352905
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.08587176040172577
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03373920917510986,
        "mean_feedback_forward_cosine": 0.9997570918452355,
        "mean_feedback_forward_relative_error": 0.02170650997469502,
        "min_feedback_forward_cosine": 0.9994320869445801
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.064969905834198
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0337340384721756,
        "mean_feedback_forward_cosine": 0.9997572072090641,
        "mean_feedback_forward_relative_error": 0.021701117451753345,
        "min_feedback_forward_cosine": 0.9994323253631592
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.054667386298179625
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03372880443930626,
        "mean_feedback_forward_cosine": 0.9997573129592403,
        "mean_feedback_forward_relative_error": 0.021695784263072476,
        "min_feedback_forward_cosine": 0.9994325637817383
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.04887079746007919
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0337241068482399,
        "mean_feedback_forward_cosine": 0.9997573667956937,
        "mean_feedback_forward_relative_error": 0.021690729569884076,
        "min_feedback_forward_cosine": 0.9994326233863831
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.0453762242937088
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03371940553188324,
        "mean_feedback_forward_cosine": 0.9997574360139908,
        "mean_feedback_forward_relative_error": 0.021685759657092633,
        "min_feedback_forward_cosine": 0.9994326829910278
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.041495048912763594
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033714789897203445,
        "mean_feedback_forward_cosine": 0.9997575321505147,
        "mean_feedback_forward_relative_error": 0.02168095063778662,
        "min_feedback_forward_cosine": 0.9994329214096069
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.037941184784173965
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033710163086652756,
        "mean_feedback_forward_cosine": 0.9997576052142728,
        "mean_feedback_forward_relative_error": 0.021676265259063052,
        "min_feedback_forward_cosine": 0.9994330406188965
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.03754971484422684
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03370567783713341,
        "mean_feedback_forward_cosine": 0.9997577013507966,
        "mean_feedback_forward_relative_error": 0.02167169302101097,
        "min_feedback_forward_cosine": 0.9994332194328308
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.03408987087845802
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03370092064142227,
        "mean_feedback_forward_cosine": 0.9997577782600157,
        "mean_feedback_forward_relative_error": 0.021667062049551357,
        "min_feedback_forward_cosine": 0.9994333386421204
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.032719900552630424
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033696338534355164,
        "mean_feedback_forward_cosine": 0.9997578570919652,
        "mean_feedback_forward_relative_error": 0.0216625030302713,
        "min_feedback_forward_cosine": 0.9994335174560547
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.031641623229980466
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033692020922899246,
        "mean_feedback_forward_cosine": 0.9997579416921062,
        "mean_feedback_forward_relative_error": 0.021658158695866025,
        "min_feedback_forward_cosine": 0.999433696269989
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.028986293906569482
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03368794173002243,
        "mean_feedback_forward_cosine": 0.9997580224467862,
        "mean_feedback_forward_relative_error": 0.021653961570512865,
        "min_feedback_forward_cosine": 0.9994337558746338
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.026132644518613815
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033683449029922485,
        "mean_feedback_forward_cosine": 0.9997580955105443,
        "mean_feedback_forward_relative_error": 0.021649496299364874,
        "min_feedback_forward_cosine": 0.9994339346885681
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.02686688132405281
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03367915004491806,
        "mean_feedback_forward_cosine": 0.9997581685743024,
        "mean_feedback_forward_relative_error": 0.02164519934224025,
        "min_feedback_forward_cosine": 0.9994341731071472
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.025645499723553657
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03367479518055916,
        "mean_feedback_forward_cosine": 0.9997582397153301,
        "mean_feedback_forward_relative_error": 0.02164089640662555,
        "min_feedback_forward_cosine": 0.9994341731071472
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.02355453469067812
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033670708537101746,
        "mean_feedback_forward_cosine": 0.9997583147018186,
        "mean_feedback_forward_relative_error": 0.021636745396761164,
        "min_feedback_forward_cosine": 0.9994344115257263
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.02261603930771351
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0336662121117115,
        "mean_feedback_forward_cosine": 0.999758405070151,
        "mean_feedback_forward_relative_error": 0.021632374309363864,
        "min_feedback_forward_cosine": 0.9994345903396606
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.02236939556080848
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033661872148513794,
        "mean_feedback_forward_cosine": 0.9997584762111786,
        "mean_feedback_forward_relative_error": 0.02162814729156033,
        "min_feedback_forward_cosine": 0.9994347095489502
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.021888015147149564
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03365737572312355,
        "mean_feedback_forward_cosine": 0.9997585531203977,
        "mean_feedback_forward_relative_error": 0.02162377931357872,
        "min_feedback_forward_cosine": 0.9994348287582397
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.021143408103883266
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033653125166893005,
        "mean_feedback_forward_cosine": 0.9997586531023825,
        "mean_feedback_forward_relative_error": 0.021619573505895752,
        "min_feedback_forward_cosine": 0.9994350075721741
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.020178167527318002
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0336487852036953,
        "mean_feedback_forward_cosine": 0.9997587223206797,
        "mean_feedback_forward_relative_error": 0.021615374818324082,
        "min_feedback_forward_cosine": 0.9994350671768188
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.019790309482812882
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033644478768110275,
        "mean_feedback_forward_cosine": 0.9997587819253245,
        "mean_feedback_forward_relative_error": 0.021611214915831244,
        "min_feedback_forward_cosine": 0.9994351863861084
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.01804049372404814
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033640433102846146,
        "mean_feedback_forward_cosine": 0.9997588569118131,
        "mean_feedback_forward_relative_error": 0.021607064461756136,
        "min_feedback_forward_cosine": 0.999435305595398
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.017340303631424903
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03363589942455292,
        "mean_feedback_forward_cosine": 0.9997589549710674,
        "mean_feedback_forward_relative_error": 0.021602735839663975,
        "min_feedback_forward_cosine": 0.9994354248046875
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.01802509668469429
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03363114595413208,
        "mean_feedback_forward_cosine": 0.9997590453393997,
        "mean_feedback_forward_relative_error": 0.021598194245128863,
        "min_feedback_forward_cosine": 0.9994357228279114
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.018359613683819772
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03362688049674034,
        "mean_feedback_forward_cosine": 0.9997591472441151,
        "mean_feedback_forward_relative_error": 0.02159395014807101,
        "min_feedback_forward_cosine": 0.9994357824325562
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.01639207499742508
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03362235799431801,
        "mean_feedback_forward_cosine": 0.9997592241533341,
        "mean_feedback_forward_relative_error": 0.02158959495324281,
        "min_feedback_forward_cosine": 0.9994359612464905
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.01609149567514658
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03361808881163597,
        "mean_feedback_forward_cosine": 0.9997592914489007,
        "mean_feedback_forward_relative_error": 0.02158542512164962,
        "min_feedback_forward_cosine": 0.9994361400604248
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.016263673985898494
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033613912761211395,
        "mean_feedback_forward_cosine": 0.9997593664353893,
        "mean_feedback_forward_relative_error": 0.021581318950460802,
        "min_feedback_forward_cosine": 0.9994363188743591
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.014643422370404006
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033610012382268906,
        "mean_feedback_forward_cosine": 0.9997594471900694,
        "mean_feedback_forward_relative_error": 0.0215773728616055,
        "min_feedback_forward_cosine": 0.9994363784790039
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.014087463182508945
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03360601142048836,
        "mean_feedback_forward_cosine": 0.9997595356356713,
        "mean_feedback_forward_relative_error": 0.021573296342525753,
        "min_feedback_forward_cosine": 0.9994365572929382
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.013209543904662132
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033601053059101105,
        "mean_feedback_forward_cosine": 0.9997596106221599,
        "mean_feedback_forward_relative_error": 0.021568678244347533,
        "min_feedback_forward_cosine": 0.9994366765022278
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.014964622594416142
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03359664976596832,
        "mean_feedback_forward_cosine": 0.999759687531379,
        "mean_feedback_forward_relative_error": 0.021564522322507634,
        "min_feedback_forward_cosine": 0.9994367361068726
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.013909739240705966
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03359221667051315,
        "mean_feedback_forward_cosine": 0.9997597875133637,
        "mean_feedback_forward_relative_error": 0.021560262798541976,
        "min_feedback_forward_cosine": 0.9994370341300964
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.013652800155878067
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03358719125390053,
        "mean_feedback_forward_cosine": 0.9997598586543914,
        "mean_feedback_forward_relative_error": 0.021555697725665186,
        "min_feedback_forward_cosine": 0.999437153339386
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.014037335795909166
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033582653850317,
        "mean_feedback_forward_cosine": 0.9997599624818371,
        "mean_feedback_forward_relative_error": 0.021551487862222617,
        "min_feedback_forward_cosine": 0.9994373917579651
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.013184962499141693
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03357814997434616,
        "mean_feedback_forward_cosine": 0.9997600374683258,
        "mean_feedback_forward_relative_error": 0.021547226896209103,
        "min_feedback_forward_cosine": 0.9994373917579651
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.0119105725453794
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033573850989341736,
        "mean_feedback_forward_cosine": 0.9997601336048495,
        "mean_feedback_forward_relative_error": 0.021543031573415763,
        "min_feedback_forward_cosine": 0.9994375705718994
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.012246701563745736
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03356962278485298,
        "mean_feedback_forward_cosine": 0.9997602085913381,
        "mean_feedback_forward_relative_error": 0.021538912513924223,
        "min_feedback_forward_cosine": 0.9994378685951233
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.012045851690918208
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03356557339429855,
        "mean_feedback_forward_cosine": 0.9997602720414439,
        "mean_feedback_forward_relative_error": 0.02153489657587582,
        "min_feedback_forward_cosine": 0.9994379281997681
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.011227504568099976
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03356126323342323,
        "mean_feedback_forward_cosine": 0.99976033741428,
        "mean_feedback_forward_relative_error": 0.02153074730848593,
        "min_feedback_forward_cosine": 0.9994380474090576
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.011131018073707819
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03355677053332329,
        "mean_feedback_forward_cosine": 0.9997604354735343,
        "mean_feedback_forward_relative_error": 0.02152619996078072,
        "min_feedback_forward_cosine": 0.9994382262229919
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.011638076515942812
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033552322536706924,
        "mean_feedback_forward_cosine": 0.9997605604510154,
        "mean_feedback_forward_relative_error": 0.0215218830553274,
        "min_feedback_forward_cosine": 0.9994382858276367
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.010579599971473216
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03354793041944504,
        "mean_feedback_forward_cosine": 0.9997606412056954,
        "mean_feedback_forward_relative_error": 0.021517586593906724,
        "min_feedback_forward_cosine": 0.999438464641571
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.01081768324844539
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03354306146502495,
        "mean_feedback_forward_cosine": 0.9997607065785316,
        "mean_feedback_forward_relative_error": 0.02151310071349144,
        "min_feedback_forward_cosine": 0.9994386434555054
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.010473076796680688
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03353889286518097,
        "mean_feedback_forward_cosine": 0.9997607854104811,
        "mean_feedback_forward_relative_error": 0.021509090183122505,
        "min_feedback_forward_cosine": 0.9994387626647949
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.009850473795235157
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03353442624211311,
        "mean_feedback_forward_cosine": 0.999760877701544,
        "mean_feedback_forward_relative_error": 0.021504730932534702,
        "min_feedback_forward_cosine": 0.9994389414787292
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.010320796567499638
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352973610162735,
        "mean_feedback_forward_cosine": 0.9997609488425716,
        "mean_feedback_forward_relative_error": 0.021500151529307326,
        "min_feedback_forward_cosine": 0.9994391202926636
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.01045252704411745
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0335252583026886,
        "mean_feedback_forward_cosine": 0.9997610526700174,
        "mean_feedback_forward_relative_error": 0.021495793149956772,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.009837141685783863
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352520614862442,
        "mean_feedback_forward_cosine": 0.9997610411336345,
        "mean_feedback_forward_relative_error": 0.02149569653275032,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.009407439221739769
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352515771985054,
        "mean_feedback_forward_cosine": 0.9997610507472869,
        "mean_feedback_forward_relative_error": 0.021495603745983492,
        "min_feedback_forward_cosine": 0.9994391202926636
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.009692396873384714
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352513164281845,
        "mean_feedback_forward_cosine": 0.9997610776655136,
        "mean_feedback_forward_relative_error": 0.02149552321662345,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.00818301939830184
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033525098115205765,
        "mean_feedback_forward_cosine": 0.9997610372881736,
        "mean_feedback_forward_relative_error": 0.021495435477024125,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.007863032380789519
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033525072038173676,
        "mean_feedback_forward_cosine": 0.9997610680518612,
        "mean_feedback_forward_relative_error": 0.02149535407642684,
        "min_feedback_forward_cosine": 0.9994391202926636
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.007916177957579493
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352503851056099,
        "mean_feedback_forward_cosine": 0.9997610680518612,
        "mean_feedback_forward_relative_error": 0.021495273757365443,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.007483635274618864
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0335250049829483,
        "mean_feedback_forward_cosine": 0.9997610584382088,
        "mean_feedback_forward_relative_error": 0.02149519364860269,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.007406631565839052
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352498635649681,
        "mean_feedback_forward_cosine": 0.9997610526700174,
        "mean_feedback_forward_relative_error": 0.02149511546257042,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.006852073890566826
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352494537830353,
        "mean_feedback_forward_cosine": 0.999761079588244,
        "mean_feedback_forward_relative_error": 0.021495032935373245,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.007527244059741497
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352490812540054,
        "mean_feedback_forward_cosine": 0.9997610565154783,
        "mean_feedback_forward_relative_error": 0.02149495154979729,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.00713704448774457
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524878323078156,
        "mean_feedback_forward_cosine": 0.9997610680518612,
        "mean_feedback_forward_relative_error": 0.021494870119157336,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.0073489983118325474
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352484479546547,
        "mean_feedback_forward_cosine": 0.9997610815109745,
        "mean_feedback_forward_relative_error": 0.021494792654148994,
        "min_feedback_forward_cosine": 0.9994391798973083
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.006711921958327293
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352481499314308,
        "mean_feedback_forward_cosine": 0.9997610661291307,
        "mean_feedback_forward_relative_error": 0.0214947137020288,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.006963134993463755
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352478891611099,
        "mean_feedback_forward_cosine": 0.9997610776655136,
        "mean_feedback_forward_relative_error": 0.02149463832498558,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.006833728794082999
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524755388498306,
        "mean_feedback_forward_cosine": 0.9997610622836698,
        "mean_feedback_forward_relative_error": 0.02149456064967859,
        "min_feedback_forward_cosine": 0.9994393587112427
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.0069753591714054345
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352471813559532,
        "mean_feedback_forward_cosine": 0.9997610872791659,
        "mean_feedback_forward_relative_error": 0.02149448093147047,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.006809330035448075
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352469205856323,
        "mean_feedback_forward_cosine": 0.9997610642064002,
        "mean_feedback_forward_relative_error": 0.021494409384866877,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.006547005600109696
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524662256240845,
        "mean_feedback_forward_cosine": 0.9997610776655136,
        "mean_feedback_forward_relative_error": 0.021494334443442283,
        "min_feedback_forward_cosine": 0.9994392991065979
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.006588688519597053
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352462872862816,
        "mean_feedback_forward_cosine": 0.9997610603609393,
        "mean_feedback_forward_relative_error": 0.021494263437606635,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.0064581088467687365
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352459892630577,
        "mean_feedback_forward_cosine": 0.9997610776655136,
        "mean_feedback_forward_relative_error": 0.021494189517632607,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.006974443473666906
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352456912398338,
        "mean_feedback_forward_cosine": 0.9997610815109745,
        "mean_feedback_forward_relative_error": 0.02149411502684797,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.006474740914031863
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524543046951294,
        "mean_feedback_forward_cosine": 0.999761079588244,
        "mean_feedback_forward_relative_error": 0.021494041302151257,
        "min_feedback_forward_cosine": 0.9994391798973083
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.005849080839082599
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0335245244204998,
        "mean_feedback_forward_cosine": 0.9997610872791659,
        "mean_feedback_forward_relative_error": 0.021493971332787506,
        "min_feedback_forward_cosine": 0.9994393587112427
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.005923768717944622
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352448716759682,
        "mean_feedback_forward_cosine": 0.9997610680518612,
        "mean_feedback_forward_relative_error": 0.02149389556518966,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.006921178120821715
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352445363998413,
        "mean_feedback_forward_cosine": 0.9997610699745917,
        "mean_feedback_forward_relative_error": 0.02149382048857308,
        "min_feedback_forward_cosine": 0.9994391798973083
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.006746948901340365
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352442756295204,
        "mean_feedback_forward_cosine": 0.9997610757427831,
        "mean_feedback_forward_relative_error": 0.02149374701923901,
        "min_feedback_forward_cosine": 0.9994391798973083
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.006347868871837854
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524394035339355,
        "mean_feedback_forward_cosine": 0.999761079588244,
        "mean_feedback_forward_relative_error": 0.02149367281385968,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.006598144030719995
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352436423301697,
        "mean_feedback_forward_cosine": 0.9997610815109745,
        "mean_feedback_forward_relative_error": 0.021493595514085986,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.0065184420828521255
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352434188127518,
        "mean_feedback_forward_cosine": 0.9997610853564355,
        "mean_feedback_forward_relative_error": 0.021493520813002702,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.005984624702259898
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352431207895279,
        "mean_feedback_forward_cosine": 0.9997610988155488,
        "mean_feedback_forward_relative_error": 0.02149344841018319,
        "min_feedback_forward_cosine": 0.9994391798973083
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.006538475824892521
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0335242785513401,
        "mean_feedback_forward_cosine": 0.9997610930473574,
        "mean_feedback_forward_relative_error": 0.02149337551165973,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.0066132744911313055
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524252474308014,
        "mean_feedback_forward_cosine": 0.9997610949700878,
        "mean_feedback_forward_relative_error": 0.0214933046560374,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.006083347657397389
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524226397275925,
        "mean_feedback_forward_cosine": 0.9997610988155488,
        "mean_feedback_forward_relative_error": 0.021493231066532673,
        "min_feedback_forward_cosine": 0.9994392991065979
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.006523385307714343
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352418914437294,
        "mean_feedback_forward_cosine": 0.999761083433705,
        "mean_feedback_forward_relative_error": 0.02149315140840988,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.006543258281052113
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352416306734085,
        "mean_feedback_forward_cosine": 0.9997610872791659,
        "mean_feedback_forward_relative_error": 0.021493081318875475,
        "min_feedback_forward_cosine": 0.9994392991065979
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.005880762688852847
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352414444088936,
        "mean_feedback_forward_cosine": 0.9997610853564355,
        "mean_feedback_forward_relative_error": 0.021493008180010702,
        "min_feedback_forward_cosine": 0.9994391798973083
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.006099545330628752
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352411836385727,
        "mean_feedback_forward_cosine": 0.9997610949700878,
        "mean_feedback_forward_relative_error": 0.021492930985386332,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.006418848850876093
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352409228682518,
        "mean_feedback_forward_cosine": 0.9997610968928183,
        "mean_feedback_forward_relative_error": 0.021492855668428443,
        "min_feedback_forward_cosine": 0.9994393587112427
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.0057901674166321755
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352406248450279,
        "mean_feedback_forward_cosine": 0.9997610872791659,
        "mean_feedback_forward_relative_error": 0.021492780261342565,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.0058912814247608185
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033524032682180405,
        "mean_feedback_forward_cosine": 0.9997611065064708,
        "mean_feedback_forward_relative_error": 0.021492710802704096,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.006571254019960761
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352399542927742,
        "mean_feedback_forward_cosine": 0.9997610930473574,
        "mean_feedback_forward_relative_error": 0.021492637423498016,
        "min_feedback_forward_cosine": 0.9994391798973083
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.006014063582122326
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352397680282593,
        "mean_feedback_forward_cosine": 0.9997610892018964,
        "mean_feedback_forward_relative_error": 0.021492564209526586,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.006366922727227211
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352394700050354,
        "mean_feedback_forward_cosine": 0.9997610872791659,
        "mean_feedback_forward_relative_error": 0.021492496373191956,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.006320757586359978
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352392092347145,
        "mean_feedback_forward_cosine": 0.9997610930473574,
        "mean_feedback_forward_relative_error": 0.021492425412420305,
        "min_feedback_forward_cosine": 0.9994392991065979
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.005957902250476182
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352389484643936,
        "mean_feedback_forward_cosine": 0.9997611045837402,
        "mean_feedback_forward_relative_error": 0.02149235218342754,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.006235929216071964
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352387621998787,
        "mean_feedback_forward_cosine": 0.9997610911246269,
        "mean_feedback_forward_relative_error": 0.021492283460834333,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.005720531714558602
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352384641766548,
        "mean_feedback_forward_cosine": 0.9997610892018964,
        "mean_feedback_forward_relative_error": 0.021492208384217754,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.0063629729913175104
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352382034063339,
        "mean_feedback_forward_cosine": 0.9997611045837402,
        "mean_feedback_forward_relative_error": 0.021492136056504903,
        "min_feedback_forward_cosine": 0.9994392991065979
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.0061088988988101485
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.033523790538311005,
        "mean_feedback_forward_cosine": 0.9997610930473574,
        "mean_feedback_forward_relative_error": 0.021492065621479865,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.006140658489614725
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03352376073598862,
        "mean_feedback_forward_cosine": 0.9997610988155488,
        "mean_feedback_forward_relative_error": 0.021491994029812275,
        "min_feedback_forward_cosine": 0.9994392395019531
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.006112837376818061
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9258,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.3920034271240234,
    "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": 2525847040,
    "peak_memory_reserved_bytes": 3393191936,
    "torch_version": "2.3.1+cu118"
  },
  "protocol_family": "oral_a_cifar_local_resnet_development",
  "provenance": {
    "git_commit": "8f9364e8ee2e2a19eb83b58452fa0e8cb155353c",
    "git_tracked_dirty": false
  },
  "schema_version": 1,
  "split": {
    "cifar_source_files": [
      {
        "bytes": 31035704,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_1",
        "sha256": "54636561a3ce25bd3e19253c6b0d8538147b0ae398331ac4a2d86c6d987368cd"
      },
      {
        "bytes": 31035320,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_2",
        "sha256": "766b2cef9fbc745cf056b3152224f7cf77163b330ea9a15f9392beb8b89bc5a8"
      },
      {
        "bytes": 31035999,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_3",
        "sha256": "0f00d98ebfb30b3ec0ad19f9756dc2630b89003e10525f5e148445e82aa6a1f9"
      },
      {
        "bytes": 31035696,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_4",
        "sha256": "3f7bb240661948b8f4d53e36ec720d8306f5668bd0071dcb4e6c947f78e9682b"
      },
      {
        "bytes": 31035623,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/data_batch_5",
        "sha256": "d91802434d8376bbaeeadf58a737e3a1b12ac839077e931237e0dcd43adcb154"
      },
      {
        "bytes": 31035526,
        "path": "/home/yurenh2/sdrn/data/cifar-10-batches-py/test_batch",
        "sha256": "f53d8d457504f7cff4ea9e021afcf0e0ad8e24a91f3fc42091b8adef61157831"
      }
    ],
    "dataset": "cifar10",
    "evaluation_split": "test",
    "input_layout": "NCHW",
    "input_shape": [
      3,
      32,
      32
    ],
    "loader_seed": 14,
    "normalization_mean": [
      0.49140000343322754,
      0.4821999967098236,
      0.4465000033378601
    ],
    "normalization_std": [
      0.24699999392032623,
      0.2434999942779541,
      0.26159998774528503
    ],
    "split_from_training_only": true,
    "split_seed": null,
    "test_examples": 10000,
    "train_examples": 50000,
    "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5",
    "validation_class_counts": {},
    "validation_examples": 0,
    "validation_index_sha256": null
  },
  "timing": {
    "apical_warmup_wall_s": 0.0,
    "evaluation_wall_s": 1.1602163314819336,
    "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": 10196.078639745712,
    "train_wall_s": 10192.868887662888
  },
  "work": {
    "apical_macs_per_example": 68420224,
    "causal_scalar_observations": 0,
    "components": {
      "apical_projection_macs": 684202240000000,
      "apical_regression_macs": 0,
      "bp_reverse_macs_estimate": 0,
      "kp_reciprocal_correlation_macs": 684202240000000,
      "local_weight_correlation_macs": 688625920000000,
      "mirror_feedback_prediction_macs": 0,
      "mirror_local_correlation_macs": 0,
      "mirror_response_macs": 0,
      "ordinary_forward_macs": 688625920000000,
      "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": 68862592,
    "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": 2745656320000000
  }
}