{
  "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": 10,
    "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_s10.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": 10,
    "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": 0.9996285140514374,
    "feedback_forward_cosine": [
      0.9996448755264282,
      0.9996241927146912,
      0.9996662139892578,
      0.9996505379676819,
      0.9997282028198242,
      0.9997563362121582,
      0.9997356534004211,
      0.9996742010116577,
      0.9997925162315369,
      0.9997165203094482,
      0.9997353553771973,
      0.9998100996017456,
      0.9997087717056274,
      0.9996922016143799,
      0.999758243560791,
      0.9997119903564453,
      0.9997909665107727,
      0.9997221231460571,
      0.9998180866241455,
      0.999744713306427,
      0.9998014569282532,
      0.9998766183853149,
      0.9998487830162048,
      0.9998272657394409,
      0.9998286962509155,
      0.9997783899307251,
      0.9998400211334229,
      0.9997010231018066,
      0.9996979832649231,
      0.9994108080863953,
      0.9999839067459106
    ],
    "feedback_forward_norm_ratio": [
      1.0001192092895508,
      1.0009658336639404,
      0.9994824528694153,
      1.0007823705673218,
      0.9986081719398499,
      1.0005202293395996,
      0.9998824000358582,
      1.0015058517456055,
      0.9998427033424377,
      1.0006579160690308,
      0.9985700249671936,
      1.0006816387176514,
      0.9992146492004395,
      1.0012266635894775,
      0.9997763633728027,
      1.0012537240982056,
      0.9999839663505554,
      1.0013760328292847,
      0.9994275569915771,
      1.0008691549301147,
      0.9993569254875183,
      1.0003502368927002,
      0.9998308420181274,
      1.0005744695663452,
      0.9997363090515137,
      1.0010242462158203,
      0.9999058842658997,
      1.0010095834732056,
      0.9994451999664307,
      1.0014487504959106,
      1.0020092725753784
    ],
    "feedback_forward_relative_error": [
      0.026650002226233482,
      0.027447007596492767,
      0.025834619998931885,
      0.0264601930975914,
      0.02333914116024971,
      0.02209114469587803,
      0.02299085073173046,
      0.02558811567723751,
      0.020373323932290077,
      0.023823890835046768,
      0.023036517202854156,
      0.019504692405462265,
      0.02413417026400566,
      0.02485603094100952,
      0.02198750525712967,
      0.024049103260040283,
      0.020448429509997368,
      0.02362847700715065,
      0.019075661897659302,
      0.022619416937232018,
      0.01992790214717388,
      0.01571011357009411,
      0.017386458814144135,
      0.018599554896354675,
      0.018510015681385994,
      0.021088147535920143,
      0.017884954810142517,
      0.024482930079102516,
      0.024578392505645752,
      0.034382302314043045,
      0.006041166838258505
    ],
    "innovation_negative_gradient_cosine": [
      0.9996632933616638,
      0.9995239973068237,
      0.9996660947799683,
      0.9995747804641724,
      0.9996673464775085,
      0.9996167421340942,
      0.9996652007102966,
      0.9995874166488647,
      0.9996679425239563,
      0.9996523261070251,
      0.9996814727783203,
      0.9996957778930664,
      0.9996981620788574,
      0.9996000528335571,
      0.9997052550315857,
      0.9996254444122314,
      0.9997178316116333,
      0.9996660947799683,
      0.9997333884239197,
      0.9997081756591797,
      0.9997397065162659,
      0.9997773766517639,
      0.9997949004173279,
      0.9997736215591431,
      0.9997889995574951,
      0.99980628490448,
      0.9998396635055542,
      0.9998664259910583,
      0.9998698234558105,
      0.9999217987060547,
      0.7832923531532288
    ],
    "mean_feedback_forward_cosine": 0.9997444114377422,
    "mean_feedback_forward_relative_error": 0.02214613657504801,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.9996632933616638,
      0.9995239973068237,
      0.9996660947799683,
      0.9995747804641724,
      0.9996673464775085,
      0.9996167421340942,
      0.9996652007102966,
      0.9995874166488647,
      0.9996679425239563,
      0.9996523261070251,
      0.9996814727783203,
      0.9996957778930664,
      0.9996981620788574,
      0.9996000528335571,
      0.9997052550315857,
      0.9996254444122314,
      0.9997178316116333,
      0.9996660947799683,
      0.9997333884239197,
      0.9997081756591797,
      0.9997397065162659,
      0.9997773766517639,
      0.9997949004173279,
      0.9997736215591431,
      0.9997889995574951,
      0.99980628490448,
      0.9998396635055542,
      0.9998664259910583,
      0.9998698234558105,
      0.9999217987060547,
      0.7832923531532288
    ],
    "teaching_negative_gradient_cosine": [
      0.9996632933616638,
      0.9995239973068237,
      0.9996660947799683,
      0.9995747804641724,
      0.9996673464775085,
      0.9996167421340942,
      0.9996652007102966,
      0.9995874166488647,
      0.9996679425239563,
      0.9996523261070251,
      0.9996814727783203,
      0.9996957778930664,
      0.9996981620788574,
      0.9996000528335571,
      0.9997052550315857,
      0.9996254444122314,
      0.9997178316116333,
      0.9996660947799683,
      0.9997333884239197,
      0.9997081756591797,
      0.9997397065162659,
      0.9997773766517639,
      0.9997949004173279,
      0.9997736215591431,
      0.9997889995574951,
      0.99980628490448,
      0.9998396635055542,
      0.9998664259910583,
      0.9998698234558105,
      0.9999217987060547,
      0.7832923531532288
    ],
    "wall_s": 0.13054966926574707
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3585952520370483,
        "mean_feedback_forward_cosine": 0.3046831685689188,
        "mean_feedback_forward_relative_error": 1.1702621463806397,
        "min_feedback_forward_cosine": 0.08090806752443314
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 2.1066284007263185
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3304003477096558,
        "mean_feedback_forward_cosine": 0.36645406773013456,
        "mean_feedback_forward_relative_error": 1.1181111816437013,
        "min_feedback_forward_cosine": 0.11908140778541565
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 1.508727656364441
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.299444317817688,
        "mean_feedback_forward_cosine": 0.4346453778205379,
        "mean_feedback_forward_relative_error": 1.057305653249064,
        "min_feedback_forward_cosine": 0.16057761013507843
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.2519019562149047
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2654600143432617,
        "mean_feedback_forward_cosine": 0.4959667310599358,
        "mean_feedback_forward_relative_error": 0.9992632356382185,
        "min_feedback_forward_cosine": 0.20524783432483673
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 1.0251145127868653
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2304916381835938,
        "mean_feedback_forward_cosine": 0.5455621259827768,
        "mean_feedback_forward_relative_error": 0.9495349939792387,
        "min_feedback_forward_cosine": 0.2493714690208435
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 0.8638715417480469
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.199565052986145,
        "mean_feedback_forward_cosine": 0.5886594960766454,
        "mean_feedback_forward_relative_error": 0.9039133075744875,
        "min_feedback_forward_cosine": 0.29433685541152954
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 0.7554565207672119
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.171512484550476,
        "mean_feedback_forward_cosine": 0.6259271387130984,
        "mean_feedback_forward_relative_error": 0.8622603945193752,
        "min_feedback_forward_cosine": 0.33613449335098267
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.6826584720993042
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1432757377624512,
        "mean_feedback_forward_cosine": 0.6577128892944705,
        "mean_feedback_forward_relative_error": 0.8246829057893446,
        "min_feedback_forward_cosine": 0.3702732026576996
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.6287693549728394
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1165492534637451,
        "mean_feedback_forward_cosine": 0.685499252811555,
        "mean_feedback_forward_relative_error": 0.7903105520432995,
        "min_feedback_forward_cosine": 0.40048497915267944
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.5804982432174682
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0891292095184326,
        "mean_feedback_forward_cosine": 0.7110802377423933,
        "mean_feedback_forward_relative_error": 0.7574431088662916,
        "min_feedback_forward_cosine": 0.43083637952804565
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.5510316325569152
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0645053386688232,
        "mean_feedback_forward_cosine": 0.7334197105899933,
        "mean_feedback_forward_relative_error": 0.7274134264838311,
        "min_feedback_forward_cosine": 0.45720577239990234
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.5122311639595032
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0398348569869995,
        "mean_feedback_forward_cosine": 0.753688100845583,
        "mean_feedback_forward_relative_error": 0.6989298068707989,
        "min_feedback_forward_cosine": 0.4827224612236023
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.49010992847442625
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0150221586227417,
        "mean_feedback_forward_cosine": 0.7723052020995848,
        "mean_feedback_forward_relative_error": 0.6717068211686227,
        "min_feedback_forward_cosine": 0.5074095726013184
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.4641452179527283
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9889848232269287,
        "mean_feedback_forward_cosine": 0.7893762050136444,
        "mean_feedback_forward_relative_error": 0.6456553873515898,
        "min_feedback_forward_cosine": 0.5324095487594604
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.4428069670677185
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9631386995315552,
        "mean_feedback_forward_cosine": 0.8049058818048046,
        "mean_feedback_forward_relative_error": 0.6210459845681344,
        "min_feedback_forward_cosine": 0.5565786957740784
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.4246428520202637
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.937900722026825,
        "mean_feedback_forward_cosine": 0.8192884499026883,
        "mean_feedback_forward_relative_error": 0.5973106839003102,
        "min_feedback_forward_cosine": 0.5796732902526855
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.4130655108165741
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9116066694259644,
        "mean_feedback_forward_cosine": 0.8326400441508139,
        "mean_feedback_forward_relative_error": 0.5743048070899902,
        "min_feedback_forward_cosine": 0.6022407412528992
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.39465926347732544
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8865914344787598,
        "mean_feedback_forward_cosine": 0.8449586822140601,
        "mean_feedback_forward_relative_error": 0.5525295075870329,
        "min_feedback_forward_cosine": 0.6241908073425293
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.38214210516929625
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8610144257545471,
        "mean_feedback_forward_cosine": 0.856250032301872,
        "mean_feedback_forward_relative_error": 0.5316746835747073,
        "min_feedback_forward_cosine": 0.6453561186790466
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.37219429008483884
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8360261917114258,
        "mean_feedback_forward_cosine": 0.8666744674405744,
        "mean_feedback_forward_relative_error": 0.5117127260854167,
        "min_feedback_forward_cosine": 0.6658006906509399
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.355763973941803
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8105968832969666,
        "mean_feedback_forward_cosine": 0.8765667311606868,
        "mean_feedback_forward_relative_error": 0.4920242529722952,
        "min_feedback_forward_cosine": 0.6857298016548157
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.3471485916996002
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7856993675231934,
        "mean_feedback_forward_cosine": 0.8855374013223956,
        "mean_feedback_forward_relative_error": 0.4736229320687632,
        "min_feedback_forward_cosine": 0.7045443654060364
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.3388717658805847
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.761393666267395,
        "mean_feedback_forward_cosine": 0.893738733183953,
        "mean_feedback_forward_relative_error": 0.45595797127293003,
        "min_feedback_forward_cosine": 0.7220780253410339
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.32842747619628904
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7373024821281433,
        "mean_feedback_forward_cosine": 0.9013767088613203,
        "mean_feedback_forward_relative_error": 0.4389302769014912,
        "min_feedback_forward_cosine": 0.7390848398208618
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.31618609956741334
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7132046818733215,
        "mean_feedback_forward_cosine": 0.9085880940960299,
        "mean_feedback_forward_relative_error": 0.4223930532893827,
        "min_feedback_forward_cosine": 0.7556705474853516
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.31259035013198855
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6883693337440491,
        "mean_feedback_forward_cosine": 0.9153294005701619,
        "mean_feedback_forward_relative_error": 0.4062719056683202,
        "min_feedback_forward_cosine": 0.7722120881080627
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.31096653012275693
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6657470464706421,
        "mean_feedback_forward_cosine": 0.9213306019383092,
        "mean_feedback_forward_relative_error": 0.3914064000691137,
        "min_feedback_forward_cosine": 0.7868608236312866
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.2952206775283813
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6425599455833435,
        "mean_feedback_forward_cosine": 0.9272541884453066,
        "mean_feedback_forward_relative_error": 0.3762246250144897,
        "min_feedback_forward_cosine": 0.8012421131134033
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.2912610262298584
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6199831366539001,
        "mean_feedback_forward_cosine": 0.9326016787559755,
        "mean_feedback_forward_relative_error": 0.36191003024578094,
        "min_feedback_forward_cosine": 0.8147948384284973
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.28817487318992613
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.597847044467926,
        "mean_feedback_forward_cosine": 0.9374953354558637,
        "mean_feedback_forward_relative_error": 0.3483095572840783,
        "min_feedback_forward_cosine": 0.8276832699775696
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.2810961621761322
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5762627124786377,
        "mean_feedback_forward_cosine": 0.9419867396354675,
        "mean_feedback_forward_relative_error": 0.3354184043503577,
        "min_feedback_forward_cosine": 0.8397367596626282
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.27481924999713897
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.554566502571106,
        "mean_feedback_forward_cosine": 0.9463227083606105,
        "mean_feedback_forward_relative_error": 0.32253597099934855,
        "min_feedback_forward_cosine": 0.8514574766159058
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.2712509515380859
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5340067744255066,
        "mean_feedback_forward_cosine": 0.9503983489928707,
        "mean_feedback_forward_relative_error": 0.3099248527038482,
        "min_feedback_forward_cosine": 0.8621289730072021
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.2722747924566269
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5158315300941467,
        "mean_feedback_forward_cosine": 0.9539481401443481,
        "mean_feedback_forward_relative_error": 0.2984740154877786,
        "min_feedback_forward_cosine": 0.8712555170059204
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.25798445239067075
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.496427059173584,
        "mean_feedback_forward_cosine": 0.9573978377926734,
        "mean_feedback_forward_relative_error": 0.28698549996460637,
        "min_feedback_forward_cosine": 0.8806583881378174
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.2554719933319092
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4773749113082886,
        "mean_feedback_forward_cosine": 0.9606026795602614,
        "mean_feedback_forward_relative_error": 0.27592927049244603,
        "min_feedback_forward_cosine": 0.8895782232284546
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.25545837335586546
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4594625234603882,
        "mean_feedback_forward_cosine": 0.9635426498228504,
        "mean_feedback_forward_relative_error": 0.26538106630886754,
        "min_feedback_forward_cosine": 0.8976370692253113
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.2484088640499115
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4422893226146698,
        "mean_feedback_forward_cosine": 0.9661906438489114,
        "mean_feedback_forward_relative_error": 0.25547523916729037,
        "min_feedback_forward_cosine": 0.9050342440605164
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.2440171889591217
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.42448484897613525,
        "mean_feedback_forward_cosine": 0.9687469697767689,
        "mean_feedback_forward_relative_error": 0.24559208870895446,
        "min_feedback_forward_cosine": 0.912443220615387
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.24323479767799377
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4080718159675598,
        "mean_feedback_forward_cosine": 0.9710481493703781,
        "mean_feedback_forward_relative_error": 0.23629709813864,
        "min_feedback_forward_cosine": 0.9190430641174316
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.24011920569419862
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3921796679496765,
        "mean_feedback_forward_cosine": 0.9732197542344371,
        "mean_feedback_forward_relative_error": 0.22721118720308428,
        "min_feedback_forward_cosine": 0.9250966906547546
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.2373897479915619
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3771727383136749,
        "mean_feedback_forward_cosine": 0.9752226568037464,
        "mean_feedback_forward_relative_error": 0.21850634558546927,
        "min_feedback_forward_cosine": 0.9306812882423401
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.23271506587028504
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.36220383644104004,
        "mean_feedback_forward_cosine": 0.9770623618556608,
        "mean_feedback_forward_relative_error": 0.21018394178921176,
        "min_feedback_forward_cosine": 0.9360086917877197
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.23216551981925965
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.34794294834136963,
        "mean_feedback_forward_cosine": 0.9787843784978313,
        "mean_feedback_forward_relative_error": 0.20208276856330135,
        "min_feedback_forward_cosine": 0.9409137964248657
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.23053674859046935
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.33451753854751587,
        "mean_feedback_forward_cosine": 0.9803864609810614,
        "mean_feedback_forward_relative_error": 0.1942895285544857,
        "min_feedback_forward_cosine": 0.9453545808792114
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.2259896070957184
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.32157811522483826,
        "mean_feedback_forward_cosine": 0.9818134846225861,
        "mean_feedback_forward_relative_error": 0.1870680811905092,
        "min_feedback_forward_cosine": 0.9494665265083313
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.2214436098384857
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3088374733924866,
        "mean_feedback_forward_cosine": 0.9831757333970839,
        "mean_feedback_forward_relative_error": 0.17989249154925346,
        "min_feedback_forward_cosine": 0.9533370733261108
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.21873262013435363
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.29687994718551636,
        "mean_feedback_forward_cosine": 0.9844553701339229,
        "mean_feedback_forward_relative_error": 0.17289376955839894,
        "min_feedback_forward_cosine": 0.9568438529968262
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.21603742006778717
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.28457173705101013,
        "mean_feedback_forward_cosine": 0.9856576804191836,
        "mean_feedback_forward_relative_error": 0.1660312604038946,
        "min_feedback_forward_cosine": 0.9602963924407959
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.21823587814807893
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.27344414591789246,
        "mean_feedback_forward_cosine": 0.9867185546505836,
        "mean_feedback_forward_relative_error": 0.1597499662349301,
        "min_feedback_forward_cosine": 0.9633457660675049
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.21540137950897217
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.26242485642433167,
        "mean_feedback_forward_cosine": 0.9877236908481967,
        "mean_feedback_forward_relative_error": 0.15357495339647417,
        "min_feedback_forward_cosine": 0.9662054777145386
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.21133698526382447
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.25207939743995667,
        "mean_feedback_forward_cosine": 0.9886252514777645,
        "mean_feedback_forward_relative_error": 0.14780485077250388,
        "min_feedback_forward_cosine": 0.9687892198562622
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.2037308853340149
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.24176177382469177,
        "mean_feedback_forward_cosine": 0.9894983691553916,
        "mean_feedback_forward_relative_error": 0.1420233254471133,
        "min_feedback_forward_cosine": 0.9712689518928528
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.21174614361763
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.23217424750328064,
        "mean_feedback_forward_cosine": 0.9902935451076876,
        "mean_feedback_forward_relative_error": 0.13653598281164323,
        "min_feedback_forward_cosine": 0.97348552942276
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.20778529538154603
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22279824316501617,
        "mean_feedback_forward_cosine": 0.9910120848686464,
        "mean_feedback_forward_relative_error": 0.13139018248165807,
        "min_feedback_forward_cosine": 0.9755691885948181
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.20543151748657226
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.21348775923252106,
        "mean_feedback_forward_cosine": 0.9916909317816457,
        "mean_feedback_forward_relative_error": 0.12633225069411338,
        "min_feedback_forward_cosine": 0.9775524139404297
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.20323132181167602
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20458975434303284,
        "mean_feedback_forward_cosine": 0.9923476730623553,
        "mean_feedback_forward_relative_error": 0.12124136775251358,
        "min_feedback_forward_cosine": 0.9793746471405029
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.20717589849948884
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19673946499824524,
        "mean_feedback_forward_cosine": 0.9928949167651515,
        "mean_feedback_forward_relative_error": 0.11681427746530502,
        "min_feedback_forward_cosine": 0.9809081554412842
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.19442318168640138
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18847723305225372,
        "mean_feedback_forward_cosine": 0.9934470076714793,
        "mean_feedback_forward_relative_error": 0.11220145477883277,
        "min_feedback_forward_cosine": 0.9824700355529785
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.2030546386051178
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18114864826202393,
        "mean_feedback_forward_cosine": 0.9939348832253487,
        "mean_feedback_forward_relative_error": 0.10793772243684338,
        "min_feedback_forward_cosine": 0.9837960600852966
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.19493575808525085
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17370857298374176,
        "mean_feedback_forward_cosine": 0.9943953079562033,
        "mean_feedback_forward_relative_error": 0.10375529179169286,
        "min_feedback_forward_cosine": 0.9850887060165405
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.19727207220554352
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.166924849152565,
        "mean_feedback_forward_cosine": 0.9948073298700394,
        "mean_feedback_forward_relative_error": 0.09986100598208365,
        "min_feedback_forward_cosine": 0.9862250089645386
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.18774492676734925
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16006633639335632,
        "mean_feedback_forward_cosine": 0.9952024394466031,
        "mean_feedback_forward_relative_error": 0.09599625006798775,
        "min_feedback_forward_cosine": 0.9873306751251221
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.19298322952270508
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15336787700653076,
        "mean_feedback_forward_cosine": 0.9955765701109364,
        "mean_feedback_forward_relative_error": 0.09218034304438098,
        "min_feedback_forward_cosine": 0.9883619546890259
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.19527318564414978
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1468193382024765,
        "mean_feedback_forward_cosine": 0.9959285663020226,
        "mean_feedback_forward_relative_error": 0.08844295378413893,
        "min_feedback_forward_cosine": 0.9893286824226379
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.1948135542345047
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14080898463726044,
        "mean_feedback_forward_cosine": 0.9962240880535495,
        "mean_feedback_forward_relative_error": 0.08517507415625357,
        "min_feedback_forward_cosine": 0.9901782274246216
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.1880095457839966
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1351540982723236,
        "mean_feedback_forward_cosine": 0.9965042529567596,
        "mean_feedback_forward_relative_error": 0.08195698453534034,
        "min_feedback_forward_cosine": 0.990946888923645
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.18653369692325592
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1296241581439972,
        "mean_feedback_forward_cosine": 0.9967624025960122,
        "mean_feedback_forward_relative_error": 0.07887501467860514,
        "min_feedback_forward_cosine": 0.9916706681251526
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.18662398671150207
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12416495382785797,
        "mean_feedback_forward_cosine": 0.9970199235023991,
        "mean_feedback_forward_relative_error": 0.07567659170637207,
        "min_feedback_forward_cosine": 0.9923562407493591
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.19312430601119995
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11924664676189423,
        "mean_feedback_forward_cosine": 0.9972401665103051,
        "mean_feedback_forward_relative_error": 0.07282290667776138,
        "min_feedback_forward_cosine": 0.9929476976394653
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.18571868690490723
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11436894536018372,
        "mean_feedback_forward_cosine": 0.9974491730813058,
        "mean_feedback_forward_relative_error": 0.07000911301903186,
        "min_feedback_forward_cosine": 0.9935097694396973
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.18454137994766234
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10968337208032608,
        "mean_feedback_forward_cosine": 0.9976482679766994,
        "mean_feedback_forward_relative_error": 0.06722739471062537,
        "min_feedback_forward_cosine": 0.9940293431282043
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.189456130194664
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10538037121295929,
        "mean_feedback_forward_cosine": 0.9978226711673122,
        "mean_feedback_forward_relative_error": 0.06468745921888659,
        "min_feedback_forward_cosine": 0.9944861531257629
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.18165870635032655
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10099851340055466,
        "mean_feedback_forward_cosine": 0.9979852380291108,
        "mean_feedback_forward_relative_error": 0.06222016592660258,
        "min_feedback_forward_cosine": 0.9949323534965515
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.17878232320785523
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09699861705303192,
        "mean_feedback_forward_cosine": 0.9981338477903797,
        "mean_feedback_forward_relative_error": 0.05988543576771213,
        "min_feedback_forward_cosine": 0.9953258037567139
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.17694817928314208
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09301231056451797,
        "mean_feedback_forward_cosine": 0.998277444993296,
        "mean_feedback_forward_relative_error": 0.057541606046499744,
        "min_feedback_forward_cosine": 0.9957007765769958
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.18061958540916442
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08925621211528778,
        "mean_feedback_forward_cosine": 0.9984065351947662,
        "mean_feedback_forward_relative_error": 0.05534977417799734,
        "min_feedback_forward_cosine": 0.9960404634475708
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.17779186904907227
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08562029153108597,
        "mean_feedback_forward_cosine": 0.9985263193807294,
        "mean_feedback_forward_relative_error": 0.05322876013815403,
        "min_feedback_forward_cosine": 0.9963556528091431
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.17973715352058411
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08209428936243057,
        "mean_feedback_forward_cosine": 0.9986398066243818,
        "mean_feedback_forward_relative_error": 0.051138524807268576,
        "min_feedback_forward_cosine": 0.9966488480567932
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.17942093858718872
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07883550971746445,
        "mean_feedback_forward_cosine": 0.998742280467864,
        "mean_feedback_forward_relative_error": 0.049174504955449415,
        "min_feedback_forward_cosine": 0.9969099760055542
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.1764600699710846
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07566260546445847,
        "mean_feedback_forward_cosine": 0.9988359059056928,
        "mean_feedback_forward_relative_error": 0.04730717038675662,
        "min_feedback_forward_cosine": 0.99715256690979
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.17309480888843537
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07258174568414688,
        "mean_feedback_forward_cosine": 0.9989236785519507,
        "mean_feedback_forward_relative_error": 0.04549225762246117,
        "min_feedback_forward_cosine": 0.997380256652832
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.17379588204622268
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06965117156505585,
        "mean_feedback_forward_cosine": 0.9990047389461149,
        "mean_feedback_forward_relative_error": 0.043746734849147255,
        "min_feedback_forward_cosine": 0.9975866079330444
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.17552788778781891
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06684258580207825,
        "mean_feedback_forward_cosine": 0.9990783602960648,
        "mean_feedback_forward_relative_error": 0.04210145669358392,
        "min_feedback_forward_cosine": 0.997776985168457
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.1720303683567047
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06409264355897903,
        "mean_feedback_forward_cosine": 0.9991503915479106,
        "mean_feedback_forward_relative_error": 0.040425917464158224,
        "min_feedback_forward_cosine": 0.9979556202888489
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.1756158047270775
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0615408718585968,
        "mean_feedback_forward_cosine": 0.9992139647083897,
        "mean_feedback_forward_relative_error": 0.0388862966229358,
        "min_feedback_forward_cosine": 0.9981147646903992
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.16958078844070434
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05904845520853996,
        "mean_feedback_forward_cosine": 0.99927399235387,
        "mean_feedback_forward_relative_error": 0.03737387319485987,
        "min_feedback_forward_cosine": 0.9982644319534302
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.1693227526807785
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05658656731247902,
        "mean_feedback_forward_cosine": 0.9993299745744274,
        "mean_feedback_forward_relative_error": 0.03590853398125018,
        "min_feedback_forward_cosine": 0.9984058737754822
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.17388515575170516
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05434687063097954,
        "mean_feedback_forward_cosine": 0.9993810269140428,
        "mean_feedback_forward_relative_error": 0.03451352797809147,
        "min_feedback_forward_cosine": 0.9985294938087463
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.16667771428108216
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05209857597947121,
        "mean_feedback_forward_cosine": 0.9994277108100152,
        "mean_feedback_forward_relative_error": 0.03318805929513708,
        "min_feedback_forward_cosine": 0.9986482858657837
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.16874846799850463
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04997039586305618,
        "mean_feedback_forward_cosine": 0.9994707511317346,
        "mean_feedback_forward_relative_error": 0.03191607415435776,
        "min_feedback_forward_cosine": 0.9987563490867615
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.17165495090007782
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04805481806397438,
        "mean_feedback_forward_cosine": 0.9995101863338102,
        "mean_feedback_forward_relative_error": 0.030704457402950334,
        "min_feedback_forward_cosine": 0.9988495111465454
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.16439861055374144
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04608337581157684,
        "mean_feedback_forward_cosine": 0.9995480871969654,
        "mean_feedback_forward_relative_error": 0.029492572520769412,
        "min_feedback_forward_cosine": 0.9989418983459473
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.16936889218330384
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.044296909123659134,
        "mean_feedback_forward_cosine": 0.9995824367769303,
        "mean_feedback_forward_relative_error": 0.028349369043304075,
        "min_feedback_forward_cosine": 0.9990223050117493
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.1659735676765442
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04257083311676979,
        "mean_feedback_forward_cosine": 0.9996139830158602,
        "mean_feedback_forward_relative_error": 0.027255154545268705,
        "min_feedback_forward_cosine": 0.9990971088409424
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.16362881194591522
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04086729884147644,
        "mean_feedback_forward_cosine": 0.9996434719331803,
        "mean_feedback_forward_relative_error": 0.026196662096246597,
        "min_feedback_forward_cosine": 0.9991677403450012
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.1682988141012192
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039153311401605606,
        "mean_feedback_forward_cosine": 0.9996712015521142,
        "mean_feedback_forward_relative_error": 0.025161175898486567,
        "min_feedback_forward_cosine": 0.999235987663269
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.16988255014419557
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0376581996679306,
        "mean_feedback_forward_cosine": 0.9996954664107291,
        "mean_feedback_forward_relative_error": 0.024214518917424065,
        "min_feedback_forward_cosine": 0.9992931485176086
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.16026985026359558
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.036105673760175705,
        "mean_feedback_forward_cosine": 0.999718306526061,
        "mean_feedback_forward_relative_error": 0.023289367586614623,
        "min_feedback_forward_cosine": 0.9993501901626587
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.16757982426166534
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034648407250642776,
        "mean_feedback_forward_cosine": 0.9997396526798126,
        "mean_feedback_forward_relative_error": 0.022391358210194494,
        "min_feedback_forward_cosine": 0.9994015693664551
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.16521677165985107
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034636080265045166,
        "mean_feedback_forward_cosine": 0.9997398238028249,
        "mean_feedback_forward_relative_error": 0.022381956839273052,
        "min_feedback_forward_cosine": 0.9994021654129028
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.09520915712833404
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0346299484372139,
        "mean_feedback_forward_cosine": 0.9997399507030365,
        "mean_feedback_forward_relative_error": 0.022375703503888462,
        "min_feedback_forward_cosine": 0.9994022846221924
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.06890638122200966
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034624312072992325,
        "mean_feedback_forward_cosine": 0.9997400160758726,
        "mean_feedback_forward_relative_error": 0.0223699755996706,
        "min_feedback_forward_cosine": 0.9994024634361267
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.06154520476341248
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03461897745728493,
        "mean_feedback_forward_cosine": 0.9997401256715098,
        "mean_feedback_forward_relative_error": 0.02236448127716299,
        "min_feedback_forward_cosine": 0.9994025230407715
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.05448996238946915
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03461356833577156,
        "mean_feedback_forward_cosine": 0.9997402468035298,
        "mean_feedback_forward_relative_error": 0.022359185491598422,
        "min_feedback_forward_cosine": 0.9994028210639954
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.04973544316411018
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034608371555805206,
        "mean_feedback_forward_cosine": 0.9997403102536355,
        "mean_feedback_forward_relative_error": 0.022353943377252546,
        "min_feedback_forward_cosine": 0.9994028806686401
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.045988887414932254
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034603506326675415,
        "mean_feedback_forward_cosine": 0.999740389085585,
        "mean_feedback_forward_relative_error": 0.02234901749198475,
        "min_feedback_forward_cosine": 0.9994032979011536
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.04122027599394321
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034598417580127716,
        "mean_feedback_forward_cosine": 0.9997404756084565,
        "mean_feedback_forward_relative_error": 0.022344095572348562,
        "min_feedback_forward_cosine": 0.9994032382965088
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.03964277899980545
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034593500196933746,
        "mean_feedback_forward_cosine": 0.9997405717449803,
        "mean_feedback_forward_relative_error": 0.022339207615943685,
        "min_feedback_forward_cosine": 0.9994034767150879
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.03581520953893662
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03458848223090172,
        "mean_feedback_forward_cosine": 0.999740683263348,
        "mean_feedback_forward_relative_error": 0.022334380976615414,
        "min_feedback_forward_cosine": 0.9994037747383118
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.0351124484539032
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03458338975906372,
        "mean_feedback_forward_cosine": 0.9997407447907233,
        "mean_feedback_forward_relative_error": 0.022329576583879608,
        "min_feedback_forward_cosine": 0.9994038343429565
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.034428428021669386
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034578461199998856,
        "mean_feedback_forward_cosine": 0.9997408543863604,
        "mean_feedback_forward_relative_error": 0.022324893022737196,
        "min_feedback_forward_cosine": 0.9994041323661804
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.03098803948402405
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03457332402467728,
        "mean_feedback_forward_cosine": 0.9997409409092318,
        "mean_feedback_forward_relative_error": 0.022320216416471427,
        "min_feedback_forward_cosine": 0.9994043111801147
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.029546013591885568
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034568533301353455,
        "mean_feedback_forward_cosine": 0.9997410428139472,
        "mean_feedback_forward_relative_error": 0.022315731632613366,
        "min_feedback_forward_cosine": 0.9994044303894043
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.028411184762120248
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03456365689635277,
        "mean_feedback_forward_cosine": 0.9997411178004357,
        "mean_feedback_forward_relative_error": 0.022311257078282295,
        "min_feedback_forward_cosine": 0.9994045495986938
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.02664868856549263
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03455884009599686,
        "mean_feedback_forward_cosine": 0.9997411927869243,
        "mean_feedback_forward_relative_error": 0.022306785378004273,
        "min_feedback_forward_cosine": 0.9994048476219177
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.026244018102288245
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03455414995551109,
        "mean_feedback_forward_cosine": 0.9997412850779872,
        "mean_feedback_forward_relative_error": 0.022302366192302397,
        "min_feedback_forward_cosine": 0.9994049072265625
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.02401714214324951
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03454875573515892,
        "mean_feedback_forward_cosine": 0.999741373523589,
        "mean_feedback_forward_relative_error": 0.022297664245049798,
        "min_feedback_forward_cosine": 0.9994050860404968
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.025217011668682097
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03454356640577316,
        "mean_feedback_forward_cosine": 0.9997414523555387,
        "mean_feedback_forward_relative_error": 0.02229296860675658,
        "min_feedback_forward_cosine": 0.9994050860404968
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.0238461730080843
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034538786858320236,
        "mean_feedback_forward_cosine": 0.999741550414793,
        "mean_feedback_forward_relative_error": 0.022288713769446456,
        "min_feedback_forward_cosine": 0.9994054436683655
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.021576540098786356
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03453381359577179,
        "mean_feedback_forward_cosine": 0.9997416157876292,
        "mean_feedback_forward_relative_error": 0.02228425509266315,
        "min_feedback_forward_cosine": 0.9994056224822998
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.022200865411758423
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034529585391283035,
        "mean_feedback_forward_cosine": 0.9997417292287273,
        "mean_feedback_forward_relative_error": 0.022280036637018764,
        "min_feedback_forward_cosine": 0.9994057416915894
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.019916686898469926
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03452460095286369,
        "mean_feedback_forward_cosine": 0.9997417946015635,
        "mean_feedback_forward_relative_error": 0.022275654343707908,
        "min_feedback_forward_cosine": 0.9994059801101685
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.01987594199478626
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03451919928193092,
        "mean_feedback_forward_cosine": 0.9997418984290092,
        "mean_feedback_forward_relative_error": 0.022271059047911438,
        "min_feedback_forward_cosine": 0.9994060397148132
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.020289315200299026
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034514255821704865,
        "mean_feedback_forward_cosine": 0.9997419599563845,
        "mean_feedback_forward_relative_error": 0.02226658763303872,
        "min_feedback_forward_cosine": 0.9994062781333923
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.0189817143394053
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034508705139160156,
        "mean_feedback_forward_cosine": 0.9997420676292912,
        "mean_feedback_forward_relative_error": 0.022261766972200524,
        "min_feedback_forward_cosine": 0.9994063973426819
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.020050239784419537
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03450360521674156,
        "mean_feedback_forward_cosine": 0.9997421733794674,
        "mean_feedback_forward_relative_error": 0.022257326411143426,
        "min_feedback_forward_cosine": 0.9994065761566162
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.01772526730298996
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03449811786413193,
        "mean_feedback_forward_cosine": 0.9997422695159912,
        "mean_feedback_forward_relative_error": 0.0222526325153247,
        "min_feedback_forward_cosine": 0.9994068741798401
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.01855657254755497
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03449344262480736,
        "mean_feedback_forward_cosine": 0.9997423541161322,
        "mean_feedback_forward_relative_error": 0.022248169557461814,
        "min_feedback_forward_cosine": 0.9994069337844849
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.01728311818420887
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03448819741606712,
        "mean_feedback_forward_cosine": 0.9997424502526561,
        "mean_feedback_forward_relative_error": 0.022243508783679818,
        "min_feedback_forward_cosine": 0.999407172203064
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.017123812686502933
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03448304161429405,
        "mean_feedback_forward_cosine": 0.9997425502346408,
        "mean_feedback_forward_relative_error": 0.022238834580827142,
        "min_feedback_forward_cosine": 0.9994072914123535
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.01670280195534229
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034477971494197845,
        "mean_feedback_forward_cosine": 0.9997426271438599,
        "mean_feedback_forward_relative_error": 0.02223430791749589,
        "min_feedback_forward_cosine": 0.9994074702262878
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.01564242881298065
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03447258844971657,
        "mean_feedback_forward_cosine": 0.999742736739497,
        "mean_feedback_forward_relative_error": 0.022229604843643405,
        "min_feedback_forward_cosine": 0.9994077682495117
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.015944313647747038
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034467972815036774,
        "mean_feedback_forward_cosine": 0.9997428155714466,
        "mean_feedback_forward_relative_error": 0.022225436378991412,
        "min_feedback_forward_cosine": 0.9994078874588013
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.014028937001600862
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034463029354810715,
        "mean_feedback_forward_cosine": 0.9997429078625094,
        "mean_feedback_forward_relative_error": 0.022220811205765894,
        "min_feedback_forward_cosine": 0.9994081258773804
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.014854636296331882
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03445770964026451,
        "mean_feedback_forward_cosine": 0.999742982848998,
        "mean_feedback_forward_relative_error": 0.02221618534155911,
        "min_feedback_forward_cosine": 0.9994082450866699
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.01487931094944477
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034452635794878006,
        "mean_feedback_forward_cosine": 0.9997430924446352,
        "mean_feedback_forward_relative_error": 0.022211641779229525,
        "min_feedback_forward_cosine": 0.9994083642959595
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.014188854694962501
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034447718411684036,
        "mean_feedback_forward_cosine": 0.9997431905038895,
        "mean_feedback_forward_relative_error": 0.022207115972114187,
        "min_feedback_forward_cosine": 0.9994085431098938
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.013264953768849373
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03444201499223709,
        "mean_feedback_forward_cosine": 0.9997433000995267,
        "mean_feedback_forward_relative_error": 0.022202054278985146,
        "min_feedback_forward_cosine": 0.9994086623191833
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.015078886508345603
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034436747431755066,
        "mean_feedback_forward_cosine": 0.9997433731632848,
        "mean_feedback_forward_relative_error": 0.02219739014042481,
        "min_feedback_forward_cosine": 0.9994089603424072
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.01349074018061161
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03443177789449692,
        "mean_feedback_forward_cosine": 0.9997434654543477,
        "mean_feedback_forward_relative_error": 0.022192851820540045,
        "min_feedback_forward_cosine": 0.9994090795516968
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.0124130152207613
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03442655876278877,
        "mean_feedback_forward_cosine": 0.9997435808181763,
        "mean_feedback_forward_relative_error": 0.022188144811098615,
        "min_feedback_forward_cosine": 0.9994092583656311
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.012687791421413422
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034421082586050034,
        "mean_feedback_forward_cosine": 0.9997436827228915,
        "mean_feedback_forward_relative_error": 0.022183355559865312,
        "min_feedback_forward_cosine": 0.9994094371795654
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.012818007640689612
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034415971487760544,
        "mean_feedback_forward_cosine": 0.9997438173140248,
        "mean_feedback_forward_relative_error": 0.022178630044143045,
        "min_feedback_forward_cosine": 0.9994096159934998
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.012154696775600314
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03441084176301956,
        "mean_feedback_forward_cosine": 0.9997438807641307,
        "mean_feedback_forward_relative_error": 0.02217404039636735,
        "min_feedback_forward_cosine": 0.9994098544120789
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.011863310490995645
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03440537303686142,
        "mean_feedback_forward_cosine": 0.9997439903597678,
        "mean_feedback_forward_relative_error": 0.022169174971960245,
        "min_feedback_forward_cosine": 0.9994099736213684
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.012201981163918973
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03439989686012268,
        "mean_feedback_forward_cosine": 0.9997440807281002,
        "mean_feedback_forward_relative_error": 0.02216440129784807,
        "min_feedback_forward_cosine": 0.9994102120399475
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.011703400059193373
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034394387155771255,
        "mean_feedback_forward_cosine": 0.9997441787873546,
        "mean_feedback_forward_relative_error": 0.022159446226132494,
        "min_feedback_forward_cosine": 0.9994103908538818
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.011733281977027655
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034389425069093704,
        "mean_feedback_forward_cosine": 0.9997442749238783,
        "mean_feedback_forward_relative_error": 0.02215483523304424,
        "min_feedback_forward_cosine": 0.9994105100631714
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.011098120585232973
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034384094178676605,
        "mean_feedback_forward_cosine": 0.9997443902877069,
        "mean_feedback_forward_relative_error": 0.02215007922401832,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.010892319923341274
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034384019672870636,
        "mean_feedback_forward_cosine": 0.999744378751324,
        "mean_feedback_forward_relative_error": 0.022149969958850452,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.008965941136330367
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438396751880646,
        "mean_feedback_forward_cosine": 0.9997443922104374,
        "mean_feedback_forward_relative_error": 0.022149880581925952,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.009145139462947845
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034383926540613174,
        "mean_feedback_forward_cosine": 0.999744386442246,
        "mean_feedback_forward_relative_error": 0.022149790243636214,
        "min_feedback_forward_cosine": 0.9994106888771057
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.00857024108633399
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438388183712959,
        "mean_feedback_forward_cosine": 0.9997443845195155,
        "mean_feedback_forward_relative_error": 0.022149701377437,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.00860008357539773
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438383713364601,
        "mean_feedback_forward_cosine": 0.9997443768285936,
        "mean_feedback_forward_relative_error": 0.022149615951122775,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.008307939018383621
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034383803606033325,
        "mean_feedback_forward_cosine": 0.9997443768285936,
        "mean_feedback_forward_relative_error": 0.022149527475478187,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.008682450590580702
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438376262784004,
        "mean_feedback_forward_cosine": 0.9997443941331678,
        "mean_feedback_forward_relative_error": 0.022149444407513066,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.008061909982338548
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438372537493706,
        "mean_feedback_forward_cosine": 0.9997444037468203,
        "mean_feedback_forward_relative_error": 0.022149356923276377,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.008938052902668714
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438368812203407,
        "mean_feedback_forward_cosine": 0.9997443922104374,
        "mean_feedback_forward_relative_error": 0.022149273209393985,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.00822726613894105
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438365086913109,
        "mean_feedback_forward_cosine": 0.9997443922104374,
        "mean_feedback_forward_relative_error": 0.022149191207943425,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.007490905041694641
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0343836210668087,
        "mean_feedback_forward_cosine": 0.9997444056695507,
        "mean_feedback_forward_relative_error": 0.02214910886100223,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.00793412824228406
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438359126448631,
        "mean_feedback_forward_cosine": 0.9997443960558984,
        "mean_feedback_forward_relative_error": 0.022149030990417925,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.007137769331932068
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438355401158333,
        "mean_feedback_forward_cosine": 0.9997444037468203,
        "mean_feedback_forward_relative_error": 0.022148951332295132,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.007553048278689385
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034383516758680344,
        "mean_feedback_forward_cosine": 0.9997443922104374,
        "mean_feedback_forward_relative_error": 0.022148873446689497,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.007888392472825944
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438347950577736,
        "mean_feedback_forward_cosine": 0.9997443960558984,
        "mean_feedback_forward_relative_error": 0.02214879309758544,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.007774316803961992
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438344597816467,
        "mean_feedback_forward_cosine": 0.9997444037468203,
        "mean_feedback_forward_relative_error": 0.02214871615832371,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.007526577415838838
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438340499997139,
        "mean_feedback_forward_cosine": 0.9997443979786288,
        "mean_feedback_forward_relative_error": 0.02214863417189448,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.007590677718669176
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034383371472358704,
        "mean_feedback_forward_cosine": 0.999744386442246,
        "mean_feedback_forward_relative_error": 0.022148555129646294,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.007188964843824506
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438333794474602,
        "mean_feedback_forward_cosine": 0.9997444152832031,
        "mean_feedback_forward_relative_error": 0.02214847773974461,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.006880649688988924
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438330069184303,
        "mean_feedback_forward_cosine": 0.9997443922104374,
        "mean_feedback_forward_relative_error": 0.02214839567820872,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.007505728708356619
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438326343894005,
        "mean_feedback_forward_cosine": 0.9997444056695507,
        "mean_feedback_forward_relative_error": 0.022148319580141577,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.0073278361514955755
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438323363661766,
        "mean_feedback_forward_cosine": 0.9997444095150116,
        "mean_feedback_forward_relative_error": 0.02214824267092251,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.006741334313601255
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438319265842438,
        "mean_feedback_forward_cosine": 0.9997444114377422,
        "mean_feedback_forward_relative_error": 0.022148165776724776,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.0072374745182693006
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438316285610199,
        "mean_feedback_forward_cosine": 0.9997444018240897,
        "mean_feedback_forward_relative_error": 0.02214809227734804,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.007213199753910303
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0343831330537796,
        "mean_feedback_forward_cosine": 0.9997444075922812,
        "mean_feedback_forward_relative_error": 0.02214801611919557,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.0068370764482021335
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034383099526166916,
        "mean_feedback_forward_cosine": 0.9997443999013593,
        "mean_feedback_forward_relative_error": 0.022147933141358437,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.007554607146382332
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438306227326393,
        "mean_feedback_forward_cosine": 0.9997444037468203,
        "mean_feedback_forward_relative_error": 0.022147857238568606,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.007335109049528837
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034383028745651245,
        "mean_feedback_forward_cosine": 0.9997444037468203,
        "mean_feedback_forward_relative_error": 0.02214777983364559,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.007077484371960163
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438299894332886,
        "mean_feedback_forward_cosine": 0.9997444056695507,
        "mean_feedback_forward_relative_error": 0.022147701302122687,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.0068565161567926405
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438296169042587,
        "mean_feedback_forward_cosine": 0.9997444056695507,
        "mean_feedback_forward_relative_error": 0.022147619616120093,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.007023111615553498
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382931888103485,
        "mean_feedback_forward_cosine": 0.9997444172059337,
        "mean_feedback_forward_relative_error": 0.022147543172562314,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.006798136008307338
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382905811071396,
        "mean_feedback_forward_cosine": 0.9997443999013593,
        "mean_feedback_forward_relative_error": 0.02214746941782294,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.006447751925587654
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438286855816841,
        "mean_feedback_forward_cosine": 0.9997444287423165,
        "mean_feedback_forward_relative_error": 0.022147395182400942,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.007190746357142925
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382835030555725,
        "mean_feedback_forward_cosine": 0.9997444075922812,
        "mean_feedback_forward_relative_error": 0.0221473200757417,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.006664027765244246
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382808953523636,
        "mean_feedback_forward_cosine": 0.9997444172059337,
        "mean_feedback_forward_relative_error": 0.022147241664389448,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.006577246445342898
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438278287649155,
        "mean_feedback_forward_cosine": 0.9997444191286641,
        "mean_feedback_forward_relative_error": 0.022147166377474225,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.007075923771187663
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438274934887886,
        "mean_feedback_forward_cosine": 0.9997444114377422,
        "mean_feedback_forward_relative_error": 0.02214709352401476,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.006495891350656748
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382712095975876,
        "mean_feedback_forward_cosine": 0.9997444075922812,
        "mean_feedback_forward_relative_error": 0.022147022638349764,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.006212037367112935
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438268229365349,
        "mean_feedback_forward_cosine": 0.9997444018240897,
        "mean_feedback_forward_relative_error": 0.022146949514506326,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.006336527991443872
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0343826524913311,
        "mean_feedback_forward_cosine": 0.9997444133604726,
        "mean_feedback_forward_relative_error": 0.02214687269541525,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.0067001722053438425
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438260778784752,
        "mean_feedback_forward_cosine": 0.9997444114377422,
        "mean_feedback_forward_relative_error": 0.022146797949267973,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.007536201997920871
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438257798552513,
        "mean_feedback_forward_cosine": 0.9997444172059337,
        "mean_feedback_forward_relative_error": 0.022146721941328818,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.006198162413164973
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382544457912445,
        "mean_feedback_forward_cosine": 0.9997444172059337,
        "mean_feedback_forward_relative_error": 0.022146650259533236,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.005991386559242383
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438250720500946,
        "mean_feedback_forward_cosine": 0.9997444152832031,
        "mean_feedback_forward_relative_error": 0.022146574777340697,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.0070398429016023875
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438247740268707,
        "mean_feedback_forward_cosine": 0.9997444095150116,
        "mean_feedback_forward_relative_error": 0.02214649931016949,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.006825379098318517
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438243642449379,
        "mean_feedback_forward_cosine": 0.9997444172059337,
        "mean_feedback_forward_relative_error": 0.022146423632699634,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.00649112942956388
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0343824066221714,
        "mean_feedback_forward_cosine": 0.9997444287423165,
        "mean_feedback_forward_relative_error": 0.022146351635456085,
        "min_feedback_forward_cosine": 0.9994107484817505
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.006362385395392776
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382373094558716,
        "mean_feedback_forward_cosine": 0.9997444306650469,
        "mean_feedback_forward_relative_error": 0.022146280028767163,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.00654777805916965
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03438233584165573,
        "mean_feedback_forward_cosine": 0.9997444345105079,
        "mean_feedback_forward_relative_error": 0.02214620744569167,
        "min_feedback_forward_cosine": 0.99941086769104
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.006772128741443157
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.034382302314043045,
        "mean_feedback_forward_cosine": 0.9997444114377422,
        "mean_feedback_forward_relative_error": 0.02214613657504801,
        "min_feedback_forward_cosine": 0.9994108080863953
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.006358181383684277
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9223,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.40806366271972655,
    "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": 10,
    "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": 1.1715624332427979,
    "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": 10222.548335790634,
    "train_wall_s": 10219.140581607819
  },
  "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
  }
}