{
  "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": 12,
    "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_s12.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": 12,
    "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": 0.9996226906776429,
    "feedback_forward_cosine": [
      0.9998250007629395,
      0.9997978210449219,
      0.999758780002594,
      0.9997544884681702,
      0.9997519254684448,
      0.9997063875198364,
      0.9997261166572571,
      0.9996814131736755,
      0.9998026490211487,
      0.9997053146362305,
      0.9997532367706299,
      0.9998248815536499,
      0.9997637867927551,
      0.9997279047966003,
      0.9997736215591431,
      0.9996997117996216,
      0.9997929334640503,
      0.999712347984314,
      0.9997790455818176,
      0.9996895790100098,
      0.999791145324707,
      0.9998652935028076,
      0.9998343586921692,
      0.9998188018798828,
      0.9998387694358826,
      0.9997784495353699,
      0.9998390674591064,
      0.9996950626373291,
      0.9995816946029663,
      0.9992325305938721,
      0.9999851584434509
    ],
    "feedback_forward_norm_ratio": [
      0.9995194673538208,
      1.0007712841033936,
      1.000235676765442,
      1.0012747049331665,
      0.9986703991889954,
      1.0010285377502441,
      0.9991890788078308,
      1.0012438297271729,
      0.999777615070343,
      1.0017437934875488,
      0.9986539483070374,
      1.0003631114959717,
      0.9993255734443665,
      1.0008310079574585,
      0.999375581741333,
      1.0016775131225586,
      0.9995991587638855,
      1.0015108585357666,
      0.9994063973426819,
      1.0014119148254395,
      0.9992547035217285,
      1.0004510879516602,
      1.0000050067901611,
      1.0006206035614014,
      0.9998761415481567,
      1.0008827447891235,
      0.9999377727508545,
      1.0009171962738037,
      0.9990943074226379,
      1.0017184019088745,
      1.0025044679641724
    ],
    "feedback_forward_relative_error": [
      0.018711449578404427,
      0.0201310645788908,
      0.021970022469758987,
      0.02221059985458851,
      0.022298455238342285,
      0.024271324276924133,
      0.023405304178595543,
      0.025290008634328842,
      0.019864164292812347,
      0.0243570227175951,
      0.022242913022637367,
      0.018726056441664696,
      0.02173955924808979,
      0.023355519399046898,
      0.021280093118548393,
      0.02458626963198185,
      0.020351605489850044,
      0.02405170165002346,
      0.02102426066994667,
      0.0249736737459898,
      0.020445581525564194,
      0.01641915924847126,
      0.018208028748631477,
      0.019054405391216278,
      0.01795344054698944,
      0.021074611693620682,
      0.017935434356331825,
      0.024722829461097717,
      0.028928402811288834,
      0.039251793175935745,
      0.005994729697704315
    ],
    "innovation_negative_gradient_cosine": [
      0.9996582269668579,
      0.9996131062507629,
      0.9996418356895447,
      0.9995585680007935,
      0.9996452331542969,
      0.9995571374893188,
      0.9996507167816162,
      0.9995861053466797,
      0.9996627569198608,
      0.9996532201766968,
      0.9996786713600159,
      0.9996798038482666,
      0.99970543384552,
      0.9996029138565063,
      0.9997167587280273,
      0.9996203780174255,
      0.9997353553771973,
      0.9996602535247803,
      0.9997515082359314,
      0.9996718168258667,
      0.9997636675834656,
      0.9998162388801575,
      0.9998011589050293,
      0.9997760057449341,
      0.999826192855835,
      0.9998064041137695,
      0.999868631362915,
      0.9999010562896729,
      0.9998774528503418,
      0.9999389052391052,
      0.7473083734512329
    ],
    "mean_feedback_forward_cosine": 0.9997512025217856,
    "mean_feedback_forward_relative_error": 0.021768693061124895,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.9996582269668579,
      0.9996131062507629,
      0.9996418356895447,
      0.9995585680007935,
      0.9996452331542969,
      0.9995571374893188,
      0.9996507167816162,
      0.9995861053466797,
      0.9996627569198608,
      0.9996532201766968,
      0.9996786713600159,
      0.9996798038482666,
      0.99970543384552,
      0.9996029138565063,
      0.9997167587280273,
      0.9996203780174255,
      0.9997353553771973,
      0.9996602535247803,
      0.9997515082359314,
      0.9996718168258667,
      0.9997636675834656,
      0.9998162388801575,
      0.9998011589050293,
      0.9997760057449341,
      0.999826192855835,
      0.9998064041137695,
      0.999868631362915,
      0.9999010562896729,
      0.9998774528503418,
      0.9999389052391052,
      0.7473083734512329
    ],
    "teaching_negative_gradient_cosine": [
      0.9996582269668579,
      0.9996131062507629,
      0.9996418356895447,
      0.9995585680007935,
      0.9996452331542969,
      0.9995571374893188,
      0.9996507167816162,
      0.9995861053466797,
      0.9996627569198608,
      0.9996532201766968,
      0.9996786713600159,
      0.9996798038482666,
      0.99970543384552,
      0.9996029138565063,
      0.9997167587280273,
      0.9996203780174255,
      0.9997353553771973,
      0.9996602535247803,
      0.9997515082359314,
      0.9996718168258667,
      0.9997636675834656,
      0.9998162388801575,
      0.9998011589050293,
      0.9997760057449341,
      0.999826192855835,
      0.9998064041137695,
      0.999868631362915,
      0.9999010562896729,
      0.9998774528503418,
      0.9999389052391052,
      0.7473083734512329
    ],
    "wall_s": 0.11745548248291016
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3349723815917969,
        "mean_feedback_forward_cosine": 0.25990806856463033,
        "mean_feedback_forward_relative_error": 1.2173136838020817,
        "min_feedback_forward_cosine": 0.11319214105606079
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 1.884043900756836
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3094651699066162,
        "mean_feedback_forward_cosine": 0.35645093456391364,
        "mean_feedback_forward_relative_error": 1.135014666664985,
        "min_feedback_forward_cosine": 0.14877766370773315
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 1.2980965474700927
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2844831943511963,
        "mean_feedback_forward_cosine": 0.43406235979449365,
        "mean_feedback_forward_relative_error": 1.0651125821375078,
        "min_feedback_forward_cosine": 0.18230189383029938
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.0457662846374511
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.257002353668213,
        "mean_feedback_forward_cosine": 0.49849749909293267,
        "mean_feedback_forward_relative_error": 1.003033523598025,
        "min_feedback_forward_cosine": 0.21811392903327942
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 0.8763960188293457
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2295236587524414,
        "mean_feedback_forward_cosine": 0.5516888105100201,
        "mean_feedback_forward_relative_error": 0.9484026335900829,
        "min_feedback_forward_cosine": 0.2541167736053467
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 0.7647238268280029
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2094680070877075,
        "mean_feedback_forward_cosine": 0.5949192950802464,
        "mean_feedback_forward_relative_error": 0.9011446881678796,
        "min_feedback_forward_cosine": 0.2851111888885498
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 0.6871170258522034
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1871272325515747,
        "mean_feedback_forward_cosine": 0.6318055352857036,
        "mean_feedback_forward_relative_error": 0.8588615530921567,
        "min_feedback_forward_cosine": 0.31310462951660156
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.6291039484786988
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1641530990600586,
        "mean_feedback_forward_cosine": 0.6638145408322734,
        "mean_feedback_forward_relative_error": 0.8201376699632213,
        "min_feedback_forward_cosine": 0.34058821201324463
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.586492281703949
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1416432857513428,
        "mean_feedback_forward_cosine": 0.692069120945469,
        "mean_feedback_forward_relative_error": 0.7841187965485358,
        "min_feedback_forward_cosine": 0.366452693939209
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.5476818246841431
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1194217205047607,
        "mean_feedback_forward_cosine": 0.7173972206730996,
        "mean_feedback_forward_relative_error": 0.7505404612710399,
        "min_feedback_forward_cosine": 0.39185771346092224
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.5080024182033539
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.09666907787323,
        "mean_feedback_forward_cosine": 0.7399982973452537,
        "mean_feedback_forward_relative_error": 0.7192039864678537,
        "min_feedback_forward_cosine": 0.4173521399497986
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.48509688959121705
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0738338232040405,
        "mean_feedback_forward_cosine": 0.7605622458842493,
        "mean_feedback_forward_relative_error": 0.6895902368330187,
        "min_feedback_forward_cosine": 0.44172707200050354
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.4655131332588196
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0519495010375977,
        "mean_feedback_forward_cosine": 0.7789299305408232,
        "mean_feedback_forward_relative_error": 0.6620327253495494,
        "min_feedback_forward_cosine": 0.4650299847126007
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.43883312452316287
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.029091238975525,
        "mean_feedback_forward_cosine": 0.7959011793136597,
        "mean_feedback_forward_relative_error": 0.6354356696528773,
        "min_feedback_forward_cosine": 0.488097608089447
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.419735015001297
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0058889389038086,
        "mean_feedback_forward_cosine": 0.8113211816357028,
        "mean_feedback_forward_relative_error": 0.6103085418862682,
        "min_feedback_forward_cosine": 0.5115365386009216
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.4059742552089691
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9821122288703918,
        "mean_feedback_forward_cosine": 0.8253101264276812,
        "mean_feedback_forward_relative_error": 0.5864743201001998,
        "min_feedback_forward_cosine": 0.5349057912826538
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.3919783997154236
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9576853513717651,
        "mean_feedback_forward_cosine": 0.8383109781049913,
        "mean_feedback_forward_relative_error": 0.5636168324178265,
        "min_feedback_forward_cosine": 0.5582489967346191
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.379794213848114
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9333812594413757,
        "mean_feedback_forward_cosine": 0.85005771921527,
        "mean_feedback_forward_relative_error": 0.5421345978975296,
        "min_feedback_forward_cosine": 0.5805637836456299
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.36320647946357726
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9090774059295654,
        "mean_feedback_forward_cosine": 0.8611171495529913,
        "mean_feedback_forward_relative_error": 0.5212958705040717,
        "min_feedback_forward_cosine": 0.6023303270339966
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.353560187292099
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8852086067199707,
        "mean_feedback_forward_cosine": 0.8712598642995281,
        "mean_feedback_forward_relative_error": 0.5012798193962343,
        "min_feedback_forward_cosine": 0.6232945322990417
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.34309122184753416
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8593584895133972,
        "mean_feedback_forward_cosine": 0.88098979573096,
        "mean_feedback_forward_relative_error": 0.4814876466989517,
        "min_feedback_forward_cosine": 0.6450381278991699
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.33722845947265623
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8355581760406494,
        "mean_feedback_forward_cosine": 0.8898247191982884,
        "mean_feedback_forward_relative_error": 0.462748872656976,
        "min_feedback_forward_cosine": 0.6645044684410095
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.33217840114593505
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8125510811805725,
        "mean_feedback_forward_cosine": 0.8976805344704659,
        "mean_feedback_forward_relative_error": 0.4454753009542342,
        "min_feedback_forward_cosine": 0.6828471422195435
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.3130986285972595
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7896296977996826,
        "mean_feedback_forward_cosine": 0.9050561285788014,
        "mean_feedback_forward_relative_error": 0.42864807623047985,
        "min_feedback_forward_cosine": 0.700555682182312
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.309137520236969
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7661307454109192,
        "mean_feedback_forward_cosine": 0.9120974483028534,
        "mean_feedback_forward_relative_error": 0.41205001982950396,
        "min_feedback_forward_cosine": 0.7182934880256653
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.3054877910041809
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7443239688873291,
        "mean_feedback_forward_cosine": 0.918309542440599,
        "mean_feedback_forward_relative_error": 0.3968265316178722,
        "min_feedback_forward_cosine": 0.7340960502624512
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.29228196764945985
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7201147079467773,
        "mean_feedback_forward_cosine": 0.9244374152152769,
        "mean_feedback_forward_relative_error": 0.3812758355371414,
        "min_feedback_forward_cosine": 0.7511111497879028
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.29109219769001005
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6977641582489014,
        "mean_feedback_forward_cosine": 0.9298505129352692,
        "mean_feedback_forward_relative_error": 0.3670203181043748,
        "min_feedback_forward_cosine": 0.7661524415016174
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.2771931728935242
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.67481929063797,
        "mean_feedback_forward_cosine": 0.9349553546597881,
        "mean_feedback_forward_relative_error": 0.3530991913810853,
        "min_feedback_forward_cosine": 0.7812139391899109
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.279468224697113
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6530283689498901,
        "mean_feedback_forward_cosine": 0.9397015263957362,
        "mean_feedback_forward_relative_error": 0.3397381053336205,
        "min_feedback_forward_cosine": 0.7948744893074036
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.2755542581796646
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6322877407073975,
        "mean_feedback_forward_cosine": 0.9440194733681218,
        "mean_feedback_forward_relative_error": 0.3271398010753816,
        "min_feedback_forward_cosine": 0.8076740503311157
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.2665172424221039
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6114274263381958,
        "mean_feedback_forward_cosine": 0.9480447096209372,
        "mean_feedback_forward_relative_error": 0.31496808125126746,
        "min_feedback_forward_cosine": 0.8199342489242554
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.2590750113105774
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5898762941360474,
        "mean_feedback_forward_cosine": 0.9518554499072414,
        "mean_feedback_forward_relative_error": 0.3030531540993721,
        "min_feedback_forward_cosine": 0.8323711156845093
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.25794359429359437
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5693039894104004,
        "mean_feedback_forward_cosine": 0.9554501272016956,
        "mean_feedback_forward_relative_error": 0.2913327791517781,
        "min_feedback_forward_cosine": 0.8437407612800598
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.2553165550708771
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.548820972442627,
        "mean_feedback_forward_cosine": 0.9587425454970329,
        "mean_feedback_forward_relative_error": 0.2802302210081008,
        "min_feedback_forward_cosine": 0.8547742366790771
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.2507095138835907
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5298911333084106,
        "mean_feedback_forward_cosine": 0.9617095435819318,
        "mean_feedback_forward_relative_error": 0.26985916903903406,
        "min_feedback_forward_cosine": 0.8644689321517944
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.24415148473739623
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5102511048316956,
        "mean_feedback_forward_cosine": 0.9645711414275631,
        "mean_feedback_forward_relative_error": 0.25947804532704816,
        "min_feedback_forward_cosine": 0.8741466999053955
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.24547552498340608
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4911539554595947,
        "mean_feedback_forward_cosine": 0.9672493127084547,
        "mean_feedback_forward_relative_error": 0.24944197963322362,
        "min_feedback_forward_cosine": 0.8833778500556946
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.2402279775428772
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4727519154548645,
        "mean_feedback_forward_cosine": 0.9697137782650609,
        "mean_feedback_forward_relative_error": 0.2398086798287207,
        "min_feedback_forward_cosine": 0.891808271408081
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.23981824064254761
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.45590081810951233,
        "mean_feedback_forward_cosine": 0.9719306768909577,
        "mean_feedback_forward_relative_error": 0.23079736578849055,
        "min_feedback_forward_cosine": 0.8993138074874878
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.22953084101200102
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4394187033176422,
        "mean_feedback_forward_cosine": 0.9739993022334191,
        "mean_feedback_forward_relative_error": 0.22205424044401414,
        "min_feedback_forward_cosine": 0.9063918590545654
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.2279604758644104
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.42284128069877625,
        "mean_feedback_forward_cosine": 0.975942728980895,
        "mean_feedback_forward_relative_error": 0.21353178952009447,
        "min_feedback_forward_cosine": 0.9132277965545654
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.22551626369476319
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4067952036857605,
        "mean_feedback_forward_cosine": 0.9777397224980016,
        "mean_feedback_forward_relative_error": 0.20538089159996278,
        "min_feedback_forward_cosine": 0.9196112751960754
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.22367396668434142
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3910422623157501,
        "mean_feedback_forward_cosine": 0.9794375992590382,
        "mean_feedback_forward_relative_error": 0.19732192158699036,
        "min_feedback_forward_cosine": 0.9256578683853149
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.22419112510681152
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.37633296847343445,
        "mean_feedback_forward_cosine": 0.9809703115494021,
        "mean_feedback_forward_relative_error": 0.18982052658834764,
        "min_feedback_forward_cosine": 0.9310845732688904
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.22211388847827912
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3624928593635559,
        "mean_feedback_forward_cosine": 0.9823756525593419,
        "mean_feedback_forward_relative_error": 0.18263642261585883,
        "min_feedback_forward_cosine": 0.9360270500183105
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.21705526567935943
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3488244116306305,
        "mean_feedback_forward_cosine": 0.9836807289431172,
        "mean_feedback_forward_relative_error": 0.1757090908865775,
        "min_feedback_forward_cosine": 0.9407151937484741
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.2158580997800827
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.33559462428092957,
        "mean_feedback_forward_cosine": 0.9848859194786318,
        "mean_feedback_forward_relative_error": 0.16906738485540113,
        "min_feedback_forward_cosine": 0.9450581073760986
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.20908656046390534
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3229137659072876,
        "mean_feedback_forward_cosine": 0.985992964236967,
        "mean_feedback_forward_relative_error": 0.1627447511399946,
        "min_feedback_forward_cosine": 0.9490985870361328
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.2055481154346466
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.31016919016838074,
        "mean_feedback_forward_cosine": 0.9870364396802841,
        "mean_feedback_forward_relative_error": 0.1565605353924536,
        "min_feedback_forward_cosine": 0.9530023336410522
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.20949283747196198
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.29767873883247375,
        "mean_feedback_forward_cosine": 0.9880552368779336,
        "mean_feedback_forward_relative_error": 0.15028610104514706,
        "min_feedback_forward_cosine": 0.9566665887832642
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.2089853828907013
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.28635820746421814,
        "mean_feedback_forward_cosine": 0.988928867924598,
        "mean_feedback_forward_relative_error": 0.14466954539379767,
        "min_feedback_forward_cosine": 0.9598673582077026
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.2060586162662506
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2751348316669464,
        "mean_feedback_forward_cosine": 0.9897554882111088,
        "mean_feedback_forward_relative_error": 0.1391651259074288,
        "min_feedback_forward_cosine": 0.9629354476928711
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.20208517848968505
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2640094459056854,
        "mean_feedback_forward_cosine": 0.9905401083730883,
        "mean_feedback_forward_relative_error": 0.13374620291494554,
        "min_feedback_forward_cosine": 0.9658530950546265
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.2008401836013794
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.25394725799560547,
        "mean_feedback_forward_cosine": 0.9912350619992902,
        "mean_feedback_forward_relative_error": 0.12873109333938168,
        "min_feedback_forward_cosine": 0.9683619141578674
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.1994219180393219
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.24407565593719482,
        "mean_feedback_forward_cosine": 0.9918945912391909,
        "mean_feedback_forward_relative_error": 0.12379781685529216,
        "min_feedback_forward_cosine": 0.9707534909248352
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.19782180733680724
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.23445096611976624,
        "mean_feedback_forward_cosine": 0.9924968769473415,
        "mean_feedback_forward_relative_error": 0.11911300689943376,
        "min_feedback_forward_cosine": 0.9729940295219421
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.19733688462734222
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22489391267299652,
        "mean_feedback_forward_cosine": 0.9930661935960093,
        "mean_feedback_forward_relative_error": 0.11452286486183444,
        "min_feedback_forward_cosine": 0.9751298427581787
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.19768200149536133
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2162090688943863,
        "mean_feedback_forward_cosine": 0.9935784993633148,
        "mean_feedback_forward_relative_error": 0.11020198860956777,
        "min_feedback_forward_cosine": 0.9769930243492126
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.1947573218727112
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20737668871879578,
        "mean_feedback_forward_cosine": 0.9940720200538635,
        "mean_feedback_forward_relative_error": 0.10589810833334923,
        "min_feedback_forward_cosine": 0.9788262844085693
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.1977950550365448
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19919180870056152,
        "mean_feedback_forward_cosine": 0.9945168860497013,
        "mean_feedback_forward_relative_error": 0.1018484249230354,
        "min_feedback_forward_cosine": 0.980451226234436
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.18924341267585754
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19078901410102844,
        "mean_feedback_forward_cosine": 0.9949399925047352,
        "mean_feedback_forward_relative_error": 0.09785510803903302,
        "min_feedback_forward_cosine": 0.9820586442947388
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.19221665453910827
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18314522504806519,
        "mean_feedback_forward_cosine": 0.995325994107031,
        "mean_feedback_forward_relative_error": 0.09406068644696666,
        "min_feedback_forward_cosine": 0.9834667444229126
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.18745509509086608
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17564021050930023,
        "mean_feedback_forward_cosine": 0.995681478131202,
        "mean_feedback_forward_relative_error": 0.09041920724895693,
        "min_feedback_forward_cosine": 0.9847860932350159
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.1897912615942955
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16852696239948273,
        "mean_feedback_forward_cosine": 0.9960067464459327,
        "mean_feedback_forward_relative_error": 0.08695372224094407,
        "min_feedback_forward_cosine": 0.9859851598739624
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.1850361007261276
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.16218271851539612,
        "mean_feedback_forward_cosine": 0.9962990860785207,
        "mean_feedback_forward_relative_error": 0.08371402034836431,
        "min_feedback_forward_cosine": 0.987011194229126
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.18250540663719178
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15562225878238678,
        "mean_feedback_forward_cosine": 0.9965787145399279,
        "mean_feedback_forward_relative_error": 0.08049640516119619,
        "min_feedback_forward_cosine": 0.9880364537239075
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.18282743438720703
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14947938919067383,
        "mean_feedback_forward_cosine": 0.9968364181057099,
        "mean_feedback_forward_relative_error": 0.07741088065649232,
        "min_feedback_forward_cosine": 0.9889557361602783
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.183723773355484
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1434447169303894,
        "mean_feedback_forward_cosine": 0.9970758730365384,
        "mean_feedback_forward_relative_error": 0.07443589410714564,
        "min_feedback_forward_cosine": 0.9898234009742737
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.18212517969369887
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13758878409862518,
        "mean_feedback_forward_cosine": 0.9972983521799887,
        "mean_feedback_forward_relative_error": 0.07155920655256318,
        "min_feedback_forward_cosine": 0.9906343221664429
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.18086355746269225
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13201770186424255,
        "mean_feedback_forward_cosine": 0.9975032037304293,
        "mean_feedback_forward_relative_error": 0.06879527920917157,
        "min_feedback_forward_cosine": 0.9913747310638428
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.17983569902420043
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12645262479782104,
        "mean_feedback_forward_cosine": 0.997694036652965,
        "mean_feedback_forward_relative_error": 0.0661307120275113,
        "min_feedback_forward_cosine": 0.9920851588249207
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.18159530221939088
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12123321741819382,
        "mean_feedback_forward_cosine": 0.99786796492915,
        "mean_feedback_forward_relative_error": 0.06359651883042627,
        "min_feedback_forward_cosine": 0.9927232265472412
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.1807903523159027
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11612481623888016,
        "mean_feedback_forward_cosine": 0.9980316719701213,
        "mean_feedback_forward_relative_error": 0.06111879724889032,
        "min_feedback_forward_cosine": 0.9933205842971802
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.18251712452888488
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11155027896165848,
        "mean_feedback_forward_cosine": 0.9981785474285003,
        "mean_feedback_forward_relative_error": 0.058794992585336006,
        "min_feedback_forward_cosine": 0.9938347339630127
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.17368552847385407
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10704890638589859,
        "mean_feedback_forward_cosine": 0.9983175185418898,
        "mean_feedback_forward_relative_error": 0.05651775657409622,
        "min_feedback_forward_cosine": 0.994320273399353
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.17487866772651672
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10279550403356552,
        "mean_feedback_forward_cosine": 0.998444099580088,
        "mean_feedback_forward_relative_error": 0.054354885953568646,
        "min_feedback_forward_cosine": 0.9947606325149536
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.1732005853843689
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09852416068315506,
        "mean_feedback_forward_cosine": 0.9985643740623228,
        "mean_feedback_forward_relative_error": 0.05222089306241082,
        "min_feedback_forward_cosine": 0.9951850771903992
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.17545873102188111
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09449601918458939,
        "mean_feedback_forward_cosine": 0.9986736409125789,
        "mean_feedback_forward_relative_error": 0.050204635207210815,
        "min_feedback_forward_cosine": 0.995569109916687
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.17256650600910187
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0906398594379425,
        "mean_feedback_forward_cosine": 0.9987740881981388,
        "mean_feedback_forward_relative_error": 0.04827549933425842,
        "min_feedback_forward_cosine": 0.9959231615066528
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.17154016873359682
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08689554035663605,
        "mean_feedback_forward_cosine": 0.9988671041304066,
        "mean_feedback_forward_relative_error": 0.04641639303055502,
        "min_feedback_forward_cosine": 0.9962524771690369
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.16824539291858673
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08307069540023804,
        "mean_feedback_forward_cosine": 0.9989579704500013,
        "mean_feedback_forward_relative_error": 0.04452944841356047,
        "min_feedback_forward_cosine": 0.9965732097625732
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.1754700458049774
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0797470211982727,
        "mean_feedback_forward_cosine": 0.9990356256884914,
        "mean_feedback_forward_relative_error": 0.04284472327919737,
        "min_feedback_forward_cosine": 0.996841311454773
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.1721455662059784
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07657480239868164,
        "mean_feedback_forward_cosine": 0.9991070608938893,
        "mean_feedback_forward_relative_error": 0.04123145305821972,
        "min_feedback_forward_cosine": 0.9970871210098267
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.16954058514118195
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07354815304279327,
        "mean_feedback_forward_cosine": 0.9991730297765424,
        "mean_feedback_forward_relative_error": 0.039683275405437715,
        "min_feedback_forward_cosine": 0.9973117709159851
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.1681050411939621
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07066227495670319,
        "mean_feedback_forward_cosine": 0.9992358280766395,
        "mean_feedback_forward_relative_error": 0.038148833378668756,
        "min_feedback_forward_cosine": 0.9975184202194214
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.16526709414482116
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06776793301105499,
        "mean_feedback_forward_cosine": 0.9992965325232475,
        "mean_feedback_forward_relative_error": 0.03660518351581789,
        "min_feedback_forward_cosine": 0.9977167844772339
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.17173692792892456
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06512687355279922,
        "mean_feedback_forward_cosine": 0.9993474541171905,
        "mean_feedback_forward_relative_error": 0.03525468307517229,
        "min_feedback_forward_cosine": 0.9978905916213989
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.1647601191496849
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06248985603451729,
        "mean_feedback_forward_cosine": 0.9993976912190837,
        "mean_feedback_forward_relative_error": 0.03387635940265271,
        "min_feedback_forward_cosine": 0.9980573058128357
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.16422447180747987
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05999685078859329,
        "mean_feedback_forward_cosine": 0.999443123417516,
        "mean_feedback_forward_relative_error": 0.03257730176612254,
        "min_feedback_forward_cosine": 0.9982086420059204
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.16520101567268372
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.057540878653526306,
        "mean_feedback_forward_cosine": 0.9994849216553473,
        "mean_feedback_forward_relative_error": 0.031336988262351484,
        "min_feedback_forward_cosine": 0.9983521699905396
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.16500281002998352
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05531076341867447,
        "mean_feedback_forward_cosine": 0.9995224148996414,
        "mean_feedback_forward_relative_error": 0.0301763474821083,
        "min_feedback_forward_cosine": 0.9984773397445679
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.1594796088385582
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.053001850843429565,
        "mean_feedback_forward_cosine": 0.9995596620344347,
        "mean_feedback_forward_relative_error": 0.02898126714412243,
        "min_feedback_forward_cosine": 0.9986019730567932
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.16494166981697084
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05076294764876366,
        "mean_feedback_forward_cosine": 0.9995936539865309,
        "mean_feedback_forward_relative_error": 0.02784653329440663,
        "min_feedback_forward_cosine": 0.9987173080444336
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.16599833611011505
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.048671893775463104,
        "mean_feedback_forward_cosine": 0.9996238696959711,
        "mean_feedback_forward_relative_error": 0.026796938941603708,
        "min_feedback_forward_cosine": 0.9988206028938293
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.16241411400318145
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.046662647277116776,
        "mean_feedback_forward_cosine": 0.999652999062692,
        "mean_feedback_forward_relative_error": 0.025740397973887382,
        "min_feedback_forward_cosine": 0.9989158511161804
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.16649711584091187
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04476156458258629,
        "mean_feedback_forward_cosine": 0.9996800115031581,
        "mean_feedback_forward_relative_error": 0.024722059707968466,
        "min_feedback_forward_cosine": 0.9990021586418152
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.1602955338382721
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04304143413901329,
        "mean_feedback_forward_cosine": 0.9997028458502984,
        "mean_feedback_forward_relative_error": 0.023824551503264135,
        "min_feedback_forward_cosine": 0.9990774989128113
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.15782464220046996
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04127826541662216,
        "mean_feedback_forward_cosine": 0.9997257974839979,
        "mean_feedback_forward_relative_error": 0.022887508444968733,
        "min_feedback_forward_cosine": 0.9991511702537537
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.16260396190643311
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03956778347492218,
        "mean_feedback_forward_cosine": 0.9997466398823646,
        "mean_feedback_forward_relative_error": 0.022003036307831927,
        "min_feedback_forward_cosine": 0.9992200136184692
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.1612073497080803
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03955405205488205,
        "mean_feedback_forward_cosine": 0.9997468052371856,
        "mean_feedback_forward_relative_error": 0.021993932133961107,
        "min_feedback_forward_cosine": 0.9992206692695618
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.08828540871143341
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039547037333250046,
        "mean_feedback_forward_cosine": 0.9997469148328227,
        "mean_feedback_forward_relative_error": 0.021988024349294363,
        "min_feedback_forward_cosine": 0.9992209672927856
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.06392889021635055
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039540424942970276,
        "mean_feedback_forward_cosine": 0.9997469936647723,
        "mean_feedback_forward_relative_error": 0.02198252376289137,
        "min_feedback_forward_cosine": 0.9992212057113647
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.05590131332099438
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03953411057591438,
        "mean_feedback_forward_cosine": 0.9997471071058704,
        "mean_feedback_forward_relative_error": 0.02197735247412516,
        "min_feedback_forward_cosine": 0.9992215037345886
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.04892089093506336
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03952792286872864,
        "mean_feedback_forward_cosine": 0.9997472128560466,
        "mean_feedback_forward_relative_error": 0.0219724063701447,
        "min_feedback_forward_cosine": 0.9992217421531677
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.043798088161945345
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03952197730541229,
        "mean_feedback_forward_cosine": 0.9997472705379609,
        "mean_feedback_forward_relative_error": 0.021967468467811422,
        "min_feedback_forward_cosine": 0.999221920967102
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.04162949063181877
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039516009390354156,
        "mean_feedback_forward_cosine": 0.99974734744718,
        "mean_feedback_forward_relative_error": 0.021962743012174483,
        "min_feedback_forward_cosine": 0.9992221593856812
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.03833890787601471
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03951011225581169,
        "mean_feedback_forward_cosine": 0.9997474416609733,
        "mean_feedback_forward_relative_error": 0.02195797023934222,
        "min_feedback_forward_cosine": 0.9992223978042603
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.03592932426214218
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03950371965765953,
        "mean_feedback_forward_cosine": 0.9997475224156533,
        "mean_feedback_forward_relative_error": 0.021953170142707326,
        "min_feedback_forward_cosine": 0.9992227554321289
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.035366734300255775
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0394979789853096,
        "mean_feedback_forward_cosine": 0.9997476012476029,
        "mean_feedback_forward_relative_error": 0.021948661625144944,
        "min_feedback_forward_cosine": 0.9992228746414185
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.030628358398377895
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039492469280958176,
        "mean_feedback_forward_cosine": 0.9997476935386658,
        "mean_feedback_forward_relative_error": 0.02194429343686469,
        "min_feedback_forward_cosine": 0.9992231130599976
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.029325532891750335
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03948645293712616,
        "mean_feedback_forward_cosine": 0.9997477839069981,
        "mean_feedback_forward_relative_error": 0.021939773322834123,
        "min_feedback_forward_cosine": 0.9992232918739319
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.029357269192934035
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03948098048567772,
        "mean_feedback_forward_cosine": 0.9997478608162172,
        "mean_feedback_forward_relative_error": 0.021935472595355204,
        "min_feedback_forward_cosine": 0.9992235898971558
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.026094600776433944
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03947504609823227,
        "mean_feedback_forward_cosine": 0.9997479415708973,
        "mean_feedback_forward_relative_error": 0.021931041062118545,
        "min_feedback_forward_cosine": 0.9992238283157349
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.02651489243924618
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039468780159950256,
        "mean_feedback_forward_cosine": 0.9997480242483078,
        "mean_feedback_forward_relative_error": 0.02192651267133413,
        "min_feedback_forward_cosine": 0.999224066734314
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.02624331370830536
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039462897926568985,
        "mean_feedback_forward_cosine": 0.9997481069257182,
        "mean_feedback_forward_relative_error": 0.021922080432094874,
        "min_feedback_forward_cosine": 0.9992242455482483
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.024278797640800476
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039457205682992935,
        "mean_feedback_forward_cosine": 0.9997481857576678,
        "mean_feedback_forward_relative_error": 0.02191784540792146,
        "min_feedback_forward_cosine": 0.9992244243621826
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.022229577784240246
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03945150226354599,
        "mean_feedback_forward_cosine": 0.9997482799714611,
        "mean_feedback_forward_relative_error": 0.021913597675701304,
        "min_feedback_forward_cosine": 0.9992246031761169
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.02159910560131073
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0394456572830677,
        "mean_feedback_forward_cosine": 0.9997483549579498,
        "mean_feedback_forward_relative_error": 0.02190916364892356,
        "min_feedback_forward_cosine": 0.999224841594696
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.021546049241423608
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03943952918052673,
        "mean_feedback_forward_cosine": 0.9997484472490126,
        "mean_feedback_forward_relative_error": 0.02190469577908516,
        "min_feedback_forward_cosine": 0.9992251396179199
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.021970328575372695
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03943442925810814,
        "mean_feedback_forward_cosine": 0.9997485318491536,
        "mean_feedback_forward_relative_error": 0.021900775301600655,
        "min_feedback_forward_cosine": 0.999225378036499
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.017999292779564857
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039428383111953735,
        "mean_feedback_forward_cosine": 0.9997486126038336,
        "mean_feedback_forward_relative_error": 0.021896474138503114,
        "min_feedback_forward_cosine": 0.9992256164550781
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.018960947802364826
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0394226498901844,
        "mean_feedback_forward_cosine": 0.9997487202767403,
        "mean_feedback_forward_relative_error": 0.0218921925632223,
        "min_feedback_forward_cosine": 0.999225914478302
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.018468987340182066
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039417196065187454,
        "mean_feedback_forward_cosine": 0.9997487798813851,
        "mean_feedback_forward_relative_error": 0.02188799973397005,
        "min_feedback_forward_cosine": 0.9992260932922363
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.016994322579205037
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039411451667547226,
        "mean_feedback_forward_cosine": 0.9997488490996822,
        "mean_feedback_forward_relative_error": 0.02188394600225072,
        "min_feedback_forward_cosine": 0.9992262721061707
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.017177284666597844
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039405569434165955,
        "mean_feedback_forward_cosine": 0.999748949081667,
        "mean_feedback_forward_relative_error": 0.021879751670865283,
        "min_feedback_forward_cosine": 0.999226450920105
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.01637926060795784
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03939993306994438,
        "mean_feedback_forward_cosine": 0.9997490086863118,
        "mean_feedback_forward_relative_error": 0.02187567081062063,
        "min_feedback_forward_cosine": 0.9992266893386841
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.015048764585107566
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039394695311784744,
        "mean_feedback_forward_cosine": 0.9997491086682966,
        "mean_feedback_forward_relative_error": 0.021871750483349446,
        "min_feedback_forward_cosine": 0.9992269277572632
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.014844148950874806
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03938871994614601,
        "mean_feedback_forward_cosine": 0.9997491605820195,
        "mean_feedback_forward_relative_error": 0.02186735495624523,
        "min_feedback_forward_cosine": 0.9992271661758423
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.015637938914895056
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03938230872154236,
        "mean_feedback_forward_cosine": 0.9997492721003871,
        "mean_feedback_forward_relative_error": 0.021862772293388844,
        "min_feedback_forward_cosine": 0.9992274045944214
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.015976422900632024
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03937559574842453,
        "mean_feedback_forward_cosine": 0.9997494028460595,
        "mean_feedback_forward_relative_error": 0.021858133405687347,
        "min_feedback_forward_cosine": 0.9992275834083557
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.015033096481263638
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03936991095542908,
        "mean_feedback_forward_cosine": 0.9997494566825128,
        "mean_feedback_forward_relative_error": 0.021853979556791244,
        "min_feedback_forward_cosine": 0.9992278218269348
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.013741299457252026
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03936389833688736,
        "mean_feedback_forward_cosine": 0.9997495451281148,
        "mean_feedback_forward_relative_error": 0.021849748873782736,
        "min_feedback_forward_cosine": 0.9992280602455139
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.013263638440668583
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03935781866312027,
        "mean_feedback_forward_cosine": 0.9997496181918729,
        "mean_feedback_forward_relative_error": 0.02184542185897308,
        "min_feedback_forward_cosine": 0.9992283582687378
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.013716067549288273
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03935127705335617,
        "mean_feedback_forward_cosine": 0.9997497297102406,
        "mean_feedback_forward_relative_error": 0.02184095559641719,
        "min_feedback_forward_cosine": 0.9992285966873169
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.01336184063643217
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03934511914849281,
        "mean_feedback_forward_cosine": 0.9997498085421901,
        "mean_feedback_forward_relative_error": 0.021836617841355262,
        "min_feedback_forward_cosine": 0.9992287158966064
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.013175583146512508
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039338357746601105,
        "mean_feedback_forward_cosine": 0.99974988929687,
        "mean_feedback_forward_relative_error": 0.021831967462334904,
        "min_feedback_forward_cosine": 0.9992291331291199
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.01325496376425028
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03933264687657356,
        "mean_feedback_forward_cosine": 0.9997499758197416,
        "mean_feedback_forward_relative_error": 0.02182770337188436,
        "min_feedback_forward_cosine": 0.999229371547699
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.011746058116555214
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03932664915919304,
        "mean_feedback_forward_cosine": 0.9997500834926483,
        "mean_feedback_forward_relative_error": 0.021823468017241648,
        "min_feedback_forward_cosine": 0.9992295503616333
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.011617636746764182
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03932031989097595,
        "mean_feedback_forward_cosine": 0.9997501873200939,
        "mean_feedback_forward_relative_error": 0.021819020591435895,
        "min_feedback_forward_cosine": 0.9992297887802124
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.01183700292006135
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03931359946727753,
        "mean_feedback_forward_cosine": 0.9997502776884264,
        "mean_feedback_forward_relative_error": 0.02181434832633503,
        "min_feedback_forward_cosine": 0.9992299675941467
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.012418612503856421
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03930603340268135,
        "mean_feedback_forward_cosine": 0.9997503795931416,
        "mean_feedback_forward_relative_error": 0.021809443335739837,
        "min_feedback_forward_cosine": 0.9992303848266602
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.01267631758466363
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03929959982633591,
        "mean_feedback_forward_cosine": 0.999750466116013,
        "mean_feedback_forward_relative_error": 0.021804695378147786,
        "min_feedback_forward_cosine": 0.9992306232452393
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.01170698377624154
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03929335996508598,
        "mean_feedback_forward_cosine": 0.999750550716154,
        "mean_feedback_forward_relative_error": 0.02180038312930734,
        "min_feedback_forward_cosine": 0.9992308020591736
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.010822395001798869
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03928709775209427,
        "mean_feedback_forward_cosine": 0.9997506680027131,
        "mean_feedback_forward_relative_error": 0.021795969892052874,
        "min_feedback_forward_cosine": 0.9992311000823975
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.010739987428039312
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03928085416555405,
        "mean_feedback_forward_cosine": 0.9997507718301588,
        "mean_feedback_forward_relative_error": 0.021791406035903963,
        "min_feedback_forward_cosine": 0.9992313385009766
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.010964248820766806
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03927396982908249,
        "mean_feedback_forward_cosine": 0.9997508621984913,
        "mean_feedback_forward_relative_error": 0.02178660711093295,
        "min_feedback_forward_cosine": 0.9992315769195557
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.011047148571312428
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03926721587777138,
        "mean_feedback_forward_cosine": 0.999750962180476,
        "mean_feedback_forward_relative_error": 0.02178180754004467,
        "min_feedback_forward_cosine": 0.9992318749427795
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.01048615313872695
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03926093131303787,
        "mean_feedback_forward_cosine": 0.9997510583169998,
        "mean_feedback_forward_relative_error": 0.021777256256750515,
        "min_feedback_forward_cosine": 0.9992320537567139
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.009832702446728944
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253976196050644,
        "mean_feedback_forward_cosine": 0.9997511563762542,
        "mean_feedback_forward_relative_error": 0.021772383486912135,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.010426850525736808
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925387188792229,
        "mean_feedback_forward_cosine": 0.9997511506080627,
        "mean_feedback_forward_relative_error": 0.021772278577930505,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.009929457244426012
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925381973385811,
        "mean_feedback_forward_cosine": 0.9997511506080627,
        "mean_feedback_forward_relative_error": 0.021772196531415947,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.008251808605119586
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925376012921333,
        "mean_feedback_forward_cosine": 0.9997511525307933,
        "mean_feedback_forward_relative_error": 0.021772107770366055,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.008543098836168647
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925371170043945,
        "mean_feedback_forward_cosine": 0.9997511582989846,
        "mean_feedback_forward_relative_error": 0.02177202533329687,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.00813495438784361
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253655821084976,
        "mean_feedback_forward_cosine": 0.9997511544535237,
        "mean_feedback_forward_relative_error": 0.021771943076483665,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.007969489766955376
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925361484289169,
        "mean_feedback_forward_cosine": 0.9997511525307933,
        "mean_feedback_forward_relative_error": 0.02177186980242691,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.007356664264649153
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253562688827515,
        "mean_feedback_forward_cosine": 0.9997511486853322,
        "mean_feedback_forward_relative_error": 0.021771795101343625,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.007573647451475263
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253514260053635,
        "mean_feedback_forward_cosine": 0.9997511640671761,
        "mean_feedback_forward_relative_error": 0.021771720129876368,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.007154074359536171
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253465831279755,
        "mean_feedback_forward_cosine": 0.9997511602217152,
        "mean_feedback_forward_relative_error": 0.021771639885921634,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.0077800188639760015
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925343230366707,
        "mean_feedback_forward_cosine": 0.9997511698353675,
        "mean_feedback_forward_relative_error": 0.021771564493857084,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.006432479541748762
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925338014960289,
        "mean_feedback_forward_cosine": 0.999751171758098,
        "mean_feedback_forward_relative_error": 0.021771485857184854,
        "min_feedback_forward_cosine": 0.9992324709892273
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.007193801200687885
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253346621990204,
        "mean_feedback_forward_cosine": 0.9997511602217152,
        "mean_feedback_forward_relative_error": 0.02177141136640022,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.006947582246512174
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925330564379692,
        "mean_feedback_forward_cosine": 0.999751171758098,
        "mean_feedback_forward_relative_error": 0.021771336409954295,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.006525301996283233
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925326094031334,
        "mean_feedback_forward_cosine": 0.9997511390716799,
        "mean_feedback_forward_relative_error": 0.021771262655214917,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.007157000288069248
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925321251153946,
        "mean_feedback_forward_cosine": 0.9997511659899065,
        "mean_feedback_forward_relative_error": 0.021771182366196188,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.00680718580968678
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253171533346176,
        "mean_feedback_forward_cosine": 0.9997511563762542,
        "mean_feedback_forward_relative_error": 0.021771103459139986,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.007068686625491828
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039253126829862595,
        "mean_feedback_forward_cosine": 0.9997511640671761,
        "mean_feedback_forward_relative_error": 0.021771032303091017,
        "min_feedback_forward_cosine": 0.999232292175293
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.006626084240972996
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925308212637901,
        "mean_feedback_forward_cosine": 0.9997511563762542,
        "mean_feedback_forward_relative_error": 0.021770962003257968,
        "min_feedback_forward_cosine": 0.999232292175293
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.006670517491921783
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925303742289543,
        "mean_feedback_forward_cosine": 0.999751175603559,
        "mean_feedback_forward_relative_error": 0.021770884643398946,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.0069802900040894745
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925300016999245,
        "mean_feedback_forward_cosine": 0.9997511659899065,
        "mean_feedback_forward_relative_error": 0.02177081188006747,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.006564090129733085
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925294801592827,
        "mean_feedback_forward_cosine": 0.9997511621444456,
        "mean_feedback_forward_relative_error": 0.021770732687605966,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.007198657158911229
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925291821360588,
        "mean_feedback_forward_cosine": 0.9997511775262894,
        "mean_feedback_forward_relative_error": 0.021770659879210494,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.006527402885854244
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0392528772354126,
        "mean_feedback_forward_cosine": 0.9997511602217152,
        "mean_feedback_forward_relative_error": 0.02177059237334517,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.006383647970482707
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039252836257219315,
        "mean_feedback_forward_cosine": 0.999751171758098,
        "mean_feedback_forward_relative_error": 0.021770518468392473,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.0067850247360765935
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925279527902603,
        "mean_feedback_forward_cosine": 0.9997511659899065,
        "mean_feedback_forward_relative_error": 0.021770449295159307,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.005945792019739747
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039252739399671555,
        "mean_feedback_forward_cosine": 0.9997511813717503,
        "mean_feedback_forward_relative_error": 0.02177037408335074,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.007371395756900311
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925269842147827,
        "mean_feedback_forward_cosine": 0.9997511775262894,
        "mean_feedback_forward_relative_error": 0.02177030530067221,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.006610088579505682
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925265371799469,
        "mean_feedback_forward_cosine": 0.9997511852172113,
        "mean_feedback_forward_relative_error": 0.021770233543769966,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.006482609926313162
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925260901451111,
        "mean_feedback_forward_cosine": 0.9997511794490199,
        "mean_feedback_forward_relative_error": 0.021770161351249102,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.006312023378238082
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039252568036317825,
        "mean_feedback_forward_cosine": 0.9997511775262894,
        "mean_feedback_forward_relative_error": 0.021770090195200136,
        "min_feedback_forward_cosine": 0.9992324709892273
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.006113427438139915
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925253078341484,
        "mean_feedback_forward_cosine": 0.9997511948308637,
        "mean_feedback_forward_relative_error": 0.021770016981228706,
        "min_feedback_forward_cosine": 0.9992324709892273
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.006099065319225192
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925248980522156,
        "mean_feedback_forward_cosine": 0.9997511640671761,
        "mean_feedback_forward_relative_error": 0.021769947252206264,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.006097415303662419
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925245255231857,
        "mean_feedback_forward_cosine": 0.9997511602217152,
        "mean_feedback_forward_relative_error": 0.021769877207735853,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.005813903130888939
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925241157412529,
        "mean_feedback_forward_cosine": 0.999751175603559,
        "mean_feedback_forward_relative_error": 0.021769803137548508,
        "min_feedback_forward_cosine": 0.9992324709892273
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.006274959384575486
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039252378046512604,
        "mean_feedback_forward_cosine": 0.9997511659899065,
        "mean_feedback_forward_relative_error": 0.021769732972907442,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.00568594042815268
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925233706831932,
        "mean_feedback_forward_cosine": 0.9997511852172113,
        "mean_feedback_forward_relative_error": 0.02176965893276276,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.006060547957420349
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925229609012604,
        "mean_feedback_forward_cosine": 0.9997511852172113,
        "mean_feedback_forward_relative_error": 0.021769587866841786,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.006110777246356011
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039252255111932755,
        "mean_feedback_forward_cosine": 0.9997511929081332,
        "mean_feedback_forward_relative_error": 0.021769519489739212,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.006131593030355871
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925221785902977,
        "mean_feedback_forward_cosine": 0.9997511813717503,
        "mean_feedback_forward_relative_error": 0.02176945082723133,
        "min_feedback_forward_cosine": 0.9992325305938721
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.0064618229238130155
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039252180606126785,
        "mean_feedback_forward_cosine": 0.9997511929081332,
        "mean_feedback_forward_relative_error": 0.021769387962957545,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.005390891584306955
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0392521396279335,
        "mean_feedback_forward_cosine": 0.9997511967535941,
        "mean_feedback_forward_relative_error": 0.021769315169583404,
        "min_feedback_forward_cosine": 0.9992323517799377
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.006207142435461283
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925210237503052,
        "mean_feedback_forward_cosine": 0.9997511909854028,
        "mean_feedback_forward_relative_error": 0.021769242871913216,
        "min_feedback_forward_cosine": 0.9992324709892273
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.005915933236479759
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039252061396837234,
        "mean_feedback_forward_cosine": 0.9997511602217152,
        "mean_feedback_forward_relative_error": 0.021769171655778925,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.005669431590288878
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925202786922455,
        "mean_feedback_forward_cosine": 0.9997511890626722,
        "mean_feedback_forward_relative_error": 0.021769103158505693,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.005420221343487501
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039251990616321564,
        "mean_feedback_forward_cosine": 0.9997511813717503,
        "mean_feedback_forward_relative_error": 0.02176903739511486,
        "min_feedback_forward_cosine": 0.9992324709892273
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.005751352581679821
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925195336341858,
        "mean_feedback_forward_cosine": 0.9997511852172113,
        "mean_feedback_forward_relative_error": 0.02176897052014547,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.005610074019469321
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039251916110515594,
        "mean_feedback_forward_cosine": 0.9997511890626722,
        "mean_feedback_forward_relative_error": 0.021768901887680252,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.005644427023380995
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925187885761261,
        "mean_feedback_forward_cosine": 0.999751204444516,
        "mean_feedback_forward_relative_error": 0.021768835012710864,
        "min_feedback_forward_cosine": 0.9992324113845825
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.005606720813140273
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03925183787941933,
        "mean_feedback_forward_cosine": 0.9997511813717503,
        "mean_feedback_forward_relative_error": 0.02176876387168323,
        "min_feedback_forward_cosine": 0.9992324709892273
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.0059318994602561
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.039251793175935745,
        "mean_feedback_forward_cosine": 0.9997512025217856,
        "mean_feedback_forward_relative_error": 0.021768693061124895,
        "min_feedback_forward_cosine": 0.9992325305938721
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.0059052299538999795
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9209,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.4008916320800781,
    "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": 12,
    "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": 1.15913987159729,
    "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": 10225.321983337402,
    "train_wall_s": 10222.131490945816
  },
  "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
  }
}