{
  "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": 13,
    "lr": 0.1,
    "lr_gamma": 0.1,
    "lr_milestones": "100,150",
    "lr_schedule": "step",
    "max_steps": 0,
    "mirror_batch_size": 1,
    "mirror_eta": 0.1,
    "mirror_every": 16,
    "mirror_noise_std": 1.0,
    "mirror_seed": 3000,
    "mirror_warmup_steps": 0,
    "mode": "kp",
    "momentum": 0.9,
    "neutral_projection": 0,
    "normalization": "batchnorm",
    "nuisance_scale": 0.0,
    "out": "results/oral_a_dynamic_scaling_v2/clean_kp_d32_s13.json",
    "output_lr": 0.1,
    "pert_directions": 1,
    "pert_every": 4,
    "pert_sigma": 0.01,
    "perturb_seed": 1000,
    "predictor_every": 0,
    "predictor_mode": "nlms",
    "predictor_warmup_steps": 0,
    "residual_scale": null,
    "seed": 13,
    "split_seed": 2027,
    "traffic_calibration_examples": 64,
    "traffic_ratio": 4.0,
    "traffic_rule": "innovation",
    "traffic_seed": 4000,
    "train_limit": 0,
    "use_residual": 1,
    "val_examples": 0,
    "vectorizer_mode": "spatial_template",
    "warmup_epochs": 0,
    "weight_decay": 0.0001,
    "weight_scale": 1.0,
    "width": 16
  },
  "calibration_metric_space": "reciprocal_local_activity_products",
  "counters": {
    "apical_warmup_examples": 0,
    "calibration_event_examples": 0,
    "causal_scalar_observations": 0,
    "logical_batch_loss_queries": 0,
    "mirror_conv_examples": 0,
    "mirror_events": 0,
    "mirror_readout_examples": 0,
    "neutral_projection_examples": 0,
    "ordinary_examples": 10000000,
    "per_example_loss_terms": 0,
    "perturbation_events": 0,
    "perturbation_forward_examples": 0,
    "predictor_update_examples": 0,
    "predictor_warmup_examples": 0,
    "traffic_audit_examples": 0,
    "traffic_calibration_examples": 0
  },
  "diagnostics": {
    "early_third_mean": 0.9996044099330902,
    "feedback_forward_cosine": [
      0.9998302459716797,
      0.9998152256011963,
      0.9996967315673828,
      0.9996758699417114,
      0.9997156262397766,
      0.9996615648269653,
      0.999754011631012,
      0.9996933341026306,
      0.9997844099998474,
      0.9996684789657593,
      0.9997125864028931,
      0.9997879266738892,
      0.9997588396072388,
      0.9997154474258423,
      0.9997693300247192,
      0.9997178316116333,
      0.9997802972793579,
      0.9997060298919678,
      0.9997598528862,
      0.9996607303619385,
      0.9997869729995728,
      0.9998668432235718,
      0.9998339414596558,
      0.9998186230659485,
      0.9998288154602051,
      0.9997850060462952,
      0.9998162984848022,
      0.9997274279594421,
      0.999793291091919,
      0.9995461106300354,
      0.9999839663505554
    ],
    "feedback_forward_norm_ratio": [
      0.9998724460601807,
      1.0005196332931519,
      0.9995152950286865,
      1.001641869544983,
      0.999545693397522,
      1.0011926889419556,
      0.9998229146003723,
      1.0016809701919556,
      0.9991549253463745,
      1.0004255771636963,
      0.9986014366149902,
      1.0009064674377441,
      0.999552845954895,
      1.0012096166610718,
      0.9988930225372314,
      1.0015215873718262,
      0.9993488788604736,
      1.001158356666565,
      0.9994612336158752,
      1.001421332359314,
      0.9993331432342529,
      1.000417709350586,
      0.9997079968452454,
      1.0008820295333862,
      0.9995848536491394,
      1.001044511795044,
      0.9995478987693787,
      1.001582384109497,
      0.9999879598617554,
      1.0011634826660156,
      1.0025697946548462
    ],
    "feedback_forward_relative_error": [
      0.018421653658151627,
      0.01923324540257454,
      0.024623440578579903,
      0.02553260512650013,
      0.0238448865711689,
      0.02605927363038063,
      0.022181058302521706,
      0.024846022948622704,
      0.020773712545633316,
      0.025759179145097733,
      0.024001268669962883,
      0.020627934485673904,
      0.02196275070309639,
      0.02390001341700554,
      0.021497350186109543,
      0.02382095344364643,
      0.02096632681787014,
      0.024285070598125458,
      0.02191520668566227,
      0.026104094460606575,
      0.020649723708629608,
      0.016331132501363754,
      0.01822110265493393,
      0.019073737785220146,
      0.018504101783037186,
      0.020774221047759056,
      0.019169632345438004,
      0.023417465388774872,
      0.020336169749498367,
      0.030168624594807625,
      0.006229218095541
    ],
    "innovation_negative_gradient_cosine": [
      0.9996152520179749,
      0.9995561242103577,
      0.9996229410171509,
      0.9995366930961609,
      0.9996248483657837,
      0.9995720386505127,
      0.9996442198753357,
      0.999592661857605,
      0.9996510744094849,
      0.9996282458305359,
      0.9996576905250549,
      0.9996254444122314,
      0.9996774792671204,
      0.9995726943016052,
      0.999693751335144,
      0.9995989799499512,
      0.9997098445892334,
      0.9996525645256042,
      0.9997299909591675,
      0.999668300151825,
      0.9997332692146301,
      0.9997650980949402,
      0.9997707009315491,
      0.9997556209564209,
      0.9997986555099487,
      0.9997751116752625,
      0.9998295307159424,
      0.9998553991317749,
      0.9998940229415894,
      0.9999567270278931,
      0.8186337351799011
    ],
    "mean_feedback_forward_cosine": 0.9997565054124401,
    "mean_feedback_forward_relative_error": 0.021717134742967544,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.9996152520179749,
      0.9995561242103577,
      0.9996229410171509,
      0.9995366930961609,
      0.9996248483657837,
      0.9995720386505127,
      0.9996442198753357,
      0.999592661857605,
      0.9996510744094849,
      0.9996282458305359,
      0.9996576905250549,
      0.9996254444122314,
      0.9996774792671204,
      0.9995726943016052,
      0.999693751335144,
      0.9995989799499512,
      0.9997098445892334,
      0.9996525645256042,
      0.9997299909591675,
      0.999668300151825,
      0.9997332692146301,
      0.9997650980949402,
      0.9997707009315491,
      0.9997556209564209,
      0.9997986555099487,
      0.9997751116752625,
      0.9998295307159424,
      0.9998553991317749,
      0.9998940229415894,
      0.9999567270278931,
      0.8186337351799011
    ],
    "teaching_negative_gradient_cosine": [
      0.9996152520179749,
      0.9995561242103577,
      0.9996229410171509,
      0.9995366930961609,
      0.9996248483657837,
      0.9995720386505127,
      0.9996442198753357,
      0.999592661857605,
      0.9996510744094849,
      0.9996282458305359,
      0.9996576905250549,
      0.9996254444122314,
      0.9996774792671204,
      0.9995726943016052,
      0.999693751335144,
      0.9995989799499512,
      0.9997098445892334,
      0.9996525645256042,
      0.9997299909591675,
      0.999668300151825,
      0.9997332692146301,
      0.9997650980949402,
      0.9997707009315491,
      0.9997556209564209,
      0.9997986555099487,
      0.9997751116752625,
      0.9998295307159424,
      0.9998553991317749,
      0.9998940229415894,
      0.9999567270278931,
      0.8186337351799011
    ],
    "wall_s": 0.11170530319213867
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3389378786087036,
        "mean_feedback_forward_cosine": 0.23121365664466734,
        "mean_feedback_forward_relative_error": 1.240019892492602,
        "min_feedback_forward_cosine": 0.1128765344619751
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 1.9838821460723877
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2985464334487915,
        "mean_feedback_forward_cosine": 0.32630072918630415,
        "mean_feedback_forward_relative_error": 1.1610169679887834,
        "min_feedback_forward_cosine": 0.16839854419231415
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 1.3787971196746827
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2493191957473755,
        "mean_feedback_forward_cosine": 0.41258884005008206,
        "mean_feedback_forward_relative_error": 1.0848919903078387,
        "min_feedback_forward_cosine": 0.23476368188858032
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.0900519537353515
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1967850923538208,
        "mean_feedback_forward_cosine": 0.48254168033599854,
        "mean_feedback_forward_relative_error": 1.0189406583386083,
        "min_feedback_forward_cosine": 0.30199190974235535
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 0.906940527381897
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.160981297492981,
        "mean_feedback_forward_cosine": 0.538169567623446,
        "mean_feedback_forward_relative_error": 0.9634415920703642,
        "min_feedback_forward_cosine": 0.3558718264102936
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 0.77196553440094
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1286283731460571,
        "mean_feedback_forward_cosine": 0.5838505808384188,
        "mean_feedback_forward_relative_error": 0.9149952463565334,
        "min_feedback_forward_cosine": 0.39263737201690674
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 0.6873360015869141
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0962344408035278,
        "mean_feedback_forward_cosine": 0.6231732560742286,
        "mean_feedback_forward_relative_error": 0.8710511913222652,
        "min_feedback_forward_cosine": 0.4282245635986328
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.6295898027706146
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0658029317855835,
        "mean_feedback_forward_cosine": 0.658060630483012,
        "mean_feedback_forward_relative_error": 0.8297388909324523,
        "min_feedback_forward_cosine": 0.460544615983963
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.5772971540260315
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0356392860412598,
        "mean_feedback_forward_cosine": 0.6879042973441463,
        "mean_feedback_forward_relative_error": 0.7924710483320297,
        "min_feedback_forward_cosine": 0.4916229546070099
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.538321427745819
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0061553716659546,
        "mean_feedback_forward_cosine": 0.7145854273150044,
        "mean_feedback_forward_relative_error": 0.7575919608916005,
        "min_feedback_forward_cosine": 0.5206334590911865
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.5051441788768768
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.976460874080658,
        "mean_feedback_forward_cosine": 0.738428444631638,
        "mean_feedback_forward_relative_error": 0.7249820914960676,
        "min_feedback_forward_cosine": 0.5487823486328125
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.48193352809906004
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9464977383613586,
        "mean_feedback_forward_cosine": 0.7600205559884349,
        "mean_feedback_forward_relative_error": 0.6940322133802599,
        "min_feedback_forward_cosine": 0.5761566162109375
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.4617367621898651
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9177645444869995,
        "mean_feedback_forward_cosine": 0.779275276968556,
        "mean_feedback_forward_relative_error": 0.6655332003870318,
        "min_feedback_forward_cosine": 0.6014045476913452
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.4340008737373352
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8895062208175659,
        "mean_feedback_forward_cosine": 0.7972497728563124,
        "mean_feedback_forward_relative_error": 0.6375727105525232,
        "min_feedback_forward_cosine": 0.6256616115570068
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.4249182174873352
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8613275289535522,
        "mean_feedback_forward_cosine": 0.8134563046116983,
        "mean_feedback_forward_relative_error": 0.6113178268555672,
        "min_feedback_forward_cosine": 0.6489408016204834
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.40700346908569335
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8324474692344666,
        "mean_feedback_forward_cosine": 0.8282331266710835,
        "mean_feedback_forward_relative_error": 0.5862936689969032,
        "min_feedback_forward_cosine": 0.6718162298202515
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.3925573239707947
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8057132363319397,
        "mean_feedback_forward_cosine": 0.8414294950423702,
        "mean_feedback_forward_relative_error": 0.5631513062023348,
        "min_feedback_forward_cosine": 0.692278265953064
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.37800764578819274
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.780025064945221,
        "mean_feedback_forward_cosine": 0.8532825554570844,
        "mean_feedback_forward_relative_error": 0.5414041290360112,
        "min_feedback_forward_cosine": 0.7115788459777832
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.35983910435676575
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7540488243103027,
        "mean_feedback_forward_cosine": 0.8645286733104337,
        "mean_feedback_forward_relative_error": 0.5199659630175559,
        "min_feedback_forward_cosine": 0.7304697632789612
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.3523705890464783
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7285679578781128,
        "mean_feedback_forward_cosine": 0.8747497585511976,
        "mean_feedback_forward_relative_error": 0.49976667088847004,
        "min_feedback_forward_cosine": 0.7482064962387085
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.3404514719581604
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7023472785949707,
        "mean_feedback_forward_cosine": 0.8845681375072848,
        "mean_feedback_forward_relative_error": 0.47956819447778887,
        "min_feedback_forward_cosine": 0.7659227848052979
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.3370584138774872
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6784579157829285,
        "mean_feedback_forward_cosine": 0.8931465956472582,
        "mean_feedback_forward_relative_error": 0.4611505793948327,
        "min_feedback_forward_cosine": 0.7814069986343384
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.32296759687423704
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6543031930923462,
        "mean_feedback_forward_cosine": 0.9012511468702747,
        "mean_feedback_forward_relative_error": 0.4430700550156255,
        "min_feedback_forward_cosine": 0.7964164018630981
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.31611102496147153
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6313464641571045,
        "mean_feedback_forward_cosine": 0.9085270089487876,
        "mean_feedback_forward_relative_error": 0.4261899431866984,
        "min_feedback_forward_cosine": 0.8101376295089722
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.3068838739681244
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6094977855682373,
        "mean_feedback_forward_cosine": 0.9153859999872023,
        "mean_feedback_forward_relative_error": 0.4096834736485635,
        "min_feedback_forward_cosine": 0.822838306427002
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.3061916961765289
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5876675844192505,
        "mean_feedback_forward_cosine": 0.9217286821334593,
        "mean_feedback_forward_relative_error": 0.393863130961695,
        "min_feedback_forward_cosine": 0.8350555300712585
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.2910228369140625
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5656070113182068,
        "mean_feedback_forward_cosine": 0.9276147388642834,
        "mean_feedback_forward_relative_error": 0.37859375582587335,
        "min_feedback_forward_cosine": 0.8469791412353516
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.28525240130901336
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5441515445709229,
        "mean_feedback_forward_cosine": 0.9331032537644909,
        "mean_feedback_forward_relative_error": 0.36385476204656786,
        "min_feedback_forward_cosine": 0.8581732511520386
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.28182858699798585
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5238852500915527,
        "mean_feedback_forward_cosine": 0.9381125915435052,
        "mean_feedback_forward_relative_error": 0.34983458105594883,
        "min_feedback_forward_cosine": 0.8683308362960815
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.27555020067214964
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5047932267189026,
        "mean_feedback_forward_cosine": 0.9428141578551261,
        "mean_feedback_forward_relative_error": 0.33617491443311015,
        "min_feedback_forward_cosine": 0.8776401281356812
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.2702294630479813
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.48609381914138794,
        "mean_feedback_forward_cosine": 0.9470732750431183,
        "mean_feedback_forward_relative_error": 0.3233049591222117,
        "min_feedback_forward_cosine": 0.8864260911941528
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.26724182792663576
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.46777960658073425,
        "mean_feedback_forward_cosine": 0.9510572341180616,
        "mean_feedback_forward_relative_error": 0.31077751204852133,
        "min_feedback_forward_cosine": 0.8946413993835449
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.2642540150499344
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4505240321159363,
        "mean_feedback_forward_cosine": 0.954611559068003,
        "mean_feedback_forward_relative_error": 0.29910426995446604,
        "min_feedback_forward_cosine": 0.9021570086479187
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.2528244213104248
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4330817759037018,
        "mean_feedback_forward_cosine": 0.9580516219139099,
        "mean_feedback_forward_relative_error": 0.28740497630450035,
        "min_feedback_forward_cosine": 0.9094547033309937
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.2537894388771057
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4169538915157318,
        "mean_feedback_forward_cosine": 0.9611637938407159,
        "mean_feedback_forward_relative_error": 0.27640540441197736,
        "min_feedback_forward_cosine": 0.9159486293792725
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.2468545224380493
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4000706374645233,
        "mean_feedback_forward_cosine": 0.9641662413074125,
        "mean_feedback_forward_relative_error": 0.26542171835899353,
        "min_feedback_forward_cosine": 0.9225339889526367
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.24769609283447266
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3847455084323883,
        "mean_feedback_forward_cosine": 0.9668317110307755,
        "mean_feedback_forward_relative_error": 0.25530476531674784,
        "min_feedback_forward_cosine": 0.9282823801040649
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.24280201545238494
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.36994194984436035,
        "mean_feedback_forward_cosine": 0.9692680258904735,
        "mean_feedback_forward_relative_error": 0.24568897389596508,
        "min_feedback_forward_cosine": 0.9336087703704834
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.23785523980140685
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.35572385787963867,
        "mean_feedback_forward_cosine": 0.9715897787001825,
        "mean_feedback_forward_relative_error": 0.2361768770121759,
        "min_feedback_forward_cosine": 0.9385188817977905
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.23314270420074462
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3419283628463745,
        "mean_feedback_forward_cosine": 0.9737087891947839,
        "mean_feedback_forward_relative_error": 0.22715740242312032,
        "min_feedback_forward_cosine": 0.9431262016296387
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.22954041654586793
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3289780020713806,
        "mean_feedback_forward_cosine": 0.9756121846937365,
        "mean_feedback_forward_relative_error": 0.21870481823721238,
        "min_feedback_forward_cosine": 0.9472616910934448
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.22527263911247253
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.31575897336006165,
        "mean_feedback_forward_cosine": 0.9774580540195588,
        "mean_feedback_forward_relative_error": 0.21023723386949109,
        "min_feedback_forward_cosine": 0.9514015316963196
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.22636460013389587
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3028050661087036,
        "mean_feedback_forward_cosine": 0.9791820549195812,
        "mean_feedback_forward_relative_error": 0.201970340263459,
        "min_feedback_forward_cosine": 0.9552354216575623
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.2263233102607727
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2906534969806671,
        "mean_feedback_forward_cosine": 0.980764873566166,
        "mean_feedback_forward_relative_error": 0.1941150179313075,
        "min_feedback_forward_cosine": 0.9587188959121704
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.22183988755226136
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2791547179222107,
        "mean_feedback_forward_cosine": 0.9821782688940724,
        "mean_feedback_forward_relative_error": 0.1868205111353628,
        "min_feedback_forward_cosine": 0.9618908166885376
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.21705163217544557
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.26807472109794617,
        "mean_feedback_forward_cosine": 0.9835057720061271,
        "mean_feedback_forward_relative_error": 0.17969327756474096,
        "min_feedback_forward_cosine": 0.9648392200469971
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.21400939811706543
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2570335268974304,
        "mean_feedback_forward_cosine": 0.9847611227343159,
        "mean_feedback_forward_relative_error": 0.17269634118964594,
        "min_feedback_forward_cosine": 0.9676309823989868
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.21496672592163085
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2468288391828537,
        "mean_feedback_forward_cosine": 0.9858841088510328,
        "mean_feedback_forward_relative_error": 0.1661904861609782,
        "min_feedback_forward_cosine": 0.9701247811317444
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.21069812502861024
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.23741383850574493,
        "mean_feedback_forward_cosine": 0.9869118948136607,
        "mean_feedback_forward_relative_error": 0.15998160106039816,
        "min_feedback_forward_cosine": 0.9723424911499023
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.20564152919769288
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22796526551246643,
        "mean_feedback_forward_cosine": 0.9878933891173332,
        "mean_feedback_forward_relative_error": 0.15384173345181248,
        "min_feedback_forward_cosine": 0.9744757413864136
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.20788719606876374
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.21927353739738464,
        "mean_feedback_forward_cosine": 0.9887850899850169,
        "mean_feedback_forward_relative_error": 0.14804859411331914,
        "min_feedback_forward_cosine": 0.9763659238815308
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.20169600876808166
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2106843739748001,
        "mean_feedback_forward_cosine": 0.9896110103976342,
        "mean_feedback_forward_relative_error": 0.14246293746175304,
        "min_feedback_forward_cosine": 0.9781721234321594
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.20488964691638947
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.20236633718013763,
        "mean_feedback_forward_cosine": 0.9903933617376512,
        "mean_feedback_forward_relative_error": 0.13697626093222248,
        "min_feedback_forward_cosine": 0.9798451066017151
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.20032569314479828
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1942722648382187,
        "mean_feedback_forward_cosine": 0.9911144433483001,
        "mean_feedback_forward_relative_error": 0.13171462486347846,
        "min_feedback_forward_cosine": 0.9814128875732422
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.1995404206466675
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18633601069450378,
        "mean_feedback_forward_cosine": 0.9917949649595446,
        "mean_feedback_forward_relative_error": 0.126550595967039,
        "min_feedback_forward_cosine": 0.9828876256942749
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.20286603238105774
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17927993834018707,
        "mean_feedback_forward_cosine": 0.9923889463947665,
        "mean_feedback_forward_relative_error": 0.12187032990397946,
        "min_feedback_forward_cosine": 0.9841449856758118
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.19383209038257598
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17214931547641754,
        "mean_feedback_forward_cosine": 0.9929645119174835,
        "mean_feedback_forward_relative_error": 0.11715754241712632,
        "min_feedback_forward_cosine": 0.9853746891021729
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.1954141134405136
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1658857762813568,
        "mean_feedback_forward_cosine": 0.9934704149923017,
        "mean_feedback_forward_relative_error": 0.11285452989320602,
        "min_feedback_forward_cosine": 0.9864099621772766
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.19057123153686523
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15934672951698303,
        "mean_feedback_forward_cosine": 0.9939606555046574,
        "mean_feedback_forward_relative_error": 0.10851789398058768,
        "min_feedback_forward_cosine": 0.9874546527862549
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.1898304710149765
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15314742922782898,
        "mean_feedback_forward_cosine": 0.9944075519038785,
        "mean_feedback_forward_relative_error": 0.10440802309782274,
        "min_feedback_forward_cosine": 0.9884060621261597
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.19179069291591644
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.14699344336986542,
        "mean_feedback_forward_cosine": 0.9948373225427443,
        "mean_feedback_forward_relative_error": 0.1003009861996097,
        "min_feedback_forward_cosine": 0.9893105030059814
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.1871666722202301
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1410326063632965,
        "mean_feedback_forward_cosine": 0.9952297114556835,
        "mean_feedback_forward_relative_error": 0.09641332095188479,
        "min_feedback_forward_cosine": 0.9901542067527771
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.18810378356218338
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13551799952983856,
        "mean_feedback_forward_cosine": 0.9955768777478126,
        "mean_feedback_forward_relative_error": 0.09282779801757104,
        "min_feedback_forward_cosine": 0.9909045696258545
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.1841782896709442
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12996059656143188,
        "mean_feedback_forward_cosine": 0.9959184700442899,
        "mean_feedback_forward_relative_error": 0.08917618771233866,
        "min_feedback_forward_cosine": 0.991633415222168
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.1905988346195221
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12496453523635864,
        "mean_feedback_forward_cosine": 0.9962086273777869,
        "mean_feedback_forward_relative_error": 0.0859366656551438,
        "min_feedback_forward_cosine": 0.9922622442245483
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.18150609948158264
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1199355497956276,
        "mean_feedback_forward_cosine": 0.9964944797177469,
        "mean_feedback_forward_relative_error": 0.08262782597974423,
        "min_feedback_forward_cosine": 0.9928679466247559
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.18521738644123079
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11520607024431229,
        "mean_feedback_forward_cosine": 0.9967465746787286,
        "mean_feedback_forward_relative_error": 0.07959193477947865,
        "min_feedback_forward_cosine": 0.9934182167053223
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.1773731155395508
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11053729802370071,
        "mean_feedback_forward_cosine": 0.9969997386778554,
        "mean_feedback_forward_relative_error": 0.07642521360708822,
        "min_feedback_forward_cosine": 0.9939394593238831
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.18311363577842713
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10619539022445679,
        "mean_feedback_forward_cosine": 0.9972271823113964,
        "mean_feedback_forward_relative_error": 0.0734694192726766,
        "min_feedback_forward_cosine": 0.9944032430648804
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.18016735112667084
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10208764672279358,
        "mean_feedback_forward_cosine": 0.9974322626667638,
        "mean_feedback_forward_relative_error": 0.07069606442124612,
        "min_feedback_forward_cosine": 0.9948270916938782
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.1764145338010788
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09786710888147354,
        "mean_feedback_forward_cosine": 0.9976302096920628,
        "mean_feedback_forward_relative_error": 0.06791745264443659,
        "min_feedback_forward_cosine": 0.9952437877655029
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.18184925278186798
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09394015371799469,
        "mean_feedback_forward_cosine": 0.9978082237705108,
        "mean_feedback_forward_relative_error": 0.06531861268224255,
        "min_feedback_forward_cosine": 0.9956168532371521
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.17577416330337525
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09005498141050339,
        "mean_feedback_forward_cosine": 0.9979749706483656,
        "mean_feedback_forward_relative_error": 0.06278794499174241,
        "min_feedback_forward_cosine": 0.9959706664085388
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.1818134422826767
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08653854578733444,
        "mean_feedback_forward_cosine": 0.998121888406815,
        "mean_feedback_forward_relative_error": 0.06045928212904161,
        "min_feedback_forward_cosine": 0.9962779879570007
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.17373866063117982
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0829930230975151,
        "mean_feedback_forward_cosine": 0.9982667949891859,
        "mean_feedback_forward_relative_error": 0.058078636325174765,
        "min_feedback_forward_cosine": 0.9965764880180359
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.17920592172145844
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07985182106494904,
        "mean_feedback_forward_cosine": 0.9983929434130269,
        "mean_feedback_forward_relative_error": 0.055919353279375264,
        "min_feedback_forward_cosine": 0.9968297481536865
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.17177996276378632
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07659643143415451,
        "mean_feedback_forward_cosine": 0.9985156078492442,
        "mean_feedback_forward_relative_error": 0.05374190770089626,
        "min_feedback_forward_cosine": 0.9970822334289551
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.17647761666297912
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07359126955270767,
        "mean_feedback_forward_cosine": 0.9986273915536942,
        "mean_feedback_forward_relative_error": 0.05167639658095375,
        "min_feedback_forward_cosine": 0.997306227684021
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.17503470136404037
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07059352844953537,
        "mean_feedback_forward_cosine": 0.9987327283428561,
        "mean_feedback_forward_relative_error": 0.04965663653227591,
        "min_feedback_forward_cosine": 0.9975208044052124
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.1740077546787262
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06786178797483444,
        "mean_feedback_forward_cosine": 0.9988272247775909,
        "mean_feedback_forward_relative_error": 0.04776773012934193,
        "min_feedback_forward_cosine": 0.9977082014083862
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.16993277304172516
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06507941335439682,
        "mean_feedback_forward_cosine": 0.9989168317087235,
        "mean_feedback_forward_relative_error": 0.04590913677407849,
        "min_feedback_forward_cosine": 0.9978917837142944
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.17373484318494797
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0625593364238739,
        "mean_feedback_forward_cosine": 0.9989959405314538,
        "mean_feedback_forward_relative_error": 0.044198656394597025,
        "min_feedback_forward_cosine": 0.9980515837669373
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.16842556659698488
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.060116615146398544,
        "mean_feedback_forward_cosine": 0.9990705097875288,
        "mean_feedback_forward_relative_error": 0.042526453883657535,
        "min_feedback_forward_cosine": 0.9982008934020996
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.1669404342365265
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.057767800986766815,
        "mean_feedback_forward_cosine": 0.9991391704928491,
        "mean_feedback_forward_relative_error": 0.04092263841941472,
        "min_feedback_forward_cosine": 0.9983383417129517
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.16348597554206848
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.055436767637729645,
        "mean_feedback_forward_cosine": 0.9992048394295477,
        "mean_feedback_forward_relative_error": 0.0393322111377793,
        "min_feedback_forward_cosine": 0.9984695911407471
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.17286750542640686
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05322682857513428,
        "mean_feedback_forward_cosine": 0.9992634538681276,
        "mean_feedback_forward_relative_error": 0.037855942883799155,
        "min_feedback_forward_cosine": 0.9985889196395874
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.16786843843221663
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05116841569542885,
        "mean_feedback_forward_cosine": 0.9993203743811576,
        "mean_feedback_forward_relative_error": 0.03636146069414193,
        "min_feedback_forward_cosine": 0.998695969581604
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.16632413774490357
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.049093734472990036,
        "mean_feedback_forward_cosine": 0.9993729649051544,
        "mean_feedback_forward_relative_error": 0.03492417012251193,
        "min_feedback_forward_cosine": 0.9987992644309998
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.16653296973705292
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.047124844044446945,
        "mean_feedback_forward_cosine": 0.9994193950007039,
        "mean_feedback_forward_relative_error": 0.0336092067581992,
        "min_feedback_forward_cosine": 0.9988933801651001
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.16568008406162263
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04536622390151024,
        "mean_feedback_forward_cosine": 0.99946129322052,
        "mean_feedback_forward_relative_error": 0.03237246349453926,
        "min_feedback_forward_cosine": 0.9989744424819946
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.16156356994628907
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04356296733021736,
        "mean_feedback_forward_cosine": 0.9995024434981807,
        "mean_feedback_forward_relative_error": 0.03111092064289316,
        "min_feedback_forward_cosine": 0.9990541338920593
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.16308865074634551
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0419180728495121,
        "mean_feedback_forward_cosine": 0.9995387408041185,
        "mean_feedback_forward_relative_error": 0.029956305850177042,
        "min_feedback_forward_cosine": 0.9991241693496704
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.15842433814048767
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04021833837032318,
        "mean_feedback_forward_cosine": 0.9995746266457343,
        "mean_feedback_forward_relative_error": 0.028767979643758265,
        "min_feedback_forward_cosine": 0.9991938471794128
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.16590015714645387
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03862576559185982,
        "mean_feedback_forward_cosine": 0.9996064747533491,
        "mean_feedback_forward_relative_error": 0.02766890376205406,
        "min_feedback_forward_cosine": 0.999256432056427
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.16425485748291016
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03712113946676254,
        "mean_feedback_forward_cosine": 0.999635394542448,
        "mean_feedback_forward_relative_error": 0.026631538575935747,
        "min_feedback_forward_cosine": 0.9993130564689636
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.16208505117416383
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03564485162496567,
        "mean_feedback_forward_cosine": 0.9996625915650399,
        "mean_feedback_forward_relative_error": 0.02561724678643288,
        "min_feedback_forward_cosine": 0.9993666410446167
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.16008623097419739
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03427170217037201,
        "mean_feedback_forward_cosine": 0.999687377483614,
        "mean_feedback_forward_relative_error": 0.024657781386087017,
        "min_feedback_forward_cosine": 0.9994142651557922
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.1584984017133713
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03291971608996391,
        "mean_feedback_forward_cosine": 0.9997108097999327,
        "mean_feedback_forward_relative_error": 0.0237135432900921,
        "min_feedback_forward_cosine": 0.9994597434997559
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.15919874101161957
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03161061182618141,
        "mean_feedback_forward_cosine": 0.9997326616317995,
        "mean_feedback_forward_relative_error": 0.02280093639368011,
        "min_feedback_forward_cosine": 0.9995017647743225
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.16238198031425477
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030365271493792534,
        "mean_feedback_forward_cosine": 0.999752281173583,
        "mean_feedback_forward_relative_error": 0.021947178478923537,
        "min_feedback_forward_cosine": 0.9995402097702026
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.1577615320110321
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03035474568605423,
        "mean_feedback_forward_cosine": 0.9997524522965954,
        "mean_feedback_forward_relative_error": 0.02193744862151723,
        "min_feedback_forward_cosine": 0.9995405673980713
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.09112625179052353
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03034961223602295,
        "mean_feedback_forward_cosine": 0.9997525426649279,
        "mean_feedback_forward_relative_error": 0.02193117661461715,
        "min_feedback_forward_cosine": 0.9995406270027161
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.0651662121462822
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030345071107149124,
        "mean_feedback_forward_cosine": 0.9997526368787212,
        "mean_feedback_forward_relative_error": 0.021925508570406708,
        "min_feedback_forward_cosine": 0.9995408654212952
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.05518694655299187
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030340731143951416,
        "mean_feedback_forward_cosine": 0.9997527310925145,
        "mean_feedback_forward_relative_error": 0.02192018811981524,
        "min_feedback_forward_cosine": 0.9995409250259399
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.05016247516512871
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03033665381371975,
        "mean_feedback_forward_cosine": 0.999752809924464,
        "mean_feedback_forward_relative_error": 0.02191504654324343,
        "min_feedback_forward_cosine": 0.9995411038398743
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.045509788563251496
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0303326603025198,
        "mean_feedback_forward_cosine": 0.9997529002927965,
        "mean_feedback_forward_relative_error": 0.021910079619697025,
        "min_feedback_forward_cosine": 0.999541163444519
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.04068453124523163
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030328577384352684,
        "mean_feedback_forward_cosine": 0.9997529579747108,
        "mean_feedback_forward_relative_error": 0.02190520210311778,
        "min_feedback_forward_cosine": 0.9995412826538086
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.03963619037270546
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030324583873152733,
        "mean_feedback_forward_cosine": 0.9997530579566956,
        "mean_feedback_forward_relative_error": 0.02190041650206812,
        "min_feedback_forward_cosine": 0.9995415210723877
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.03669267584860325
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030320720747113228,
        "mean_feedback_forward_cosine": 0.9997531310204537,
        "mean_feedback_forward_relative_error": 0.02189569534253209,
        "min_feedback_forward_cosine": 0.9995415210723877
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.03461138196527958
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030317235738039017,
        "mean_feedback_forward_cosine": 0.9997532002387508,
        "mean_feedback_forward_relative_error": 0.021891336106965618,
        "min_feedback_forward_cosine": 0.9995415806770325
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.03089582992076874
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03031330741941929,
        "mean_feedback_forward_cosine": 0.9997532829161613,
        "mean_feedback_forward_relative_error": 0.021886823428494316,
        "min_feedback_forward_cosine": 0.9995417594909668
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.02949810125529766
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030309543013572693,
        "mean_feedback_forward_cosine": 0.9997533540571889,
        "mean_feedback_forward_relative_error": 0.021882356520021153,
        "min_feedback_forward_cosine": 0.9995419979095459
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.02843426930308342
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030305659398436546,
        "mean_feedback_forward_cosine": 0.9997534617300956,
        "mean_feedback_forward_relative_error": 0.021877802427737944,
        "min_feedback_forward_cosine": 0.9995419979095459
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.027714148516058922
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030302129685878754,
        "mean_feedback_forward_cosine": 0.9997535290256623,
        "mean_feedback_forward_relative_error": 0.021873503232434872,
        "min_feedback_forward_cosine": 0.9995420575141907
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.02623457520365715
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030298389494419098,
        "mean_feedback_forward_cosine": 0.9997536097803423,
        "mean_feedback_forward_relative_error": 0.021869151507534327,
        "min_feedback_forward_cosine": 0.9995421767234802
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.024785830131173133
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03029458038508892,
        "mean_feedback_forward_cosine": 0.9997536751531786,
        "mean_feedback_forward_relative_error": 0.021864845177098628,
        "min_feedback_forward_cosine": 0.9995423555374146
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.024115232612490652
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03029109351336956,
        "mean_feedback_forward_cosine": 0.9997537635987804,
        "mean_feedback_forward_relative_error": 0.021860653369296944,
        "min_feedback_forward_cosine": 0.9995425939559937
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.022689599386453627
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0302873644977808,
        "mean_feedback_forward_cosine": 0.9997538347398082,
        "mean_feedback_forward_relative_error": 0.021856272938631235,
        "min_feedback_forward_cosine": 0.9995425343513489
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.02323253556996584
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03028363548219204,
        "mean_feedback_forward_cosine": 0.9997539212626796,
        "mean_feedback_forward_relative_error": 0.02185203465482881,
        "min_feedback_forward_cosine": 0.9995427131652832
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.02146149538576603
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030279796570539474,
        "mean_feedback_forward_cosine": 0.9997540058628205,
        "mean_feedback_forward_relative_error": 0.021847640765049765,
        "min_feedback_forward_cosine": 0.9995428323745728
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.021298135353326798
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0302763469517231,
        "mean_feedback_forward_cosine": 0.9997540866175005,
        "mean_feedback_forward_relative_error": 0.021843459547287034,
        "min_feedback_forward_cosine": 0.9995429515838623
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.019140538485646248
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030272936448454857,
        "mean_feedback_forward_cosine": 0.9997541308403015,
        "mean_feedback_forward_relative_error": 0.021839403006578644,
        "min_feedback_forward_cosine": 0.9995430111885071
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.01843281641423702
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030269691720604897,
        "mean_feedback_forward_cosine": 0.999754253895052,
        "mean_feedback_forward_relative_error": 0.021835425523139777,
        "min_feedback_forward_cosine": 0.9995430707931519
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.017320715464651585
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030265865847468376,
        "mean_feedback_forward_cosine": 0.9997542846587396,
        "mean_feedback_forward_relative_error": 0.02183128496812236,
        "min_feedback_forward_cosine": 0.999543309211731
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.018616918771862983
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03026205487549305,
        "mean_feedback_forward_cosine": 0.9997543692588806,
        "mean_feedback_forward_relative_error": 0.021826918447209943,
        "min_feedback_forward_cosine": 0.999543309211731
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.01787962683171034
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03025832027196884,
        "mean_feedback_forward_cosine": 0.9997544730863264,
        "mean_feedback_forward_relative_error": 0.021822680538940813,
        "min_feedback_forward_cosine": 0.9995434880256653
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.016799339402318002
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030255036428570747,
        "mean_feedback_forward_cosine": 0.9997545269227797,
        "mean_feedback_forward_relative_error": 0.021818770065663323,
        "min_feedback_forward_cosine": 0.9995435476303101
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.015555499247312546
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030251365154981613,
        "mean_feedback_forward_cosine": 0.9997546096001902,
        "mean_feedback_forward_relative_error": 0.02181462037767614,
        "min_feedback_forward_cosine": 0.9995437264442444
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.01577266572535038
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030247338116168976,
        "mean_feedback_forward_cosine": 0.9997547076594445,
        "mean_feedback_forward_relative_error": 0.021810064212449135,
        "min_feedback_forward_cosine": 0.9995439052581787
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.016338454892113805
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030243637040257454,
        "mean_feedback_forward_cosine": 0.9997547807232026,
        "mean_feedback_forward_relative_error": 0.021805861138648564,
        "min_feedback_forward_cosine": 0.9995437860488892
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.015070238414108754
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030239706858992577,
        "mean_feedback_forward_cosine": 0.999754878782457,
        "mean_feedback_forward_relative_error": 0.02180133913193018,
        "min_feedback_forward_cosine": 0.9995439052581787
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.015652655007243155
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030236007645726204,
        "mean_feedback_forward_cosine": 0.9997549729962503,
        "mean_feedback_forward_relative_error": 0.021797153978578506,
        "min_feedback_forward_cosine": 0.9995441436767578
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.014617297267764806
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030232984572649002,
        "mean_feedback_forward_cosine": 0.9997550345236256,
        "mean_feedback_forward_relative_error": 0.021793444460678484,
        "min_feedback_forward_cosine": 0.9995443224906921
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.013100630081892014
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030228734016418457,
        "mean_feedback_forward_cosine": 0.9997551095101141,
        "mean_feedback_forward_relative_error": 0.02178893143671655,
        "min_feedback_forward_cosine": 0.9995443224906921
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.014554911104142665
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030225105583667755,
        "mean_feedback_forward_cosine": 0.9997551729602199,
        "mean_feedback_forward_relative_error": 0.021784932487794468,
        "min_feedback_forward_cosine": 0.9995445013046265
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.012763612243533135
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030221687629818916,
        "mean_feedback_forward_cosine": 0.9997552633285522,
        "mean_feedback_forward_relative_error": 0.021780829591255997,
        "min_feedback_forward_cosine": 0.9995444416999817
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.012405773229599
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030217928811907768,
        "mean_feedback_forward_cosine": 0.999755367155998,
        "mean_feedback_forward_relative_error": 0.021776424675819375,
        "min_feedback_forward_cosine": 0.9995446801185608
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.013631196454167367
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030214227735996246,
        "mean_feedback_forward_cosine": 0.9997554498334085,
        "mean_feedback_forward_relative_error": 0.021772182321236018,
        "min_feedback_forward_cosine": 0.9995447397232056
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.012717415338754654
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030210770666599274,
        "mean_feedback_forward_cosine": 0.999755528665358,
        "mean_feedback_forward_relative_error": 0.02176817787450648,
        "min_feedback_forward_cosine": 0.9995449185371399
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.011400041757076979
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030206751078367233,
        "mean_feedback_forward_cosine": 0.99975561711096,
        "mean_feedback_forward_relative_error": 0.021763561578888084,
        "min_feedback_forward_cosine": 0.9995450377464294
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.012817322442829609
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03020297735929489,
        "mean_feedback_forward_cosine": 0.9997557074792923,
        "mean_feedback_forward_relative_error": 0.02175934123055589,
        "min_feedback_forward_cosine": 0.999545156955719
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.01158828136779368
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030199110507965088,
        "mean_feedback_forward_cosine": 0.999755774774859,
        "mean_feedback_forward_relative_error": 0.021755064341930613,
        "min_feedback_forward_cosine": 0.9995452761650085
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.011865053795874119
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030195636674761772,
        "mean_feedback_forward_cosine": 0.9997558612977305,
        "mean_feedback_forward_relative_error": 0.021750994792748845,
        "min_feedback_forward_cosine": 0.9995453357696533
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.010737874366939067
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030192187055945396,
        "mean_feedback_forward_cosine": 0.9997559285932972,
        "mean_feedback_forward_relative_error": 0.021747009242854772,
        "min_feedback_forward_cosine": 0.9995453357696533
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.010735899466872215
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030188128352165222,
        "mean_feedback_forward_cosine": 0.9997560497253172,
        "mean_feedback_forward_relative_error": 0.02174242093197761,
        "min_feedback_forward_cosine": 0.9995455741882324
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.01134729971215129
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03018438071012497,
        "mean_feedback_forward_cosine": 0.9997561420163801,
        "mean_feedback_forward_relative_error": 0.021738111041486263,
        "min_feedback_forward_cosine": 0.9995457530021667
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.010523225945830345
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030180975794792175,
        "mean_feedback_forward_cosine": 0.9997562112346772,
        "mean_feedback_forward_relative_error": 0.02173394923128428,
        "min_feedback_forward_cosine": 0.999545693397522
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.009820032843351364
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03017698973417282,
        "mean_feedback_forward_cosine": 0.999756278530244,
        "mean_feedback_forward_relative_error": 0.021729366147830602,
        "min_feedback_forward_cosine": 0.9995459318161011
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.010756156692318618
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030173545703291893,
        "mean_feedback_forward_cosine": 0.9997563862031505,
        "mean_feedback_forward_relative_error": 0.02172539821795879,
        "min_feedback_forward_cosine": 0.9995459914207458
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.009835685213804245
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169757083058357,
        "mean_feedback_forward_cosine": 0.9997564708032916,
        "mean_feedback_forward_relative_error": 0.021721043413685213,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.009867143030613661
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169716104865074,
        "mean_feedback_forward_cosine": 0.9997564631123697,
        "mean_feedback_forward_relative_error": 0.021720950416619738,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.008295226010084152
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016967698931694,
        "mean_feedback_forward_cosine": 0.9997564746487525,
        "mean_feedback_forward_relative_error": 0.021720858801516793,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.00862344180509448
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016965463757515,
        "mean_feedback_forward_cosine": 0.9997564631123697,
        "mean_feedback_forward_relative_error": 0.021720769514720285,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.008312926781252027
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016963228583336,
        "mean_feedback_forward_cosine": 0.9997564708032916,
        "mean_feedback_forward_relative_error": 0.021720683127040823,
        "min_feedback_forward_cosine": 0.9995459914207458
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.007363148895800114
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016960807144642,
        "mean_feedback_forward_cosine": 0.9997564669578306,
        "mean_feedback_forward_relative_error": 0.021720602297254147,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.00785008468285203
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169574543833733,
        "mean_feedback_forward_cosine": 0.9997564823396744,
        "mean_feedback_forward_relative_error": 0.0217205208215502,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.007802247190997005
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169550329446793,
        "mean_feedback_forward_cosine": 0.9997564727260221,
        "mean_feedback_forward_relative_error": 0.02172043769349975,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.0067719343725591895
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169522389769554,
        "mean_feedback_forward_cosine": 0.9997564746487525,
        "mean_feedback_forward_relative_error": 0.02172035832078226,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.006957013593614101
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169496312737465,
        "mean_feedback_forward_cosine": 0.999756476571483,
        "mean_feedback_forward_relative_error": 0.021720276709886327,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.006896061420664191
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169475823640823,
        "mean_feedback_forward_cosine": 0.9997564746487525,
        "mean_feedback_forward_relative_error": 0.02172020019122189,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.007256742753982544
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016945719718933,
        "mean_feedback_forward_cosine": 0.9997564881078659,
        "mean_feedback_forward_relative_error": 0.02172011966186185,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.007564757667630911
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169425532221794,
        "mean_feedback_forward_cosine": 0.999756480416944,
        "mean_feedback_forward_relative_error": 0.02172003458103826,
        "min_feedback_forward_cosine": 0.999546229839325
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.007529743331968784
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169401317834854,
        "mean_feedback_forward_cosine": 0.9997564554214478,
        "mean_feedback_forward_relative_error": 0.021719952128947742,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.007267204804420471
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169377103447914,
        "mean_feedback_forward_cosine": 0.999756480416944,
        "mean_feedback_forward_relative_error": 0.02171987280129425,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.006840419639255852
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169356614351273,
        "mean_feedback_forward_cosine": 0.9997564784942135,
        "mean_feedback_forward_relative_error": 0.021719793128150124,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.0072005226527899506
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169332399964333,
        "mean_feedback_forward_cosine": 0.999756476571483,
        "mean_feedback_forward_relative_error": 0.02171971240351277,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.006555082276463509
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016931563615799,
        "mean_feedback_forward_cosine": 0.9997564900305963,
        "mean_feedback_forward_relative_error": 0.021719636125189644,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.0067100419231597335
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169295147061348,
        "mean_feedback_forward_cosine": 0.9997564727260221,
        "mean_feedback_forward_relative_error": 0.021719558284648003,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.006859775168448687
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169276520609856,
        "mean_feedback_forward_cosine": 0.9997564727260221,
        "mean_feedback_forward_relative_error": 0.02171948000848774,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.007023929670471698
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169248580932617,
        "mean_feedback_forward_cosine": 0.9997564919533268,
        "mean_feedback_forward_relative_error": 0.021719402137903437,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.00637954917833209
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169226229190826,
        "mean_feedback_forward_cosine": 0.9997564881078659,
        "mean_feedback_forward_relative_error": 0.021719321548458067,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.006854683260917664
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169205740094185,
        "mean_feedback_forward_cosine": 0.9997564842624049,
        "mean_feedback_forward_relative_error": 0.021719245540518916,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.0066349893228709695
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169187113642693,
        "mean_feedback_forward_cosine": 0.9997565015669791,
        "mean_feedback_forward_relative_error": 0.021719168976790482,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.006290283277481794
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169174075126648,
        "mean_feedback_forward_cosine": 0.9997564957987878,
        "mean_feedback_forward_relative_error": 0.021719093764981916,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.006629161743670702
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169157311320305,
        "mean_feedback_forward_cosine": 0.9997564881078659,
        "mean_feedback_forward_relative_error": 0.021719019078919963,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.006182823036164045
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169138684868813,
        "mean_feedback_forward_cosine": 0.9997564919533268,
        "mean_feedback_forward_relative_error": 0.021718938955135884,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.006323601663410664
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169112607836723,
        "mean_feedback_forward_cosine": 0.9997565054124401,
        "mean_feedback_forward_relative_error": 0.021718856503045367,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.006986099981889129
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169080942869186,
        "mean_feedback_forward_cosine": 0.9997564919533268,
        "mean_feedback_forward_relative_error": 0.0217187755831307,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.006972612196803093
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169058591127396,
        "mean_feedback_forward_cosine": 0.999756476571483,
        "mean_feedback_forward_relative_error": 0.021718700311236804,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.0067703842197358605
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169036239385605,
        "mean_feedback_forward_cosine": 0.999756513103362,
        "mean_feedback_forward_relative_error": 0.021718626541476094,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.0063818694199621676
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030169015750288963,
        "mean_feedback_forward_cosine": 0.9997564784942135,
        "mean_feedback_forward_relative_error": 0.021718546733139985,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.006102044712156057
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016900084912777,
        "mean_feedback_forward_cosine": 0.9997564938760573,
        "mean_feedback_forward_relative_error": 0.021718474240192482,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.005858240797370672
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016897849738598,
        "mean_feedback_forward_cosine": 0.9997565034897097,
        "mean_feedback_forward_relative_error": 0.02171839940391721,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.005866434761993587
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168967321515083,
        "mean_feedback_forward_cosine": 0.9997564919533268,
        "mean_feedback_forward_relative_error": 0.021718321923887538,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.005773927080184221
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168946832418442,
        "mean_feedback_forward_cosine": 0.9997564996442487,
        "mean_feedback_forward_relative_error": 0.021718249115492066,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.0064538452906906605
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168922618031502,
        "mean_feedback_forward_cosine": 0.9997564919533268,
        "mean_feedback_forward_relative_error": 0.021718173317851558,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.006007886548340321
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168896540999413,
        "mean_feedback_forward_cosine": 0.9997564977215182,
        "mean_feedback_forward_relative_error": 0.021718097685445702,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.005968254810981452
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016887977719307,
        "mean_feedback_forward_cosine": 0.999756513103362,
        "mean_feedback_forward_relative_error": 0.02171802355517303,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.00620140129506588
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016885556280613,
        "mean_feedback_forward_cosine": 0.9997564919533268,
        "mean_feedback_forward_relative_error": 0.021717947877703175,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.005907446707040072
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016883134841919,
        "mean_feedback_forward_cosine": 0.999756513103362,
        "mean_feedback_forward_relative_error": 0.02171787343198253,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.005722253931760788
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168810859322548,
        "mean_feedback_forward_cosine": 0.999756513103362,
        "mean_feedback_forward_relative_error": 0.02171779963217916,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.005939775842446834
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168788507580757,
        "mean_feedback_forward_cosine": 0.9997565034897097,
        "mean_feedback_forward_relative_error": 0.021717723173600052,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.006015781491547823
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016876056790352,
        "mean_feedback_forward_cosine": 0.9997564977215182,
        "mean_feedback_forward_relative_error": 0.02171765173214578,
        "min_feedback_forward_cosine": 0.9995459914207458
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.005618943987675011
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016873449087143,
        "mean_feedback_forward_cosine": 0.9997564957987878,
        "mean_feedback_forward_relative_error": 0.021717575558971976,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.006377606983333826
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168725177645683,
        "mean_feedback_forward_cosine": 0.9997565073351706,
        "mean_feedback_forward_relative_error": 0.02171750329134445,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.005670744403153658
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03016870841383934,
        "mean_feedback_forward_cosine": 0.9997564861851354,
        "mean_feedback_forward_relative_error": 0.021717430452906316,
        "min_feedback_forward_cosine": 0.9995461702346802
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.005683699643909931
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168689787387848,
        "mean_feedback_forward_cosine": 0.9997565054124401,
        "mean_feedback_forward_relative_error": 0.02171735931187868,
        "min_feedback_forward_cosine": 0.9995459914207458
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.0059974849678576
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168669298291206,
        "mean_feedback_forward_cosine": 0.9997564957987878,
        "mean_feedback_forward_relative_error": 0.021717282222403635,
        "min_feedback_forward_cosine": 0.9995460510253906
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.006048647853694856
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168643221259117,
        "mean_feedback_forward_cosine": 0.9997565188715535,
        "mean_feedback_forward_relative_error": 0.021717206334635135,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.005997108866162598
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.030168624594807625,
        "mean_feedback_forward_cosine": 0.9997565054124401,
        "mean_feedback_forward_relative_error": 0.021717134742967544,
        "min_feedback_forward_cosine": 0.9995461106300354
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.005753214195370674
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9234,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.3684505432128906,
    "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": 13,
    "normalization_mean": [
      0.49140000343322754,
      0.4821999967098236,
      0.4465000033378601
    ],
    "normalization_std": [
      0.24699999392032623,
      0.2434999942779541,
      0.26159998774528503
    ],
    "split_from_training_only": true,
    "split_seed": null,
    "test_examples": 10000,
    "train_examples": 50000,
    "training_augmentation": "random_crop_32_padding_4_zero_then_horizontal_flip_p0.5",
    "validation_class_counts": {},
    "validation_examples": 0,
    "validation_index_sha256": null
  },
  "timing": {
    "apical_warmup_wall_s": 0.0,
    "evaluation_wall_s": 1.1665029525756836,
    "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": 10229.90122294426,
    "train_wall_s": 10226.761625766754
  },
  "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
  }
}