{
  "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": 11,
    "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_s11.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": 11,
    "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": 0.9995934665203094,
    "feedback_forward_cosine": [
      0.9998228549957275,
      0.9998177289962769,
      0.9997299909591675,
      0.9997137784957886,
      0.9997239112854004,
      0.9996533393859863,
      0.9997869729995728,
      0.9996950626373291,
      0.999748706817627,
      0.9996466040611267,
      0.9997207522392273,
      0.9998055696487427,
      0.9997636079788208,
      0.999703049659729,
      0.9997877478599548,
      0.9997124075889587,
      0.9997643232345581,
      0.9996864795684814,
      0.9997842311859131,
      0.9996967315673828,
      0.9997855424880981,
      0.9998668432235718,
      0.9998509883880615,
      0.9998328685760498,
      0.9998459815979004,
      0.9997791647911072,
      0.9998233318328857,
      0.9996689558029175,
      0.9995805621147156,
      0.9992669224739075,
      0.999984860420227
    ],
    "feedback_forward_norm_ratio": [
      0.9996518492698669,
      1.0002232789993286,
      0.9992798566818237,
      1.0017285346984863,
      0.9986801147460938,
      1.0010799169540405,
      0.999582052230835,
      1.0007548332214355,
      0.999423623085022,
      1.0021605491638184,
      0.9986611008644104,
      1.00065016746521,
      0.9993356466293335,
      1.0007761716842651,
      0.999326229095459,
      1.0015517473220825,
      0.9993433952331543,
      1.001268744468689,
      0.9996417760848999,
      1.0017112493515015,
      0.9993442893028259,
      1.0004926919937134,
      0.9998202323913574,
      1.0006990432739258,
      0.9999847412109375,
      1.000831961631775,
      0.9998155236244202,
      1.0010372400283813,
      0.9992583394050598,
      1.0016268491744995,
      1.0027137994766235
    ],
    "feedback_forward_relative_error": [
      0.018819428980350494,
      0.019098827615380287,
      0.023236405104398727,
      0.024010000750422478,
      0.023518972098827362,
      0.026368213817477226,
      0.020639995113015175,
      0.024714823812246323,
      0.022416267544031143,
      0.026700245216488838,
      0.023652229458093643,
      0.019735991954803467,
      0.02174312248826027,
      0.024391446262598038,
      0.020609065890312195,
      0.02405175007879734,
      0.02171194739639759,
      0.02508857101202011,
      0.020772576332092285,
      0.02470984496176243,
      0.020718853920698166,
      0.016328049823641777,
      0.017266981303691864,
      0.01830025017261505,
      0.017553480342030525,
      0.021037161350250244,
      0.018793849274516106,
      0.025765253230929375,
      0.028958963230252266,
      0.03835690766572952,
      0.0061418828554451466
    ],
    "innovation_negative_gradient_cosine": [
      0.9996030330657959,
      0.9995779991149902,
      0.9996092915534973,
      0.9995189905166626,
      0.999632716178894,
      0.9995143413543701,
      0.9996480941772461,
      0.9996263980865479,
      0.9996553659439087,
      0.9995484352111816,
      0.9996688365936279,
      0.9996404051780701,
      0.9996864795684814,
      0.9995951652526855,
      0.9996931552886963,
      0.9996488094329834,
      0.9997095465660095,
      0.9996280074119568,
      0.9997208714485168,
      0.999658465385437,
      0.9997446537017822,
      0.9997918605804443,
      0.999794602394104,
      0.9997910261154175,
      0.9998301267623901,
      0.9998152256011963,
      0.9998809099197388,
      0.9999029636383057,
      0.9998573064804077,
      0.9999117255210876,
      0.8352094292640686
    ],
    "mean_feedback_forward_cosine": 0.9997435442862972,
    "mean_feedback_forward_relative_error": 0.022103592227663723,
    "normalization_state": "training_batch_stats_without_running_update",
    "raw_negative_gradient_cosine": [
      0.9996030330657959,
      0.9995779991149902,
      0.9996092915534973,
      0.9995189905166626,
      0.999632716178894,
      0.9995143413543701,
      0.9996480941772461,
      0.9996263980865479,
      0.9996553659439087,
      0.9995484352111816,
      0.9996688365936279,
      0.9996404051780701,
      0.9996864795684814,
      0.9995951652526855,
      0.9996931552886963,
      0.9996488094329834,
      0.9997095465660095,
      0.9996280074119568,
      0.9997208714485168,
      0.999658465385437,
      0.9997446537017822,
      0.9997918605804443,
      0.999794602394104,
      0.9997910261154175,
      0.9998301267623901,
      0.9998152256011963,
      0.9998809099197388,
      0.9999029636383057,
      0.9998573064804077,
      0.9999117255210876,
      0.8352094292640686
    ],
    "teaching_negative_gradient_cosine": [
      0.9996030330657959,
      0.9995779991149902,
      0.9996092915534973,
      0.9995189905166626,
      0.999632716178894,
      0.9995143413543701,
      0.9996480941772461,
      0.9996263980865479,
      0.9996553659439087,
      0.9995484352111816,
      0.9996688365936279,
      0.9996404051780701,
      0.9996864795684814,
      0.9995951652526855,
      0.9996931552886963,
      0.9996488094329834,
      0.9997095465660095,
      0.9996280074119568,
      0.9997208714485168,
      0.999658465385437,
      0.9997446537017822,
      0.9997918605804443,
      0.999794602394104,
      0.9997910261154175,
      0.9998301267623901,
      0.9998152256011963,
      0.9998809099197388,
      0.9999029636383057,
      0.9998573064804077,
      0.9999117255210876,
      0.8352094292640686
    ],
    "wall_s": 0.10811614990234375
  },
  "epochs": [
    {
      "epoch": 1,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3494246006011963,
        "mean_feedback_forward_cosine": 0.24975507466062422,
        "mean_feedback_forward_relative_error": 1.2236636665559584,
        "min_feedback_forward_cosine": 0.09657269716262817
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 391,
      "train_examples": 50000,
      "train_loss": 1.9218929443359376
    },
    {
      "epoch": 2,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.3228424787521362,
        "mean_feedback_forward_cosine": 0.33737454731618205,
        "mean_feedback_forward_relative_error": 1.1493131826000829,
        "min_feedback_forward_cosine": 0.13145390152931213
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 782,
      "train_examples": 50000,
      "train_loss": 1.4069220595550538
    },
    {
      "epoch": 3,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2915085554122925,
        "mean_feedback_forward_cosine": 0.42328307801677334,
        "mean_feedback_forward_relative_error": 1.071990292879843,
        "min_feedback_forward_cosine": 0.17247258126735687
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1173,
      "train_examples": 50000,
      "train_loss": 1.1637197574043274
    },
    {
      "epoch": 4,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.261440396308899,
        "mean_feedback_forward_cosine": 0.49132025337988333,
        "mean_feedback_forward_relative_error": 1.0071991691666264,
        "min_feedback_forward_cosine": 0.2104986310005188
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1564,
      "train_examples": 50000,
      "train_loss": 0.9531205076599121
    },
    {
      "epoch": 5,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.2322450876235962,
        "mean_feedback_forward_cosine": 0.5464878394719093,
        "mean_feedback_forward_relative_error": 0.9507570064836933,
        "min_feedback_forward_cosine": 0.24918709695339203
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 1955,
      "train_examples": 50000,
      "train_loss": 0.8120380458831787
    },
    {
      "epoch": 6,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.209352731704712,
        "mean_feedback_forward_cosine": 0.5937906542131978,
        "mean_feedback_forward_relative_error": 0.8996120143321252,
        "min_feedback_forward_cosine": 0.2835087776184082
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2346,
      "train_examples": 50000,
      "train_loss": 0.7274218988037109
    },
    {
      "epoch": 7,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1875540018081665,
        "mean_feedback_forward_cosine": 0.6319110643479132,
        "mean_feedback_forward_relative_error": 0.8559480530600394,
        "min_feedback_forward_cosine": 0.31155192852020264
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 2737,
      "train_examples": 50000,
      "train_loss": 0.6619951996421813
    },
    {
      "epoch": 8,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1641260385513306,
        "mean_feedback_forward_cosine": 0.6644031443903523,
        "mean_feedback_forward_relative_error": 0.8167407426141924,
        "min_feedback_forward_cosine": 0.34022173285484314
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3128,
      "train_examples": 50000,
      "train_loss": 0.6074947142028808
    },
    {
      "epoch": 9,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1407192945480347,
        "mean_feedback_forward_cosine": 0.6924694247784153,
        "mean_feedback_forward_relative_error": 0.7813257488512224,
        "min_feedback_forward_cosine": 0.367876797914505
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3519,
      "train_examples": 50000,
      "train_loss": 0.561738102016449
    },
    {
      "epoch": 10,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.1186273097991943,
        "mean_feedback_forward_cosine": 0.7176954640496161,
        "mean_feedback_forward_relative_error": 0.7480449099694529,
        "min_feedback_forward_cosine": 0.3937365710735321
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 3910,
      "train_examples": 50000,
      "train_loss": 0.528632747182846
    },
    {
      "epoch": 11,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0955536365509033,
        "mean_feedback_forward_cosine": 0.7401004299040763,
        "mean_feedback_forward_relative_error": 0.7170918420437844,
        "min_feedback_forward_cosine": 0.4198266863822937
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4301,
      "train_examples": 50000,
      "train_loss": 0.49865909043312073
    },
    {
      "epoch": 12,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.072723388671875,
        "mean_feedback_forward_cosine": 0.7600079726788306,
        "mean_feedback_forward_relative_error": 0.6883098050471275,
        "min_feedback_forward_cosine": 0.4447500705718994
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 4692,
      "train_examples": 50000,
      "train_loss": 0.4733160824394226
    },
    {
      "epoch": 13,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.04936945438385,
        "mean_feedback_forward_cosine": 0.7782817290675256,
        "mean_feedback_forward_relative_error": 0.6611209915530297,
        "min_feedback_forward_cosine": 0.46964436769485474
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5083,
      "train_examples": 50000,
      "train_loss": 0.4554216194534302
    },
    {
      "epoch": 14,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0267332792282104,
        "mean_feedback_forward_cosine": 0.7951147604373193,
        "mean_feedback_forward_relative_error": 0.634850035271337,
        "min_feedback_forward_cosine": 0.4923594295978546
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5474,
      "train_examples": 50000,
      "train_loss": 0.43665811025619505
    },
    {
      "epoch": 15,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 1.0028632879257202,
        "mean_feedback_forward_cosine": 0.8101996471804958,
        "mean_feedback_forward_relative_error": 0.6104882330663742,
        "min_feedback_forward_cosine": 0.5159362554550171
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 5865,
      "train_examples": 50000,
      "train_loss": 0.41420850196838377
    },
    {
      "epoch": 16,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.980301558971405,
        "mean_feedback_forward_cosine": 0.8237882064234826,
        "mean_feedback_forward_relative_error": 0.5876467982607503,
        "min_feedback_forward_cosine": 0.5376675128936768
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6256,
      "train_examples": 50000,
      "train_loss": 0.396331945734024
    },
    {
      "epoch": 17,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9565905332565308,
        "mean_feedback_forward_cosine": 0.8366033742504735,
        "mean_feedback_forward_relative_error": 0.5652567576016149,
        "min_feedback_forward_cosine": 0.5602827668190002
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 6647,
      "train_examples": 50000,
      "train_loss": 0.3821376348876953
    },
    {
      "epoch": 18,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.933087170124054,
        "mean_feedback_forward_cosine": 0.8488402366638184,
        "mean_feedback_forward_relative_error": 0.5431818466994071,
        "min_feedback_forward_cosine": 0.5823767781257629
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7038,
      "train_examples": 50000,
      "train_loss": 0.37093611728668213
    },
    {
      "epoch": 19,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.9099282026290894,
        "mean_feedback_forward_cosine": 0.8596655476477838,
        "mean_feedback_forward_relative_error": 0.5227617837728993,
        "min_feedback_forward_cosine": 0.6027913689613342
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7429,
      "train_examples": 50000,
      "train_loss": 0.3591542248249054
    },
    {
      "epoch": 20,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.887123167514801,
        "mean_feedback_forward_cosine": 0.8697121566341769,
        "mean_feedback_forward_relative_error": 0.5031659867494337,
        "min_feedback_forward_cosine": 0.622233510017395
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 7820,
      "train_examples": 50000,
      "train_loss": 0.3470640205192566
    },
    {
      "epoch": 21,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8634553551673889,
        "mean_feedback_forward_cosine": 0.8789754836790024,
        "mean_feedback_forward_relative_error": 0.4844576029046889,
        "min_feedback_forward_cosine": 0.6421133875846863
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8211,
      "train_examples": 50000,
      "train_loss": 0.34042161994934084
    },
    {
      "epoch": 22,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.839503824710846,
        "mean_feedback_forward_cosine": 0.8877149147372092,
        "mean_feedback_forward_relative_error": 0.4661956309310852,
        "min_feedback_forward_cosine": 0.6618831753730774
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8602,
      "train_examples": 50000,
      "train_loss": 0.3278131375980377
    },
    {
      "epoch": 23,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.8167532086372375,
        "mean_feedback_forward_cosine": 0.8957876390026461,
        "mean_feedback_forward_relative_error": 0.4486217181528768,
        "min_feedback_forward_cosine": 0.679898738861084
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 8993,
      "train_examples": 50000,
      "train_loss": 0.31996051305770873
    },
    {
      "epoch": 24,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7931098341941833,
        "mean_feedback_forward_cosine": 0.9032602079453007,
        "mean_feedback_forward_relative_error": 0.43187700644616156,
        "min_feedback_forward_cosine": 0.698142945766449
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9384,
      "train_examples": 50000,
      "train_loss": 0.3156246472454071
    },
    {
      "epoch": 25,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7688338756561279,
        "mean_feedback_forward_cosine": 0.9103797116587239,
        "mean_feedback_forward_relative_error": 0.41535361591846715,
        "min_feedback_forward_cosine": 0.7163633108139038
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 9775,
      "train_examples": 50000,
      "train_loss": 0.3047918969631195
    },
    {
      "epoch": 26,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7459856271743774,
        "mean_feedback_forward_cosine": 0.9168003509121556,
        "mean_feedback_forward_relative_error": 0.39990548020409,
        "min_feedback_forward_cosine": 0.7331688404083252
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10166,
      "train_examples": 50000,
      "train_loss": 0.2932687960529327
    },
    {
      "epoch": 27,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.7222238779067993,
        "mean_feedback_forward_cosine": 0.9228495224829643,
        "mean_feedback_forward_relative_error": 0.38474732637405396,
        "min_feedback_forward_cosine": 0.7496653199195862
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10557,
      "train_examples": 50000,
      "train_loss": 0.2891711379432678
    },
    {
      "epoch": 28,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6987154483795166,
        "mean_feedback_forward_cosine": 0.9285203603006178,
        "mean_feedback_forward_relative_error": 0.37021790877465277,
        "min_feedback_forward_cosine": 0.7655463218688965
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 10948,
      "train_examples": 50000,
      "train_loss": 0.2846348499298096
    },
    {
      "epoch": 29,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6765077114105225,
        "mean_feedback_forward_cosine": 0.933629983855832,
        "mean_feedback_forward_relative_error": 0.35641476944569617,
        "min_feedback_forward_cosine": 0.7801104784011841
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11339,
      "train_examples": 50000,
      "train_loss": 0.28453598520278933
    },
    {
      "epoch": 30,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.6536064743995667,
        "mean_feedback_forward_cosine": 0.9384912733108767,
        "mean_feedback_forward_relative_error": 0.3428528928948987,
        "min_feedback_forward_cosine": 0.7945746183395386
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 11730,
      "train_examples": 50000,
      "train_loss": 0.2765980608868599
    },
    {
      "epoch": 31,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.631303608417511,
        "mean_feedback_forward_cosine": 0.9429083658802894,
        "mean_feedback_forward_relative_error": 0.33018695827453365,
        "min_feedback_forward_cosine": 0.808189868927002
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12121,
      "train_examples": 50000,
      "train_loss": 0.2714606456851959
    },
    {
      "epoch": 32,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.610146164894104,
        "mean_feedback_forward_cosine": 0.9470289849465893,
        "mean_feedback_forward_relative_error": 0.31787017180073646,
        "min_feedback_forward_cosine": 0.8205476999282837
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12512,
      "train_examples": 50000,
      "train_loss": 0.26684712250709536
    },
    {
      "epoch": 33,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5883633494377136,
        "mean_feedback_forward_cosine": 0.9509143098708122,
        "mean_feedback_forward_relative_error": 0.30590766403944264,
        "min_feedback_forward_cosine": 0.8330889344215393
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 12903,
      "train_examples": 50000,
      "train_loss": 0.2581737526512146
    },
    {
      "epoch": 34,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5682885646820068,
        "mean_feedback_forward_cosine": 0.954488661981398,
        "mean_feedback_forward_relative_error": 0.2944098338484764,
        "min_feedback_forward_cosine": 0.8441348075866699
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13294,
      "train_examples": 50000,
      "train_loss": 0.2537596423053741
    },
    {
      "epoch": 35,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5482786893844604,
        "mean_feedback_forward_cosine": 0.9578218037082303,
        "mean_feedback_forward_relative_error": 0.28327399444195533,
        "min_feedback_forward_cosine": 0.8548110723495483
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 13685,
      "train_examples": 50000,
      "train_loss": 0.24909167971611024
    },
    {
      "epoch": 36,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5286539793014526,
        "mean_feedback_forward_cosine": 0.9609254348662591,
        "mean_feedback_forward_relative_error": 0.2725322518137194,
        "min_feedback_forward_cosine": 0.8649052381515503
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14076,
      "train_examples": 50000,
      "train_loss": 0.2504457321071625
    },
    {
      "epoch": 37,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.5099877715110779,
        "mean_feedback_forward_cosine": 0.9636899290546295,
        "mean_feedback_forward_relative_error": 0.2626277352533033,
        "min_feedback_forward_cosine": 0.8742328882217407
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14467,
      "train_examples": 50000,
      "train_loss": 0.24400218109130858
    },
    {
      "epoch": 38,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4907877445220947,
        "mean_feedback_forward_cosine": 0.9664159205652052,
        "mean_feedback_forward_relative_error": 0.2525373976076803,
        "min_feedback_forward_cosine": 0.8832972645759583
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 14858,
      "train_examples": 50000,
      "train_loss": 0.24301130887031555
    },
    {
      "epoch": 39,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.47337213158607483,
        "mean_feedback_forward_cosine": 0.9688678172326857,
        "mean_feedback_forward_relative_error": 0.24305977407963045,
        "min_feedback_forward_cosine": 0.8913840055465698
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15249,
      "train_examples": 50000,
      "train_loss": 0.23847964415550232
    },
    {
      "epoch": 40,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4556553363800049,
        "mean_feedback_forward_cosine": 0.9711612674497789,
        "mean_feedback_forward_relative_error": 0.2338835645106531,
        "min_feedback_forward_cosine": 0.8993598222732544
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 15640,
      "train_examples": 50000,
      "train_loss": 0.2351317779827118
    },
    {
      "epoch": 41,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.43773984909057617,
        "mean_feedback_forward_cosine": 0.9733847475820973,
        "mean_feedback_forward_relative_error": 0.2246551436762656,
        "min_feedback_forward_cosine": 0.9070427417755127
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16031,
      "train_examples": 50000,
      "train_loss": 0.23791468878746033
    },
    {
      "epoch": 42,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.4209793508052826,
        "mean_feedback_forward_cosine": 0.9753380629324144,
        "mean_feedback_forward_relative_error": 0.21621643319245307,
        "min_feedback_forward_cosine": 0.9139285087585449
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16422,
      "train_examples": 50000,
      "train_loss": 0.22914538418769836
    },
    {
      "epoch": 43,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.40411102771759033,
        "mean_feedback_forward_cosine": 0.9771927402865502,
        "mean_feedback_forward_relative_error": 0.20793859900966769,
        "min_feedback_forward_cosine": 0.9206207394599915
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 16813,
      "train_examples": 50000,
      "train_loss": 0.22515615461349486
    },
    {
      "epoch": 44,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.38886141777038574,
        "mean_feedback_forward_cosine": 0.9788539909547375,
        "mean_feedback_forward_relative_error": 0.20017769115586434,
        "min_feedback_forward_cosine": 0.9264025688171387
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17204,
      "train_examples": 50000,
      "train_loss": 0.22249501668930055
    },
    {
      "epoch": 45,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3738791346549988,
        "mean_feedback_forward_cosine": 0.9804486267028316,
        "mean_feedback_forward_relative_error": 0.1924592797313967,
        "min_feedback_forward_cosine": 0.9318917393684387
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17595,
      "train_examples": 50000,
      "train_loss": 0.22315158581256866
    },
    {
      "epoch": 46,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3590734899044037,
        "mean_feedback_forward_cosine": 0.9819177581417945,
        "mean_feedback_forward_relative_error": 0.18506195684594492,
        "min_feedback_forward_cosine": 0.9371626973152161
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 17986,
      "train_examples": 50000,
      "train_loss": 0.2209020572423935
    },
    {
      "epoch": 47,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3451373279094696,
        "mean_feedback_forward_cosine": 0.9832455381270377,
        "mean_feedback_forward_relative_error": 0.1781211506214834,
        "min_feedback_forward_cosine": 0.9418581128120422
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18377,
      "train_examples": 50000,
      "train_loss": 0.21732004253387452
    },
    {
      "epoch": 48,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.331503301858902,
        "mean_feedback_forward_cosine": 0.9844934344291687,
        "mean_feedback_forward_relative_error": 0.1713354237858326,
        "min_feedback_forward_cosine": 0.9463133811950684
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 18768,
      "train_examples": 50000,
      "train_loss": 0.21509123004674913
    },
    {
      "epoch": 49,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.3184560537338257,
        "mean_feedback_forward_cosine": 0.9856345480488192,
        "mean_feedback_forward_relative_error": 0.16490591882217315,
        "min_feedback_forward_cosine": 0.9503976106643677
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19159,
      "train_examples": 50000,
      "train_loss": 0.21304820533752442
    },
    {
      "epoch": 50,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.30529871582984924,
        "mean_feedback_forward_cosine": 0.9867343152723005,
        "mean_feedback_forward_relative_error": 0.15851076764445152,
        "min_feedback_forward_cosine": 0.9543863534927368
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19550,
      "train_examples": 50000,
      "train_loss": 0.21148577075958253
    },
    {
      "epoch": 51,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.29255080223083496,
        "mean_feedback_forward_cosine": 0.9877838703893846,
        "mean_feedback_forward_relative_error": 0.15210549449247698,
        "min_feedback_forward_cosine": 0.9580827951431274
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 19941,
      "train_examples": 50000,
      "train_loss": 0.21210725172042846
    },
    {
      "epoch": 52,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.28086167573928833,
        "mean_feedback_forward_cosine": 0.9886988747504449,
        "mean_feedback_forward_relative_error": 0.1463090545467792,
        "min_feedback_forward_cosine": 0.9613544344902039
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20332,
      "train_examples": 50000,
      "train_loss": 0.2093979245519638
    },
    {
      "epoch": 53,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2697209119796753,
        "mean_feedback_forward_cosine": 0.9895473930143541,
        "mean_feedback_forward_relative_error": 0.14069656198543887,
        "min_feedback_forward_cosine": 0.9643323421478271
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 20723,
      "train_examples": 50000,
      "train_loss": 0.20638913254737853
    },
    {
      "epoch": 54,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2590985596179962,
        "mean_feedback_forward_cosine": 0.9903181868214761,
        "mean_feedback_forward_relative_error": 0.13540669770971422,
        "min_feedback_forward_cosine": 0.9670572280883789
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21114,
      "train_examples": 50000,
      "train_loss": 0.2027353626680374
    },
    {
      "epoch": 55,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.24897614121437073,
        "mean_feedback_forward_cosine": 0.9910333944905189,
        "mean_feedback_forward_relative_error": 0.13032094749712175,
        "min_feedback_forward_cosine": 0.9695420265197754
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21505,
      "train_examples": 50000,
      "train_loss": 0.1971484166240692
    },
    {
      "epoch": 56,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.23913848400115967,
        "mean_feedback_forward_cosine": 0.9916987149946151,
        "mean_feedback_forward_relative_error": 0.12539815325890818,
        "min_feedback_forward_cosine": 0.9718892574310303
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 21896,
      "train_examples": 50000,
      "train_loss": 0.2015540213871002
    },
    {
      "epoch": 57,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.22938434779644012,
        "mean_feedback_forward_cosine": 0.9923227448617259,
        "mean_feedback_forward_relative_error": 0.12060328981568737,
        "min_feedback_forward_cosine": 0.9741147756576538
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22287,
      "train_examples": 50000,
      "train_loss": 0.20267077679634093
    },
    {
      "epoch": 58,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.220234215259552,
        "mean_feedback_forward_cosine": 0.9928988775899333,
        "mean_feedback_forward_relative_error": 0.11598796721908354,
        "min_feedback_forward_cosine": 0.9761254787445068
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 22678,
      "train_examples": 50000,
      "train_loss": 0.19510222814559935
    },
    {
      "epoch": 59,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.21128128468990326,
        "mean_feedback_forward_cosine": 0.9934314162500443,
        "mean_feedback_forward_relative_error": 0.11157054478122343,
        "min_feedback_forward_cosine": 0.9780063629150391
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23069,
      "train_examples": 50000,
      "train_loss": 0.19726337415695192
    },
    {
      "epoch": 60,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.2025374472141266,
        "mean_feedback_forward_cosine": 0.9939339449328761,
        "mean_feedback_forward_relative_error": 0.1072343870276405,
        "min_feedback_forward_cosine": 0.9797813296318054
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23460,
      "train_examples": 50000,
      "train_loss": 0.1983313562297821
    },
    {
      "epoch": 61,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.19448445737361908,
        "mean_feedback_forward_cosine": 0.9943843233969903,
        "mean_feedback_forward_relative_error": 0.10317202777631822,
        "min_feedback_forward_cosine": 0.9813475608825684
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 23851,
      "train_examples": 50000,
      "train_loss": 0.1898783901500702
    },
    {
      "epoch": 62,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.18635331094264984,
        "mean_feedback_forward_cosine": 0.9948251804997844,
        "mean_feedback_forward_relative_error": 0.09904818537254487,
        "min_feedback_forward_cosine": 0.9828630685806274
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24242,
      "train_examples": 50000,
      "train_loss": 0.19620646570205688
    },
    {
      "epoch": 63,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1787845939397812,
        "mean_feedback_forward_cosine": 0.9952079942149501,
        "mean_feedback_forward_relative_error": 0.09532134523314814,
        "min_feedback_forward_cosine": 0.984218418598175
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 24633,
      "train_examples": 50000,
      "train_loss": 0.19244548971176148
    },
    {
      "epoch": 64,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.17157258093357086,
        "mean_feedback_forward_cosine": 0.995572809250124,
        "mean_feedback_forward_relative_error": 0.09162182679339763,
        "min_feedback_forward_cosine": 0.9854565858840942
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25024,
      "train_examples": 50000,
      "train_loss": 0.18765464908599855
    },
    {
      "epoch": 65,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1646691858768463,
        "mean_feedback_forward_cosine": 0.9959113328687607,
        "mean_feedback_forward_relative_error": 0.08806144157724996,
        "min_feedback_forward_cosine": 0.9865995645523071
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25415,
      "train_examples": 50000,
      "train_loss": 0.18908571057319642
    },
    {
      "epoch": 66,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15825730562210083,
        "mean_feedback_forward_cosine": 0.9962189274449502,
        "mean_feedback_forward_relative_error": 0.08468100735016408,
        "min_feedback_forward_cosine": 0.9876154661178589
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 25806,
      "train_examples": 50000,
      "train_loss": 0.18638300099372865
    },
    {
      "epoch": 67,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.15199866890907288,
        "mean_feedback_forward_cosine": 0.996499728771948,
        "mean_feedback_forward_relative_error": 0.08147101512839718,
        "min_feedback_forward_cosine": 0.9885689616203308
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26197,
      "train_examples": 50000,
      "train_loss": 0.1860867762851715
    },
    {
      "epoch": 68,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1457824558019638,
        "mean_feedback_forward_cosine": 0.9967687264565499,
        "mean_feedback_forward_relative_error": 0.07829492887662302,
        "min_feedback_forward_cosine": 0.9894801378250122
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26588,
      "train_examples": 50000,
      "train_loss": 0.18519178348064422
    },
    {
      "epoch": 69,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.1397344470024109,
        "mean_feedback_forward_cosine": 0.9970131401092776,
        "mean_feedback_forward_relative_error": 0.07528625374599811,
        "min_feedback_forward_cosine": 0.990331768989563
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 26979,
      "train_examples": 50000,
      "train_loss": 0.18635893783569335
    },
    {
      "epoch": 70,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.13424205780029297,
        "mean_feedback_forward_cosine": 0.997237301641895,
        "mean_feedback_forward_relative_error": 0.07240463306586589,
        "min_feedback_forward_cosine": 0.9910707473754883
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27370,
      "train_examples": 50000,
      "train_loss": 0.183787780418396
    },
    {
      "epoch": 71,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12852050364017487,
        "mean_feedback_forward_cosine": 0.9974454006841106,
        "mean_feedback_forward_relative_error": 0.06964071479535872,
        "min_feedback_forward_cosine": 0.9918137788772583
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 27761,
      "train_examples": 50000,
      "train_loss": 0.1831598755478859
    },
    {
      "epoch": 72,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.12334217876195908,
        "mean_feedback_forward_cosine": 0.9976358356014374,
        "mean_feedback_forward_relative_error": 0.06700158954387711,
        "min_feedback_forward_cosine": 0.9924593567848206
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28152,
      "train_examples": 50000,
      "train_loss": 0.1804471286201477
    },
    {
      "epoch": 73,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11835367977619171,
        "mean_feedback_forward_cosine": 0.9978146033902322,
        "mean_feedback_forward_relative_error": 0.06443115107474788,
        "min_feedback_forward_cosine": 0.9930556416511536
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28543,
      "train_examples": 50000,
      "train_loss": 0.17463825380325318
    },
    {
      "epoch": 74,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.11329466849565506,
        "mean_feedback_forward_cosine": 0.9979841824500791,
        "mean_feedback_forward_relative_error": 0.06190058997561855,
        "min_feedback_forward_cosine": 0.9936355352401733
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 28934,
      "train_examples": 50000,
      "train_loss": 0.18227414899349212
    },
    {
      "epoch": 75,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10864241421222687,
        "mean_feedback_forward_cosine": 0.998138023960975,
        "mean_feedback_forward_relative_error": 0.059497022821057226,
        "min_feedback_forward_cosine": 0.9941459894180298
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29325,
      "train_examples": 50000,
      "train_loss": 0.17801999889135361
    },
    {
      "epoch": 76,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10435745120048523,
        "mean_feedback_forward_cosine": 0.9982780121987865,
        "mean_feedback_forward_relative_error": 0.05721727512296169,
        "min_feedback_forward_cosine": 0.9945970773696899
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 29716,
      "train_examples": 50000,
      "train_loss": 0.1765562902069092
    },
    {
      "epoch": 77,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.10019416362047195,
        "mean_feedback_forward_cosine": 0.9984062794716128,
        "mean_feedback_forward_relative_error": 0.05505009404113216,
        "min_feedback_forward_cosine": 0.9950186014175415
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30107,
      "train_examples": 50000,
      "train_loss": 0.17533604710578918
    },
    {
      "epoch": 78,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09605756402015686,
        "mean_feedback_forward_cosine": 0.9985280709881936,
        "mean_feedback_forward_relative_error": 0.05291419536356003,
        "min_feedback_forward_cosine": 0.9954196810722351
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30498,
      "train_examples": 50000,
      "train_loss": 0.17509580305576325
    },
    {
      "epoch": 79,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.09225407242774963,
        "mean_feedback_forward_cosine": 0.9986377935255727,
        "mean_feedback_forward_relative_error": 0.05090813531029609,
        "min_feedback_forward_cosine": 0.9957756996154785
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 30889,
      "train_examples": 50000,
      "train_loss": 0.17200049627304076
    },
    {
      "epoch": 80,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.08852459490299225,
        "mean_feedback_forward_cosine": 0.9987405211694779,
        "mean_feedback_forward_relative_error": 0.04895767881985633,
        "min_feedback_forward_cosine": 0.9961096048355103
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31280,
      "train_examples": 50000,
      "train_loss": 0.1740963316631317
    },
    {
      "epoch": 81,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0847882628440857,
        "mean_feedback_forward_cosine": 0.9988403108812147,
        "mean_feedback_forward_relative_error": 0.0469820047578504,
        "min_feedback_forward_cosine": 0.9964291453361511
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 31671,
      "train_examples": 50000,
      "train_loss": 0.17462056126117706
    },
    {
      "epoch": 82,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.081540547311306,
        "mean_feedback_forward_cosine": 0.9989246245353453,
        "mean_feedback_forward_relative_error": 0.04524247964183169,
        "min_feedback_forward_cosine": 0.9966965317726135
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32062,
      "train_examples": 50000,
      "train_loss": 0.16816279454231262
    },
    {
      "epoch": 83,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07826635986566544,
        "mean_feedback_forward_cosine": 0.9990054138245121,
        "mean_feedback_forward_relative_error": 0.043515207546372565,
        "min_feedback_forward_cosine": 0.9969558715820312
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32453,
      "train_examples": 50000,
      "train_loss": 0.16755252022266387
    },
    {
      "epoch": 84,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07517033070325851,
        "mean_feedback_forward_cosine": 0.9990779449862819,
        "mean_feedback_forward_relative_error": 0.0419047734730186,
        "min_feedback_forward_cosine": 0.9971915483474731
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 32844,
      "train_examples": 50000,
      "train_loss": 0.16846428961753845
    },
    {
      "epoch": 85,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.07203090190887451,
        "mean_feedback_forward_cosine": 0.9991485553403054,
        "mean_feedback_forward_relative_error": 0.04027359788456271,
        "min_feedback_forward_cosine": 0.9974207282066345
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33235,
      "train_examples": 50000,
      "train_loss": 0.1688399866914749
    },
    {
      "epoch": 86,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06910812854766846,
        "mean_feedback_forward_cosine": 0.999212589956099,
        "mean_feedback_forward_relative_error": 0.03873308156166346,
        "min_feedback_forward_cosine": 0.9976263046264648
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 33626,
      "train_examples": 50000,
      "train_loss": 0.16943029206752777
    },
    {
      "epoch": 87,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06635339558124542,
        "mean_feedback_forward_cosine": 0.9992720696233934,
        "mean_feedback_forward_relative_error": 0.0372469199280585,
        "min_feedback_forward_cosine": 0.9978106021881104
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34017,
      "train_examples": 50000,
      "train_loss": 0.1678766533613205
    },
    {
      "epoch": 88,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06360498815774918,
        "mean_feedback_forward_cosine": 0.9993283556353662,
        "mean_feedback_forward_relative_error": 0.035780609765600775,
        "min_feedback_forward_cosine": 0.9979878664016724
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34408,
      "train_examples": 50000,
      "train_loss": 0.16858385267257692
    },
    {
      "epoch": 89,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.06101856008172035,
        "mean_feedback_forward_cosine": 0.9993801578398673,
        "mean_feedback_forward_relative_error": 0.03437901651787181,
        "min_feedback_forward_cosine": 0.9981483221054077
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 34799,
      "train_examples": 50000,
      "train_loss": 0.16932012516021727
    },
    {
      "epoch": 90,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05856332555413246,
        "mean_feedback_forward_cosine": 0.9994261822392864,
        "mean_feedback_forward_relative_error": 0.03308020258742955,
        "min_feedback_forward_cosine": 0.9982941746711731
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35190,
      "train_examples": 50000,
      "train_loss": 0.16621017497062682
    },
    {
      "epoch": 91,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.056274961680173874,
        "mean_feedback_forward_cosine": 0.9994692244837361,
        "mean_feedback_forward_relative_error": 0.031815457938899915,
        "min_feedback_forward_cosine": 0.9984242916107178
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35581,
      "train_examples": 50000,
      "train_loss": 0.16870774178981782
    },
    {
      "epoch": 92,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05400526523590088,
        "mean_feedback_forward_cosine": 0.9995097979422538,
        "mean_feedback_forward_relative_error": 0.03057598082288619,
        "min_feedback_forward_cosine": 0.9985485672950745
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 35972,
      "train_examples": 50000,
      "train_loss": 0.16830255025863647
    },
    {
      "epoch": 93,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.05168907344341278,
        "mean_feedback_forward_cosine": 0.9995479910604416,
        "mean_feedback_forward_relative_error": 0.02936676538158809,
        "min_feedback_forward_cosine": 0.998670220375061
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36363,
      "train_examples": 50000,
      "train_loss": 0.17085209632396697
    },
    {
      "epoch": 94,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04972684755921364,
        "mean_feedback_forward_cosine": 0.9995804256008517,
        "mean_feedback_forward_relative_error": 0.028296017298294653,
        "min_feedback_forward_cosine": 0.9987689256668091
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 36754,
      "train_examples": 50000,
      "train_loss": 0.1595303696537018
    },
    {
      "epoch": 95,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04777975007891655,
        "mean_feedback_forward_cosine": 0.9996111931339386,
        "mean_feedback_forward_relative_error": 0.027240780242268118,
        "min_feedback_forward_cosine": 0.9988634586334229
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37145,
      "train_examples": 50000,
      "train_loss": 0.1591718906545639
    },
    {
      "epoch": 96,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.04582907259464264,
        "mean_feedback_forward_cosine": 0.9996409127789159,
        "mean_feedback_forward_relative_error": 0.026182132983399976,
        "min_feedback_forward_cosine": 0.9989540576934814
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37536,
      "train_examples": 50000,
      "train_loss": 0.15775668437480928
    },
    {
      "epoch": 97,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.043918970972299576,
        "mean_feedback_forward_cosine": 0.9996690192530232,
        "mean_feedback_forward_relative_error": 0.025140733878699043,
        "min_feedback_forward_cosine": 0.9990392923355103
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 37927,
      "train_examples": 50000,
      "train_loss": 0.16501512162208556
    },
    {
      "epoch": 98,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.042085129767656326,
        "mean_feedback_forward_cosine": 0.9996941474176222,
        "mean_feedback_forward_relative_error": 0.02417049887439897,
        "min_feedback_forward_cosine": 0.9991177320480347
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38318,
      "train_examples": 50000,
      "train_loss": 0.1663829922580719
    },
    {
      "epoch": 99,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.040321219712495804,
        "mean_feedback_forward_cosine": 0.9997173393926313,
        "mean_feedback_forward_relative_error": 0.023241238458262335,
        "min_feedback_forward_cosine": 0.9991900324821472
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 38709,
      "train_examples": 50000,
      "train_loss": 0.16541705914497376
    },
    {
      "epoch": 100,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038703419268131256,
        "mean_feedback_forward_cosine": 0.9997384528959951,
        "mean_feedback_forward_relative_error": 0.022359017237660386,
        "min_feedback_forward_cosine": 0.9992535710334778
      },
      "lr": 0.1,
      "output_lr": 0.1,
      "step": 39100,
      "train_examples": 50000,
      "train_loss": 0.1572692034816742
    },
    {
      "epoch": 101,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03869009017944336,
        "mean_feedback_forward_cosine": 0.9997386201735465,
        "mean_feedback_forward_relative_error": 0.022349788747247187,
        "min_feedback_forward_cosine": 0.9992542266845703
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39491,
      "train_examples": 50000,
      "train_loss": 0.09018189778804779
    },
    {
      "epoch": 102,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03868260979652405,
        "mean_feedback_forward_cosine": 0.9997387355373751,
        "mean_feedback_forward_relative_error": 0.02234361457428144,
        "min_feedback_forward_cosine": 0.9992544651031494
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 39882,
      "train_examples": 50000,
      "train_loss": 0.0702357754611969
    },
    {
      "epoch": 103,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038675710558891296,
        "mean_feedback_forward_cosine": 0.9997388489784733,
        "mean_feedback_forward_relative_error": 0.02233777842634628,
        "min_feedback_forward_cosine": 0.9992546439170837
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40273,
      "train_examples": 50000,
      "train_loss": 0.05911954815030098
    },
    {
      "epoch": 104,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038668934255838394,
        "mean_feedback_forward_cosine": 0.9997389431922666,
        "mean_feedback_forward_relative_error": 0.02233229835908259,
        "min_feedback_forward_cosine": 0.9992549419403076
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 40664,
      "train_examples": 50000,
      "train_loss": 0.053982924233675005
    },
    {
      "epoch": 105,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03866267576813698,
        "mean_feedback_forward_cosine": 0.9997390393287905,
        "mean_feedback_forward_relative_error": 0.022327132688294494,
        "min_feedback_forward_cosine": 0.9992551803588867
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41055,
      "train_examples": 50000,
      "train_loss": 0.04768824928998947
    },
    {
      "epoch": 106,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038656026124954224,
        "mean_feedback_forward_cosine": 0.9997391373880448,
        "mean_feedback_forward_relative_error": 0.022322015085768317,
        "min_feedback_forward_cosine": 0.9992554187774658
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41446,
      "train_examples": 50000,
      "train_loss": 0.044762988986968996
    },
    {
      "epoch": 107,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03864984214305878,
        "mean_feedback_forward_cosine": 0.9997392354472991,
        "mean_feedback_forward_relative_error": 0.022317053509816047,
        "min_feedback_forward_cosine": 0.9992556571960449
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 41837,
      "train_examples": 50000,
      "train_loss": 0.04090550207406282
    },
    {
      "epoch": 108,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03864375129342079,
        "mean_feedback_forward_cosine": 0.9997393162019791,
        "mean_feedback_forward_relative_error": 0.022312285679001963,
        "min_feedback_forward_cosine": 0.9992560148239136
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42228,
      "train_examples": 50000,
      "train_loss": 0.03672991137981415
    },
    {
      "epoch": 109,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03863723203539848,
        "mean_feedback_forward_cosine": 0.9997393873430067,
        "mean_feedback_forward_relative_error": 0.022307376302177866,
        "min_feedback_forward_cosine": 0.9992561340332031
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 42619,
      "train_examples": 50000,
      "train_loss": 0.03636346148490906
    },
    {
      "epoch": 110,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03863116726279259,
        "mean_feedback_forward_cosine": 0.999739489247722,
        "mean_feedback_forward_relative_error": 0.022302621720178474,
        "min_feedback_forward_cosine": 0.9992564916610718
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43010,
      "train_examples": 50000,
      "train_loss": 0.03307025036215782
    },
    {
      "epoch": 111,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0386250838637352,
        "mean_feedback_forward_cosine": 0.9997395642342106,
        "mean_feedback_forward_relative_error": 0.02229790753054042,
        "min_feedback_forward_cosine": 0.9992567300796509
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43401,
      "train_examples": 50000,
      "train_loss": 0.032560025746524335
    },
    {
      "epoch": 112,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03861873596906662,
        "mean_feedback_forward_cosine": 0.9997396680616564,
        "mean_feedback_forward_relative_error": 0.022293061754035372,
        "min_feedback_forward_cosine": 0.9992568492889404
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 43792,
      "train_examples": 50000,
      "train_loss": 0.031159462106227876
    },
    {
      "epoch": 113,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038612477481365204,
        "mean_feedback_forward_cosine": 0.9997397622754497,
        "mean_feedback_forward_relative_error": 0.02228844101210275,
        "min_feedback_forward_cosine": 0.9992570877075195
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44183,
      "train_examples": 50000,
      "train_loss": 0.02895025592982769
    },
    {
      "epoch": 114,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03860635310411453,
        "mean_feedback_forward_cosine": 0.9997398526437821,
        "mean_feedback_forward_relative_error": 0.022283784038717707,
        "min_feedback_forward_cosine": 0.9992573857307434
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44574,
      "train_examples": 50000,
      "train_loss": 0.027717424998283387
    },
    {
      "epoch": 115,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038599804043769836,
        "mean_feedback_forward_cosine": 0.9997399487803059,
        "mean_feedback_forward_relative_error": 0.022278990461340836,
        "min_feedback_forward_cosine": 0.9992576837539673
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 44965,
      "train_examples": 50000,
      "train_loss": 0.02742766058623791
    },
    {
      "epoch": 116,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0385938324034214,
        "mean_feedback_forward_cosine": 0.9997400372259079,
        "mean_feedback_forward_relative_error": 0.022274345550085266,
        "min_feedback_forward_cosine": 0.9992578029632568
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45356,
      "train_examples": 50000,
      "train_loss": 0.02480702959358692
    },
    {
      "epoch": 117,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03858798369765282,
        "mean_feedback_forward_cosine": 0.9997401179805878,
        "mean_feedback_forward_relative_error": 0.022269865693224054,
        "min_feedback_forward_cosine": 0.9992581605911255
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 45747,
      "train_examples": 50000,
      "train_loss": 0.024096279681921004
    },
    {
      "epoch": 118,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03858125954866409,
        "mean_feedback_forward_cosine": 0.9997402064261898,
        "mean_feedback_forward_relative_error": 0.0222650226506014,
        "min_feedback_forward_cosine": 0.9992583990097046
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46138,
      "train_examples": 50000,
      "train_loss": 0.02406107524037361
    },
    {
      "epoch": 119,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03857496753334999,
        "mean_feedback_forward_cosine": 0.999740319867288,
        "mean_feedback_forward_relative_error": 0.02226024840567862,
        "min_feedback_forward_cosine": 0.9992586374282837
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46529,
      "train_examples": 50000,
      "train_loss": 0.022692273498773573
    },
    {
      "epoch": 120,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03856874629855156,
        "mean_feedback_forward_cosine": 0.9997404025446984,
        "mean_feedback_forward_relative_error": 0.022255517061679594,
        "min_feedback_forward_cosine": 0.9992587566375732
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 46920,
      "train_examples": 50000,
      "train_loss": 0.02152857553243637
    },
    {
      "epoch": 121,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03856196999549866,
        "mean_feedback_forward_cosine": 0.9997404948357613,
        "mean_feedback_forward_relative_error": 0.022250843399594867,
        "min_feedback_forward_cosine": 0.9992590546607971
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47311,
      "train_examples": 50000,
      "train_loss": 0.021097033315598965
    },
    {
      "epoch": 122,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03855599835515022,
        "mean_feedback_forward_cosine": 0.9997405813586328,
        "mean_feedback_forward_relative_error": 0.022246450426117066,
        "min_feedback_forward_cosine": 0.999259352684021
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 47702,
      "train_examples": 50000,
      "train_loss": 0.019415753941237925
    },
    {
      "epoch": 123,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038549765944480896,
        "mean_feedback_forward_cosine": 0.999740679417887,
        "mean_feedback_forward_relative_error": 0.022241957680953127,
        "min_feedback_forward_cosine": 0.9992594718933105
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48093,
      "train_examples": 50000,
      "train_loss": 0.018513377546072005
    },
    {
      "epoch": 124,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038543179631233215,
        "mean_feedback_forward_cosine": 0.9997407717089499,
        "mean_feedback_forward_relative_error": 0.022237125033092116,
        "min_feedback_forward_cosine": 0.9992598295211792
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48484,
      "train_examples": 50000,
      "train_loss": 0.019798150712773206
    },
    {
      "epoch": 125,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03853682428598404,
        "mean_feedback_forward_cosine": 0.9997408832273176,
        "mean_feedback_forward_relative_error": 0.022232338395570556,
        "min_feedback_forward_cosine": 0.9992600679397583
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 48875,
      "train_examples": 50000,
      "train_loss": 0.019051756429374216
    },
    {
      "epoch": 126,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03853025287389755,
        "mean_feedback_forward_cosine": 0.9997409524456147,
        "mean_feedback_forward_relative_error": 0.022227739990358392,
        "min_feedback_forward_cosine": 0.9992603063583374
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49266,
      "train_examples": 50000,
      "train_loss": 0.01761170121088624
    },
    {
      "epoch": 127,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03852350637316704,
        "mean_feedback_forward_cosine": 0.9997410678094433,
        "mean_feedback_forward_relative_error": 0.0222228912696723,
        "min_feedback_forward_cosine": 0.999260425567627
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 49657,
      "train_examples": 50000,
      "train_loss": 0.018012608473673464
    },
    {
      "epoch": 128,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03851630166172981,
        "mean_feedback_forward_cosine": 0.9997411639459671,
        "mean_feedback_forward_relative_error": 0.02221776469941101,
        "min_feedback_forward_cosine": 0.9992607831954956
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50048,
      "train_examples": 50000,
      "train_loss": 0.018129425482228397
    },
    {
      "epoch": 129,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03850993141531944,
        "mean_feedback_forward_cosine": 0.9997412773870653,
        "mean_feedback_forward_relative_error": 0.022213073372240027,
        "min_feedback_forward_cosine": 0.9992610216140747
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50439,
      "train_examples": 50000,
      "train_loss": 0.016865740612745286
    },
    {
      "epoch": 130,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03850335255265236,
        "mean_feedback_forward_cosine": 0.9997413504508234,
        "mean_feedback_forward_relative_error": 0.022208275228378274,
        "min_feedback_forward_cosine": 0.9992613196372986
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 50830,
      "train_examples": 50000,
      "train_loss": 0.016044229831546544
    },
    {
      "epoch": 131,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03849764168262482,
        "mean_feedback_forward_cosine": 0.9997414331282338,
        "mean_feedback_forward_relative_error": 0.022204190267310027,
        "min_feedback_forward_cosine": 0.9992614984512329
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51221,
      "train_examples": 50000,
      "train_loss": 0.014237593904286623
    },
    {
      "epoch": 132,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03849118947982788,
        "mean_feedback_forward_cosine": 0.9997415408011405,
        "mean_feedback_forward_relative_error": 0.022199535321804784,
        "min_feedback_forward_cosine": 0.999261736869812
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 51612,
      "train_examples": 50000,
      "train_loss": 0.015662225911915302
    },
    {
      "epoch": 133,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03848356753587723,
        "mean_feedback_forward_cosine": 0.9997416446285863,
        "mean_feedback_forward_relative_error": 0.022194101505221858,
        "min_feedback_forward_cosine": 0.9992619752883911
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52003,
      "train_examples": 50000,
      "train_loss": 0.015363968058675528
    },
    {
      "epoch": 134,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038476429879665375,
        "mean_feedback_forward_cosine": 0.9997417638378758,
        "mean_feedback_forward_relative_error": 0.022188984173079653,
        "min_feedback_forward_cosine": 0.9992623329162598
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52394,
      "train_examples": 50000,
      "train_loss": 0.015120945150852204
    },
    {
      "epoch": 135,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0384698286652565,
        "mean_feedback_forward_cosine": 0.9997418618971302,
        "mean_feedback_forward_relative_error": 0.022184129383775494,
        "min_feedback_forward_cosine": 0.9992625713348389
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 52785,
      "train_examples": 50000,
      "train_loss": 0.013894460175037385
    },
    {
      "epoch": 136,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038462795317173004,
        "mean_feedback_forward_cosine": 0.9997419811064198,
        "mean_feedback_forward_relative_error": 0.022179397814456495,
        "min_feedback_forward_cosine": 0.999262809753418
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53176,
      "train_examples": 50000,
      "train_loss": 0.014200187674462796
    },
    {
      "epoch": 137,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03845585510134697,
        "mean_feedback_forward_cosine": 0.9997420753202131,
        "mean_feedback_forward_relative_error": 0.022174607391559308,
        "min_feedback_forward_cosine": 0.9992630481719971
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53567,
      "train_examples": 50000,
      "train_loss": 0.014059381691217423
    },
    {
      "epoch": 138,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03844912350177765,
        "mean_feedback_forward_cosine": 0.9997421599203541,
        "mean_feedback_forward_relative_error": 0.022169834213151086,
        "min_feedback_forward_cosine": 0.999263346195221
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 53958,
      "train_examples": 50000,
      "train_loss": 0.013133898133039475
    },
    {
      "epoch": 139,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038440823554992676,
        "mean_feedback_forward_cosine": 0.9997422964342179,
        "mean_feedback_forward_relative_error": 0.022164240374319976,
        "min_feedback_forward_cosine": 0.9992636442184448
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54349,
      "train_examples": 50000,
      "train_loss": 0.015220713586062192
    },
    {
      "epoch": 140,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03843340277671814,
        "mean_feedback_forward_cosine": 0.9997424233344293,
        "mean_feedback_forward_relative_error": 0.022158963921209498,
        "min_feedback_forward_cosine": 0.9992640018463135
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 54740,
      "train_examples": 50000,
      "train_loss": 0.013820327893048524
    },
    {
      "epoch": 141,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038426365703344345,
        "mean_feedback_forward_cosine": 0.9997424983209179,
        "mean_feedback_forward_relative_error": 0.02215411449452081,
        "min_feedback_forward_cosine": 0.9992642998695374
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55131,
      "train_examples": 50000,
      "train_loss": 0.012681063927710056
    },
    {
      "epoch": 142,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03841947391629219,
        "mean_feedback_forward_cosine": 0.9997426232983989,
        "mean_feedback_forward_relative_error": 0.02214897439003952,
        "min_feedback_forward_cosine": 0.9992644786834717
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55522,
      "train_examples": 50000,
      "train_loss": 0.012453560272157192
    },
    {
      "epoch": 143,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038412805646657944,
        "mean_feedback_forward_cosine": 0.9997427309713056,
        "mean_feedback_forward_relative_error": 0.02214435871029573,
        "min_feedback_forward_cosine": 0.9992647171020508
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 55913,
      "train_examples": 50000,
      "train_loss": 0.011589504393190146
    },
    {
      "epoch": 144,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03840521723031998,
        "mean_feedback_forward_cosine": 0.9997428347987514,
        "mean_feedback_forward_relative_error": 0.02213895676897899,
        "min_feedback_forward_cosine": 0.9992650747299194
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56304,
      "train_examples": 50000,
      "train_loss": 0.012354984072893859
    },
    {
      "epoch": 145,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0383986234664917,
        "mean_feedback_forward_cosine": 0.9997429443943885,
        "mean_feedback_forward_relative_error": 0.02213434368792561,
        "min_feedback_forward_cosine": 0.9992653727531433
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 56695,
      "train_examples": 50000,
      "train_loss": 0.010935525473169982
    },
    {
      "epoch": 146,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03839060664176941,
        "mean_feedback_forward_cosine": 0.9997430482218342,
        "mean_feedback_forward_relative_error": 0.022128906491543015,
        "min_feedback_forward_cosine": 0.9992656707763672
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57086,
      "train_examples": 50000,
      "train_loss": 0.012370056481957435
    },
    {
      "epoch": 147,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03838308900594711,
        "mean_feedback_forward_cosine": 0.9997431751220457,
        "mean_feedback_forward_relative_error": 0.022123666600354256,
        "min_feedback_forward_cosine": 0.9992658495903015
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57477,
      "train_examples": 50000,
      "train_loss": 0.011595427350848913
    },
    {
      "epoch": 148,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038375068455934525,
        "mean_feedback_forward_cosine": 0.9997432770267609,
        "mean_feedback_forward_relative_error": 0.02211842517698965,
        "min_feedback_forward_cosine": 0.9992662668228149
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 57868,
      "train_examples": 50000,
      "train_loss": 0.011534525494948029
    },
    {
      "epoch": 149,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03836674615740776,
        "mean_feedback_forward_cosine": 0.99974339431332,
        "mean_feedback_forward_relative_error": 0.022112570748093627,
        "min_feedback_forward_cosine": 0.999266505241394
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58259,
      "train_examples": 50000,
      "train_loss": 0.012372618612647057
    },
    {
      "epoch": 150,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835928067564964,
        "mean_feedback_forward_cosine": 0.9997435462090277,
        "mean_feedback_forward_relative_error": 0.022107558430082375,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.010000000000000002,
      "output_lr": 0.010000000000000002,
      "step": 58650,
      "train_examples": 50000,
      "train_loss": 0.0101877372437343
    },
    {
      "epoch": 151,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835919499397278,
        "mean_feedback_forward_cosine": 0.9997435173680705,
        "mean_feedback_forward_relative_error": 0.022107452905226137,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59041,
      "train_examples": 50000,
      "train_loss": 0.01016308972582221
    },
    {
      "epoch": 152,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038359135389328,
        "mean_feedback_forward_cosine": 0.9997435077544181,
        "mean_feedback_forward_relative_error": 0.022107359772968675,
        "min_feedback_forward_cosine": 0.9992667436599731
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59432,
      "train_examples": 50000,
      "train_loss": 0.009141140254288912
    },
    {
      "epoch": 153,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835907578468323,
        "mean_feedback_forward_cosine": 0.9997435000634962,
        "mean_feedback_forward_relative_error": 0.02210726635530591,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 59823,
      "train_examples": 50000,
      "train_loss": 0.008568081556633115
    },
    {
      "epoch": 154,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038359012454748154,
        "mean_feedback_forward_cosine": 0.9997435077544181,
        "mean_feedback_forward_relative_error": 0.022107177429021366,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60214,
      "train_examples": 50000,
      "train_loss": 0.00920205535236746
    },
    {
      "epoch": 155,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038358960300683975,
        "mean_feedback_forward_cosine": 0.9997435327499143,
        "mean_feedback_forward_relative_error": 0.022107091852493824,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60605,
      "train_examples": 50000,
      "train_loss": 0.008157383770346641
    },
    {
      "epoch": 156,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038358911871910095,
        "mean_feedback_forward_cosine": 0.9997435135226096,
        "mean_feedback_forward_relative_error": 0.022107006786691565,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 60996,
      "train_examples": 50000,
      "train_loss": 0.007641669589877128
    },
    {
      "epoch": 157,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038358863443136215,
        "mean_feedback_forward_cosine": 0.9997435404408362,
        "mean_feedback_forward_relative_error": 0.022106922456934568,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61387,
      "train_examples": 50000,
      "train_loss": 0.007855133966505527
    },
    {
      "epoch": 158,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038358818739652634,
        "mean_feedback_forward_cosine": 0.9997435250589924,
        "mean_feedback_forward_relative_error": 0.022106839824588068,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 61778,
      "train_examples": 50000,
      "train_loss": 0.007312930462509394
    },
    {
      "epoch": 159,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038358770310878754,
        "mean_feedback_forward_cosine": 0.99974351544534,
        "mean_feedback_forward_relative_error": 0.02210676003127329,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62169,
      "train_examples": 50000,
      "train_loss": 0.007122721250653267
    },
    {
      "epoch": 160,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038358718156814575,
        "mean_feedback_forward_cosine": 0.999743526981723,
        "mean_feedback_forward_relative_error": 0.022106672652185924,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62560,
      "train_examples": 50000,
      "train_loss": 0.008205709009766578
    },
    {
      "epoch": 161,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0383586660027504,
        "mean_feedback_forward_cosine": 0.9997435250589924,
        "mean_feedback_forward_relative_error": 0.022106589524135474,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 62951,
      "train_examples": 50000,
      "train_loss": 0.007589992175102234
    },
    {
      "epoch": 162,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835861757397652,
        "mean_feedback_forward_cosine": 0.9997435212135315,
        "mean_feedback_forward_relative_error": 0.022106507507663582,
        "min_feedback_forward_cosine": 0.9992667436599731
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63342,
      "train_examples": 50000,
      "train_loss": 0.007849129783064127
    },
    {
      "epoch": 163,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835856169462204,
        "mean_feedback_forward_cosine": 0.999743526981723,
        "mean_feedback_forward_relative_error": 0.022106419572786938,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 63733,
      "train_examples": 50000,
      "train_loss": 0.007677926018461585
    },
    {
      "epoch": 164,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835850954055786,
        "mean_feedback_forward_cosine": 0.9997435212135315,
        "mean_feedback_forward_relative_error": 0.022106339013384234,
        "min_feedback_forward_cosine": 0.9992667436599731
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64124,
      "train_examples": 50000,
      "train_loss": 0.007710860101804137
    },
    {
      "epoch": 165,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835846483707428,
        "mean_feedback_forward_cosine": 0.999743526981723,
        "mean_feedback_forward_relative_error": 0.022106256290909745,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64515,
      "train_examples": 50000,
      "train_loss": 0.00716620623677969
    },
    {
      "epoch": 166,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0383584089577198,
        "mean_feedback_forward_cosine": 0.9997435462090277,
        "mean_feedback_forward_relative_error": 0.022106170413955566,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 64906,
      "train_examples": 50000,
      "train_loss": 0.00753851116836071
    },
    {
      "epoch": 167,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835836052894592,
        "mean_feedback_forward_cosine": 0.9997435346726449,
        "mean_feedback_forward_relative_error": 0.022106088337398345,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65297,
      "train_examples": 50000,
      "train_loss": 0.007372459765523672
    },
    {
      "epoch": 168,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038358304649591446,
        "mean_feedback_forward_cosine": 0.9997435423635668,
        "mean_feedback_forward_relative_error": 0.022106010902432666,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 65688,
      "train_examples": 50000,
      "train_loss": 0.006897408326864243
    },
    {
      "epoch": 169,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835826367139816,
        "mean_feedback_forward_cosine": 0.9997435289044534,
        "mean_feedback_forward_relative_error": 0.02210593456402421,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66079,
      "train_examples": 50000,
      "train_loss": 0.006711386793404817
    },
    {
      "epoch": 170,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835821896791458,
        "mean_feedback_forward_cosine": 0.9997435442862972,
        "mean_feedback_forward_relative_error": 0.022105860493836865,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66470,
      "train_examples": 50000,
      "train_loss": 0.006610966650322079
    },
    {
      "epoch": 171,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0383581705391407,
        "mean_feedback_forward_cosine": 0.9997435385181058,
        "mean_feedback_forward_relative_error": 0.02210578015975414,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 66861,
      "train_examples": 50000,
      "train_loss": 0.007155023907870054
    },
    {
      "epoch": 172,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835812211036682,
        "mean_feedback_forward_cosine": 0.9997435308271839,
        "mean_feedback_forward_relative_error": 0.02210570598441747,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67252,
      "train_examples": 50000,
      "train_loss": 0.006892853389084339
    },
    {
      "epoch": 173,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835806995630264,
        "mean_feedback_forward_cosine": 0.9997435462090277,
        "mean_feedback_forward_relative_error": 0.022105625575228084,
        "min_feedback_forward_cosine": 0.9992669820785522
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 67643,
      "train_examples": 50000,
      "train_loss": 0.0068606297035515304
    },
    {
      "epoch": 174,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835802152752876,
        "mean_feedback_forward_cosine": 0.9997435423635668,
        "mean_feedback_forward_relative_error": 0.022105549372011615,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68034,
      "train_examples": 50000,
      "train_loss": 0.006844044959545136
    },
    {
      "epoch": 175,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835797309875488,
        "mean_feedback_forward_cosine": 0.9997435250589924,
        "mean_feedback_forward_relative_error": 0.022105472192408576,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68425,
      "train_examples": 50000,
      "train_loss": 0.0070141338916122915
    },
    {
      "epoch": 176,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.0383579358458519,
        "mean_feedback_forward_cosine": 0.9997435365953753,
        "mean_feedback_forward_relative_error": 0.022105397746687936,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 68816,
      "train_examples": 50000,
      "train_loss": 0.006231199810951949
    },
    {
      "epoch": 177,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357898592948914,
        "mean_feedback_forward_cosine": 0.9997435385181058,
        "mean_feedback_forward_relative_error": 0.02210531798341582,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69207,
      "train_examples": 50000,
      "train_loss": 0.007026433295011521
    },
    {
      "epoch": 178,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835785388946533,
        "mean_feedback_forward_cosine": 0.9997435404408362,
        "mean_feedback_forward_relative_error": 0.022105244168591116,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69598,
      "train_examples": 50000,
      "train_loss": 0.0062881049451977014
    },
    {
      "epoch": 179,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835780918598175,
        "mean_feedback_forward_cosine": 0.9997435346726449,
        "mean_feedback_forward_relative_error": 0.02210516375940173,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 69989,
      "train_examples": 50000,
      "train_loss": 0.006777262418195605
    },
    {
      "epoch": 180,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835776820778847,
        "mean_feedback_forward_cosine": 0.99974355582268,
        "mean_feedback_forward_relative_error": 0.022105090335131652,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70380,
      "train_examples": 50000,
      "train_loss": 0.0069646171709895136
    },
    {
      "epoch": 181,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835771977901459,
        "mean_feedback_forward_cosine": 0.9997435423635668,
        "mean_feedback_forward_relative_error": 0.022105010481731545,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 70771,
      "train_examples": 50000,
      "train_loss": 0.006164614319801331
    },
    {
      "epoch": 182,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357678800821304,
        "mean_feedback_forward_cosine": 0.9997435481317581,
        "mean_feedback_forward_relative_error": 0.02210493797376271,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71162,
      "train_examples": 50000,
      "train_loss": 0.006379115580543875
    },
    {
      "epoch": 183,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835761547088623,
        "mean_feedback_forward_cosine": 0.9997435519772191,
        "mean_feedback_forward_relative_error": 0.022104859502325134,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71553,
      "train_examples": 50000,
      "train_loss": 0.006984152169078588
    },
    {
      "epoch": 184,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835757076740265,
        "mean_feedback_forward_cosine": 0.9997435519772191,
        "mean_feedback_forward_relative_error": 0.022104779618882364,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 71944,
      "train_examples": 50000,
      "train_loss": 0.0067542794423550365
    },
    {
      "epoch": 185,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835752233862877,
        "mean_feedback_forward_cosine": 0.9997435308271839,
        "mean_feedback_forward_relative_error": 0.022104701553020748,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72335,
      "train_examples": 50000,
      "train_loss": 0.006420430699884892
    },
    {
      "epoch": 186,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357485085725784,
        "mean_feedback_forward_cosine": 0.9997435404408362,
        "mean_feedback_forward_relative_error": 0.022104620993618044,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 72726,
      "train_examples": 50000,
      "train_loss": 0.0062022028070688245
    },
    {
      "epoch": 187,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357436656951904,
        "mean_feedback_forward_cosine": 0.9997435538999496,
        "mean_feedback_forward_relative_error": 0.022104545676660155,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73117,
      "train_examples": 50000,
      "train_loss": 0.006654117411822081
    },
    {
      "epoch": 188,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357388228178024,
        "mean_feedback_forward_cosine": 0.9997435500544887,
        "mean_feedback_forward_relative_error": 0.022104472703030034,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73508,
      "train_examples": 50000,
      "train_loss": 0.006000056919679046
    },
    {
      "epoch": 189,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357336074113846,
        "mean_feedback_forward_cosine": 0.9997435423635668,
        "mean_feedback_forward_relative_error": 0.022104391677966042,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 73899,
      "train_examples": 50000,
      "train_loss": 0.007019033769890666
    },
    {
      "epoch": 190,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835730627179146,
        "mean_feedback_forward_cosine": 0.9997435519772191,
        "mean_feedback_forward_relative_error": 0.0221043243072927,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74290,
      "train_examples": 50000,
      "train_loss": 0.006082673249095678
    },
    {
      "epoch": 191,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835725784301758,
        "mean_feedback_forward_cosine": 0.9997435692817934,
        "mean_feedback_forward_relative_error": 0.02210424972638007,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 74681,
      "train_examples": 50000,
      "train_loss": 0.006554380866661668
    },
    {
      "epoch": 192,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835722431540489,
        "mean_feedback_forward_cosine": 0.9997435635136019,
        "mean_feedback_forward_relative_error": 0.022104179381483024,
        "min_feedback_forward_cosine": 0.9992669820785522
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75072,
      "train_examples": 50000,
      "train_loss": 0.005760983594581485
    },
    {
      "epoch": 193,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835718333721161,
        "mean_feedback_forward_cosine": 0.9997435385181058,
        "mean_feedback_forward_relative_error": 0.022104102427199963,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75463,
      "train_examples": 50000,
      "train_loss": 0.006248740261644125
    },
    {
      "epoch": 194,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357146084308624,
        "mean_feedback_forward_cosine": 0.9997435308271839,
        "mean_feedback_forward_relative_error": 0.022104029844124472,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 75854,
      "train_examples": 50000,
      "train_loss": 0.00558479504339397
    },
    {
      "epoch": 195,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835710138082504,
        "mean_feedback_forward_cosine": 0.9997435481317581,
        "mean_feedback_forward_relative_error": 0.02210396020523002,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76245,
      "train_examples": 50000,
      "train_loss": 0.00593255756765604
    },
    {
      "epoch": 196,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835706040263176,
        "mean_feedback_forward_cosine": 0.9997435712045238,
        "mean_feedback_forward_relative_error": 0.022103888974074396,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 76636,
      "train_examples": 50000,
      "train_loss": 0.005618409241810441
    },
    {
      "epoch": 197,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038357026875019073,
        "mean_feedback_forward_cosine": 0.9997435769727153,
        "mean_feedback_forward_relative_error": 0.022103817893132087,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77027,
      "train_examples": 50000,
      "train_loss": 0.005956166023015976
    },
    {
      "epoch": 198,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.038356974720954895,
        "mean_feedback_forward_cosine": 0.9997435500544887,
        "mean_feedback_forward_relative_error": 0.022103737018281413,
        "min_feedback_forward_cosine": 0.9992668032646179
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77418,
      "train_examples": 50000,
      "train_loss": 0.006610512352585793
    },
    {
      "epoch": 199,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835694119334221,
        "mean_feedback_forward_cosine": 0.9997435423635668,
        "mean_feedback_forward_relative_error": 0.022103664119757952,
        "min_feedback_forward_cosine": 0.9992668628692627
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 77809,
      "train_examples": 50000,
      "train_loss": 0.005906309532299638
    },
    {
      "epoch": 200,
      "feedback_tracking": {
        "max_feedback_forward_relative_error": 0.03835690766572952,
        "mean_feedback_forward_cosine": 0.9997435442862972,
        "mean_feedback_forward_relative_error": 0.022103592227663723,
        "min_feedback_forward_cosine": 0.9992669224739075
      },
      "lr": 0.0010000000000000002,
      "output_lr": 0.0010000000000000002,
      "step": 78200,
      "train_examples": 50000,
      "train_loss": 0.005890884000062943
    }
  ],
  "evaluation_protocol": {
    "test_evaluations": 1,
    "test_used_for_selection": false,
    "validation_evaluations": 0
  },
  "final": {
    "accuracy": 0.9242,
    "epoch": 200,
    "evaluation_split": "test",
    "finite": true,
    "loss": 0.3821306213378906,
    "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": 11,
    "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": 1.1649303436279297,
    "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": 10230.376488685608,
    "train_wall_s": 10227.093840837479
  },
  "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
  }
}