{
  "best_metric": 0.04138818383216858,
  "best_model_checkpoint": "/data/wheld3/byt5-small-cstop_artificial/checkpoint-400",
  "epoch": 375.0,
  "global_step": 3000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 25.0,
      "learning_rate": 0.0009333333333333333,
      "loss": 0.2091,
      "step": 200
    },
    {
      "epoch": 25.0,
      "eval_exact_match": 0.01073345259391771,
      "eval_loss": 0.05550273135304451,
      "eval_runtime": 14.4433,
      "eval_samples_per_second": 38.703,
      "eval_steps_per_second": 1.246,
      "step": 200
    },
    {
      "epoch": 50.0,
      "learning_rate": 0.0008666666666666667,
      "loss": 0.0129,
      "step": 400
    },
    {
      "epoch": 50.0,
      "eval_exact_match": 0.04114490161001789,
      "eval_loss": 0.04138818383216858,
      "eval_runtime": 15.5988,
      "eval_samples_per_second": 35.836,
      "eval_steps_per_second": 1.154,
      "step": 400
    },
    {
      "epoch": 75.0,
      "learning_rate": 0.0008,
      "loss": 0.004,
      "step": 600
    },
    {
      "epoch": 75.0,
      "eval_exact_match": 0.03935599284436494,
      "eval_loss": 0.04834498465061188,
      "eval_runtime": 14.4825,
      "eval_samples_per_second": 38.598,
      "eval_steps_per_second": 1.243,
      "step": 600
    },
    {
      "epoch": 100.0,
      "learning_rate": 0.0007333333333333333,
      "loss": 0.0018,
      "step": 800
    },
    {
      "epoch": 100.0,
      "eval_exact_match": 0.03935599284436494,
      "eval_loss": 0.0555691160261631,
      "eval_runtime": 14.6196,
      "eval_samples_per_second": 38.236,
      "eval_steps_per_second": 1.231,
      "step": 800
    },
    {
      "epoch": 125.0,
      "learning_rate": 0.0006666666666666666,
      "loss": 0.0011,
      "step": 1000
    },
    {
      "epoch": 125.0,
      "eval_exact_match": 0.03935599284436494,
      "eval_loss": 0.05983056500554085,
      "eval_runtime": 14.3768,
      "eval_samples_per_second": 38.882,
      "eval_steps_per_second": 1.252,
      "step": 1000
    },
    {
      "epoch": 150.0,
      "learning_rate": 0.0006,
      "loss": 0.0008,
      "step": 1200
    },
    {
      "epoch": 150.0,
      "eval_exact_match": 0.03756708407871199,
      "eval_loss": 0.05796467885375023,
      "eval_runtime": 14.8269,
      "eval_samples_per_second": 37.702,
      "eval_steps_per_second": 1.214,
      "step": 1200
    },
    {
      "epoch": 175.0,
      "learning_rate": 0.0005333333333333334,
      "loss": 0.0008,
      "step": 1400
    },
    {
      "epoch": 175.0,
      "eval_exact_match": 0.03935599284436494,
      "eval_loss": 0.06270118057727814,
      "eval_runtime": 15.0576,
      "eval_samples_per_second": 37.124,
      "eval_steps_per_second": 1.195,
      "step": 1400
    },
    {
      "epoch": 200.0,
      "learning_rate": 0.00046666666666666666,
      "loss": 0.0005,
      "step": 1600
    },
    {
      "epoch": 200.0,
      "eval_exact_match": 0.03756708407871199,
      "eval_loss": 0.0657651349902153,
      "eval_runtime": 14.712,
      "eval_samples_per_second": 37.996,
      "eval_steps_per_second": 1.223,
      "step": 1600
    },
    {
      "epoch": 225.0,
      "learning_rate": 0.0004,
      "loss": 0.0005,
      "step": 1800
    },
    {
      "epoch": 225.0,
      "eval_exact_match": 0.03756708407871199,
      "eval_loss": 0.0657225102186203,
      "eval_runtime": 14.3583,
      "eval_samples_per_second": 38.932,
      "eval_steps_per_second": 1.254,
      "step": 1800
    },
    {
      "epoch": 250.0,
      "learning_rate": 0.0003333333333333333,
      "loss": 0.0004,
      "step": 2000
    },
    {
      "epoch": 250.0,
      "eval_exact_match": 0.03756708407871199,
      "eval_loss": 0.0687028169631958,
      "eval_runtime": 14.3636,
      "eval_samples_per_second": 38.918,
      "eval_steps_per_second": 1.253,
      "step": 2000
    },
    {
      "epoch": 275.0,
      "learning_rate": 0.0002666666666666667,
      "loss": 0.0003,
      "step": 2200
    },
    {
      "epoch": 275.0,
      "eval_exact_match": 0.03756708407871199,
      "eval_loss": 0.07311432808637619,
      "eval_runtime": 14.4221,
      "eval_samples_per_second": 38.76,
      "eval_steps_per_second": 1.248,
      "step": 2200
    },
    {
      "epoch": 300.0,
      "learning_rate": 0.0002,
      "loss": 0.0003,
      "step": 2400
    },
    {
      "epoch": 300.0,
      "eval_exact_match": 0.03756708407871199,
      "eval_loss": 0.07225054502487183,
      "eval_runtime": 14.3984,
      "eval_samples_per_second": 38.824,
      "eval_steps_per_second": 1.25,
      "step": 2400
    },
    {
      "epoch": 325.0,
      "learning_rate": 0.00013333333333333334,
      "loss": 0.0002,
      "step": 2600
    },
    {
      "epoch": 325.0,
      "eval_exact_match": 0.03756708407871199,
      "eval_loss": 0.0736461654305458,
      "eval_runtime": 14.3582,
      "eval_samples_per_second": 38.932,
      "eval_steps_per_second": 1.254,
      "step": 2600
    },
    {
      "epoch": 350.0,
      "learning_rate": 6.666666666666667e-05,
      "loss": 0.0002,
      "step": 2800
    },
    {
      "epoch": 350.0,
      "eval_exact_match": 0.03935599284436494,
      "eval_loss": 0.07408309727907181,
      "eval_runtime": 14.7389,
      "eval_samples_per_second": 37.927,
      "eval_steps_per_second": 1.221,
      "step": 2800
    },
    {
      "epoch": 375.0,
      "learning_rate": 0.0,
      "loss": 0.0002,
      "step": 3000
    },
    {
      "epoch": 375.0,
      "eval_exact_match": 0.03935599284436494,
      "eval_loss": 0.07445083558559418,
      "eval_runtime": 14.3465,
      "eval_samples_per_second": 38.964,
      "eval_steps_per_second": 1.255,
      "step": 3000
    },
    {
      "epoch": 375.0,
      "step": 3000,
      "total_flos": 1.869744384549504e+17,
      "train_loss": 0.01553155255690217,
      "train_runtime": 22590.8838,
      "train_samples_per_second": 67.992,
      "train_steps_per_second": 0.133
    }
  ],
  "max_steps": 3000,
  "num_train_epochs": 375,
  "total_flos": 1.869744384549504e+17,
  "trial_name": null,
  "trial_params": null
}