{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.6627906976744184,
  "eval_steps": 500,
  "global_step": 228,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2,
      "learning_rate": 0.00017142857142857143,
      "loss": 3.8841,
      "step": 34
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001818770226537217,
      "loss": 3.5074,
      "step": 68
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.00015987055016181232,
      "loss": 3.4318,
      "step": 102
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00013786407766990292,
      "loss": 3.3266,
      "step": 136
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00011585760517799353,
      "loss": 3.3409,
      "step": 170
    },
    {
      "epoch": 1.52,
      "learning_rate": 9.385113268608415e-05,
      "loss": 3.2522,
      "step": 204
    }
  ],
  "logging_steps": 34,
  "max_steps": 344,
  "num_train_epochs": 2,
  "save_steps": 500,
  "total_flos": 5.938502977978368e+16,
  "trial_name": null,
  "trial_params": null
}