{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 39.075,
  "eval_steps": 500,
  "global_step": 840,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 10.05,
      "learning_rate": 3.982142857142857e-05,
      "loss": 0.8762,
      "step": 223
    },
    {
      "epoch": 21.02,
      "learning_rate": 7.964285714285715e-05,
      "loss": 0.3777,
      "step": 446
    },
    {
      "epoch": 31.06,
      "learning_rate": 0.00011946428571428572,
      "loss": 0.2965,
      "step": 669
    }
  ],
  "logging_steps": 223,
  "max_steps": 11200,
  "num_train_epochs": 40,
  "save_steps": 500,
  "total_flos": 5.46297808613802e+17,
  "trial_name": null,
  "trial_params": null
}