{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.25,
  "eval_steps": 500,
  "global_step": 4,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.12,
      "learning_rate": 0.0001,
      "loss": 0.7996,
      "step": 1
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 0.784,
      "step": 2
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00018571428571428572,
      "loss": 0.7719,
      "step": 3
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00017142857142857143,
      "loss": 0.7155,
      "step": 4
    }
  ],
  "logging_steps": 1,
  "max_steps": 16,
  "num_train_epochs": 2,
  "save_steps": 500,
  "total_flos": 325176671600640.0,
  "trial_name": null,
  "trial_params": null
}