{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 8.88888888888889,
  "eval_steps": 500,
  "global_step": 40,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.22,
      "learning_rate": 0.0008535533905932737,
      "loss": 2.3716,
      "step": 10
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.0005,
      "loss": 0.4248,
      "step": 20
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.00014644660940672628,
      "loss": 0.0627,
      "step": 30
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0,
      "loss": 0.0084,
      "step": 40
    },
    {
      "epoch": 8.89,
      "step": 40,
      "total_flos": 333423851839488.0,
      "train_loss": 0.7168619632720947,
      "train_runtime": 10.2793,
      "train_samples_per_second": 17.511,
      "train_steps_per_second": 3.891
    }
  ],
  "logging_steps": 10,
  "max_steps": 40,
  "num_train_epochs": 10,
  "save_steps": 1000,
  "total_flos": 333423851839488.0,
  "trial_name": null,
  "trial_params": null
}