{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2838,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.53,
      "learning_rate": 4.119097956307259e-05,
      "loss": 0.978,
      "step": 500
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.238195912614517e-05,
      "loss": 0.8605,
      "step": 1000
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.357293868921776e-05,
      "loss": 0.795,
      "step": 1500
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.4763918252290346e-05,
      "loss": 0.7404,
      "step": 2000
    },
    {
      "epoch": 2.64,
      "learning_rate": 5.954897815362931e-06,
      "loss": 0.7087,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2838,
      "total_flos": 1.0577527714510848e+16,
      "train_loss": 0.8031494646025342,
      "train_runtime": 36606.133,
      "train_samples_per_second": 0.31,
      "train_steps_per_second": 0.078
    }
  ],
  "max_steps": 2838,
  "num_train_epochs": 3,
  "total_flos": 1.0577527714510848e+16,
  "trial_name": null,
  "trial_params": null
}