{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 19.813084112149532,
  "eval_steps": 500,
  "global_step": 1060,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 9.35,
      "learning_rate": 2.641509433962264e-05,
      "loss": 1.7773,
      "step": 500
    },
    {
      "epoch": 18.69,
      "learning_rate": 2.830188679245283e-06,
      "loss": 1.082,
      "step": 1000
    },
    {
      "epoch": 19.81,
      "step": 1060,
      "total_flos": 2.9595008832425165e+17,
      "train_loss": 1.403595557302799,
      "train_runtime": 1785.2215,
      "train_samples_per_second": 9.523,
      "train_steps_per_second": 0.594
    }
  ],
  "logging_steps": 500,
  "max_steps": 1060,
  "num_train_epochs": 20,
  "save_steps": 200,
  "total_flos": 2.9595008832425165e+17,
  "trial_name": null,
  "trial_params": null
}