{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 1635,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.53,
      "learning_rate": 1.3883792048929665e-05,
      "loss": 0.4996,
      "step": 500
    },
    {
      "epoch": 3.06,
      "learning_rate": 7.767584097859327e-06,
      "loss": 0.3259,
      "step": 1000
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.6513761467889911e-06,
      "loss": 0.2073,
      "step": 1500
    },
    {
      "epoch": 5.0,
      "step": 1635,
      "total_flos": 1719430746777600.0,
      "train_loss": 0.3296431981824589,
      "train_runtime": 167.8728,
      "train_samples_per_second": 311.426,
      "train_steps_per_second": 9.74
    }
  ],
  "logging_steps": 500,
  "max_steps": 1635,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 1719430746777600.0,
  "trial_name": null,
  "trial_params": null
}