trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1146,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.31,
      "learning_rate": 2.8184991273996508e-05,
      "loss": 2.0847,
      "step": 500
    },
    {
      "epoch": 2.62,
      "learning_rate": 6.369982547993019e-06,
      "loss": 1.5616,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1146,
      "total_flos": 4264004695910400.0,
      "train_loss": 1.7737141554268243,
      "train_runtime": 927.9599,
      "train_samples_per_second": 9.86,
      "train_steps_per_second": 1.235
    }
  ],
  "max_steps": 1146,
  "num_train_epochs": 3,
  "total_flos": 4264004695910400.0,
  "trial_name": null,
  "trial_params": null
}