trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 30.0,
  "global_step": 1470,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 10.2,
      "learning_rate": 3.2993197278911564e-05,
      "loss": 1.9694,
      "step": 500
    },
    {
      "epoch": 20.41,
      "learning_rate": 1.5986394557823133e-05,
      "loss": 0.4335,
      "step": 1000
    },
    {
      "epoch": 30.0,
      "step": 1470,
      "total_flos": 3035800535040000.0,
      "train_runtime": 924.1697,
      "train_samples_per_second": 1.591
    }
  ],
  "max_steps": 1470,
  "num_train_epochs": 30,
  "total_flos": 3035800535040000.0,
  "trial_name": null,
  "trial_params": null
}