trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 1250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.0,
      "learning_rate": 1.8e-05,
      "loss": 0.1466,
      "step": 500
    },
    {
      "epoch": 4.0,
      "learning_rate": 6e-06,
      "loss": 0.0378,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "step": 1250,
      "total_flos": 4010669170133124.0,
      "train_loss": 0.07716391067504882,
      "train_runtime": 348.8673,
      "train_samples_per_second": 57.3,
      "train_steps_per_second": 3.583
    }
  ],
  "max_steps": 1250,
  "num_train_epochs": 5,
  "total_flos": 4010669170133124.0,
  "trial_name": null,
  "trial_params": null
}