trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1062,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.41,
      "learning_rate": 2.772218773524804e-05,
      "loss": 1.6278,
      "step": 500
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.280991068959739e-07,
      "loss": 0.9544,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1062,
      "total_flos": 2311098177945600.0,
      "train_loss": 1.2636232214458918,
      "train_runtime": 272.594,
      "train_samples_per_second": 3.896,
      "train_steps_per_second": 3.896
    }
  ],
  "max_steps": 1062,
  "num_train_epochs": 3,
  "total_flos": 2311098177945600.0,
  "trial_name": null,
  "trial_params": null
}