trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 860,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 5.81,
      "learning_rate": 1.3953488372093024e-05,
      "loss": 0.7424,
      "step": 500
    },
    {
      "epoch": 10.0,
      "step": 860,
      "total_flos": 7281145314017280.0,
      "train_loss": 0.458767032623291,
      "train_runtime": 808.9499,
      "train_samples_per_second": 12.658,
      "train_steps_per_second": 1.063
    }
  ],
  "max_steps": 860,
  "num_train_epochs": 10,
  "total_flos": 7281145314017280.0,
  "trial_name": null,
  "trial_params": null
}