trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1317,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.14,
      "learning_rate": 3.455696202531646e-05,
      "loss": 0.2088,
      "step": 500
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.3459915611814346e-05,
      "loss": 0.0215,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1317,
      "total_flos": 1240045901331264.0,
      "train_loss": 0.09003285648429946,
      "train_runtime": 223.9348,
      "train_samples_per_second": 188.117,
      "train_steps_per_second": 5.881
    }
  ],
  "max_steps": 1317,
  "num_train_epochs": 3,
  "total_flos": 1240045901331264.0,
  "trial_name": null,
  "trial_params": null
}