trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 1566,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.96,
      "learning_rate": 3.403575989782887e-05,
      "loss": 0.3118,
      "step": 500
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.8071519795657726e-05,
      "loss": 0.2617,
      "step": 1000
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.1072796934865904e-06,
      "loss": 0.2432,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1566,
      "total_flos": 73168104062976.0,
      "train_loss": 0.27109126508768794,
      "train_runtime": 953.4562,
      "train_samples_per_second": 6.57,
      "train_steps_per_second": 1.642
    }
  ],
  "logging_steps": 500,
  "max_steps": 1566,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 73168104062976.0,
  "trial_name": null,
  "trial_params": null
}