trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1548,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.97,
      "learning_rate": 3.3850129198966415e-05,
      "loss": 1.1055,
      "step": 500
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.7700258397932818e-05,
      "loss": 0.7112,
      "step": 1000
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.550387596899225e-06,
      "loss": 0.529,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1548,
      "total_flos": 820457454108672.0,
      "train_loss": 0.7752933292733915,
      "train_runtime": 1402.0046,
      "train_samples_per_second": 282.363,
      "train_steps_per_second": 1.104
    }
  ],
  "max_steps": 1548,
  "num_train_epochs": 3,
  "total_flos": 820457454108672.0,
  "trial_name": null,
  "trial_params": null
}