trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 1370,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.82,
      "learning_rate": 1.905109489051095e-05,
      "loss": 0.0577,
      "step": 500
    },
    {
      "epoch": 3.65,
      "learning_rate": 8.102189781021898e-06,
      "loss": 0.0156,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "step": 1370,
      "total_flos": 2718128676681120.0,
      "train_loss": 0.028748657929636268,
      "train_runtime": 270.7551,
      "train_samples_per_second": 80.959,
      "train_steps_per_second": 5.06
    }
  ],
  "max_steps": 1370,
  "num_train_epochs": 5,
  "total_flos": 2718128676681120.0,
  "trial_name": null,
  "trial_params": null
}