trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2388,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.63,
      "learning_rate": 3.953098827470687e-05,
      "loss": 2.3454,
      "step": 500
    },
    {
      "epoch": 1.26,
      "learning_rate": 2.906197654941374e-05,
      "loss": 1.6412,
      "step": 1000
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.8592964824120602e-05,
      "loss": 1.4022,
      "step": 1500
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.123953098827471e-06,
      "loss": 1.2104,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2388,
      "total_flos": 3265142518775808.0,
      "train_loss": 1.5687911194972257,
      "train_runtime": 812.5919,
      "train_samples_per_second": 29.38,
      "train_steps_per_second": 2.939
    }
  ],
  "max_steps": 2388,
  "num_train_epochs": 3,
  "total_flos": 3265142518775808.0,
  "trial_name": null,
  "trial_params": null
}