trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.995112414467253,
  "global_step": 1149,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.3,
      "learning_rate": 5.825852782764811e-06,
      "loss": 0.4089,
      "step": 500
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.3375224416517058e-06,
      "loss": 0.2804,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1149,
      "total_flos": 2.326518414840955e+17,
      "train_loss": 0.331210768046433,
      "train_runtime": 2723.726,
      "train_samples_per_second": 27.041,
      "train_steps_per_second": 0.422
    }
  ],
  "max_steps": 1149,
  "num_train_epochs": 3,
  "total_flos": 2.326518414840955e+17,
  "trial_name": null,
  "trial_params": null
}