trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 30.0,
  "global_step": 12510,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 11.99,
      "learning_rate": 4.060393227846204e-05,
      "loss": 0.7845,
      "step": 5000
    },
    {
      "epoch": 23.98,
      "learning_rate": 6.712477093047967e-06,
      "loss": 0.0208,
      "step": 10000
    },
    {
      "epoch": 30.0,
      "step": 12510,
      "total_flos": 1.1203460780602368e+17,
      "train_loss": 0.32232193622848304,
      "train_runtime": 17163.0194,
      "train_samples_per_second": 8.74,
      "train_steps_per_second": 0.729
    }
  ],
  "max_steps": 12510,
  "num_train_epochs": 30,
  "total_flos": 1.1203460780602368e+17,
  "trial_name": null,
  "trial_params": null
}