trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 2240,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.23,
      "learning_rate": 3.883928571428572e-05,
      "loss": 3.5379,
      "step": 500
    },
    {
      "epoch": 4.46,
      "learning_rate": 2.767857142857143e-05,
      "loss": 2.7518,
      "step": 1000
    },
    {
      "epoch": 6.7,
      "learning_rate": 1.6517857142857144e-05,
      "loss": 2.5251,
      "step": 1500
    },
    {
      "epoch": 8.93,
      "learning_rate": 5.357142857142857e-06,
      "loss": 2.4254,
      "step": 2000
    },
    {
      "epoch": 10.0,
      "step": 2240,
      "total_flos": 6131004327843840.0,
      "train_loss": 2.763779067993164,
      "train_runtime": 597.6126,
      "train_samples_per_second": 37.349,
      "train_steps_per_second": 3.748
    }
  ],
  "max_steps": 2240,
  "num_train_epochs": 10,
  "total_flos": 6131004327843840.0,
  "trial_name": null,
  "trial_params": null
}