trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "global_step": 40,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 0.00015000000000000001,
      "loss": 0.4049,
      "step": 10
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0001,
      "loss": 0.1204,
      "step": 20
    },
    {
      "epoch": 3.0,
      "learning_rate": 5e-05,
      "loss": 0.0492,
      "step": 30
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0,
      "loss": 0.0236,
      "step": 40
    },
    {
      "epoch": 4.0,
      "step": 40,
      "total_flos": 5.05524198703104e+16,
      "train_loss": 0.14953994750976562,
      "train_runtime": 18.1531,
      "train_samples_per_second": 35.256,
      "train_steps_per_second": 2.203
    }
  ],
  "max_steps": 40,
  "num_train_epochs": 4,
  "total_flos": 5.05524198703104e+16,
  "trial_name": null,
  "trial_params": null
}