trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 1585,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.58,
      "learning_rate": 2.053627760252366e-05,
      "loss": 1.5656,
      "step": 500
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.1072555205047318e-05,
      "loss": 0.6697,
      "step": 1000
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.608832807570978e-06,
      "loss": 0.3875,
      "step": 1500
    },
    {
      "epoch": 5.0,
      "step": 1585,
      "total_flos": 4838583106668000.0,
      "train_loss": 0.8455495816299968,
      "train_runtime": 715.3867,
      "train_samples_per_second": 66.265,
      "train_steps_per_second": 2.216
    }
  ],
  "max_steps": 1585,
  "num_train_epochs": 5,
  "total_flos": 4838583106668000.0,
  "trial_name": null,
  "trial_params": null
}