trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 1290,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 3.88,
      "learning_rate": 3.062015503875969e-05,
      "loss": 1.2544,
      "step": 500
    },
    {
      "epoch": 7.75,
      "learning_rate": 1.1240310077519382e-05,
      "loss": 0.2636,
      "step": 1000
    },
    {
      "epoch": 10.0,
      "step": 1290,
      "total_flos": 8586620760883200.0,
      "train_loss": 0.6355197699495064,
      "train_runtime": 2040.869,
      "train_samples_per_second": 10.045,
      "train_steps_per_second": 0.632
    }
  ],
  "max_steps": 1290,
  "num_train_epochs": 10,
  "total_flos": 8586620760883200.0,
  "trial_name": null,
  "trial_params": null
}