trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2181,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.69,
      "learning_rate": 3.853736817973407e-05,
      "loss": 3.1317,
      "step": 500
    },
    {
      "epoch": 1.38,
      "learning_rate": 2.7074736359468134e-05,
      "loss": 2.8754,
      "step": 1000
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.56121045392022e-05,
      "loss": 2.7749,
      "step": 1500
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.149472718936268e-06,
      "loss": 2.6897,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2181,
      "total_flos": 3332709108154368.0,
      "train_runtime": 760.1324,
      "train_samples_per_second": 2.869
    }
  ],
  "max_steps": 2181,
  "num_train_epochs": 3,
  "total_flos": 3332709108154368.0,
  "trial_name": null,
  "trial_params": null
}