trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2346,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.64,
      "learning_rate": 3.94075021312873e-05,
      "loss": 0.1202,
      "step": 500
    },
    {
      "epoch": 1.28,
      "learning_rate": 2.8751065643648767e-05,
      "loss": 0.0785,
      "step": 1000
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.8094629156010232e-05,
      "loss": 0.0481,
      "step": 1500
    },
    {
      "epoch": 2.56,
      "learning_rate": 7.438192668371696e-06,
      "loss": 0.0162,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2346,
      "total_flos": 1.4799996864e+16,
      "train_loss": 0.057570336623163375,
      "train_runtime": 650.909,
      "train_samples_per_second": 115.223,
      "train_steps_per_second": 3.604
    }
  ],
  "max_steps": 2346,
  "num_train_epochs": 3,
  "total_flos": 1.4799996864e+16,
  "trial_name": null,
  "trial_params": null
}