trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 4581,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 2.9993451211525868e-05,
      "loss": 1.8457,
      "step": 1
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.9999999999999998e-05,
      "loss": 0.4662,
      "step": 1527
    },
    {
      "epoch": 2.0,
      "learning_rate": 9.999999999999999e-06,
      "loss": 0.1822,
      "step": 3054
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0,
      "loss": 0.1067,
      "step": 4581
    },
    {
      "epoch": 3.0,
      "step": 4581,
      "total_flos": 1.7559130935918592e+18,
      "train_loss": 0.25197537243040957,
      "train_runtime": 29853.1163,
      "train_samples_per_second": 2.454,
      "train_steps_per_second": 0.153
    }
  ],
  "logging_steps": 500,
  "max_steps": 4581,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 1.7559130935918592e+18,
  "trial_name": null,
  "trial_params": null
}