trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 2766,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.54,
      "learning_rate": 4.096167751265365e-05,
      "loss": 0.0465,
      "step": 500
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.192335502530731e-05,
      "loss": 0.0003,
      "step": 1000
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.2885032537960955e-05,
      "loss": 0.0054,
      "step": 1500
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.3846710050614606e-05,
      "loss": 0.0012,
      "step": 2000
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.808387563268257e-06,
      "loss": 0.0002,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2766,
      "total_flos": 5781661149312000.0,
      "train_loss": 0.009698579121523807,
      "train_runtime": 643.1965,
      "train_samples_per_second": 34.399,
      "train_steps_per_second": 4.3
    }
  ],
  "logging_steps": 500,
  "max_steps": 2766,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 5781661149312000.0,
  "trial_name": null,
  "trial_params": null
}