trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.9985984583041345,
  "eval_steps": 500,
  "global_step": 1426,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.7,
      "learning_rate": 2.082089552238806e-05,
      "loss": 0.481,
      "step": 500
    },
    {
      "epoch": 1.4,
      "learning_rate": 9.62686567164179e-06,
      "loss": 0.1133,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "step": 1426,
      "total_flos": 6003148215168000.0,
      "train_loss": 0.23042542536496113,
      "train_runtime": 560.0994,
      "train_samples_per_second": 40.761,
      "train_steps_per_second": 2.546
    }
  ],
  "logging_steps": 500,
  "max_steps": 1426,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 500,
  "total_flos": 6003148215168000.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}