trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 1284,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.17,
      "learning_rate": 3.0529595015576326e-05,
      "loss": 0.6312,
      "step": 500
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.1059190031152649e-05,
      "loss": 0.5597,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1284,
      "total_flos": 2.1800265159020544e+17,
      "train_loss": 0.5830277178517754,
      "train_runtime": 14271.7239,
      "train_samples_per_second": 0.719,
      "train_steps_per_second": 0.09
    }
  ],
  "logging_steps": 500,
  "max_steps": 1284,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 2.1800265159020544e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}