trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "eval_steps": 500,
  "global_step": 2120,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 4.72,
      "learning_rate": 5.4439794766146746e-05,
      "loss": 0.8136,
      "step": 1000
    },
    {
      "epoch": 9.43,
      "learning_rate": 7.884711026201585e-07,
      "loss": 0.2175,
      "step": 2000
    }
  ],
  "logging_steps": 1000,
  "max_steps": 2120,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 10000,
  "total_flos": 7918421362409472.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}