trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 2015,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.2406947890818858,
      "grad_norm": 5.6448655128479,
      "learning_rate": 1.5037220843672458e-05,
      "loss": 0.4689,
      "step": 500
    },
    {
      "epoch": 2.4813895781637716,
      "grad_norm": 4.1219892501831055,
      "learning_rate": 1.0074441687344914e-05,
      "loss": 0.3429,
      "step": 1000
    },
    {
      "epoch": 3.7220843672456576,
      "grad_norm": 11.060373306274414,
      "learning_rate": 5.11166253101737e-06,
      "loss": 0.2524,
      "step": 1500
    },
    {
      "epoch": 4.962779156327543,
      "grad_norm": 13.940281867980957,
      "learning_rate": 1.488833746898263e-07,
      "loss": 0.196,
      "step": 2000
    }
  ],
  "logging_steps": 500,
  "max_steps": 2015,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 6354846821675520.0,
  "train_batch_size": 48,
  "trial_name": null,
  "trial_params": null
}