trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.0026171159382360636,
  "eval_steps": 20,
  "global_step": 20,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "grad_norm": 0.40162393450737,
      "learning_rate": 0.0,
      "log_odds_chosen": -0.10784949362277985,
      "log_odds_ratio": -0.819013774394989,
      "logits/chosen": -114.29496002197266,
      "logits/rejected": -113.46098327636719,
      "logps/chosen": -1.9617910385131836,
      "logps/rejected": -1.8746588230133057,
      "loss": 5.1729,
      "nll_loss": 5.090985298156738,
      "rewards/accuracies": 0.4312500059604645,
      "rewards/chosen": -0.1961791068315506,
      "rewards/margins": -0.008713215589523315,
      "rewards/rejected": -0.1874658763408661,
      "step": 20
    },
    {
      "epoch": 0.0,
      "eval_log_odds_chosen": -0.010702775791287422,
      "eval_log_odds_ratio": -0.7446900606155396,
      "eval_logits/chosen": -111.06063079833984,
      "eval_logits/rejected": -111.48336791992188,
      "eval_logps/chosen": -2.004408597946167,
      "eval_logps/rejected": -1.9918007850646973,
      "eval_loss": 5.372676849365234,
      "eval_nll_loss": 5.298208236694336,
      "eval_rewards/accuracies": 0.46799999475479126,
      "eval_rewards/chosen": -0.20044086873531342,
      "eval_rewards/margins": -0.0012607867829501629,
      "eval_rewards/rejected": -0.19918008148670197,
      "eval_runtime": 372.018,
      "eval_samples_per_second": 5.376,
      "eval_steps_per_second": 2.688,
      "step": 20
    }
  ],
  "logging_steps": 20,
  "max_steps": 20,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 10,
  "total_flos": 0.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}