eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.3248252868652344,
    "eval_logits/rejected": -2.1362955570220947,
    "eval_logps/chosen": -230.59922790527344,
    "eval_logps/rejected": -218.62928771972656,
    "eval_loss": 0.6897218227386475,
    "eval_rewards/accuracies": 0.6370000243186951,
    "eval_rewards/chosen": 0.014057200402021408,
    "eval_rewards/margins": 0.08423101156949997,
    "eval_rewards/rejected": -0.07017382234334946,
    "eval_runtime": 710.6006,
    "eval_samples": 2000,
    "eval_samples_per_second": 2.815,
    "eval_steps_per_second": 1.407
}