eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.8436477184295654,
    "eval_logits/rejected": -0.7463203072547913,
    "eval_logps/chosen": -395.2843933105469,
    "eval_logps/rejected": -366.02801513671875,
    "eval_loss": 0.05383266508579254,
    "eval_rewards/accuracies": 0.42899999022483826,
    "eval_rewards/chosen": 0.001020837458781898,
    "eval_rewards/margins": -0.000186597288120538,
    "eval_rewards/rejected": 0.0012074349215254188,
    "eval_runtime": 546.0387,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.663,
    "eval_steps_per_second": 0.916
}