eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.2298238277435303,
    "eval_logits/rejected": -2.717153549194336,
    "eval_logps/chosen": -136.61859130859375,
    "eval_logps/rejected": -193.9042205810547,
    "eval_loss": 0.6931706666946411,
    "eval_rewards/accuracies": 0.3333333432674408,
    "eval_rewards/chosen": -0.0010885795345529914,
    "eval_rewards/margins": -0.0014355214079841971,
    "eval_rewards/rejected": 0.0003469419898465276,
    "eval_runtime": 8.4518,
    "eval_samples": 80,
    "eval_samples_per_second": 9.465,
    "eval_steps_per_second": 0.355
}