eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.97661554813385,
    "eval_logits/rejected": -1.9212419986724854,
    "eval_logps/chosen": -690.5653686523438,
    "eval_logps/rejected": -633.3815307617188,
    "eval_loss": 0.6524845957756042,
    "eval_rewards/accuracies": 0.6080127954483032,
    "eval_rewards/chosen": 0.33147066831588745,
    "eval_rewards/margins": 0.13620509207248688,
    "eval_rewards/rejected": 0.19526559114456177,
    "eval_runtime": 812.3144,
    "eval_samples": 12451,
    "eval_samples_per_second": 15.328,
    "eval_steps_per_second": 0.48
}