eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.7963508367538452,
    "eval_logits/rejected": -1.5644817352294922,
    "eval_logps/chosen": -150.905029296875,
    "eval_logps/rejected": -171.62582397460938,
    "eval_loss": 0.6391857266426086,
    "eval_rewards/accuracies": 0.7526785731315613,
    "eval_rewards/chosen": 0.12320567667484283,
    "eval_rewards/margins": 0.1261579990386963,
    "eval_rewards/rejected": -0.0029523156117647886,
    "eval_runtime": 429.2674,
    "eval_samples": 4472,
    "eval_samples_per_second": 10.418,
    "eval_steps_per_second": 0.326
}