eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -4.396326541900635,
    "eval_logits/rejected": -4.383143901824951,
    "eval_logps/chosen": -375.64892578125,
    "eval_logps/rejected": -396.0606689453125,
    "eval_loss": 0.6749153733253479,
    "eval_rewards/accuracies": 0.5740000009536743,
    "eval_rewards/chosen": -1.5494239330291748,
    "eval_rewards/margins": 0.14782528579235077,
    "eval_rewards/rejected": -1.697249174118042,
    "eval_runtime": 1692.8429,
    "eval_samples": 2994,
    "eval_samples_per_second": 1.769,
    "eval_steps_per_second": 0.222
}