eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.023923397064209,
    "eval_logits/rejected": -3.017603874206543,
    "eval_logps/chosen": -166.78807067871094,
    "eval_logps/rejected": -187.0471954345703,
    "eval_loss": 0.6481729745864868,
    "eval_rewards/accuracies": 0.6171003580093384,
    "eval_rewards/chosen": -0.9537805914878845,
    "eval_rewards/margins": 0.1656205952167511,
    "eval_rewards/rejected": -1.1194013357162476,
    "eval_runtime": 483.9291,
    "eval_samples": 4304,
    "eval_samples_per_second": 8.894,
    "eval_steps_per_second": 1.112
}