eval_results.json

{
    "epoch": 9.955555555555556,
    "eval_logits/chosen": -3360374.568421053,
    "eval_logits/rejected": -5993503.695238095,
    "eval_logps/chosen": -18.104893092105264,
    "eval_logps/rejected": -99.21747581845239,
    "eval_loss": 0.20157214999198914,
    "eval_rewards/chosen": -0.21784908896998356,
    "eval_rewards/margins": 7.7506631645642425,
    "eval_rewards/rejected": -7.968512253534226,
    "eval_runtime": 38.9293,
    "eval_samples_per_second": 5.138,
    "eval_steps_per_second": 2.569,
    "kl": 0.0
}