eval_results.json

{
    "epoch": 0.9998706171561651,
    "eval_kl": 0.01532898098230362,
    "eval_logps/chosen": -340.4056263858093,
    "eval_logps/rejected": -394.53469279661016,
    "eval_loss": 0.40464144945144653,
    "eval_rewards/chosen": -0.636441926469824,
    "eval_rewards/margins": 1.5609318031713308,
    "eval_rewards/rejected": -2.197373729641155,
    "eval_runtime": 260.8191,
    "eval_samples": 1846,
    "eval_samples_per_second": 7.078,
    "eval_steps_per_second": 3.539
}