eval_results.json

{
    "epoch": 0.9985553308292401,
    "eval_logits/chosen": -0.5159785151481628,
    "eval_logits/rejected": -0.4926067888736725,
    "eval_logps/chosen": -154.935546875,
    "eval_logps/rejected": -162.19007873535156,
    "eval_loss": 1.5933648347854614,
    "eval_nll_loss": 0.42284756898880005,
    "eval_rewards/accuracies": 0.626086950302124,
    "eval_rewards/chosen": -15.493555068969727,
    "eval_rewards/margins": 0.7254539132118225,
    "eval_rewards/rejected": -16.219009399414062,
    "eval_runtime": 73.3207,
    "eval_samples": 1826,
    "eval_samples_per_second": 24.904,
    "eval_steps_per_second": 1.568
}