all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": NaN,
    "eval_logits/rejected": 0.96450275182724,
    "eval_logps/chosen": -362.19049072265625,
    "eval_logps/rejected": -333.4285583496094,
    "eval_loss": NaN,
    "eval_nll_loss": NaN,
    "eval_rewards/accuracies": 0.6507936716079712,
    "eval_rewards/chosen": 0.540829598903656,
    "eval_rewards/margins": 0.5512520670890808,
    "eval_rewards/rejected": -0.010456387884914875,
    "eval_runtime": 8.4003,
    "eval_samples_per_second": 119.044,
    "eval_steps_per_second": 7.5
}