eval_results.json

{
    "epoch": 0.9981298423724285,
    "eval_logits/chosen": -7.053029537200928,
    "eval_logits/rejected": -7.020767688751221,
    "eval_logps/chosen": -18186.599609375,
    "eval_logps/rejected": -17593.8046875,
    "eval_loss": 0.011397347785532475,
    "eval_rewards/accuracies": 0.4395161271095276,
    "eval_rewards/chosen": -180.98304748535156,
    "eval_rewards/margins": -5.912590503692627,
    "eval_rewards/rejected": -175.0704803466797,
    "eval_runtime": 174.1418,
    "eval_samples_per_second": 11.261,
    "eval_steps_per_second": 0.178
}