eval_results.json

{
    "epoch": 0.9997382884061764,
    "eval_logits/chosen": -1.4810994863510132,
    "eval_logits/rejected": -1.413936734199524,
    "eval_logps/chosen": -483.0326232910156,
    "eval_logps/rejected": -564.1859130859375,
    "eval_loss": 0.4935336709022522,
    "eval_rewards/accuracies": 0.7698412537574768,
    "eval_rewards/chosen": -2.1675848960876465,
    "eval_rewards/margins": 1.0059263706207275,
    "eval_rewards/rejected": -3.173511266708374,
    "eval_runtime": 276.8986,
    "eval_samples": 2000,
    "eval_samples_per_second": 7.223,
    "eval_steps_per_second": 0.228
}