eval_results.json

{
    "epoch": 0.9230769230769231,
    "eval_logits/chosen": -2.3219950199127197,
    "eval_logits/rejected": -2.219332218170166,
    "eval_logps/chosen": -277.3203430175781,
    "eval_logps/rejected": -226.97976684570312,
    "eval_loss": 0.6930660009384155,
    "eval_rewards/accuracies": 0.4642857015132904,
    "eval_rewards/chosen": -0.0008508563041687012,
    "eval_rewards/margins": -0.00032856108737178147,
    "eval_rewards/rejected": -0.0005222952459007502,
    "eval_runtime": 163.6884,
    "eval_samples": 1000,
    "eval_samples_per_second": 6.109,
    "eval_steps_per_second": 0.385
}