eval_results.json

{
    "epoch": 0.8571428571428571,
    "eval_logits/chosen": -2.6940619945526123,
    "eval_logits/rejected": -2.7640316486358643,
    "eval_logps/chosen": -248.7039794921875,
    "eval_logps/rejected": -237.63433837890625,
    "eval_loss": 0.6914129853248596,
    "eval_rewards/accuracies": 0.4375,
    "eval_rewards/chosen": 0.012479877099394798,
    "eval_rewards/margins": -0.0028497937601059675,
    "eval_rewards/rejected": 0.01532967109233141,
    "eval_runtime": 8.293,
    "eval_samples_per_second": 11.697,
    "eval_steps_per_second": 0.241
}