eval_results.json

{
    "epoch": 2.955223880597015,
    "eval_logits/chosen": -2.3867504596710205,
    "eval_logits/rejected": -2.3840601444244385,
    "eval_logps/chosen": -234.9886016845703,
    "eval_logps/rejected": -205.3957977294922,
    "eval_loss": 0.6642679572105408,
    "eval_rewards/accuracies": 0.7749999761581421,
    "eval_rewards/chosen": 1.2068389654159546,
    "eval_rewards/margins": 1.5741199254989624,
    "eval_rewards/rejected": -0.3672809302806854,
    "eval_runtime": 63.475,
    "eval_samples_per_second": 14.967,
    "eval_steps_per_second": 0.236
}