eval_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 91.33722686767578,
    "eval_logits/rejected": 94.05353546142578,
    "eval_logps/chosen": -428.1109619140625,
    "eval_logps/rejected": -515.8482666015625,
    "eval_loss": 0.4566638171672821,
    "eval_rewards/accuracies": 0.7659574747085571,
    "eval_rewards/chosen": -3.5880744457244873,
    "eval_rewards/margins": 1.7552103996276855,
    "eval_rewards/rejected": -5.343284606933594,
    "eval_runtime": 90.1839,
    "eval_samples": 750,
    "eval_samples_per_second": 8.316,
    "eval_steps_per_second": 0.521
}