eval_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": -18.044893264770508,
    "eval_logits/rejected": -25.161956787109375,
    "eval_logps/chosen": -1530.05712890625,
    "eval_logps/rejected": -1648.0323486328125,
    "eval_loss": 0.8036046624183655,
    "eval_rewards/accuracies": 0.7291666865348816,
    "eval_rewards/chosen": -0.44631412625312805,
    "eval_rewards/margins": 0.8397437930107117,
    "eval_rewards/rejected": -1.2860580682754517,
    "eval_runtime": 37.4541,
    "eval_samples": 750,
    "eval_samples_per_second": 20.025,
    "eval_steps_per_second": 0.641
}