eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.9072314500808716,
    "eval_logits/rejected": -1.686521053314209,
    "eval_logps/chosen": -320.5736083984375,
    "eval_logps/rejected": -328.9005126953125,
    "eval_loss": 0.604966402053833,
    "eval_rewards/accuracies": 0.6899999976158142,
    "eval_rewards/chosen": -0.04819466918706894,
    "eval_rewards/margins": 0.26583459973335266,
    "eval_rewards/rejected": -0.3140292465686798,
    "eval_runtime": 384.0617,
    "eval_samples": 2000,
    "eval_samples_per_second": 5.207,
    "eval_steps_per_second": 0.651
}