eval_results.json

{
    "epoch": 0.998660594695955,
    "eval_logits/chosen": -2.375,
    "eval_logits/rejected": -2.375,
    "eval_logps/chosen": -0.451171875,
    "eval_logps/rejected": -0.47265625,
    "eval_loss": 0.8194748759269714,
    "eval_rewards/accuracies": 0.5531914830207825,
    "eval_rewards/chosen": -1.125,
    "eval_rewards/margins": 0.052490234375,
    "eval_rewards/rejected": -1.1796875,
    "eval_runtime": 63.5524,
    "eval_samples": 2994,
    "eval_samples_per_second": 47.111,
    "eval_steps_per_second": 1.479
}