eval_results.json

{
    "epoch": 0.9874476987447699,
    "eval_logits/chosen": -2.7839648723602295,
    "eval_logits/rejected": -2.763737201690674,
    "eval_logps/chosen": -587.9949951171875,
    "eval_logps/rejected": -758.1629638671875,
    "eval_loss": 0.3443484902381897,
    "eval_rewards/accuracies": 0.8700980544090271,
    "eval_rewards/chosen": -1.6458674669265747,
    "eval_rewards/margins": 1.476360559463501,
    "eval_rewards/rejected": -3.122227668762207,
    "eval_runtime": 296.2496,
    "eval_samples": 6491,
    "eval_samples_per_second": 21.911,
    "eval_steps_per_second": 0.344
}