eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.35237225890159607,
    "eval_logits/rejected": -0.32851526141166687,
    "eval_logps/chosen": -679.8935546875,
    "eval_logps/rejected": -1442.5582275390625,
    "eval_loss": 0.27591243386268616,
    "eval_rewards/accuracies": 0.8920454382896423,
    "eval_rewards/chosen": -1.230484962463379,
    "eval_rewards/margins": 5.869514465332031,
    "eval_rewards/rejected": -7.099999904632568,
    "eval_runtime": 171.0958,
    "eval_samples": 1392,
    "eval_samples_per_second": 8.136,
    "eval_steps_per_second": 0.257
}