eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.243324041366577,
    "eval_logits/rejected": -2.2384440898895264,
    "eval_logps/chosen": -34.02931213378906,
    "eval_logps/rejected": -37.51406478881836,
    "eval_loss": 0.9987560510635376,
    "eval_rewards/accuracies": 0.4925249218940735,
    "eval_rewards/chosen": 0.0026186639443039894,
    "eval_rewards/margins": 0.0013388387160375714,
    "eval_rewards/rejected": 0.0012798249954357743,
    "eval_runtime": 145.7879,
    "eval_samples": 343,
    "eval_samples_per_second": 2.353,
    "eval_steps_per_second": 0.295
}