eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.0223519802093506,
    "eval_logits/rejected": -3.0161619186401367,
    "eval_logps/chosen": -167.01016235351562,
    "eval_logps/rejected": -187.9011993408203,
    "eval_loss": 0.6463221311569214,
    "eval_rewards/accuracies": 0.6203531622886658,
    "eval_rewards/chosen": -0.9560015797615051,
    "eval_rewards/margins": 0.17193979024887085,
    "eval_rewards/rejected": -1.127941370010376,
    "eval_runtime": 483.9068,
    "eval_samples": 4304,
    "eval_samples_per_second": 8.894,
    "eval_steps_per_second": 1.112
}