eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 98.34632873535156,
    "eval_logits/rejected": 98.3492202758789,
    "eval_logps/chosen": -32.23118591308594,
    "eval_logps/rejected": -35.818275451660156,
    "eval_loss": 0.4955456852912903,
    "eval_rewards/accuracies": 0.5282392501831055,
    "eval_rewards/chosen": 0.08479735255241394,
    "eval_rewards/margins": 0.025497542694211006,
    "eval_rewards/rejected": 0.059299811720848083,
    "eval_runtime": 103.8037,
    "eval_samples": 343,
    "eval_samples_per_second": 3.304,
    "eval_steps_per_second": 0.414
}