eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.553934335708618,
    "eval_logits/rejected": -2.510233163833618,
    "eval_logps/chosen": -362.36248779296875,
    "eval_logps/rejected": -395.8988952636719,
    "eval_loss": 0.5367478132247925,
    "eval_rewards/accuracies": 0.7103174328804016,
    "eval_rewards/chosen": -0.7912437319755554,
    "eval_rewards/margins": 0.6874489188194275,
    "eval_rewards/rejected": -1.4786925315856934,
    "eval_runtime": 165.9175,
    "eval_samples": 500,
    "eval_samples_per_second": 3.014,
    "eval_steps_per_second": 0.38
}