eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.3150386810302734,
    "eval_logits/rejected": -1.198398232460022,
    "eval_logps/chosen": -632.8635864257812,
    "eval_logps/rejected": -740.5719604492188,
    "eval_loss": 0.4757336974143982,
    "eval_rewards/accuracies": 0.7540000081062317,
    "eval_rewards/chosen": -3.6824865341186523,
    "eval_rewards/margins": 1.2776294946670532,
    "eval_rewards/rejected": -4.960116863250732,
    "eval_runtime": 1729.9711,
    "eval_samples": 2000,
    "eval_samples_per_second": 1.156,
    "eval_steps_per_second": 0.145
}