eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.0213913917541504,
    "eval_logits/rejected": -3.038022518157959,
    "eval_logps/chosen": -296.33209228515625,
    "eval_logps/rejected": -291.47021484375,
    "eval_loss": 0.4926324486732483,
    "eval_rewards/accuracies": 0.7570000290870667,
    "eval_rewards/chosen": -1.3098689317703247,
    "eval_rewards/margins": 1.4436919689178467,
    "eval_rewards/rejected": -2.753561019897461,
    "eval_runtime": 1173.4935,
    "eval_samples": 2000,
    "eval_samples_per_second": 1.704,
    "eval_steps_per_second": 0.213
}