eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.030522108078003,
    "eval_logits/rejected": -1.1812418699264526,
    "eval_logps/chosen": -795.1126098632812,
    "eval_logps/rejected": -652.2422485351562,
    "eval_loss": 0.6183284521102905,
    "eval_rewards/accuracies": 0.6783919334411621,
    "eval_rewards/chosen": 0.5534913539886475,
    "eval_rewards/margins": 0.214975506067276,
    "eval_rewards/rejected": 0.33851587772369385,
    "eval_runtime": 181.6928,
    "eval_samples": 398,
    "eval_samples_per_second": 2.191,
    "eval_steps_per_second": 2.191
}