eval_results.json

{
    "epoch": 0.9981298423724285,
    "eval_logits/chosen": -1.771275281906128,
    "eval_logits/rejected": -1.7774497270584106,
    "eval_logps/chosen": -1.5688886642456055,
    "eval_logps/rejected": -2.27230167388916,
    "eval_loss": 1.0684642791748047,
    "eval_rewards/accuracies": 0.8911290168762207,
    "eval_rewards/chosen": -15.688885688781738,
    "eval_rewards/margins": 7.0341291427612305,
    "eval_rewards/rejected": -22.72301483154297,
    "eval_runtime": 84.8089,
    "eval_samples": 1961,
    "eval_samples_per_second": 23.123,
    "eval_steps_per_second": 0.731
}