eval_results.json

{
    "epoch": 0.9981298423724285,
    "eval_logits/chosen": 0.1334349662065506,
    "eval_logits/rejected": 0.1487051546573639,
    "eval_logps/chosen": -137.7278594970703,
    "eval_logps/rejected": -142.43006896972656,
    "eval_loss": 0.6647459268569946,
    "eval_rewards/accuracies": 0.5967742204666138,
    "eval_rewards/chosen": -0.4943546652793884,
    "eval_rewards/margins": 0.06238555163145065,
    "eval_rewards/rejected": -0.5567402243614197,
    "eval_runtime": 174.6477,
    "eval_samples_per_second": 11.228,
    "eval_steps_per_second": 0.178
}