all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.22265625,
    "eval_logits/rejected": -3.2421875,
    "eval_logps/chosen": -297.0,
    "eval_logps/rejected": -294.0,
    "eval_loss": 0.6393749713897705,
    "eval_rewards/accuracies": 0.671875,
    "eval_rewards/chosen": -0.0892333984375,
    "eval_rewards/margins": 0.1739501953125,
    "eval_rewards/rejected": -0.263427734375,
    "eval_runtime": 2.6575,
    "eval_samples_per_second": 37.629,
    "eval_steps_per_second": 1.505
}