all_results.json

{
    "epoch": 2.999297541394882,
    "eval_logits/chosen": 0.491447389125824,
    "eval_logits/rejected": 0.5435228943824768,
    "eval_logps/chosen": -7.457221031188965,
    "eval_logps/rejected": -8.611748695373535,
    "eval_loss": 0.5195529460906982,
    "eval_rewards/accuracies": 0.7284866571426392,
    "eval_rewards/chosen": -7.457221031188965,
    "eval_rewards/margins": 1.1545281410217285,
    "eval_rewards/rejected": -8.611748695373535,
    "eval_runtime": 39.8171,
    "eval_samples": 1345,
    "eval_samples_per_second": 33.779,
    "eval_steps_per_second": 8.464,
    "total_flos": 0.0,
    "train_loss": 0.5211759163684967,
    "train_runtime": 30222.1049,
    "train_samples": 59790,
    "train_samples_per_second": 5.935,
    "train_steps_per_second": 0.185
}