all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": 3.998190402984619,
    "eval_logits/rejected": 4.15942907333374,
    "eval_logps/chosen": -399.73858642578125,
    "eval_logps/rejected": -601.5961303710938,
    "eval_loss": 1.72599458694458,
    "eval_rewards/accuracies": 0.3075396716594696,
    "eval_rewards/chosen": -5.072336673736572,
    "eval_rewards/margins": 3.0445241928100586,
    "eval_rewards/rejected": -8.116861343383789,
    "eval_runtime": 114.7048,
    "eval_samples": 2000,
    "eval_samples_per_second": 17.436,
    "eval_steps_per_second": 0.549,
    "total_flos": 0.0,
    "train_loss": 0.0,
    "train_runtime": 0.1474,
    "train_samples": 6113,
    "train_samples_per_second": 82937.448,
    "train_steps_per_second": 1302.469
}