{ "epoch": 3.0, "eval_logits/chosen": -16.375, "eval_logits/rejected": -16.125, "eval_logps/chosen": -696.0, "eval_logps/rejected": -712.0, "eval_loss": 1.1395312547683716, "eval_rewards/accuracies": 0.5, "eval_rewards/chosen": -3.796875, "eval_rewards/margins": 0.423828125, "eval_rewards/rejected": -4.21875, "eval_runtime": 46.5036, "eval_samples": 2000, "eval_samples_per_second": 43.007, "eval_steps_per_second": 0.688 }