fix dpo metrics

Former-commit-id: 57029280da825a39fbf5a05097921b861f126669

src/llamafactory/train/sft/trainer.py

@@ -87,7 +87,11 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
         """
         loss = super().compute_loss(model, inputs, return_outputs, **kwargs)
         if is_transformers_version_equal_to_4_46() and not getattr(self, "model_accepts_loss_kwargs", False):
-            loss /= self.args.gradient_accumulation_steps  # other model should not scale the loss
+            # other model should not scale the loss
+            if return_outputs:
+                return (loss[0] / self.args.gradient_accumulation_steps, *loss[1:])
+            else:
+                return loss / self.args.gradient_accumulation_steps
 
         return loss