[bugfix] fix mllm grpo liger loss (#9406)

Author: hjh0119

swift/rlhf_trainers/grpo_trainer.py

@@ -1001,7 +1001,10 @@ def compute_loss(self, model, inputs, return_outputs=False, num_items_in_batch=N
             inputs = inputs[0]
         if self.use_liger_loss:
             unwrapped_model = self.accelerator.unwrap_model(model)
-            return self._forward_redirection(model, unwrapped_model, self.compute_liger_loss, unwrapped_model, inputs)
+            forward_kwargs = self._prepare_model_inputs(inputs)
+            return self._forward_redirection(model, unwrapped_model,
+                                             lambda *_, **__: self.compute_liger_loss(unwrapped_model, inputs),
+                                             **forward_kwargs)
         else:
             return self._compute_loss(model, inputs)