[Model] Implement missing get_language_model for Keye-VL (vllm-project#20631)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

Author: DarkLight1337

vllm/model_executor/models/keye.py

@@ -1592,6 +1592,9 @@ def _parse_and_validate_multimodal_inputs(self, **kwargs: object) -> dict:
 
         return modalities
 
+    def get_language_model(self) -> torch.nn.Module:
+        return self.language_model
+
     def get_multimodal_embeddings(
             self, **kwargs: object) -> Optional[MultiModalEmbeddings]: