Fix remote code

custom_st.py

@@ -51,7 +51,7 @@ class MultiModalTransformer(BaseTransformer):
         self, features: Dict[str, torch.Tensor], **kwargs
     ) -> Dict[str, torch.Tensor]:       
         if features.get("inputs_embeds", None) is None:
-            features["inputs_embeds"] = self.auto_model.base_model.embed_tokens(features["input_ids"])
+            features["inputs_embeds"] = self.auto_model.base_model.get_input_embeddings()(features["input_ids"])
             if features.get("pixel_values", None) is not None:
                 features["pixel_values"] = features["pixel_values"].type(self.auto_model.visual.get_dtype())
                 image_embeds = self.auto_model.visual(