Fix text only inputs for MRoPE models with the Transformers modelling backend (#37055)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

vllm/model_executor/models/transformers/multimodal.py

@@ -448,20 +448,21 @@ class MultiModalMixin(SupportsMultiModal, SupportsMRoPE):
         # In v4 `get_rope_index` doesn't have wildcard `kwargs`, and
         # can't accept arbitrary args, even if its value is `None`
         kwargs = {}
-        if mm_token_type_ids:
-            if not hasattr(self, "_get_rope_index_accepts_mm_token_type_ids"):
-                import inspect
-
-                sig = inspect.signature(self.model.get_rope_index)
-                params = sig.parameters
-                self._get_rope_index_accepts_mm_token_type_ids = (
-                    "mm_token_type_ids" in params
-                    or any(
-                        p.kind == inspect.Parameter.VAR_KEYWORD for p in params.values()
-                    )
-                )
-            if self._get_rope_index_accepts_mm_token_type_ids:
+        if not hasattr(self, "_get_rope_index_accepts_mm_token_type_ids"):
+            import inspect
+
+            sig = inspect.signature(self.model.get_rope_index)
+            params = sig.parameters
+            self._get_rope_index_accepts_mm_token_type_ids = (
+                "mm_token_type_ids" in params
+                or any(p.kind == inspect.Parameter.VAR_KEYWORD for p in params.values())
+            )
+        if self._get_rope_index_accepts_mm_token_type_ids:
+            if mm_token_type_ids:
                 kwargs["mm_token_type_ids"] = torch.cat(mm_token_type_ids)
+            else:
+                shape = (1, len(input_tokens))
+                kwargs["mm_token_type_ids"] = torch.zeros(*shape, dtype=torch.int)
 
         mrope_positions, mrope_position_delta = self.model.get_rope_index(
             input_ids=torch.tensor(input_tokens).unsqueeze(0),