[BugFix] wire make_empty_intermediate_tensors on AyaVision and Voxtral (#43118)

Signed-off-by: Keyi Li <likey6688@gmail.com>
Co-authored-by: Keyi Li <likey6688@gmail.com>

vllm/model_executor/models/aya_vision.py

@@ -358,6 +358,10 @@ class AyaVisionForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsP
                 architectures=["Cohere2ForCausalLM"],
             )
 
+        self.make_empty_intermediate_tensors = (
+            self.language_model.make_empty_intermediate_tensors
+        )
+
     @property
     def dtype(self):
         return next(self.parameters()).dtype

vllm/model_executor/models/voxtral.py

@@ -349,6 +349,10 @@ class VoxtralForConditionalGeneration(
                 dim=config.text_config.hidden_size,
             )
 
+        self.make_empty_intermediate_tensors = (
+            self.language_model.make_empty_intermediate_tensors
+        )
+
     def get_mm_mapping(self) -> MultiModelKeys:
         """Get module prefix for multimodal models to filter LoRA modules."""
         return MultiModelKeys.from_string_field(