[Refactor] Define MM data parser in processing info instead of processor itself (#33260)

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

vllm/model_executor/models/molmo2.py

@@ -1860,6 +1860,12 @@ def get_frame_times_and_chosen_fps(
 
 
 class Molmo2ProcessingInfo(BaseProcessingInfo):
+    def get_data_parser(self):
+        return MultiModalDataParser(
+            video_needs_metadata=True,
+            expected_hidden_size=self._get_expected_hidden_size(),
+        )
+
     def get_hf_processor(self, **kwargs: object) -> Molmo2ProcessorWrapper:
         processor = self.ctx.get_hf_processor(**kwargs)
         hf_config = self.ctx.get_hf_config()
@@ -2183,9 +2189,6 @@ class Molmo2MultiModalProcessor(BaseMultiModalProcessor[Molmo2ProcessingInfo]):
 
         return prompt_tokens
 
-    def _get_data_parser(self) -> MultiModalDataParser:
-        return MultiModalDataParser(video_needs_metadata=True)
-
     def _call_hf_processor(
         self,
         prompt: str,