[Refactor] Move MM data parsing outside processor (#33408)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
2026-02-01 00:46:14 +08:00
parent 92924b2ddd
commit 88c3e114d8
43 changed files with 228 additions and 139 deletions
--- a/vllm/model_executor/models/gemma3_mm.py
+++ b/vllm/model_executor/models/gemma3_mm.py
@@ -290,9 +290,8 @@ class Gemma3MultiModalProcessor(BaseMultiModalProcessor[Gemma3ProcessingInfo]):

        # HF processor pops the `num_crops` kwarg, which is needed by vLLM
        if (images := mm_data.get("images")) is not None:
-            parsed_images = self.data_parser.parse_mm_data({"image": images}).get_items(
-                "image", ImageProcessorItems
-            )
+            mm_items = self.info.parse_mm_data({"image": images}, validate=False)
+            parsed_images = mm_items.get_items("image", ImageProcessorItems)
            image_sizes = [
                parsed_images.get_image_size(i) for i in range(len(parsed_images))
            ]