biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

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

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-02-01 00:46:14 +08:00
committed by GitHub
parent 92924b2ddd
commit 88c3e114d8
43 changed files with 228 additions and 139 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/model_executor/models/siglip.py
View File

@@ -187,20 +187,20 @@ class SiglipMultiModalProcessor(BaseMultiModalProcessor[SiglipProcessingInfo]):
def apply(
self,
prompt: str | list[int],
mm_data: MultiModalDataDict,
mm_items: MultiModalDataItems,
hf_processor_mm_kwargs: Mapping[str, object],
tokenization_kwargs: Mapping[str, object] | None = None,
*,
mm_uuids: MultiModalUUIDDict | None = None,
) -> MultiModalInputs:
if prompt and mm_data:
if prompt and mm_items:
raise ValueError(
"Siglip accepts text-only or image-only inputs, not both! "
"Image-only inputs means passing an image with an empty text "
"prompt."
)
if mm_data:
if mm_items:
# For multi-modal data, the prompt after processing should
# only contain the image token
tokenization_kwargs = {
@@ -210,7 +210,7 @@ class SiglipMultiModalProcessor(BaseMultiModalProcessor[SiglipProcessingInfo]):
return super().apply(
prompt=prompt,
mm_data=mm_data,
mm_items=mm_items,
hf_processor_mm_kwargs=hf_processor_mm_kwargs,
tokenization_kwargs=tokenization_kwargs,
mm_uuids=mm_uuids,