Refactor mm processors and Enable mixed modality processing (#7629)

Signed-off-by: Xinyuan Tong <justinning0323@outlook.com>
This commit is contained in:
Xinyuan Tong
2025-06-30 23:14:48 -07:00
committed by GitHub
parent 886d344964
commit 3a911b854d
28 changed files with 235 additions and 428 deletions
@@ -78,12 +78,6 @@ class PixtralProcessor(BaseMultimodalProcessor):
*args,
**kwargs,
):
if not image_data:
return None
if isinstance(image_data, str):
image_data = [image_data]
mm_data = self.load_mm_data(
prompt=input_text,
multimodal_tokens=self.multimodal_tokens,