fix(multimodal): handle tensor images in exact-token preprocessing (#30368)

Signed-off-by: Zhuangcheng(Jesse) Gu <zcgu@connect.hku.hk>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Mick <mickjagger19@icloud.com>

python/sglang/srt/multimodal/processors/base_processor.py

@@ -1732,7 +1732,14 @@ class BaseMultimodalProcessor(ABC):
 
         """
         assert images is not None
-        image_sizes = [(image.height, image.width) for image in images]
+        image_sizes = [
+            (
+                tuple(image.shape[-2:])
+                if isinstance(image, torch.Tensor)
+                else (image.height, image.width)
+            )
+            for image in images
+        ]
         num_image_tokens = self._processor._get_num_multimodal_tokens(
             image_sizes=image_sizes
         ).num_image_tokens