[Bugfix][Multimodal] Normalize direct PIL image inputs (#47566)
Signed-off-by: Ting Sun <suntcrick@gmail.com> Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
This commit is contained in:
@@ -24,6 +24,7 @@ from vllm.utils.collection_utils import is_list_of
|
||||
from vllm.utils.import_utils import LazyLoader
|
||||
|
||||
from .audio import AudioResampler, AudioSpec, normalize_audio
|
||||
from .image import convert_image_mode, normalize_image
|
||||
from .inputs import (
|
||||
AudioItem,
|
||||
HfAudioItem,
|
||||
@@ -621,6 +622,13 @@ class MultiModalDataParser:
|
||||
else:
|
||||
data_items = data
|
||||
|
||||
data_items = [
|
||||
convert_image_mode(normalize_image(item), "RGB")
|
||||
if isinstance(item, PILImage.Image)
|
||||
else item
|
||||
for item in data_items
|
||||
]
|
||||
|
||||
return ImageProcessorItems(data_items)
|
||||
|
||||
def _parse_video_data(
|
||||
|
||||
Reference in New Issue
Block a user