diff --git a/vllm/multimodal/parse.py b/vllm/multimodal/parse.py index 0cce4b2661..9d1e005da7 100644 --- a/vllm/multimodal/parse.py +++ b/vllm/multimodal/parse.py @@ -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(