Re-submit: Fix: Proper RGBA -> RGB conversion for PIL images. (#18569)

Signed-off-by: Chenheli Hua <huachenheli@outlook.com>
This commit is contained in:
Chenheli Hua
2025-05-23 01:59:18 +00:00
committed by GitHub
parent 46791e1b4b
commit 04eb88dc80
15 changed files with 89 additions and 20 deletions
@@ -11,6 +11,7 @@ from vllm import LLM, SamplingParams
from vllm.assets.audio import AudioAsset
from vllm.assets.image import ImageAsset
from vllm.assets.video import VideoAsset
from vllm.multimodal.image import convert_image_mode
from vllm.utils import FlexibleArgumentParser
@@ -45,7 +46,8 @@ def get_mixed_modalities_query() -> QueryResult:
"audio":
AudioAsset("mary_had_lamb").audio_and_sample_rate,
"image":
ImageAsset("cherry_blossom").pil_image.convert("RGB"),
convert_image_mode(
ImageAsset("cherry_blossom").pil_image, "RGB"),
"video":
VideoAsset(name="baby_reading", num_frames=16).np_ndarrays,
},
@@ -19,6 +19,7 @@ from vllm import LLM, EngineArgs, SamplingParams
from vllm.assets.image import ImageAsset
from vllm.assets.video import VideoAsset
from vllm.lora.request import LoRARequest
from vllm.multimodal.image import convert_image_mode
from vllm.utils import FlexibleArgumentParser
@@ -1096,8 +1097,8 @@ def get_multi_modal_input(args):
"""
if args.modality == "image":
# Input image and question
image = ImageAsset("cherry_blossom") \
.pil_image.convert("RGB")
image = convert_image_mode(
ImageAsset("cherry_blossom").pil_image, "RGB")
img_questions = [
"What is the content of this image?",
"Describe the content of this image in detail.",