numpy
Pillow
Requests
torch
torchvision
transformers=4.41.2
accelerate
qwen-vl-utils
av