search

yisol / IDM-VTON

IDM-VTON : Improving Diffusion Models for Authentic Virtual Try-on in the Wild
https://idm-vton.github.io/
3.04k stars 461 forks source link

Newest package versions #53

Closed whitepapercg closed 1 month ago

whitepapercg commented 1 month ago

It seems adapting the code for some web interfaces causes problems with generating all-black clothing, or with some elements (spots) of the reference clothing on generation.

Is it possible that this is related to newer versions of Python and CUDA?

python=3.11.8 pytorch=2.0.1=py3.11_cuda12.1 pytorch-cuda=12.1 torchaudio=2.0.2=py311_cu121 torchtriton (not supported on windows) torchvision=0.15.2=py311_cu121 accelerate==0.25.0 torchmetrics==1.2.1 tqdm==4.66.1 transformers==4.36.2 diffusers==0.25.0 einops==0.7.0 bitsandbytes==0.39.0 scipy==1.11.1 opencv-python gradio==4.24.0 fvcore cloudpickle omegaconf pycocotools basicsr av onnxruntime==1.16.2

Windows 10

narutodiablo commented 1 month ago

i really hope sir yisol and the team would consider porting to windows platform

FurkanGozukara commented 1 month ago

it works great on Windows 10 with Python 3.10 : https://github.com/yisol/IDM-VTON/issues/47

whitepapercg commented 1 month ago

Replaced "PositionNet" to "GLIGENTextBoundingboxProjection" for new diffusers versions