search

cloneofsimo / paint-with-words-sd

Implementation of Paint-with-words with Stable Diffusion : method from eDiff-I that let you generate image from text-labeled segmentation map.
MIT License
635 stars 52 forks source link

Torch missing #7

Closed DoctorMG closed 1 year ago

DoctorMG commented 1 year ago

there is an import of torch required, that seems to be missing in requirements.txt

should be added or documented

cloneofsimo commented 1 year ago

Well I thought torch part is kind of trivial because stable diffusion is built on torch... Im also not that aware of making appropriate setuptools to automatically handle various os-cudakernel dependent torch setups Any help would be appreciated

DoctorMG commented 1 year ago

Hm, me neither. What I did to install torch was just this: x:> pip install torch torchvision torchaudio

When I execute paint_with_words after this pip the break by missing torch is gone. There is some warning smalltalk like "no CUDA found", "will need longer time" (... sure ...) but they all start with "WARNING". So I guess these are no show stoppers and the pip command for installation is ok.

Since I get a hard exception failure shortly later I don't know if there is something else missing beyound torch.

Arron17 commented 1 year ago

Hm, me neither. What I did to install torch was just this: x:> pip install torch torchvision torchaudio

When I execute paint_with_words after this pip the break by missing torch is gone. There is some warning smalltalk like "no CUDA found", "will need longer time" (... sure ...) but they all start with "WARNING". So I guess these are no show stoppers and the pip command for installation is ok.

Since I get a hard exception failure shortly later I don't know if there is something else missing beyound torch.

You should use the Pytorch website to get the correct command for your system that includes CUDA.

https://pytorch.org/get-started/locally/

cloneofsimo commented 1 year ago

Right, like the CLIP repo, ill just add more info on installing pytorch.