[Question] Where does auto service get cuda?

[bonlime]

Hi, i'm trying to build some custom docker image with automatic inside and reading through your Dockerfiles i can't understand where CUDA is installed? I see that for hlky & lstein there is a line in Dockerfile with
RUN conda install cudatoolkit=11.6 -c conda-forge && conda clean -a -y
but there is no mention of CUDA in automatic docker and still it somehow works. Could you point me where it gets the CUDA?

[AbdBarho]

@bonlime here:

stable-diffusion-webui-docker/services/AUTOMATIC1111/Dockerfile

Line 37 in 8623c73

RUN pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116

which basically tells pip: please install pytorch that comes bundled with cuda 11.6

[bonlime]

Thanks, now I get it. I though it meant "download pytorch prebuilt for cuda 11.6" but didn't include the cuda itself.