I have run stable diffusion on my GTX 1060.
It has only 6gb of vram. So I used this version.
I tried to run the docker way. But the docker image became huge and it is based on ubuntu base image.
I actually stopped while building that image and found out the image was 14gb. So i run it in a simple virtual environment. And I have tensorflow==2.9 installed. So i have to use conda to match cuda version for the required pytorch cuda version.