Run StableDiffusion in WSL

Tested on WSL with ubuntu22.04 and RTX 3090 Ti graphics card

• Download weights from here (5.2GB)

Now clone the repo

git clone https://github.com/Stability-AI/stablediffusion.git
cd stablediffusion

Change python from 3.8.5 to 3.9

nano environment.yml

#Create the conda env and install requirements

conda env create -f environment.yaml
conda activate ldm
pip install -r requirements.txt

Install xformers

conda install xformers -c xformers/label/dev

Run script with ref to weights and given prompt

python scripts/txt2img.py --prompt "tomato onion and garlic on a wooden surface" --ckpt /path_to_checkpoint/v2-1_768-ema-pruned.ckpt --config configs/stable-diffusion/v2-inference-v.yaml --H 768 --W 768 --device cuda

Dont forget --device cuda flag otherwise you will get RuntimeError: expected scalar type BFloat16 but found Float

Sometime an error is thrown for some prompts. Some user on Github found out why:

I discovered that this runtime issue stems from a shape mismatch between 'q' and 'k' in the cross-attention module. The batch sizes differ between 'q' and 'k' due to the unequal batch sizes between the cues and images. Kindly make sure that the number of your prompts is divisible by 'n_sample' solves the problem.