A very hacky Stable Difussion Serving
Using virtualenv is highly recommended.
pip install -r requirements.txt
- Run
fetch.sh
andxformers.sh
(takes a long time) in parallel
Start server. If --addr
isn't provided, default to localhost:9001
.
python server.py --addr <addr>:<port>
Run inference, results will be saved to current directory as JPEG files.
curl -vJO -d "prompt=" http://<addr>:<port>/sd
TODO
MIT