Added fastapi server and Docker files.

[Banbury]

I needed a REST API, that I can deploy with Docker, for my workflow. So I wrote a simple one and the necessary Docker files.
Since the model is fast enough for me even on CPU, I didn't make the container compatible with CUDA. It allows me to keep Stable Diffusion on the GPU at the same time as stable-fast-3d.
Since I have already done the work for it, I thought you might be interested in it. License is CC0, no strings attached.

Start the server with:

fastapi run server.py

If you have trouble starting the FastAPI server, the reason is most likely the __init_.py file in the root directory. FastAPI tries to import server.py as a module, and that executes __init__.py. I'm not sure, why this is there, but you need to delete or rename it, for the server to start.

[wiz-inc-a28a8b7b4c]

Wiz Scan Summary

Scan Module						Total
IaC Misconfigurations	0	1	6	2	2	11
Vulnerabilities	0	0	0	0	0	0
Sensitive Data	0	0	0	0	0	0
Secrets	0	0	0	0	0	0
Total	0	1	6	2	2	11

View scan details in Wiz

[Banbury]

UVUnwrapper uses std::exchange, that doesn't seem to exist in newer versions of gcc. That forced me to downgrade the version of Debian.

[Akrobate]

FROM python:3.10-slim-bullseye

pytorchvision cannot be installed if the version of python is > 3.10

[Banbury]

I'm using this Docker image myself, and it works quite fine. I will try your change, when I have the time, but could you tell me, what the difference is?