Hey all! Diffusers is magic
I’m the author of docker-diffusers-api, an HTTP REST API around diffusers running in a docker container. You can run it anywhere, but it’s architected to run on banana.dev who offer serverless GPU. It’s used in production for https://kiri.art/ (which is also open source).
Features
- Pipelines: txt2img, img2img and inpainting in a single container
- Models: stable-diffusion, waifu-diffusion, and easy to add others (e.g. jp-sd)
- All model inputs supported, including setting nsfw filter per request
- Permute base config to multiple forks based on yaml config with vars
- Optionally send signed event logs / performance data to a REST endpoint
- Can automatically download a checkpoint file and convert to diffusers.
The dev
branch is a little ahead than main
for now, with diffusers 0.9.0, SDv2 support, S3 support and dreambooth training.
Maybe it will help others here too.