docker-diffusers-api-runpod

Copyright (c) 2022 by Gadi Cohen dragon@wastelands.net MIT Licensed.

Quick Start

1. Clone the repo

2. Set any necessary environment variables and then build with e.g.

./build.sh -t user/runpod:sd-v2-1-512 \
  --build-arg MODEL_ID="stabilityai/stable-diffusion-2-1-base"

(or just use docker build as you see fit).

3. Upload to your repository of choice, e.g.

$ docker push user/runpod:sd-v2-1-512

See https://forums.kiri.art/t/running-on-runpod-io/102?u=gadicc for deployment instructions.

More full examples

Download AnimeAnything v3, from diffusers branch / revision on HuggingFace, which is an fp16 model, and is already prebuilt on our own S3-storage in the default location.

$ ./build.sh -t user/runpod:sd-v2-1-512 \
  --build-arg MODEL_ID="Linaqruf/anything-v3.0" \
  --build-arg MODEL_PRECISION="fp16" \
  --build-arg MODEL_REVISION="diffusers" \
  --build-arg MODEL_URL="s3://"