catnipwastaken/dooocus repository overview

⁠ Fooocus on docker

⁠ Quickstart

docker-compose file:

version: "3"
services:
  dooocus:
    container_name: dooocus
    image: catnipwastaken/dooocus:latest
    ports:
      - "7865:7865/tcp"
    volumes:
      - </path/to/models>:/fooocus/models
      - </path/to/saved/images>:/fooocus/outputs
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia # Amd cards have not been tested
            count: 1
            capabilities: [gpu]
    restart: unless-stopped

Keep in mind that it will take a while to download the required models to make this program work, the first time using this will require a bit of waiting.
The uploaded builds have been tested on a mobile RTX2060, an RTX3080 and a RTX3070.
It's recommended to place the models in a fast drive, otherwise it will take a bit to load the tensors. This will result in "Waiting for task to start", however, some patiance will result in the image processing to start regardless. Moving the model to the gpu will be done only once per session, therefore the talked waiting will only affect the image creation only once: after having generated one image, the successive ones will start almost instantly.