Add offline option to better support stateless and offline environments

[qingfengfenga]

Add offline option to better support stateless and offline environments

https://github.com/qingfengfenga/stable-diffusion-webui-docker

docker compose --profile download up --build
# wait until its done, then:

docker compose --profile [ui] up --build
# where [ui] is one of: invoke | auto | auto-cpu | sygil | sygil-sl

docker compose --profile [ui]-offline up --build
# where [ui] is one of: invoke | auto | auto-cpu | sygil | sygil-sl
# Afterwards, all you need is a docker-compose.yml and offline image to start the service in the new environment

docker compose --profile [ui]-offline up -d
# where [ui] is one of: invoke | auto | auto-cpu | sygil | sygil-sl

[AbdBarho]

unfortunately, I don't have control over what the UIs do, this repo just packages them in a container.

This should be a feature request in the corresponding UI, not here.

[qingfengfenga]

unfortunately, I don't have control over what the UIs do, this repo just packages them in a container.

This should be a feature request in the corresponding UI, not here.

Perhaps the use of the ui offline name here is ambiguous and can be further modified.

During the process of packaging the UI into a container, this repository separately processes the data directory, allowing multiple UIs to share the data directory.

The generated images do not contain data. If I want to use the generated images elsewhere (in a stateless or offline environment) and cannot mount the data directory, I need to download the data from the data directory again every time I start.

The specific operation of the ui-offline option is to copy the downloaded data directory from the download container to the corresponding image, and generate an image containing data, that's all.

commit

qingfengfenga@a09d510

yaml part

  auto-offline: &automatic-offline
    <<: *base_service_offline
    profiles: ["auto-offline"]
    build:
      context: .
      dockerfile: ./services/AUTOMATIC1111/Dockerfile-offline
    image: sd-auto-offline:49
    environment:
      - CLI_ARGS=--allow-code --medvram --xformers --enable-insecure-extension-access --api

Dockerfile-offline examples

FROM sd-auto:49

RUN rm -rf /data && mkdir -p /data

COPY ./data /data

[AbdBarho]

Yeah, you would think that this should be enough, but the UIs will still download additional data on first interaction, and this data is usually cached in /data/.cache if I knew what it is, but that is always changing.

So you would need to run the image at least once, click all of the buttons so everything gets downloaded, and then you can copy the /data folder into the container.

Not a good solution, I know, and from the way the uis are implemented, there's nothing more I can do.

[qingfengfenga]

Yeah, you would think that this should be enough, but the UIs will still download additional data on first interaction, and this data is usually cached in if I knew what it is, but that is always changing./data/.cache

So you would need to run the image at least once, click all of the buttons so everything gets downloaded, and then you can copy the folder into the container./data

Not a good solution, I know, and from the way the uis are implemented, there's nothing more I can do.

Yes, I also tried to directly generate images that contain all the data, as you mentioned, the images still need to be downloaded at startup.

What I am currently doing is copying the data directory from the first successful run into the image. After installing the extension and debugging the model locally, I can easily share it with others or use it in other environments. Before UI support, I think this is a solution.

In addition, even if it is not possible to pre download all the files, it is beneficial for many people to minimize the number of files downloaded each time they start. (Shorter waiting time and websites that require proxies to access)

Perhaps it's more appropriate to call it ui release or ui full.