Can you only run different model types in the same image?

[kilimchoi]

Based on readme, it sounds like we have to build a separate docker image for different model type.
so if we want to support sd2+, sdxl, sd3 models, do we need to build 3 separate docker images?

[TimPietrusky]

@kilimchoi thanks for getting in touch!

so if we want to support sd2+, sdxl, sd3 models, do we need to build 3 separate docker images?

Exactly, this is how it works right now. The idea was, that people can pick one of the images for their use case and use it without the worry about any other model. This makes sure, that the image sizes are "small" and only contain the things that are totally needed for the specific model.

If you want to have multiple models in one worker, you have two options:

1. You bake them into one image yourself
2. You attach a network-storage and use the base image (as it contains no models at all). On the network-storage, you can then add all the models you want.