Multiple models loaded on startup (SDXL and Refiner).

[the-dusky]

I'm running a serverless worker that is basically just a docker image. Adding the refiner is too slow if switching back and forth between models. I can make multiple models setting = 2 in the UI and after on generation it is super fast. But I would like to be able to get to that state out of the box. I don't see any multiple models in memory launch flag and I don't know how to load 2 models to start either. Any advice, thoughts?

[w-e-w]

there's no option to load multiple models
currently the best thing you can do (without resorting to modify the code) is to use
use a script waite unitl webui has launched then use an API call to switch the model, doing so makes it load a socnond model

[the-dusky]

got it, I wonder if I can do this during the build phase of a docker image? Thanks getting back!

If I wanted to update the code to say allow something like --ckpt1 sdxl --ckpt2 refiner at launch, is there anything I could look at to get started?