Is it possible to train a MultiControlNetModel(line, color) to reproduce the result?

[RaccoonDML]

Thanks for the author's excellent work. The results are very impressive! Here is my question when reproduce the result:
I find that the adapter is another ControlNet encoder. So can I just use MultiControlNetModel in diffusers to wrap two controlnet models for both training and inference?

Code may be like this:
controlnet = MultiControlNetModel([line_controlnet, color_controlnet])