yahoozoo

Are there any open source models in this space?

Look up UVR (Ultimate Vocal Remover). I believe most of the models predict frequency masks to split out stems.