I would love to port the diffusion upscaler to a MLProgram. It seems like the UNET architecture is very similar to the diffusion models already ported. I have a lot of experience porting models to CoreML, but, looking at the conversion script for this project, it seems a bit outside my ability.
I'm willing to work on it though. Any pointers? Anyone else interested in working on it?