Curious if there's a way to use GGUF models?

[wilvancleve]

This is a really great project, and I've had a lot of fun playing with this using schnell on my m1. I'm curious about other flux-derived models, in particular, Shuttle 3 Diffusion. Is there a way that quantized GGUF models can be converted to run a little faster on Apple Silicon using the mflux library?

[filipstrand]

Thank you! Currently, the only supported models and formats are the diffusers version of flux schnell and dev or the ones saved from this project. But we have an open PR from @anthonywu which I'm planing on including in the next release, and it looks like that one works with shuttle-3-diffusion among others. As for GGUF format support, that would be a nice feature to support, but probably something that won't be prioritised for a while.