MLX models unavailable on M1 Max (64GB) in LM Studio 0.3.5? Seems only gguf (metal llama.cpp) is recognised, and search doesn't include mlx models..
Posted by dogfighter75@reddit | LocalLLaMA | View on Reddit | 5 comments
mayo551@reddit
You need to drag on studio to the applications folder on the installer.
Then launch through the application menu.
Otherwise it can’t scan your hardware.
dogfighter75@reddit (OP)
... incredible that this is the solution, I'll try to get used to macOS first. Deleting the thread as it's not helping anyone else probably. Thank you!!
mayo551@reddit
Iirc macOS sandboxes the application otherwise.
macOS is really good at sandboxing things. You can even make your own sandbox scripts and restrict applications that way.
krisirk@reddit
I see there's a message saying "Error surveying hardware" for the MLX runtime in your first screenshot. I'd guess that it didn't load up something correctly. Maybe a reboot/reinstall of LM Studio will help.
It works just fine on my M1 Macbook Air. Lists both GGUF and MLX under configure runtimes at the top of your first screenshot.
dogfighter75@reddit (OP)
It seems the second image didn't upload, here it is:
Enthused by this thread this thread I've been trying to set up LM Studio on my M1 Max, but mlx models such as those uploaded on the Hugging Face MLX community didn't show up in search. I had a look, and I can only tick gguf in search. I think based on the screenshots that mlx isn't recognised as compatible. Does anyone know how I can fix this?