Awni Hannun

So for the first one you are probably just copying it into terminal incorrectly. Try putting it in a single line: ``` python3 -m mlx_lm.lora --model ~/Models/mixtral-8x7B-v0.1 --adapter-file ~/Models/adapters/mixstral-8x7B-ft.npz --prompt...

> python3 -m mlx_lm.fuse --model ~/Models/mixtral-8x7B-v0.1 --adapter-file ~/Models/adapters/mixstral-8x7B-ft.npz --save-path ~/Models/mixtral-ft I can't reproduce the problem you got with the fusion, it runs fine for me. My recommendation is to update...

Thanks for the help debugging all this @mzbac, I missed the fact that @lin72h was using `mlx_lm.lora` to do generation. We removed that a while ago in order to simplify...

Also @mzbac do you plan to send a fix for the Python version issue?

How did you install the macOS library? I just tried it via pip and it worked fine. Could you share the output of: ``` pip show mlx ``` And ```...

Can you do `pip show mlx-metal` as well?

Thanks. Could you also list these two directories: /Users/michael/Dev/yirgacheffe/venv2/lib/python3.13/site-packages/mlx /Users/michael/Dev/yirgacheffe/venv2/lib/python3.13/site-packages/mlx/lib

Strange it didn’t include the library in the old one.. I don’t know why that is. I’ll double check that it is in the wheel and get back to you

Ok I think I understand the issue. We changed the way MLX packaging works in 0.26.5. It now installs the MLX binary via `mlx-metal`. What's happening is if you `pip...

Right if you already installed `mlx==0.26.5` you'd have to uninstall both `mlx` and `mlx-metal`. If you haven't yet installed it then uninstall mlx before installing the new version.