GPU isn't used on MacOS

[alryks]

I've got M2 Pro chip, running on MacOS Sequoia 15.1. I built sd with the flag -DSD_METAL=ON. I am using FLUX.1-dev-Q4_K.gguf model.

When I start generation with command sd --diffusion-model model.gguf --vae ae.safetensors --clip_l clip_l.safetensors --t5xxl t5xxl.safetensors --cfg-scale 1.0 --sampling-method euler -o ~/Downloads/img.png -p "a lovely cat holding a sign says 'flux.cpp'" the generation runs and the resulting image is meeting my needs, but it is SO slow.

I started wondering if it wasn't running on GPU, so I opened GPU History in Activity Monitor and noticed, that while generation the GPU almost isn't used.

I also get a huge amount of errors:

run.log

Why so? Isn't that how I should install stable-diffusion.cpp on MacOS?