llama.cpp

ci : use macos-latest for arm64 webgpu build

#16029

Merged