llama.cpp

ggml webgpu: update Vulkan backend CI to use self-hosted runner

#21052

Open