llama.cpp

ggml-webgpu: Add the support of `MUL_MAT_ID`

#21147

Open