llama.cpp

ggml-webgpu: Support GPU profiling beyond the maximum query count

#22995

Merged