llama.cpp

59377a6c - ggml-backend: fix async set/get fallback sync (#19179)