llama.cpp

fa4a9f2a - quantize : handle user-defined pruning of whole layers (blocks) (#13037)