llama.cpp

2776db6c - Revert "ggml-cpu: handle 3d tensors in repack mat_mul (#17030)" (#17233)