llama.cpp

feat(ggml-cpu): add VLEN1024 Q4_K/Q5_K 64x1 CPU_REPACK path

#25553

Closed