llama.cpp

ac17f8ac - ggml : use ggml_vqtbl1q_u8 for 32-bit compat (whisper/0)