llama.cpp

246d622f - fast-quant: add skeleton & single tensor quant modes