llama.cpp

Convert vector to f16 for dequantize mul mat vec

#1913

Merged