llama.cpp

745cbcf2 - llama-quant : fix the verification of attention layers for encoder-decoder models (#16023)