llama.cpp

ef4c5b87 - presets : fix pooling param for embedding models (#16455)