llama.cpp

2f567611 - llama-model : support Qwen2 embedding models and pooling_mode_lasttoken (#13245)