llama.cpp

convert: fix LoRA conversion crash for Qwen3.5 V-head reorder

#28324

Open