llama.cpp

ggml: support concat for scalar types at cuda backend

#24011

Open