wadestuart

It looks like llama.cpp ran into some of the same types of issues pulling it in to their interfaces -- here is a pr that shows how they are working...