There are ROCm versions of llama.cpp, ollama, and kobold.cpp that work well, although they’ll have to add support for this model before they could run it.