Pure Go hardware accelerated local inference on VLMs using llama.cppgithub.com/hybridgroup1 pointdeadprogram7 months ago