An open-source project enabling local inference of large AI models on consumer hardware via mmap layer streaming and Top-K sparsity, built b