How would I compile an entirely custom-made GPT model and run it locally on an Intel-based Mac? Most of the sources that I've consulted have yielded unsatisfactory methods, so I'm hoping that someone here knows.