[ubutler]

Great work! This is the fastest inference I have ever seen of any truly large language model (>=70b parameters).

Just FYI, you might want to fix autocorrect on iOS, your textbox seems to suppress it (at least for me).