Yeah, I have cross attention in my separate articles (the multi-modal and understanding attention from-scratch) ones. This article was focused on text LLMs to keep the scope reasonable.

→ View original post on X — @rasbt,