Transformers

Attention Mechanisms in Transformers

Working through scaled dot-product attention from scratch — the query/key/value framework, why it works, and the one scaling detail I kept overlooking.