A brief introduction to attention in the transformer architecture.

Altering the tokens to which a token in the input sequence attends.

Reducing the number of calculations to compute attention.