Skip to content

Latest commit

 

History

History
18 lines (13 loc) · 683 Bytes

File metadata and controls

18 lines (13 loc) · 683 Bytes

Attention Workshop

Learn how the attention mechanism works, from intuition to implementation.

Getting started

Open Exercise 0 and follow along.

Exercises

Exercise Description
Exercise 0 The problem attention solves
Exercise 1 Words as vectors
Exercise 2 Dot product as similarity
Exercise 3 The attention mechanism
Exercise 4 Softmax: fixing the scores
Exercise 5 The weighted sum