Architectures · intermediate · concept 53 of 176
Attention Mechanism
A technique allowing models to focus on relevant parts of input when producing output. Self-attention (in Transformers) lets each position attend to all other positions in a sequence.
Interactive · 3D
Watch attention poll the sentence →
The journey step where every token asks every earlier token what matters.
Key terms
QueryKeyValueAttention weightsSoftmax
Learn these first
Videos
▶ Attention in transformers, step-by-step | Deep Learning Chapter 6 ↗
3Blue1Brown · YouTube
▶ Transformers, the tech behind LLMs | Deep Learning Chapter 5 ↗
3Blue1Brown · YouTube
Guides and articles
Courses, papers, and more