Think Together · Start with a Short Story
The robot gets a sentence. It can combine information from different places instead of looking only at neighboring words. How does a Transformer help?
Lay Out the Process
- Input representationTurn text into numbers
- AttentionConnect different positions
- Other layersContinue transforming the data
- OutputUse it to understand or generate
A Transformer is a kind of neural network architecture that uses components such as attention.
It combines information from the context, then processes it through layers. It works for language and other tasks. Different models may use encoders, decoders, or related variations; they do not all have exactly the same full structure.
The name describes a model architecture, not a program that automatically knows everything.
Now You Try
If an image model uses a Transformer, can you say it only knows how to process text?
Ready? Open this to see the explanation
No. Similar structures can also be used for images, sound, and other data. The input representation and training task matter.
One Idea to Take Away
A Transformer is a kind of neural network architecture that uses components such as attention.
Try covering the diagram above and explaining each step to someone in your family. If one step is hard to explain, return to that card and look again.