# Attention, trained end to end

Follow a tiny trained attention pointer through embeddings, Q/K/V, residuals, a second attention stage, cross-entropy, backpropagation and Adam updates.

Fourteen interactive steps with Coach–Zon dialogue connect a prediction to the gradients and optimizer updates behind learning. This separate deep dive complements the original multi-head attention overview.

Canonical page: https://21ifm.com/ai/transformers/attention-trained-end-to-end

## Explore

- [Original multi-head attention overview](https://21ifm.com/ai/transformers/multi-head-attention)
- [Backpropagation foundations](https://21ifm.com/ai/backpropagation)
- [Transformer learning path](https://21ifm.com/ai/transformers)