Self-attention, explained without the heavy math

The intuition behind self-attention and transformers — query/key/value, multi-head, and why it beat RNNs — minus the algebra.

Raccontata dadev.to