Encoder-Decoder Architecture
Domain: LLM Concepts
🎯 Motivation
The original transformer had two parts: an encoder that understands the input, and a decoder that generates the output. Good for translation.
📋 Concrete Example
For translation: encoder reads Hello in English and creates a rich understanding. Decoder uses that to generate Bonjour in French, token by token.
🔗 Analogy
A two-person translation team - one person reads and deeply understands the source text (encoder), then briefs the other person who writes the translation (decoder). Each specializes in their role.