Transformer Architecture

Domain: NLP Fundamentals

🎯 Motivation

The neural network design based entirely on attention, no recurrence. Powers all modern LLMs.

📋 Concrete Example

A transformer uses self-attention layers and feed-forward layers, processing all tokens at once in parallel.

🔗 Analogy

Like a round-table meeting where everyone speaks and listens at once, instead of a phone chain where messages pass one by one.

Related Concepts

• Self-Attention
• Positional Encoding
• Encoder
• Decoder