Transformer-Based Models

Tutorial sessions · Research Masters (ReMa) · Radboud University · 2024

Tutorial sessions conducted as part of the Transformer-Based Models (ReMa) course at Radboud University. Seven weekly sessions of 2 hours each, building up a transformer decoder from scratch using the tiny-Shakespeare dataset.

Role: Tutorial instructor  ·  Level: Research Masters (ReMa)  ·  Year: 2024


Tutorial Topic Notebook
Tutorials 1 & 2 Why Transformers? Tokenization & Data Preparation Open notebook ↗
Tutorial 3 Building a Bigram Language Model Open notebook ↗
Tutorial 4 Attention Mechanism & Transformer Components Open notebook ↗
Tutorial 5 Full Decoder-Only Transformer Architecture Open notebook ↗