LLM Fundamentals: Build a Tiny LLM From Scratch

Ten full chapters, tested Python and PyTorch labs, worked calculations, and narrated walkthroughs. Build a tokenizer, attention, a transformer, training and generation.

1. 1Tokens: teaching a computer to read5 q
2. 2Embeddings: giving each token a vector of meaning5 q
3. 3The bigram model: your first real predictor5 q
4. 4Self attention: the idea that changed everything4 q
5. 5Multi head attention: many perspectives at once4 q
6. 6The transformer block: residuals, layer norm, and an MLP4 q
7. 7Assemble the GPT: stacking blocks into a model5 q
8. 8The training loop: how a model actually learns5 q
9. 9Generation: sampling text from your model5 q
10. 10Scaling, and what comes next5 q