Catalogue · Faculty of Machine Learning
ML 220
Neural Networks: Zero to Hero
Andrej Karpathy builds everything live, from a scalar autograd engine (micrograd) through character-level language models to a working GPT and a tokenizer, in pure Python and PyTorch. Nothing is imported that has not been built. The single best bridge between “I can code” and “I understand transformers” — and the recommended gateway into the AI faculty.
Syllabus
- backpropagation
- autograd
- language modelling
- transformers
- GPT from scratch
Held at Andrej Karpathy. The materials remain theirs; the structure is ours. Finished it? Record it in your transcript.