graykode/nlp-tutorial

Implements foundational NLP architectures from word embeddings (Word2Vec, FastText) through modern transformers (BERT, Seq2Seq with attention) in PyTorch, with each model constrained to under 100 lines of code. Features hands-on Jupyter notebooks paired with original papers, covering CNNs and RNNs for text classification, machine translation, and language modeling tasks. Progresses systematically from basic embedding models to attention mechanisms and transformer-based architectures.

14,870 stars. No commits in the last 6 months.