Books

This is a short shelf, not a catalogue. I only keep a title here when I can say what it changed in my work.

Natural Language Processing with Transformers

Lewis Tunstall, Leandro von Werra, and Thomas Wolf. O’Reilly.

The most useful part is the end-to-end treatment of transformer workflows: datasets, tokenization, fine-tuning, error analysis, and model sharing in one system. It is particularly good at showing where library abstractions help and where they hide decisions that still need to be understood.

Machine Learning with PyTorch and Scikit-Learn

Sebastian Raschka, Yuxi Liu, and Vahid Mirjalili. Packt.

I use this as a bridge between classical baselines and neural models, particularly for its treatment of evaluation pipelines before model complexity increases.