The canonical ML workbook swaps TensorFlow for PyTorch

What it does

This repo is the engine room for Aurélien Géron’s O’Reilly book Hands-On Machine Learning with Scikit-Learn and PyTorch. It contains every example and exercise solution from the first edition, turning printed chapters into runnable Jupyter notebooks. The curriculum covers classical ML with Scikit-Learn and deep learning with PyTorch and Hugging Face.

The interesting bit

It is a deliberate fork in the road from the author’s older TensorFlow/Keras edition (handson-ml3). The notebooks are built to run in Google Colab, so you can break things in a disposable environment without wrestling with local CUDA drivers or dependency hell.

Key highlights

• Complete, book-aligned code and exercise solutions for a structured ML curriculum.
• Built on Scikit-Learn, PyTorch, and Hugging Face rather than TensorFlow/Keras.
• Designed for one-click Colab execution; nbviewer is available for read-only browsing.
• Targets Python 3.12, with 3.10 and 3.11 supported; 3.13 is explicitly not recommended due to missing library support.

Caveats

• Colab environments are temporary; the README warns that your work will be deleted after a while, so download anything you want to keep.
• macOS users may hit SSL certificate errors depending on how Python was installed.
• The GitHub-native notebook viewer is described as slower, occasionally broken for math rendering, and prone to failing on large notebooks.

Verdict

A good choice if you need a structured, textbook-grade introduction to ML with a current PyTorch stack. Look elsewhere if you are hunting for reference architectures or if you are already past the fundamentals.