ageron/handson-ml2 ? reverse-engineered prompt

Build me a GitHub repo of Jupyter notebooks that teaches machine learning step by step, with a clean landing notebook, chapters that cover the basics all the way through neural nets, computer vision, NLP, and reinforcement learning, and a few extra notebooks for the math and tools people need along the way.

I want it to feel like a complete learning companion, with example code, exercise solutions, dataset loading, and simple instructions for getting started locally or in Docker. Please make sure the notebooks are actually runnable, the environment setup is clear, and the project works nicely for someone opening it in Jupyter, Colab, or a similar notebook viewer. If anything depends on current library behavior, look up the latest docs online and adjust the code so it runs cleanly.