The max-llm-book has gotten a couple of big updates in the past month. We trimmed the python steps that (while nicely broken up to pace the tutorial) made it harder to picture the whole architecture you were building toward. This streamlined the book, but made it less interactive.
To improve the interactivity and visual support, we replaced these steps with a Jupyter notebook. You can now execute the cells of the various steps to build GPT2 in a familiar format for many model developers. The cells include matplotlib visualizations of the output so it’s much easier to conceptualize what each step achieves.
Finally, the tutorial now shows you how to use MAX serve to serve the model (thanks to @rachfop) which is a much more satisfying result than a simple text completion loop. Check it out, and please continue to create issues and PRs where you see opportunities to improve and extend the tutorial.