Book Update #2 - Hands-On Large Language Models
A Christmas Update!
Happy (Early) Christmas! 🎅
Over the last few months, Jay and I have been working hard on writing our upcoming “Hands-On Large Language Models” book.
It is progressing well and Jay already did a book update but it was high time you also heard from me!
So here it is, a book update filled with visualizations 😉
Early Releases
Publishing with O’Reilly has been an amazing experience. There are so many wonderful people involved that it would take an update by itself to name them all.
Part of this process is regular early releases of the book. Whenever Jay or I have finished the first draft of a chapter, and after some initial feedback, we can release the chapter on their platform.
This process allows us to involve potential readers of the book early and get feedback through various channels.
Thus far, we have released 5 chapters with many more coming:
We aim to add a chapter to this early release every two months or so. It often depends on the complexity of the chapter we are currently writing.
And at times, writer’s block happens!
However, we already have several other chapters almost finished and ready to ship to the early release.
Although the first draft is in sight, there is still so much to be done!
Content
The book will be filled to the brim with images! This might not come as a surprise to those following Jay and I.
We love to explain content using, what we call, visual language.
“A picture says more than a thousand words.”
Honestly, we believe images are tremendously helpful in communicating complex subjects and sometimes, text is not sufficient to get to a state of intuition.
I quickly counted the number of custom images that we have released in those five early released chapters:
We already have 89 visualizations!
We expect to have at least 200 custom visualizations throughout the entire book and are likely to average around 20 images per chapter.
Things can always change of course but our intention has always been to create a highly visual book!
So… what does this mean for the content of the book? In practice, we like to have a good balance between using text, visuals, and code to explain and use concepts.
Going through the first five chapters, a single page would be, on average, distributed like so:
This means that 50% of the book is currently textual explanations, 25% visual explanations, and 25% hands-on coding examples.
Some chapters will contain more code than visuals and vice versa so this representation is just a rough proxy.
There are even some chapters without a single line of code!
Did I Spend My Time Wisely?
Each person tackles writing a book differently and I wanted a way to track how I would write one.
I figured, why not track the process of writing chapters?
I always make sure to work on a single chapter at a time. To keep it simple, I only track the following:
Date
Total number of words and pages (for that specific chapter)
Time spent writing (on that specific chapter)
Type of writing (text, visuals, coding, or review)
Tracking more metadata would take away from the writing process. At least, for me.
As a data scientist, what do you do with that data? You visualize it!
In the graph below you will see my progress writing on some of the chapters that are currently in the early release.
You might notice my lack of writing in October and November!
Well… my wife, daughter and I were all in a competition to see who was sick the longest. (Un)fortunately, I won 😅
With each chapter, I started to get more into the flow of writing and understand how to approach the book. Some chapters are straightforward and others require a bit more time.
Either way, it is fun to see how my progression has been since we started writing. I have more visualizations and data but I will save those for the next book update!
What is Next?
Thanks for making it this far!
There is so much more that I want to share. A detailed account of my experience writing the book, the collaboration with Jay (SPOILER: It’s great!), hurdles we faced (LLM-field is moving way too fast 😅), etc.
For now, I’m happy with the current state of the book but there is also so much to be done before we are finished.
Stay tuned for more updates!
Great book Thanks Maarten
Is the code open publicly? like hosted on Github?
LLM requires high computational power, is it feasible to run on personal devices by following this handbook?