Just Do It

A trip to Barnes & Noble with my children inspired me to pick up another new technical book (I got my non-technical fix from Kindle Unlimited), Build a Large Language Model (From Scratch) by Sebastian Raschka. Getting into bed at 9:30PM, I went through the book carefully (by my estimate) and finished the first two chapters in an hour. The book was fascinating, as the author went through the in-depth details of the why and how of fundamental LLM tasks such as tokenization and embedding. I loved it then, and in the morning, I had a disappointing feeling that I had not learned much about anything at all.

Recalling my lesson from Learning How to Learn, the only Coursera course that I completed, I realized that I ran into an illusion of learning the previous evening. My background allowed me to understand and appreciate the beautiful details and simplicities in Raschka’s book. At the same time, these have remained Raschka’s knowledge and not yet mine. I need to do something to make it mine.

In the remainder of this essay, I will record my progress of doing things as they are described and taught by the book. This helps me to both record and reflect on my own learning activities for future references. It should be noted that both this blog post and the coding activities are/will be done completely on LazyVim as part of my other training process.

Just Do It (Something, Anything)

Raschka maintains a GitHub repository accompanying the book. The repository is frequently updated (latest commit was less than a month prior to the date of this blog). I forked this repository to keep a copy, but decided against simple cloning and running the codes as is since it still felt like demonstrative surface learning. Instead, I created a new learning repository in which each directory will be dedicated to all the coding activities created from scratch for each technical book going forward.

Track Your Progress

Showcase It at the End

Conclusion