Book Review and Challenge Review of Chapter 7 of *Building an LLM from Scratch*

Chapter 7 covers the process of fine-tuning a model to follow instructions.

In other words, making it give the desired response to a given question.

As always, what you need is data.

1. Instruction Fine-tuning Procedure

The key here is to prepare a question–answer dataset and train it as input–output pairs.

This is called a prompt style.

Aside from that, the steps are similar to the previous chapters: tokenize the content, train, and evaluate.

2. Challenge Review

Looking at the long six-week challenge period, I started with the confidence of I can finish a single book like this in no time! but ended up defeated.

It took quite a lot of time just to understand the meaning.

To be honest, I was also busy.

For about two weeks I was preparing an astronomy class, so I was coming home after 10 p.m. every day, and when I got home, the kids were waiting for me.

Still, I praise myself for continuing to take on challenges like this and give myself another push, resolving to work hard and finish the remaining code lectures on Inflearn.