Finetuning LLM models

One of the most useful technical skills in the job market is fine tuning existing models, especially llama3, the most capable open source model. Fine tuning is a way to influence the model's style in responses. You can make it witty, funny, serious and everything in between.

Setup

1.Intro and Setup

What is finetuning? In this lecture, that is exactly what I'll explain, and I'll show you how to get started installing all the required libraries

Llama3

2.Finetuning vs RAG

Instead of influencing the content a model has access to (RAG), with finetuning we can change the style of responses we get from a model. Not only that, we can also turn a model into an expert at pretty much anything, by having access to the right data.

Finetune

3.How to finetune a model

Let's fine tune llama3!

Hugging Face

4.How to save a finetuned model

In this lesson you'll learn how to save a model after finetuning and run it.

More Data!!

5.The real bottleneck is Data!

The main thing problem in the case of finetuning is the lack of data. The training aspect is actually the easy part. How do we solve this?

Quick DataSet

6.How to quickly build a dataset for finetuning

There is a way to build a dataset quickly and cheaply, we explore it in this video.