Daniel Lyons' Notes

19 Tips to Better AI Fine Tuning

Description

🔥 Want to make your LLMs smarter? Discover the truth about fine-tuning - it's not what most people think! Learn when to use it, when to avoid it, and how to prepare for success with popular tools like Axolotl, Unsloth, and MLX.

🎯 In this video, you'll learn:
• What fine-tuning actually does (hint: it's not about new knowledge)
• The key differences between full fine-tuning, LoRA, and QLoRA
• How to prepare high-quality training data
• Which base model to choose for your project
• Common pitfalls to avoid

💡 Want to implement fine-tuning yourself? Subscribe and hit the notification bell to catch our upcoming tutorials on Axolotl, unsloth, and MLX!

Drop a comment below: Which fine-tuning tool should we cover first? 👇

Want to sponsor this channel? Let me know what your plans are here: https://www.technovangelist.com/sponsor

My Notes

00:00 - Intro

00:29 - The first tip

00:44 - The second tip...what it does

01:17 - But wait

02:49 - The next one

03:02 - The approaches

04:33 - The first example

05:04 - The parameters

06:40 - Another tip

07:12 - Second great use case

07:39 - This tip is a mistake to avoid

07:55 - Another mistake to avoid

08:12 - RAG is often better

09:09 - Patterns

09:31 - The most crucial piece

10:13 - Another tip

10:31 - Good Data

11:01 - High quality data

11:34 - What your training data must have

12:29 - Foundations

12:55 - Model size

13:47 - Something often overlooked

14:56 - But what do i DO

Transcript

19 Tips to Better AI Fine Tuning
Interactive graph
On this page
Description
My Notes
00:00 - Intro
00:29 - The first tip
00:44 - The second tip...what it does
01:17 - But wait
02:49 - The next one
03:02 - The approaches
04:33 - The first example
05:04 - The parameters
06:40 - Another tip
07:12 - Second great use case
07:39 - This tip is a mistake to avoid
07:55 - Another mistake to avoid
08:12 - RAG is often better
09:09 - Patterns
09:31 - The most crucial piece
10:13 - Another tip
10:31 - Good Data
11:01 - High quality data
11:34 - What your training data must have
12:29 - Foundations
12:55 - Model size
13:47 - Something often overlooked
14:56 - But what do i DO
Transcript