What should I learn first before I fine-tune?

I’m interested in making a fine-tuned model so I can quickly make high-quality flashcards that follow a very specific format. but I have zero fine-tuning experience, let alone programming skills.

I am currently learning the basics of Python, but I’m not sure if it would be enough to even train my first fine-tuned model. Will I have to learn machine learning fundamentals, deep learning basics, NLP, data preparation, and model fine-tuning first? Is there a simpler learning path? What do you recommend?

Also, I will eventually cover these topics as part of my education, so this isn’t all just for fine-tuning a model.

Thank you in advance for any guidance!

1 Like

Background knowledge will help you a lot. There are many small decisions that can make your project successful or not. If you are working on a project to explore, you can go explore, try, fail, and learn. Learning machine learning and deep learning basics will equip you with a great foundation. However, you can still do fine-tunning with openai. You should start reading this : Fine-tuning - OpenAI API

Also, there are some papers showing that in-context learning is often better than fine-tuning small models. So, you may want to work on the prompt engineering parts to see whether you can solve your problem with prompting.

Good luck!

2 Likes

You’ll learn what you need to know along the way. Read the finetuning documentation and stay curious. I finetuned 10 models before I got the outputs I wanted. Just know that actually finetuning will be about 10-20% of the work. Most of it it data gathering, cleaning, sorting, creating, etc.

But again, while it is possible to finetune a model to generate flashcards, this is a task that on-shelf models can complete with proper instruction and prompting.

1 Like