How models are trained
The same network, with the same billions of numbers, can be an autocomplete that rambles or an assistant that answers. The difference is a much smaller round of training at the end.
In this lesson you'll learn to
- Describe pretraining, what it teaches a model, and roughly what it takes.
- Describe supervised fine-tuning and preference tuning, and what each one adds.
- Tell a base model from an instruct model, and pick the right one for a job.
- Explain knowledge cutoffs and why a model doesn't learn from your API calls.
Create a free account to open this lesson
It's free. You'll get every free lesson, quizzes that save your progress, daily flashcards and streaks.