Deep Papers cover image

Toolformer: Training LLMs To Use Tools

Deep Papers

CHAPTER

The Limitations of Self Supervised Learning

So you've generated this massive amount of API calls to annotate your data set. They're not very good. You filter them using this kind of clever mechanism. Now you have a data set that is annotated with these API calls. What do you do after that? Like, how do you actually use that to get the language model to use the tools? So there's two steps here. And then the final step is actually very, very simple. We just train the language model for a bit longer on this new data set with API calls. It won't like unlearn things that it learned during pre-training because it's still the exact same data distribution.

00:00
Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner