Peter & Boris — Fine-tuning OpenAI's GPT-3
Gradient Dissent: Conversations on AI
00:00
The Challenge of Fine Tuning Models
When we started out, the way we did finting was basically, you know, in some way you almost grented a set of depus where the models ran on. The problem with that is, like, if you're like trying something new, debes hours are expensive so you don't want to really pay for to reserve a debe for like, even a fraction of an hour. We have, like, an amazing engineering team at openings cand of reade figured out lots of tricks around balancing where these models end up and cashing them in the right way. Om curseif you you find tuneda ant modon or you'll find tun as totof it
Transcript
Play full episode
