How EleutherAI Trains and Releases LLMs: Interview with Stella Biderman
Gradient Dissent: Conversations on AI
00:00
How to Use a T0 and MT0 Model to Improve Performance
The T0 and MT0 models are the best-performing. They're relatively small but on standard NLP benchmark stuff they substantially outperform even like the base G2B3 models, OPT and many other much larger models. In particular niche applications, there are specific fine-tuned models that are extremely extremely compelling. The ones I've personally been most impressed with are from novel AI. And so it kind of comes down to what you want to do with the models and what you care about. But in general, if there is something that's been fine- Tuned to your application context, that's probably going to be the best. If there isn't, then I would generally
Transcript
Play full episode
