Connor Shorten - Research Scientist, Weaviate - ChatGPT, LLMs, Form vs Meaning
Vector Podcast
00:00
How to Debug a Search System
The beer benchmarks is it's about diverse text retrieval so you know it's like arguana nf corpus trek covid. The idea would be that we could test several different models and you could maybe say like okay well i'm building a nutrition uh I'm building a Nutrition search apps or something like that but then fine tuning is going to be a super powerful lever for us. We're using open AI models cohere models hugging face models and it's just such an interesting thing to tune with gnais fine tuner. You can feed some manual examples in your basketball example how far you've been manually labeling in your basketball game to get the best results out of this system.
Transcript
Play full episode
