The LLM-era of data science has begun. But before you jump in head first, it’s important to understand LLM fundamentals, key concepts, and strategies for building production AI applications.
In this episode of ML Real Talk, Daliana Liu, Data Scientist and Host of ML Real Talk, and Arnav Garg, ML Engineer at Predibase, delve into all things LLMs from how these complex models are trained to exploring LLM customization from few shot learning to finetuning.
Whether you're an AI enthusiast, a seasoned data scientist, or a curious engineer, this talk gets you ready for the sophisticated LLM technology that's shaping our future. This is more than just a talk-it's an engaging journey into the core of AI's most fascinating capabilities.
These are the hot topics we dive into:
- How are LLMs trained?
- What are Alpaca and Llama models and other popular open-source LLMs?
- What is few shot and zero shot learning?
- How does finetuning work and how do you do it?
- How to iterate and conduct error analysis?
- What you can do to improve data quality / data labeling?
- Which set of tools should you select for your use case?
Start querying and customizing open-source LLMs on your data on managed infrastructure with the Predibase free trial: predibase.com/....
Негізгі бет ML Real Talk: 10 Things You Must Know About LLMs
Пікірлер