How to Fine-Tune LLMs for Specialized Enterprise Tasks - Curating Data and Emerging Methods
Snorkel AI via YouTube
Overview
Discover how to fine-tune Large Language Models (LLMs) for specialized enterprise tasks in this 51-minute webinar by Snorkel AI experts. Learn about emerging fine-tuning and alignment methods like DPO, ORPO, and SPIN, and explore techniques for rapidly curating high-quality instruction and preference data. Gain insights into evaluating LLM accuracy for production deployment and see a practical demonstration of the fine-tuning, alignment, and evaluation process. Understand the importance of domain-specific knowledge and high-quality training data in transforming foundation models like Meta's Llama 3 into specialized LLMs. Explore topics such as data considerations, the development process, and creating effective data slices for model training. Enhance your understanding of enterprise AI and LLM fine-tuning through this comprehensive webinar.
Syllabus
Introduction
When and why finetune LLMs
Data considerations
Recent methods
Training data
Outline
Mission
Development Process
Domain Expert
Quality Model
Quality Model Example
Data Slices
Writing Data Slices
QA Session
Taught by
Snorkel AI