InstructLab is an open source project that allows anyone to enhance large language models (LLMs) for generative AI. Created by IBM and Red Hat, the InstructLab community provides a cost-effective solution to improve the alignment of LLMs and opens the doors for those with minimal machine learning experience to contribute. Join Red Hat’s Senior Director, Grant Shipley as he demonstrates step-by-step how to add new knowledge to an LLM.
00:00 Introduction
00:20 Serving the LLM with the iLab command
00:44 What is a quantized model?
01:47 Chatting with the model and finding a hallucination
03:28 Training the model with questions and answers
05:00 Generating synthetic data
06:15 Business use case
07:02 Vetting synthetic data with the critic model
07:50 Training the model with the synthetic data
08:09 Checking the newly trained model
09:45 Applying the model for a job task
11:06 Wrap up
Learn more about the InstructLab project at:
https://github.com/instructlab
Subscribe to Red Hat’s YouTube channel: https://www.youtube.com/redhat/?sub_confirmation=1
#RedHat #InstructLab #AI
