Creating Your Own LLM Tuning Platform with Open Source Technologies

Explore the world of Large Language Model (LLM) tuning in this 35-minute conference talk from DevConf.US 2024. Dive into FMS HF Tuning, an open-source package by IBM that leverages HuggingFace's Supervised Fine-tuning Trainer to support multiple LLM tuning techniques. Gain insights into when, why, and where to use this library, and understand its architectural integration with Open Data Hub and Red Hat OpenShift AI platform. Discover various tuning techniques, including Low-rank adaptation (LoRA), prompt tuning, fine-tuning, and inference. Learn how to deploy and run production-ready LLM model tuning and inference on ODH. By the end of this talk, presented by James Busche and Kelly Abuelsaad, acquire a deeper understanding of LLM tuning complexities, benefits, and the open-source tools available to enhance your AI solutions.

Creating Your Own LLM Tuning Platform with Open Source Technologies - DevConf.US 2024