Setting up LLMs on CURC Resources (A Short Course)

Large language models (LLMs) are rapidly gaining popularity due to their impressive performance across a wide range of disciplines, from the humanities to the physical sciences. However, effectively utilizing LLMs often requires significant computational resources. This Short Course* is designed to help researchers take full advantage of CU Research Computing (CURC) resources to deploy and run LLMs efficiently. This session will provide a high-level overview of LLMs, along with key considerations when working with them. It will offer guidance on using the popular LLM frameworks, Ollama and Hugging Face’s Transformers, to download and run state-of-the-art open-source models on CURC resources. Additionally, we will cover CURC’s new LLM chat interface, available in Open OnDemand. This session is intended for users with a basic understanding of programming and some familiarity with Linux command-line environments.