The LLM tutorial is optional, with those not attending the tutorial arriving for lunch on 19-June at 14:00, followed by an opening plenary at 15:00 and a welcome reception in the evening. The workshop will run until 13:45 on Friday 21-June.
Large Language Models (LLMs) for Science and Engineering Applications
This half-day tutorial will cover the wide range of LLM use across the scientific community, highlighting opportunities for cross-disciplinary and multi-institutional collaboration. It will include introductory hands-on exercises for researchers of all backgrounds interested in using LLMs for their research. The materials have been refined through four iterations at Argonne National Laboratory, each attracting more than 100 participants from 2019 to 2024. The tutorial has been expanded with materials from Pacific Northwest National Laboratory. This enhanced curriculum was presented to over 100 participants at Supercomputing Asia in February 2024.
Tutorial instructors include Ian Foster, Kyle Hippe, Alexander Brace, Carlo Siebenschuh, and Murali Emani from Argonne National Laboratory and The University of Chicago and Neeraj Kumar from the Pacific Northwest National Laboratory.
The specific curriculum modules for the tutorial has been optimized based on input from workshop participants.
| 09:00-09:45 | Introduction to Large Language Models (LLMs) for Science (PDF) | Ian Foster |
| 09:45-10:30 | Prompt Engineering | Alexander Brace |
| 10:30-11:00 | Retrieval Augmented Generation (RAG) (PDF) | Kyle Hippe |
| 11:00-11:30 | Coffee Break | |
| 11:30-12:15 | Agents | Neeraj Kumar |
| 12:15-12:45 | Beyond Transformers (PDF) | Carlo Siebenschuh |
| 12:45-13:30 | Performance Measurement and Optimization (PDF) | Murali Emani |
Please provide feedback using THIS FORM. We apologize for the projection issues — please factor that out of your assessment (we are working with the hotel on that, but it’s no fault of the instructors!). Special kudos to Carlo for switching from PPT to paper mid-presentation!
Before You Arrive: Prepare for Hands-On
While much of the tutorial materials can be absorbed by listening and following along, the steps below will prepare you to do hands-on. It is best for you to then complete the steps below BEFORE you come to the tutorial. To do this, please follow the instructions from a recent 2-day Argonne LLMs tutorial, in particular setting up a Google Colab account and a Hugging Face account and access token. The TPC LLM tutorial draws from the modules for this half-day tutorial, so there you can also find additional modules at that site that will be useful for further study and exercise. For example, many tutorial attendees expressed interest in building LLMs from scratch, and this is one of the hands-on modules included in the 2-day curriculum you can access there.