Workshop Talk
Presenters: Mazal Bethany, Emet Bethany, Mohammad Bahrami Karkevandi
We delivered a training workshop on multi-agent LLMs as part of international summer program at the School of Data Science (SDS) hosting 12 international students from June 24 to July 27 for the first annual SDS Summer Immersion Experience. The students, from two private Mexican universities- Tecnológico de Monterrey (TEC) and the Instituto Tecnológico Autónomo de México (ITAM)- spent a month in San Antonio assisting SDS core faculty with their research.
Day 1: High Performance Computing
This workshop covers the fundamentals of running jobs on UTSA’s ARC HPC systems that offer GPU computing capabilities to students. Attendees were introduced to networking, storage, job submission, and best practices for using common software with LLMs. With assistance from ARC’s administrative staff, students were given access to GPUs on ARC. During the workshop, attendees connected to ARC and ran commands to ensure everyone could access and use ARC. Utilizing GPU access, attendees participated in a hands-on assignment to fine-tune a small LLM, Microsoft Phi-2, on an instruction-based dataset.
Day 2: HuggingFace Workshop
This workshop covers leveraging the Hugging Face platform for machine learning tasks. Attendees learn the basics of getting started with Hugging Face, exploring the Model and Dataset Hub, and selecting appropriate models, particularly Large Language Models. The workshop includes fine-tuning techniques and provides code examples for loading models/datasets, creating training scripts, and utilizing other useful functions. Attendees also participate in a practical demonstration, creating a simple chatbot using an LLM (Llama3) from Hugging Face, complete with chat history functionality.
Talks & Presentation
- Information Coming Soon