Speakers
Description
This workshop provides a practical introduction to leveraging open-source Large Language Models (LLMs) with a focus on accessibility and customization. The first half will guide participants through the installation and setup of Ollama, a tool for running LLMs locally. We will then use the basic Ollama installation to explore pulling pre-trained models and customizing their behaviour to create useful activities for learners. We will cover how to utilize modelfiles to fine-tune model parameters, craft effective system prompts, and shape the model’s responses to specific tasks. The workshop will conclude with a discussion on envisioning practical educational deployments and applications of these customized LLMs, fostering ideas for innovative teaching and learning experiences. No prior experience with LLMs is required, though running your own LLM on your own hardware WILL require a tinkerer's heart and some troubleshooting skills (i.e. Googling, reading, thinking, trying to tweak the settings).
Author Bio
Adam Jenkins works at the Shizuoka Institute of Science and Technology (SIST) and a free and open-source software advocate. At SIST, he is the system administrator for the iLearn@SIST platform which has integrated a self-hosted Ollama AI.
Luc Gougeon is an educator at Sojo University in Kumamoto. He is a researcher in computational thinking and programming for children. Luc is a tinkerer in charge of the IoT / AI center, where he encourages students to use open-source software.