05-30, 13:10–14:00 (Europe/Amsterdam), Fluor podium
Open source large language models and surrounding tools ecosystem have lowered the barriers of entry to easily experiment with LLM/AI models. Workshop for both technical and non-technical researchers and support staff.
Together with the wide-spread development of Large Language Models (LLMs), an ecosystem of software packages and applications is emerging to run these models locally. While the biggest models still require big datacenter infrastructure, there is a clear accompanying trend in creating “smaller”* models which still achieve great performance on a wide variety of tasks for a small fraction of the compute requirements & cost. Resulting in lower barriers of entry to explore, test and run these models in practice.
With recent tooling, open-source LLMs have become accessible to run on your laptop or desktop. Making the models available locally in user-friendly GUI interfaces or more developer-oriented APIs.
In this workshop we will explore how to use some of these tools to explore the capabilities of open-source LLMs. Helping researchers answer questions or dilemmas such as:
Do I need to rely on big tech LLM APIs?
When and on how could I use an open source LLM?
When to train a new model or when to fine-tune an existing model?
Relevant for research support to host & run models for researchers and RDM around datasets for e.g. fine-tuning or testing of models.
Workshop outline:
• Introduction presentation about open-source LLMs and tooling ecosystem
• Demo of open source tools & common models
• Group exercises: first interactions with a local, open source LLM
No laptop required, but feel free to bring one!
Hundreds of billions of parameters 10^11
*Several billion parameters 10^9
Service Manager, Research Data Infrastructure lab at Eindhoven University of Technology