NRIS GPU Tutorials Series

The Norwegian Research Infrastructure Services (NRIS) is offering a series of GPU Tutorial events given every now and then on relevant topics in the area of GPU utilization, Machine Learning, Large Language models and related topics.

The course is open to all and free of charge, but registration is mandatory.

Targeted Audience: The course is aimed at students, researchers and employees who want to upskill their knowledge on GPU utilization and usage.

Practical Information

  • This is an online course via zoom. Participants require access to a computer (not provided by the course organisers) with internet connectivity.

Note

Access to necessary GPU resources will be provided by the course organizers, but access is based upon request. Please thoroughly read information given for each individual events and follow the instructions for necessary registrations given there.

Events and Schedule

All times are given in CET/CEST depending on date.

Fine-Tuning LLMs with Multi-GPU Training - A Hands-on Workshop

Date: Tuesday September 2nd 2025. Time: 09:30-15:30

Content: This hands-on workshop, titled Fine-Tuning LLMs with Multi-GPU Training, is designed to provide participants with practical experience in fine-tuning large language models (LLMs) with a special focus on LLaMA on high-performance computing (HPC) systems.

Learning Outcomes:

By the end of this workshop, participants will be able to:

  • Learn the basics of LLaMA.

  • Set up a distributed training environment using PyTorch, Torchao, and Torchtune on an HPC system.

  • Fine-tune a LLaMA model using LoRA techniques on a single GPU and then scale up to multiple GPUs.

  • Perform inference tasks like summarization using the fine-tuned model.

  • Monitor GPU usage and GPU memory utilization.

This workshop is ideal for researchers, developers, and students who are familiar with Python and want to gain practical insights into scalable LLM training with LLaMA.

Registration: Registration is soon open

Note

  • Link to video to be announced.

  • Link to Q&A document to be announced.