AI Resources at MSI
The Minnesota Supercomputing Institute (MSI) at the University of Minnesota provides compute resources for researchers within the university and other Minnesota institutions. The research we support increasingly involves AI and GPU-accelerated work. MSI offers some resources to for working with and learning to use AI tools and compute tasks. This is our collection of compute, software, and educational resources to help you get started.
Compute Resources at MSI
Here are the GPU compute partitions we have available.
| GPU type | GPU Partition Names | Node sharing? | GPU Memory | Cores per node | Walltime limit | Total node memory | Local scratch per node | Max. nodes per job |
|---|---|---|---|---|---|---|---|---|
| H100 | msigpu | Yes | 80GB | 128 | 24:00:00 | 768 GB | 850 GB | 4 |
| A100 | a100-4, msigpu | Yes | 40GB | 64 | 96:00:00 | 499 GB | 850 GB | 4 |
| A100 | a100-8, msigpu | Yes | 40GB | 128 | 24:00:00 | 1002 GB | 850 GB | 1 |
| L40S | msigpu | Yes | 48GB | 128 | 24:00:00 | 768 GB | 850 GB | 4 |
| A40 | preempt-gpu | Yes | 48GB | 128 | 24:00:00 | 499 GB | 850 GB | 1 |
| A40, L40S | interactive-gpu | Yes | 48GB | 128 | 24:00:00 | 60 GB | 228 GB | 1 |
| V100 | v100, msigpu | Yes | 32GB | 24 | 24:00:00 | 374 GB | 859 GB | 1 |
Storage Resources at MSI
What are some of the ways that Tier 1,2 and global scratch can be used for work with AI?
Project spaces have the following free resources available to use:
- Primary Storage (Tier 1) up to 20TB & 5 Million file count
- Secondary Storage (Tier 2) 120TB for buckets owned by the PI of the group
- Global Scratch 40TB & 13.2 Million file count
- Personal Home directory 200GB & 1 Million file count
Software at MSI
MSI offers ready to use software, but software can also be installed by users into home or project directories. Ask us if you’re not sure whether we offer some software you’re interested in. Here is a list of AI-related software that has been used on MSI resources.
General Deep Learning
Libraries that allow you to implement or use off-the-shelf models.
Bioinformatics
Software specialized for working with biological sequences.
Computer vision and image analysis
Software specialized for working with images.
Generative AI
Large Language Models (LLMs) have become household items but they are not relegated to the cloud; there are models that can be downloaded and run locally. These LLM-related software are not part of our offerings but can be downloaded and run on MSI resources:
Software and libraries for working with large language models.
Learning Materials
Internal
- MSI offers an Introduction to Deep Learning tutorial every fall/spring/summer: Intro to deep learning tutorial
- Knowledge Base
External
Here are some external educational resources that we’ve found helpful: