Introduction to supercomputing for AI

High performance computing (HPC) resources can be used to accelerate AI workflows. The EuroHPC Joint Undertaking (JU) offers free access to such resources to SMEs as well as larger companies. In this hands-on, you will learn:

• What is an HPC resource and how it is different from a cloud environment;

• How to connect to a cluster and explore resources;

• How to load necessary software and how workloads are managed;

• How to run a demo AI workflow based on Singularity.

Prerequisites

You will need to have credentials to access the PDC cluster. A working SSH client is needed: it is included on macOS and most Linux flavours; it is also available on Windows in the Powershell or under the Windows Subsystem for Linux (WSL).

Who is the course for?

This course is intended for data scientists that want to take advantage of higher computing power to perform their workflows. Some degree of familiarity with a command-line shell is recommended, but no expertise is required. No previous knowledge of supercomputing environments is required.

About the course

We will train a Unet model to be able to recognise water in satellite pictures. The source data comes from the Sentinel-2 satellite. Let us take the following picture for example:

The left panel shows the original image from Sentinel-2, whereas the right one was manually masked to highlight water regions.

The objective is to train a Unet model to perform this automatically. The source code can be found at this repo. The example is based on Tensorflow and will be run using Singularity. The structure of the example is the following:

./supercomputing4ai_demo
├── build_singularity.def
├── images
│   └── generated-images
├── models
│   ├── serving
│   │   └── main.py
│   └── unet
│       ├── data
│       │   └── water
│       │       ├── Images
│       │       └── Masks
│       ├── main.py
│       └── result
│           ├── models
│           └── training
└── README.md

The models subfolder contains the model to be trained (unet) and the inference code (serving). Under data, the training dataset can be found, with the Images being some satellite images and Masks being the water-covered areas in those images. Upon running main.py in the unet folder, a Unet will be trained, producing a set of weights in the models/ subfolder and training statistics (binary cross-entropy loss and accuracy). Inference is then performed with the models/serving/main.py script, which takes as an input an image and generates a mask of the water parts.

The lesson

• What is a supercomputer and how is it different from cloud?
• Connecting to a HPC resource
• Move between folders, ls, transferring to/from local storage
• Available software and modules
• Queueing jobs and running a Singularity container

Reference

• Quick Reference
• Instructor’s guide

See also

Further introductory material can be found on the Introduction to LUMI and HPC carpentry pages.

Credits

The lesson file structure and browsing layout is inspired by and derived from work by CodeRefinery licensed under the MIT license. We have copied and adapted most of their license text.

Instructional Material

This instructional material is made available under the Creative Commons Attribution license (CC-BY-4.0). The following is a human-readable summary of (and not a substitute for) the full legal text of the CC-BY-4.0 license. You are free to:

• share - copy and redistribute the material in any medium or format

• adapt - remix, transform, and build upon the material for any purpose, even commercially.

The licensor cannot revoke these freedoms as long as you follow these license terms:

• Attribution - You must give appropriate credit (mentioning that your work is derived from work that is Copyright (c) ENCCS and individual contributors and, where practical, linking to https://enccs.github.io/sphinx-lesson-template), provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.

• No additional restrictions - You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

With the understanding that:

• You do not have to comply with the license for elements of the material in the public domain or where your use is permitted by an applicable exception or limitation.

• No warranties are given. The license may not give you all of the permissions necessary for your intended use. For example, other rights such as publicity, privacy, or moral rights may limit how you use the material.

Software

Except where otherwise noted, the example programs and other software provided with this repository are made available under the OSI-approved MIT license.