Overview HPC for AI

In recent years artificial intelligence became an importance tool for many scientific domains. Various problems are tackled using methods from both classical machine learning and modern deep learning. These include but are not limited to challenges from

The data types used for artificial intelligence methods are vast, encompassing images, videos, audio, graphs and point clouds. HPC systems enable the analysis of large data sets and the training of state-of-the-art models. Our software stack and hardware is optimized to increase the performance (speed) of the training while keeping the quality of the results the same.

Services

We offer various services connected to artificial intelligence. Please contact us if you have a use case and we will discuss how to proceed individually.

Software and Libraries

Our HPC systems have a number of software tools installed. Examples are:

  • Anaconda for Python users
  • MPI
  • Cuda
  • Tensorflow
  • PyTorch

If your desired software libraries is not available, feel free to contact us.

KISSKI

The central goal of KISSKI is to carry out research on AI methods and their provision in a highly available AI service centre for critical and sensitive infrastructures. The focus is on the socially highly relevant fields of health and energy. These are among the future fields of application-oriented AI research in Germany. The increasing complexity of tasks as well as the availability of ever more extensive amounts of data through new sensor and actuator technology offer great potential for the application of AI methods in these fields.

Investments in the necessary hardware infrastructure and access to expert knowledge are a major challenge, especially for small and medium-sized enterprises. A central concern of KISSKI is therefore to establish effective access to AI infrastructure, methods and expertise for such companies.

Both the energy and health sector represent critical and sensitive infrastructures, which places high demands on the use of AI methods. These include guaranteed response times (e.g. for the automated control of power grids or anomaly detection in monitoring data of intensive care units), strict requirements for data security, resilience against cyber attacks, and the interpretability of AI models. KISSKI aims to answer how AI models for critical infrastructure can be effectively applied to these demanding circumstances.

Make many of the developed concepts and services we will make usable for other application areas. In this way, numerous areas such as mobility, e-commerce, Industry 4.0 or biodiversity research can benefit from KISSKI.

Support

We can support you on a wide range of common issues:

  • Exploring the potential benefits of HPC systems from AI workflows
  • Transferring AI workflows to HPC systems (e.g. parallelization and optimization)
  • Profiling applications to identify bottlenecks

Additionally, you can check out our other domain specific services linked above.

Infrastructure

The GWDG operates multiple HPC systems. Two systems are available to universities and publicly funded research institutes:

  • SCC: The SCC is available to members of the University of Göttingen and the Max Planck Institutes.
  • Emmy: Access to Emmy is granted based on a project basis
  • Grete: Access to Grete is granted based on a project basis

Information on how to access the HPC systems can be found here.

If you have any questions, feel free to contact us.