The working group's FLAIR@Scale AI lab offers students and researchers web- based access to powerful hardware for classes, theses and projects.
A total of five servers are available in the AI lab for students and researchers to implement classes, theses and projects in the research area Data Science and AI . The servers have a total of 18 physical GPUs. Through virtualization, these can be expanded to up to 52 virtual GPUs so that they can also be used in large groups.
Kubeflow is used to manage the servers. A software framework provides Jupyter notebooks for prototyping and allows the implementation of complex workflows in the area of hyperparameter optimization and MLOps.
In addition to the servers, embedded hardware in the form of Nvidia Jetson development boards and Coral TPUs is available for local inference tasks.
Further information on the equipment and use of the lab is available here.