Account management of the LRZ AI Systems is associated with the LRZ Linux Cluster. Access permissions are handled individually on per account basis (not on the project level). If you already have an account for the Linux Cluster, you will additionally have to request access to the LRZ AI Systems for this account. Once access is granted, you will be able to use the corresponding user name and password to access the LRZ AI Systems. Otherwise, you need to set up access to the LRZ Linux Cluster first (see Access and Login to the Linux Cluster).
Access to the LRZ AI Systems can be requested by submitting a service request to LRZ Servicedesk. By using this link, you will open a ticket for the appropriate service. Then, please select "AI topics" and "LRZ AI Systems - Request for Access" from the drop-down lists. The web form will ask for some mandatory information, such as the Linux Cluster account username. Provided that your account has access permissions for the Linux Cluster, your service request can be submitted successfully. Further, the request has to include a description of the intended usage, i.e. a short summary of the use case as well as the expected workload, including a description of system resource requirements. Please specify which system(s) are intended to be used (see 1. General Description and Resources) and whether there is the intention to use the web-based frontend (see 10. Interactive Web Servers on the LRZ AI Systems). Access will be granted only after review of this application.
Please bear in mind that due to the limited amount of current resources, the LRZ AI Systems should not be considered as a development or debugging environment. Any programs and scripts defining your production jobs should have been already tested with other (GPU-)resources before.
In order to access the LRZ AI Systems, make sure you are connected to the Munich Scientific Network (MWN), then use SSH to contact login.ai.lrz.de. Provide your LRZ Linux Cluster credentials to log in.
For example, a user with an LRZ Linux Cluster account xyyyyzz and a workstation connected to the Munich Scientific Network (MWN), can perform the login using the OpenSSH command line tools as follows:
ssh login.ai.lrz.de -l xxyyyzz
This login node is meant for preparing and submitting your jobs to one or more of the available resources (see General Description and Resources), which are managed by the Slurm Workload Manager. As an example, a comprehensive guide on submitting jobs via Slurm on SuperMUC-NG can be consulted here https://doku.lrz.de/display/PUBLIC/Job+Processing+with+SLURM+on+SuperMUC-NG.
Alternatively, a web-based frontend is accessible from within the Munich Scientific Network (MWN) at https://datalab3.srv.lrz.de and the same credentials can be used for login. For futher details see Interactive Web Servers on the LRZ AI Systems.