This page covers the administrative side of LRZ's AI Systems Service and is most relevant to Principal Investigators who would like to enable their research group with access to modern Compute Resources. There is further information for Master Users on how they manage user access for the group. Regular Users may proceed reading 1.1 AI Systems Login detailing how to login and are also recommended to work through the pages Enroot and Slurm as they explain how to use the AI Systems.

Prerequisites

Prospective users of the AI Systems must demonstrate that their intended work or use case has a clear research focus. This can be done by associating their user account with an LRZ project that is AI-enabled. The following sections explain the different procedures to be granted access depending on whether a projects already exists and if yes whether it is already AI-enabled. Projects may become inactive after a year and need to be renewed annually. For this, the Principal Investigators will receive automatically a reminder and Email with further instructions before the project expires.  

Additional information for MCML.

User Accounts and AI-enabled Projects

LRZ User Account

A valid LRZ user account is required to access the AI Systems. An overview of institutions that already provide LRZ user accounts and the corresponding services can be found on the following page: Accounts for LRZ Services.

Master users are encouraged to reuse existing LRZ user IDs whenever possible but may create a new one if a user does not yet have a valid account. More information can be found here: Creating new User Accounts

Access through an Existing Project

For individual users: If you know of an existing AI-enabled LRZ project that aligns with your research, contact the project’s master user(s) to request that your account be granted access to the AI Systems. (If you do not yet have an LRZ account, the master user can create one for you within the project.)

For master users: If your LRZ project is not yet AI-enabled but could qualify, submit a service request to the LRZ Servicedesk using the following form. (or choose “Predefined service request” → “AI Systems Project Activation/Update”). Based on the information provided, LRZ will assess whether the project is suitable for AI System access. (Please note: in general, older projects with many members do not qualify for AI enablement.)

Apply for a New AI-enabled Project

Prospective master users can apply for a new AI-enabled project as described in Application for an LRZ Project by completing, signing, and submitting the provided PDF form. (In the section “Sonstige Bemerkungen”, include “AI Systems”.)

Once the LRZ project has been created, it must be AI-enabled. To do so, submit a follow-up service request to the LRZ Servicedesk using the following form. (or select “Predefined service request” → “AI Systems Project Activation/Update”).

Alternatively, access can be requested via the BayernKI initiative: https://www.ki-in-bayern.de/

Managing User Access within a Project

Access to the LRZ AI Systems is controlled through membership in the <projectID>-ai-c group. Designated Master Users of AI-enabled LRZ projects manage individual user access by adding or removing accounts from the -ai-c group via the IDM 2 Portal.

Membership in this group provides: Access to LRZ AI Systems, use of general system resources, VPN access to the Munich Scientific Network (MWN). For detailed instructions, see How to add or remove members from groups?

The responsibility for managing access lies with the Master User. There is no limit to the number of users that can be added to the -ai-c group.

Usage Guidelines

Guidelines for Development and Testing

Please note that LRZ AI Systems provide limited computational resources and are not primarily intended for development, prototyping, or debugging. Programs and scripts used for production jobs should be tested on other systems beforehand whenever possible.

If you choose to perform development or testing on the AI Systems, keep your resource usage to the minimum necessary. Preferably by using https://login.ai.lrz.de (restricted to single GPUs/MIG devices; see section 6. Interactive Apps) or by requesting appropriately sized interactive allocations (see section 5.1 Slurm Interactive Jobs). In general, minimize idle time and avoid inefficiencies to ensure fair resource availability for all users.