GPU Infrastructure

• Getting Started
• Working with Docker
• Docker and SSH
• Nice Scripts
• BASH Configuration
• Docker Volumes for Datasets
• Optimize STT training with Huggingface
• SSH Configuration
• Tips and Tricks

Important notes

Resource allocation

Only allocate resources as you need it. E.g. avoid allocate GPU’s if you only run CPU workload.

Docker containerization

Always containerize your tasks inside a Docker container.

Fair use

For all jobs to be processed, resources should not be reserved if nothing is running. So whenever possible use the basic submit command, so the resources get freed after the script is finished. Also batch jobs can be used. Furthermore you can set time limits with the –time option to make sure a job is not running to long without any work.

Port Allocations

If you have any services running within your job, where you want to connect to (e.g. Jupyter Notebook) you need to specify a port for the communication. In order there are no conflict between different users, every user has his own port range to work with. The table below shows which ports are allocated for which user.

Allocations: Link to XLS