Fair Use Policy

JupyterHub on GPULab should be used for interactive development and short training runs. Therefore, the following fair use guidelines have been implemented:

Resource usage

The following resource usage is considered ‘fair use’ for JupyterHub jobs:

• 1 GPU, or 2 if you need to run concurrent code
• 4 CPU cores
• 40GB CPU RAM

If you do need to exceed these ‘fair use’ limits, you will need to explain why you do so. Consider running it directly and non-interactively on GPULab instead.

Duration of JupyterHub sessions

JupyterHub sessions should only last as long as you are interactively using them.

Code execution should not exceed more than a few hours, otherwise it should be run directly and non-interactively on GPULab.

By default, we set the maximum duration of a JupyterHub session to 24 hours. If you need to exceed this ‘fair use’ limit, you will need to explain why you do so. It is not possible to exceed a maximum duration 72 hours.