title | nav_order |
---|---|
Getting started |
2 |
The main JupyterHub server can be reached at jupyterhub.hpc.kuleuven.be. You will be asked to authenticate with your central KU Leuven credentials.
For vsctmp
accounts, the server address is jupyterhub2.hpc.kuleuven.be, where you will be prompted for a
username and password.
After logging in, you will be able to launching a Jupyter Notebook App by selecting a job profile and a compute project.
NOTE: This assumes that you have at least one compute project (with credits) associated with your VSC account.
The list of job profiles allows you to choose between different numbers of cores, numbers of GPUs, and walltime durations. Keep in mind that 'heavier' jobs profiles may lead to queue times when the cluster (Genius) is being heavily used.
After submitting your request, you will need to wait a moment until the job
has started on the cluster. Currently, the best way to check this is to
login to Genius with your VSC account and execute the showq
command.
Then, you can navigate to the "Home" tab and access the Dashboard of your Jupyter Notebook App via the "My Server" button.
In the Dashboard, you can either open one of your existing Notebook documents (if any) or create a new one. For new Notebook documents, you will have the choice between ones with a Python kernel or with an R kernel. These are then powered by the CPU cores (and possibly GPU cards) allocated to your compute job on the cluster.
A detailed explanation and step-by-step tutorial of the JupyterHub can be found here. The two example notebooks used in the tutorial can be found here. The example conda environments created as on the software installation page can be used for these notebooks.