This is JupyberLab online service supporting GPU cards, user can log in to our web-portal (https://lab-ubda.polyu.edu.hk) and quickly run interactive block of code in the Jupyter Notebook directly.
After logging into JupyterLab, users can load or switch their python environment (ipykernel) and access their UBDA HPC data drives in the JupyterLab dashboard web-interface. After use, users simply need to shut down the container (Files/Hub Control Panel/Stop My Server) to release the resources for others. The default number of GPU cards is 1, and the maximum wall-time of the service is 48 hours. To check the available free-seats, log into the HPC platform (ubdaplatform.polyu.edu.hk) and type "lab-ubda" to quote.
The UBDA user home directory "/ubda/home/NetID" is automatically mounted to the JupyterLab container home directory "/home/NetID". Therefore, users can upload/download their scripts, datasets and results through the JupyterLab dashboard web-interface or through the SFTP client software (such as WinSCP, FileZilla).
To switch python environments in JupyterLab, users should use the Conda manager to create the required new environments and libraries. Make sure that the iPhython Kernel (ipykernel) is also be installed in the environment. Then register those newly created environments in the kernelspec under JupyterLab. After that you can select the newly created kernel in JupyterLab start-up dashboard.
Below is an example:
conda create --name py38 python=3.8.12
conda activate py38
conda install ipykernel
python -m ipykernel install --user --name py38 --display-name "Python (py38)"
For more information, please email to ubdae@polyu.edu.hk
