Jupyter Notebooks, JupyterLab, TensorFlow, PyTorch

We recommend to use Open OnDemand https://apps.hpc.ntnu.no . There are several alternatives to start Jupyter.

Start Jupyter from Anaconda environment module

Anaconda Environment Module has preinstalled Jupyter and other python modules. You can change "Environment Setup" by adding this command:

module load Anaconda3/2022.10

Use python environment module and your locally installed Jupyter

You will use one of preinstalled Python environment modules. And you need to installed install Jupyter to your home directory. Example commands:

module load Python/3.11.3-GCCcore-12.3.0
pip install --user notebook
pip install --user jupyterlab

These python modules will be installed to your cluster home directory: /cluster/home/USERNAME/.local/lib/python3.11/
You will need to update "Environment Setup" to load the same environment module:

Use your own setup, virtual environment, ...

Do you already have your environment with Jupyter and other python modules?
Steps for Python.
Or you can leave "Environment Setup" empty but you will need to configure your cluster shell environment (.bashrc file). So Open OnDemand can start command: "jupyter notebook" or "jupyter lab".

Example with Virtual Environment:

Run Jupyter Notebook and TensorFlow (CUDA, GPU)

There are several TensorFlow Environment Modules installed on IDUN:

$ module avail TensorFlow
TensorFlow/2.11.0-foss-2022a-CUDA-11.7.0
TensorFlow/2.11.0-foss-2022a
TensorFlow/2.13.0-foss-2023a

This is example with GPU/CUDA:

Check how many GPUs available:

Run Jupyter Notebook and PyTorch

There are several PyTorch Environment Modules installed on IDUN:

$ module avail PyTorch
PyTorch/1.12.0-foss-2022a-CUDA-11.7.0
PyTorch/1.12.1-foss-2022a-CUDA-11.7.0
PyTorch/1.13.1-foss-2022b
PyTorch/2.0.1-foss-2022a

Example with torchvision and GPU:

Check available GPU: