JupyterLab
Introduction
JupyterLab is a web-based user interface for Project Jupyter which allows interaction with documents and activities such as Jupyter notebooks, text editors, and terminals.
JupyterLab should not be started on the login node. Instead it should be run on a compute node. This can be done using the following approach.
Create a password
The following needs to be done only once.
On the login node of the HPC cluster, run the commands below. You will be prompted to generate a password, which you will use later to access JupterLab.
module add JupyterLab/4.0.5-GCCcore-12.3.0
jupyter lab --generate-config
jupyter lab passwordSubmit batch job
Use the following script as a template and submit it via sbatch as a normal job.
#!/bin/bash
#SBATCH --job-name=jupyter-notebook
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=8
#SBATCH --mem=20G
#SBATCH --gres=gpu:1
#SBATCH --qos=hiprio
#SBATCH --partition=gpu
#SBATCH --time=3:00:00
# Get tunneling information
XDG_RUNTIME_DIR=""
node=$(hostname -s)
port=`python -c 'import socket; s=socket.socket(); s.bind(("", 0)); print(s.getsockname()[1]); s.close()'`
# Tunneling instructions
echo -e "
On your local computer, create an SSH tunnel:
ssh -N -f -L ${port}:${node}:${port} ${USER}@curta.zedat.fu-berlin.de
On your local computer, point a browser to:
http://localhost:${port}
and enter the password you created earlier.
"
# Load modules and run JupyterLab
module add JupyterLab/4.0.5-GCCcore-12.3.0
jupyter lab path
jupyter-lab --no-browser --port=${port} --ip=${node}