When you work in JupyterLab, you may find it convenient to customize the JupyterLab interface by adding a conda environment to the list of available kernels. To accomplish this, complete the procedures described below.

Note: Some of the command lines below are too long to be formatted as one line, so they
are broken with a backslash (\).

Step 1: Activate the Conda Environment

1. Launch an interactive job on a Cabeus compute node, such as mil_a100 or gh200:

   cfe01 % qsub -I -q gpu_devel -lselect=1:model=mil_a100:mem=64GB:ncpus=16:ngpus=1 \ 
   -l walltime=1:00:00

2. Load the appropriate miniconda module:
   • On the A100 nodes:

     module use -a /swbuild/analytix/tools/modulefiles
     module load miniconda3/v4
     

   • On the Grace Hopper nodes:

     module use -a /swbuild/analytix/tools/modulefiles
     module load miniconda3/gh2

3. View available conda environments:

   conda info –envs

4. Activate the environment you want to add to the JupyterLab kernel list:

   source activate conda_environment
   

   where conda_environment is environment name (for example, mamba_ssm).

Step 2: Add the Conda Environment to Available Kernels in JupyterLab

1. Register the conda environment with Jupyter:

   python -m ipykernel install --user --name=conda_environment \ 
   --display-name "Python (conda_environment)"

2. Start JupyterLab using a Unix socket:

   jupyter lab --sock /tmp/user.sock --no-browser

3. Open a new terminal and create an SSH tunnel from your local workstation:

   ssh -l username -o "StrictHostKeyChecking ask" \ 
   -L 8080:/tmp/user.sock -o ProxyJump=sfe,front_end compute_node 

   where username is your NAS user name; front_end is the front end system you submitted from (for example, pfe22); and compute_node is the actual node name (for example, r101i2n6).

4. Access JupyterLab in your browser:
   Open http://localhost:8080
5. Select your conda kernel in JupyterLab:
   In JupyterLab, open or create a notebook; navigate to Kernel > Change Kernel; and select your conda environment (for example, Python conda_environment).

Now you are all set to run JupyterLab with your custom conda environment.