How to Customize JupyterLab with Anaconda Environments

1. Open a Delta terminal (Open OnDemand shell or direct SSH).

2. Load anaconda_<cpu, gpu, mi100>. (Pick which one you want to use as your base installation, the example uses cpu.)

   [arnoldg@dt-login03 ~]$ module load anaconda3_cpu
   

3. Initialize your default login shell to use conda environments.

   [arnoldg@dt-login03 ~]$ conda init bash
   ...
   [arnoldg@dt-login03 ~]$ bash
   (base) [arnoldg@dt-login03 ~]$
   

   • If you see error messages when you run conda init bash, ctrl+c through them and continue. As long as conda added code to the end of your .bashrc (or similar for other shells), things will work properly.

   • After you run conda init bash, you will not need to load anaconda3_cpu (or gpu) modules again. Just use your new custom environment.

4. Start a new shell with bash or a new terminal or login session on Delta. The prompt should show that you are within the conda environment you chose. If you want to change environments later (say to anaconda3_mi100) you can edit your .bashrc and do another conda init bash with that new module loaded.

   (base) [arnoldg@dt-login03 ~]$
   

5. Create your new custom environment by making a new empty environment or cloning your chosen module:

   New empty environment (setup time ~10 minutes)New clone of chosen module (setup time ~30 minutes)

   Install Jupyter into the environment to use it with Open OnDemand. This option adds about 150 Python modules to your environment and requires about 1.3 GB of space in your $HOME.

   1. Create a new conda environment. (The example environment name is mynewenv.)

      Note

      If you will be making custom environments for more than one partition type (cpu, gpu, mi100), it may be helpful to include that metadata in the name of your environment.

      (base) [arnoldg@dt-login03 ~]$ conda create --name mynewenv
      
      Collecting package metadata (current_repodata.json): done
      Solving environment: done
      
      ## Package Plan ##
      
        environment location: /u/arnoldg/.conda/envs/mynewenv
      
      Proceed ([y]/n)? y
      
      Preparing transaction: done
      Verifying transaction: done
      Executing transaction: done
      #
      # To activate this environment, use
      #
      #     $ conda activate mynewenv
      #
      # To deactivate an active environment, use
      #
      #     $ conda deactivate
      

   2. Activate the new environment.

      (base) [arnoldg@dt-login03 ~]$ conda activate mynewenv
      

   3. Install Jupyter into the new environment.

      (mynewenv) [arnoldg@dt-login03 ~]$ conda install jupyter
      
      Collecting package metadata (current_repodata.json): done
      Solving environment: done
      
      ## Package Plan ##
      
        environment location: /u/arnoldg/.conda/envs/mynewenv
      
        added / updated specs:
          - jupyter
      
      
      The following NEW packages will be INSTALLED:
      
        _libgcc_mutex      pkgs/main/linux-64::_libgcc_mutex-0.1-main None
        _openmp_mutex      pkgs/main/linux-64::_openmp_mutex-5.1-1_gnu None
        anyio              pkgs/main/linux-64::anyio-3.5.0-py310h06a4308_0 None
        argon2-cffi        pkgs/main/noarch::argon2-cffi-21.3.0-pyhd3eb1b0_0 None
      ...
      
      Proceed ([y]/n)? y
      

   4. Verify Jupyter installs.

      (mynewenv) [arnoldg@dt-login03 ~]$ conda list | grep jupyter
      jupyter                   1.0.0           py310h06a4308_8
      jupyter_client            7.3.5           py310h06a4308_0
      jupyter_console           6.4.3              pyhd3eb1b0_0
      jupyter_core              4.11.1          py310h06a4308_0
      jupyter_server            1.18.1          py310h06a4308_0
      jupyterlab                3.4.4           py310h06a4308_0
      jupyterlab_pygments       0.1.2                      py_0
      jupyterlab_server         2.15.2          py310h06a4308_0
      jupyterlab_widgets        1.0.0              pyhd3eb1b0_1
      

      (mynewenv) [arnoldg@dt-login03 ~]$ conda list | wc -l
      152
      

      (mynewenv) [arnoldg@dt-login03 ~]$ du -sh $HOME/.conda/envs/mynewenv
      1.3G    /u/arnoldg/.conda/envs/mynewenv
      

6. Start an Open OnDemand Jupyter session.

7. In JupyterLab, you can hover over items in the Launcher to see which environment will be used; select the one you want for this session.

   

   You can also change your kernel to match if you are opening a notebook from a different environment.