Singularity Kernel

The are two steps for using a singularity image as custom kernel on JupyterLab

1. Prepare the image with python packages
2. create a kernel file for launching the container

Prepare the container image

The container must have installed the jupyter-console and ipykernel.

The followin snipet shows an example of a container def image:

Bootstrap: docker
From: jupyter/base-notebook:latest

%help
    JupyterLab.

%environment
    export LC_ALL=C

%post
    . /.singularity.d/env/10-docker*.sh

%post
    export DEBIAN_FRONTEND=noninteractive

    apt update

    pip install jupyter-console
    pip install ipykernel

JupyterLab kernel file

The following snippets can be used as kernel.json for yours custom singularity image.

Generic (CPU nodes)GPU nodes

{
  "argv": [
    "singularity",
    "exec",
    "<image-path>",
    "python3",
    "-m",
    "ipykernel_launcher",
    "-f",
    "{connection_file}"
  ],
  "display_name": "<name>",
  "language": "python",
  "metadata": {
    "debugger": true
  }
}

{
  "argv": [
    "singularity",
    "exec",
    "--nv",
    "<image-path>",
    "python3",
    "-m",
    "ipykernel_launcher",
    "-f",
    "{connection_file}"
  ],
  "display_name": "<name>",
  "language": "python",
  "metadata": {
    "debugger": true
  }
}