Skip to content

1. Jupyterlab Setup in Data Scientists's own Laptop

This page describes how you can setup your own laptop or desktop's Jupyterlab for accessing InfinStor services.

1.1. Prerequisites

  • Install Node.js (npm) 14.x using these instructions
  • Install aws cli for your operating system
    • after installation, run aws configure to setup AWS access key id and access key secret to allow access to AWS. This configuration is needed to access the mlflow artifact S3 bucket.
  • Download and install miniconda3 for your operating system.
    • The 'JupyterLab' server will be setup in your laptop or desktop in a conda environment.
  • Create a new conda environment named infinstor. Open a new terminal and run the following commands:

    conda activate base
    conda create --name infinstor python=3.8
  • Activate this new 'conda environment' infinstor

    • run the command conda activate infinstor
    • once activated, the prompt has the word infinstor in it, similar to what is shown below. The prompt will vary based on the operating system used, but will contain the word infinstor if the conda environment is successfully activated. Verify this before proceeding further.
      • (infinstor) user@laptop $
  • Next, install pre-requisites as shown below in this infinstor conda environment

    conda activate infinstor
    pip install --upgrade mlflow
    conda install -c conda-forge -y ipywidgets
    pip install --upgrade jupyterlab
    jupyter labextension install @jupyter-widgets/jupyterlab-manager
    pip install --upgrade aiohttp

1.2. InfinStor Python Packages

The InfinStor serivce is augmented by three python packages that are distributed through Pypi and installing using pip. These packages need to be installed in the Data Scientist laptop or desktop, the Cloud VM where transforms are executed and Cloud VMs that are part of distributed computing frameworks such as EMR

  • infinstor: The InfinStor SDK. It is installing using 'pip install infinstor'
  • infinstor-mlflow-plugin: The MLflow plugin required for MLflow tracking, projects, transforms etc.
  • jupyterlab_infinstor: This package supports the InfinStor jupyterlab sidebar. This is loaded in the jupyterlab server process and needs to be installed in that python environment using the command 'pip install jupyterlab_infinstor'

The following commands uninstall older versions and install new versions of the infinstor pip packages.

conda activate infinstor
pip uninstall -y infinstor
pip install infinstor
pip uninstall -y infinstor-mlflow-plugin
pip install infinstor-mlflow-plugin
pip uninstall -y jupyterlab_infinstor
pip install jupyterlab_infinstor

1.3. InfinStor npm packages

The InfinStor service is aided by the jupyterlab sidebar npm package jupyterlab-infinstor. This package is installed in the machine where the jupyterlab server runs.


This package is compatible with version 2.x and 3.x of jupyterlab. Ensure that you are running version 2.x or 3.x of jupyterlab by typing the command jupyter --version and looking for the line jupyter lab. The jupyter lab version should be greater than 2.x

conda activate infinstor
jupyter labextension install jupyterlab-infinstor

1.4. Start Jupyterlab server

  • Start Jupyterlab server using the command below.

    • Add --ip= to the command if you want the JupyterLab server to be accessible from other computers.
    • Ask your Infinstor platform administrator for the value of MFLOW_TRACKING_URI. It is usually similar to infinstor://mlflow.infinstor.<your_domain>. Use this value in the export command below

      export MLFLOW_TRACKING_URI=<value from administrator>
      jupyter lab --log-level INFO --no-browser 
  • After successful startup of JupyterLab, will see an URL similar to below in the output. Use this URL in your browser to access your JupyterLab server

    To access the server, copy and paste one of these URLs: