aws-sagemaker-custom-jupyter-kernel

Introduction

There are several built-in Juptyer Kernels ready-to-use when working on an AWS Sagemaker Jupyter Notebook Instance. They are easy to use but generally not convenient to customize:

• If you want to install new packages to the built-in Juptyer Kernels, there is no guarantee that the new packages are compatible with the existing ones.
• Even if you can install new packages to the built-in Juptyer Kernels, you will lose the modified/custom kernels when restarting the Notebook Instance.

This repo provides a couple of easy-to-use template scripts to help you set up a custom jupyter kernel on a AWS Sagemaker Jupyter Notebook Instance.

Build/Start Custom Jupyter Kernel

When first create the Jupyter notebook instance.

1. Need to edit environment.yml to specify your custom environment if you want to build a different Python Kernel to this example in this repo.

2. Run the command below to build the custom Python Kernel named Conda_my-custom-jupyter-kernel.

   ./build_env.sh

(Optional): Every time re-start the Jupyter Notebook instance, run the command below to add the custom Python Kernel.

./start_env.sh

To use the custom Kernel, create a new Jupyter notebook, and select Conda_my-custom-jupyter-kernel as the Python Kernel.