How to build Anaconda virtual environment used in Azure Machine Learning and link with Jupyter

Introduction

Build a machine learning environment using Jupyter notebooks on Azure Machine Learning It is a summary of the methods for.

It is assumed that Azure Machine Learning Studio and Jupyter notebooks can be started. Use Anaconda to build and switch virtual environments.

What is Anaconda?

A collection of Python packages for data science --Supports over 1,500 packages --Supports both GUI and command line --Can be operated using the conda command

By using Anaconda, it is possible to easily build and switch the development environment without taking time to install the library.

Reference: https://www.creativevillage.ne.jp/72837

procedure

1. Launch Terminal on jupyter notebooks

2. Anaconda virtual environment setup and Jupyter integration

Follow the steps below in Terminal on Jupyter.

2.1 Building a virtual environment with conda

$ conda create --name my_notebook_env python=3.7 -y

-name my_notebook_env: The name of the virtual environment. You can set whatever you like. python = 3.7: Python version used in the virtual environment -y: Automatically respond to confirmation items with yes

• Display a list of conda environments: jupyter kernelspec list

2.2 Enabling the virtual environment

$ conda activate my_notebook_env

• To disable: conda deactivate

2.3 Installing libraries in a virtual environment

Install the Azure Machine Learning SDK for Python.

$ pip install --upgrade azureml-sdk[explain,automl,interpret,notebooks]

See below for details on the installed components. Azure Machine Learning SDK for Python

2.4 Adding a virtual environment kernel to Jupyter

Execute the following command with the virtual environment activated.

$ ipython kernel install --user --name=my_notebook_env --display-name=my_notebook_env

--user: Installed in the current user environment --name: Specify the name of the kernel --display-name: The display name of the kernel. This time, it is the same as the kernel name, but you can set an easy-to-understand display name here.

• Jupyter kernel list display: jupyter kernelspec list
• Kernel removal: jupyter kernelspec uninstall my_notebook_env

3. Kernel switching on Jupyter notebooks

3.1 Jupyter reload

In order to reflect the change, if there is an open node book, reload it by reloading the browser.

3.2 Kernel switching

Select Kernel → Change kernel on the node book, and select the virtual environment (my_notebook_env) created this time.

It is successful if the display on the upper right of the node book screen is switched correctly.

Summary

You can easily build and switch the development environment by using Anaconda. By using Azure Machine Learning, you can perform machine learning in the familiar Jupyter environment.

reference

What is Azure Machine Learning Azure Machine Learning SDK for Python Easy usage memo of Anaconda (conda)