EnCortex-Framework

Setup

In this section, we cover setting up your AzureML environment to run EnCortex. AzureML empowers data scientists and developers to build, deploy, and manage high-quality models faster and with confidence. AML also helps accelerate time to value with industry-leading machine learning operations (MLOps), open-source interoperability, and integrated tools. Hence, we recommend running EnCortex on Azure ML.

In the next few pages and sections, we cover:

  1. We cover setting up your AzureML studio.
  2. Setup AzureML notebooks
  3. Installing EnCortex
  4. Tutorials

Please proceed next if this is your first time using EnCortex or you can head onto the Advanced section to start developing with EnCortex.

(setup/azureml)=

AzureML(AML)

Setup your AzureML studio. To do this, you’ll need access to your resource group in Azure. In this setup, we cover setting up AzureML through the portal, but other setup methods can be found here.

  1. Sign in to your Azure Portal -> + Create a Resource -> Search “Azure Machine Learning” and follow the instructions to setup your AzureML studio.
  2. From the Overview tab on the sidebar of your now created Azure ML workspace, click on Download config.json button to download your AML config.json file.
  3. Head onto ml.azure.com and sign in to your account to view the AML studio. Alternatively, you can directly access your studio from the Overview tab by clicking on the Studio Web URL.

(setup/azureml/compute)=

Compute Resources

Now that you’ve setup your AML studio, we now move on to creating a compute resource. A compute resource is VM-like entity that allows you to run code. Having a compute resource is mandatory to run any EnCortex experiment. Click on the Compute tab from the sidebar inside your AML studio and click on + New button of the type of compute resource you want to spin up. We recommend selecting the following compute resources based on requirements:

  1. For simple EDA and small workloads, we recommend a CPU Virtual Machine. From the sidebar, Compute -> Compute Instance -> Enter Compute Name -> CPU -> STANDARD_D13_V2
  2. For larger workloads involving multiple parallel experiments, we recommend a GPU Virtual Machine. From the sidebar, Compute -> Compute Cluster -> Enter Cluster name -> Dedicated tier -> Standard_DS3_v2 .

You’re now ready to setup AML notebooks.

(run/azureml-notebooks)=

AzureML notebooks

AzureML notebooks provides a jupyter notebook interface to play and explore EnCortex. To run EnCortex on AzureML notebooks, follow the instructiosn below:

  1. Make sure your compute is instantiated(see previous section)
  2. From your studio homepage, navigate to the Notebooks tab and click on the + icon to create a notebook.
  3. Connect to the above created compute on the top of notebook through the dropdown titled Compute Instance and select the Python 3.8 - AzureML environment from the far-right dropdown.
  4. Run the the installation instructions mentioned in Setup through Pip and you’re ready to run EnCortex.

Installation

Your AML notebook environment is now set. Now, in order to install EnCortex, we recommend the following approach:

(setup/pip)=

Approach 1: Pip(AML Notebook)

Accessible by @Microsoft email aliases only.

  1. Download the latest wheel files from here(Go to the wheels folder)
  2. Create a folder on AML called encortex_wheels through the UI.
  3. Upload the downloaded wheels to the encortex_wheels folder.
  4. Run the following command in an AML notebook cell:
!pip install encortex_wheels/*.whl

To verify installation, run the following code in your AML notebook cell and it should give out a string like 0.1.23:

import encortex
print(encortex.__version__)

You’re now ready to use AzureML notebooks. To run pre-defined scenarios, head on to Tutorials and follow the AML Notebook instructions for each of the scenarios there.

This site is open source. Improve this page.