Machine Learning environment in Azure

Objective

This document contains basic details on the resources required for setting up the Machine Learning environment in Azure

Although there are number of ways we can set up the environment, basically we have emphasised on doing ML by using Azure Machine Learning studio.

1. Azure Machine Learning Studio
2. Azure Databricks
3. Machine Learning with Azure Functions
4. Azure Chatbots

Azure Resources

Organising your cloud-based resources is critical to securing, managing, and tracking the costs related to your workloads. To organize your resources, define a management group hierarchy, follow a well-considered naming convention and apply resource tagging.

Subscription - An Azure subscription is linked to a single account which is used for billing purposes. Within the subscription, resources can be provisioned as instances of the many Azure products and services. Each subscription has limits or quotas on the amount of resources you can create and use. Organizations can use subscriptions to manage costs and the resources that are created by users, teams, or projects. Free Azure accounts can be converted to pay-as-you-go accounts.

Resource groups - A resource group is a logical container into which Azure resources like azureml, databases,  storage accounts, key vaults are deployed and managed.

Resources - Resources are instances of services that you create, like virtual machines, storage, SQL databases , Azure Machine Learning Studio etc.

Azure Machine Learning Studio 

Once we have created a Azure ML Machine Learning Resource. We need to provision an AzureML workspace for our Machine Learning Activities

Azure Machine Learning workspace -

The workspace is the top-level resource for Azure Machine Learning, providing a centralised place to work with all the artifacts you create when you use Azure Machine Learning. The workspace keeps a history of all training runs, including logs, metrics, output, and a snapshot of your scripts. You use this information to determine which training run produces the best model.

Once you have a model you like, you register it with the workspace. You then use the registered model and scoring scripts to deploy to Azure Container Instances, Azure Kubernetes Service, or to a field-programmable gate array (FPGA) as a REST-based HTTP endpoint. You can also deploy the model to an Azure IoT Edge device as a module.

Workspace Hierarchy -

 The above diagram shows the following components of a workspace:

• A workspace can contain Azure Machine Learning compute instances, cloud resources configured with the Python environment necessary to run Azure Machine Learning.
• User roles enable you to share your workspace with other users, teams, or projects.
• Compute targets are used to run your experiments.
• When you create the workspace, associated resources are also created for you.
• Experiments are training runs you use to build your models.
• Pipelines are reusable workflows for training and retraining your model.
• Datasets aid in management of the data you use for model training and pipeline creation.
• Once you have a model you want to deploy, you create a registered model.
• Use the registered model and a scoring script to create a deployment endpoint.

Associated resources

When you create a new workspace, it automatically creates several Azure resources that are used by the workspace:

Azure Storage account: Is used as the default datastore for the workspace. Jupyter notebooks that are used with your Azure Machine Learning compute instances are stored here as well. 

Important By default, the storage account is a general-purpose v1 account. You can upgrade this to general-purpose v2 after the workspace has been created. Do not enable hierarchical namespace on the storage account after upgrading to general-purpose v2.

To use an existing Azure Storage account, it cannot be a premium account (Premium_LRS and Premium_GRS). It also cannot have a hierarchical namespace (used with Azure Data Lake Storage Gen2). Neither premium storage or hierarchical namespaces are supported with the default storage account of the workspace. You can use premium storage or hierarchical namespace with non-default storage accounts.

Azure Key Vault: Stores secrets that are used by compute targets and other sensitive information that's needed by the workspace.

Basic Architecture -