Manage Azure ML Resources with Python SDK

Azure Machine Learning SDK v2 for Python is a client library for managing Azure ML resources. The guide shows how to connect with MLClient and DefaultAzureCredential, use workspace configuration, and operate on workspaces, data assets, models, compute, jobs, pipelines, environments, and datastores.

Workspace operations include creating and listing workspaces. Data examples register file and folder assets, while model examples register and list models. Compute guidance covers creating and listing clusters, including minimum and maximum instances and idle scale-down. Job patterns include submitting a command job, streaming its logs, and building a multi-step pipeline from components. Environment examples create a custom environment from an image and conda file.

The SDK exposes resource operations through MLClient properties such as workspaces, jobs, models, data, compute, environments, datastores, and components. Practices include versioning data, models, and environments; using environments for reproducible training; streaming job logs; registering models after successful training; tagging resources; and scaling idle compute down to manage costs.

Use this SDK when building Python workflows that manage Azure Machine Learning resources or submit training work to Azure ML compute. It supports resource registration, command jobs, pipeline orchestration, and monitoring job output.

The examples use placeholder names, paths, compute settings, and training inputs. Adapt them to the target workspace, registered assets, available compute, and environment. The source demonstrates management operations and does not guarantee that an example job or pipeline will run unchanged in a particular project.

Connection inputs include a subscription ID, resource group, workspace name, and credential, or a configuration file available in the current or parent directory. Operations take resource names, versions, paths, asset types, compute parameters, code locations, commands, inputs, and environment references.

Outputs include workspace and asset records, job submissions and logs, compute resources, pipeline runs, environment registrations, and datastore information. The client can stream a job's logs and return the job URL after submission.

Install the package with:

pip install azure-ai-ml

The library integrates with Azure Machine Learning and Azure Identity. The examples use Azure ML datastores, URI file and folder data assets, custom models, AmlCompute, command jobs, pipeline components, environments, and the default datastore.

Python developers and machine learning engineers can use the SDK to automate workspace administration and organize training workflows. Platform teams can use its resource operations for versioned data and models, reproducible environments, job monitoring, and compute lifecycle management.