Machine Learning operations brings data science to the world of devops. Data scientists create models on their workstations. MLOps adds automation, validation and monitoring to any environment including machine learning on kubernetes. In this session you hear about latest developments and see it in action.
5. Cloud-hosted pipelines for Linux, Windows and
macOS, with unlimited minutes for open source
Azure Pipelines
Any language, any platform, any cloud
Build, test, and deploy Node.js, Python, Java, PHP, Ruby,
C/C++, .NET, Android, and iOS apps. Run in parallel on
Linux, macOS, and Windows. Deploy to Azure, AWS,
GCP or on-premises
Extensible
Explore and implement a wide range of community-
built build, test, and deployment tasks, along with
hundreds of extensions from Slack to SonarCloud.
Support for YAML, reporting and more
Best-in-class for open source
Ensure fast continuous integration/continuous delivery
(CI/CD) pipelines for every open source project. Get
unlimited build minutes for all open source projects with
up to 10 free parallel jobs across Linux, macOS and
Windows
Containers and Kubernetes
Easily build and push images to container registries like
Docker Hub and Azure Container Registry. Deploy
containers to individual hosts or Kubernetes.
10. Machine Learning on Azure
Domain specific pretrained models
To reduce time to market
Azure
Databricks
Machine
Learning VMs
Popular frameworks
To build advanced deep learning solutions
TensorFlowPytorch Onnx
Azure Machine
Learning
LanguageSpeech
…
SearchVision
Productive services
To empower data science and development teams
Powerful infrastructure
To accelerate deep learning
Scikit-Learn
PyCharm Jupyter
Familiar Data Science tools
To simplify model development
Visual Studio Code Command line
CPU GPU FPGA
11. What is the E2E ML lifecycle?
• Develop & train model with reusable ML pipelines
• Package model using containers to capture runtime dependencies for inference
• Validate model behavior—functionally, in terms of responsiveness, in terms of regulatory compliance
• Deploy model—to cloud & edge, for use in real-time/streaming/batch processing
• Monitor model behavior & business value, know when to replace/deprecate a stale model
Train Model Validate Model Deploy ModelPackage Model Monitor Model
Retrain Model
12. App developer
using Azure DevOps
MLOps Workflow
Build appCollaborate Test app Release app Monitor app
Model reproducibility Model retrainingModel deploymentModel validation
Data scientist using
Azure Machine Learning
13. MLOps with Azure Machine Learning
Train model Validate
model
Deploy
model
Monitor
model
Retrain model
Model reproducibility Model retrainingModel deploymentModel validation
Build appCollaborate Test app Release app Monitor app
Azure DevOps integration
App developer
using Azure DevOps
Data scientist using
Azure Machine Learning
15. Azure Services Supporting MLOps
Infrastructure as Code CI/CD Testing / Release / Monitoring
• Azure Resource Manager Templates
• Azure ML Python SDK & CLI
• Azure SDK’s
• Azure DevOps Pipelines
• Azure ML Training Services
• Azure Repos / GitHub
• Azure Boards
• Azure DevOps for automated testing
• R - Runit and testthat
• Python - PyUnit, pytest, nose, …
• Azure ML & MLFlow Tracking
• Azure Data Prep SDK (analyse/profile)
• Azure ML Model Management
(Instrumentation, Telemetry)
• Azure Monitor for app telemetry
16. First Class Model Training Tasks
CI pipeline captures:
1. Create sandbox
2. Run unit tests and code quality checks
3. Attach to compute
4. Run training pipeline
5. Evaluate model
6. Register model
18. Automated Deployment
CD pipeline captures:
1. Package model into container
image
2. Validate and profile model
3. Deploy model to DevTest (ACI)
4. If all is well, proceed to rollout to
AKS
Everything is done via the CLI
Every great organization needs an enduring mission – a noble goal, a challenge worth the effort. Not long after Satya became CEO, we announced ours. We found it hiding right in plain sight. It connects back to our earliest days as a company.
We will empower every person and every organization on the planet to achieve more.
Empower is a word that gets thrown around quite a bit in our industry. When we talk about it we mean with the right tools – anyone can do anything. “On the planet,” is the most interesting piece – we’re at a unique moment in history – a moment that is thrilling and uncertain. We are seeing unprecedented power and untapped potential. Our mission is to bring people together and help people achieve their potential.
Azure Pipelines is the perfect launchpad for your code – automating your builds and deployments so you spend less time with the nuts and bolts and more time being creative
8
13
15
The overall approach is to orchestrate continuous integration and continuous delivery Azure Pipelines from Azure DevOps. These pipelines are triggered by changes to artifacts that describe a machine learning pipeline, that is created with the Azure Machine Learning SDK. For example, checking in a change to the model training script executes the Azure Pipelines Build Pipeline, which trains the model and creates the container image. Then this triggers an Azure Pipelines Release pipeline that deploys the model as a web service, by using the Docker image that was created in the Build pipeline. Once in production, the scoring web service is monitored using a combination of Application Insights and Azure Storage.