The document provides an overview of machine learning capabilities on Azure, including Azure Machine Learning service, popular frameworks, managed compute, devops for machine learning, pipelines, deployment options, and tool agnostic Python SDK. Key features highlighted are support for open source frameworks, bringing AI to everyone with an end-to-end scalable platform, boosting data science productivity, and seamless integration with Azure services. Resources and links are also provided for further learning.
2. Overview
● Machine Learning on Azure
● Custom AI
● Compute Targets (DSVMs and Managed Compute)
● DevOps for Machine Learning
● Azure Machine Learning Pipelines
● Flexible and Support for Open Source Frameworks
● Deployment
● Tool Agnostic Python SDK
3. Building blocks for a Data Science Project
Data
sources
Classical ML
Deep learning
Build and train
models
Experimentation
and pipelines
Hyperparameter
tuning
DevOps for data
science
Deployment
4. Machine Learning on Azure
Domain Specific Pretrained Models
To reduce time to market
Azure
Databricks
Machine
Learning VMs
Popular Frameworks
To build machine learning and deep learning solutions TensorFlow
PyTorch ONNX
Azure Machine
Learning
Language
Speech
…
Search
Vision
Productive Services
To empower data science and development teams
Powerful Hardware
To accelerate deep learning
Scikit-Learn
PyCharm Jupyter
Familiar Data Science Tools
To simplify model development Visual Studio Code Command line
CPU GPU FPGA
From the Intelligent Cloud to the Intelligent Edge
5. Azure Machine Learning Service
Set of Azure Cloud
Services
Python
SDK
✓ Prepare Data
✓ Build Models
✓ Train Models
✓ Manage Models
✓ Track Experiments
✓ Deploy Models
That enables you to:
7. Productive Services
Empower data science and development teams
Individual data scientists
Desktop solutions adequate
Need cloud for sporadic compute needs
Integrated data science & data engineering teams
Desktop solutions not adequate
Need a unified big data & machine learning solution
Azure Machine Learning
service
Azure Databricks
+
Machine Learning VMs
11. Distributed training on managed compute
Gather results
Secure Access
Scale resources
Schedule jobs
Dependencies and Containers
Provision VM clusters
Distribute data
Handling failures
12. Training Infrastructure
Scale resources
Autoscale resources to only pay while
running a job
Schedule jobs
Train at cloud scale using a
framework of choice
Dependencies and Containers
Leverage system-managed AML
compute or bring your own
compute
Provision VM clusters
Use the latest NDv2 series VMs
with the NVIDIA V100 GPUs
Distribute data
Manage and share resources across
a workspace
13. Powerful Infrastructure
Accelerate deep learning
General purpose machine
learning
D, F, L, M, H Series
CPUs
Optimized for flexibility Optimized for performance
GPUs FPGAs
Deep learning
N Series
Specialized hardware
accelerated deep learning
Project Brainwave
Support for image classification and recognition scenarios
ResNet 50, ResNet 152, VGG-16, SSD-VGG, DenseNet-121
FPGA NEW UPDATES:
15. Prepare
Data
Register and
Manage Model
Train &
Test Model
Build
Image
…
Build model
(your favorite IDE)
Deploy Service
Monitor Model
Prepare Experiment Deploy
DevOps loop for data science
17. Model Management in detail
Create/Retrain Model
Enable DevOps with full CI/CD
integration with VSTS
Register Model
Track model versions with a
central model registry
Monitor
Oversea deployments through
Azure AppInsights
18. Use leaderboards, side by side run
comparison and model selection
Conduct a hyperparameter search on
traditional ML or DNN
Leverage service-side capture of run metrics,
output logs and models
Manage training jobs locally, scaled-up or
scaled-out
Experimentation
95%
80%
75%
90%
85%
20. Azure Machine Learning Pipelines
Prepare data Build & train models Deploy & predict
Data storage
locations
Data ingestion
Data Preparation Model building & training Model deployment
Normalization
Transformation
Validation
Featurization
Hyper-parameter tuning
Automatic model selection
Model testing
Model validation
Deployment
Batch scoring
21. Prepare data Build & train models Deploy & predict
Data storage
locations
Data ingestion
Data Preparation Model building & training Model deployment
Normalization
Transformation
Validation
Featurization
Hyper-parameter tuning
Automatic model selection
Model testing
Model validation
Deployment
Batch scoring
Normalization
Transformation
Validation
Featurization
Hyper-parameter tuning
Automatic model selection
Model testing Testing error
Azure Machine Learning Pipelines
22. Prepare data Build & train models Deploy & predict
Data storage
locations
Data ingestion
Data Preparation Model building & training Model deployment
Normalization
Transformation
Validation
Featurization
Hyper-parameter tuning
Automatic model selection
Model testing
Model validation
Deployment
Batch scoring
Normalization
Transformation
Validation
Featurization
Hyper-parameter tuning
Automatic model selection
Model testing Testing error
Model testing
Model validation
Deployment
Batch scoring
Error
resolved
Azure Machine Learning Pipelines
23. Azure Machine Learning Pipelines with new
data
Prepare data Build & train models Deploy & predict
Data storage
locations
Data ingestion
Data Preparation Model building & training Model deployment
Normalization
Transformation
Validation
Featurization
Hyper-parameter tuning
Automatic model selection
Model testing
Model validation
Deployment
Batch scoring
Normalization
Transformation
Validation
Featurization
Hyper-parameter tuning
Automatic model selection
Model testing
Model validation
New
data
Deployment
Batch scoring
24. Advantages of Azure ML Pipelines
Unattended runs
Schedule a few steps to run in parallel or
in sequence to focus on other tasks while
your pipeline runs
Mixed and diverse compute
Use multiple pipelines that are reliably
coordinated across heterogeneous and
scalable computes and storages
Reusability
Create templates of pipelines for specific
scenarios such as retraining and batch
scoring
Tracking and versioning
Name and version your data sources,
inputs and outputs with the pipelines SDK
26. Popular Frameworks
Use your favorite machine learning frameworks without getting locked into one framework
ONNX
Community project created by Facebook and Microsoft
Use the best tool for the job. Train in one framework and
transfer to another for inference
TensorFlow PyTorch Scikit-Learn
MXNet Chainer Keras
29. Flexible Deployment
Deploy and manage models on intelligent cloud and edge
Train & deploy Train & deploy
Deploy
Model optimization for cloud & edge
Manage models in production
Capture model telemetry
Retrain models
30. Deploy Azure ML models at scale
Azure Machine Learning Service
Azure Machine
Learning
Experimentation
Cognitive
Services
External
Model
Model Registry
Register Model
Cloud Service
Heavy Edge
Light Edge
Image Registry
Create &
Register Image
Your IDE
Scoring File
Image Registry
Deployment & Model
Monitoring
33. Tool Agnostic Python SDK
Use your favorite IDEs, editors, notebooks,
and frameworks
Flexibility of your local environment or
curated cloud environment
Integrate with other services like
Azure Databricks
Get started quickly without any complex
pre-requisites
PyCharm Jupyter Visual Studio Code
35. Azure Machine Learning service
Bring AI to everyone with an end-to-end, scalable, trusted platform
Built with your needs in mind
Support for open source frameworks
Managed compute
DevOps for machine learning
Simple deployment
Automated machine learning
Seamlessly integrated with the Azure Portfolio
Boost your data science productivity
Increase your rate of experimentation
Deploy and manage your models everywhere
Tool agnostic Python SDK
36. Azure Machine Learning with Azure Databricks
https://aka.ms/aml-notebook-databricks-e2e
Azure Notebooks
https://notebooks.azure.com/azureml/projects/azureml-getting-started
Azure ML Docs
https://docs.microsoft.com/en-us/azure/machine-learning/service/
Resources beyond this AI Airlift