Applied Data Science with Azure Machine Learning - for online purpose
Material for: "Applied Data Science with Azure Machine Learning in a day"; March SQLBits 2023
Applied data science with Azure Machine Learning (AML) will give you a thorough look into the world of data scientists using Azure Machine Learning. This training will give data scientists information on where and how to use AML, how to build, deploy, and maintain predictive models with dedicated built-in tools, open-source frameworks and the powerful Python SDK. This training will provide attendees with knowledge and power to use the platform for finding and making better decisions and models in your organization.
With a fully-managed and enterprise-ready platform, Azure Machine Learning is a powerful platform that will suit basically every data science project and address different framework flavours.
Training consists of 7 modules and will explore all the important assets, tools and machine learning frameworks.
Applied data science with Azure machine learning (AML) full-day training will deliver a thorough exploration and look into to use AML, how to build, deploy, and maintain predictive models with dedicated built-in tools, open-source frameworks and powerful Python SDK and PyTorch.
- Understand the basic concepts of data science processes and what cloud services bring to the table
- Get on board with designing and preparing Machine Learning solutions with AML
- Explore data, train the models and use tune your prediction models
- Learn how to prepare the model for deployment by creating pipelines and use of MLFlow
- Deploy monitor and retrain your models
- Learn how to use Python SDK
- Use your models with other Azure Services (Storage, SQL Database)
The time outline for the training is designed in 7 modules each for 45 - 50 minutes. Times are displayed in UTC. Coffee and lunch breaks will be aligned with the organizers on the day of the workshop.
- 08.45 – Gathering and preps[^1]
- 09.00 - 09.55 - Module 1
- 10.00 - 10.50 - Module 2
- 10.50 - 11.00 – Coffee Break
- 11.00 - 11.50 - Module 3
- 12.00 - 12.50 - Module 4
- 12.50 - 13.50 - LunchTime
- 14.00 - 15.00 - Module 5
- 15.00 - 15.50 - Module 6
- 15.50 - 16.00 – Coffee Break
- 16.00 - 17.00 - Module 7
- 17.00 + Gathering and wrap-up
Module 1 - Starting with Azure Portal, Azure Machine Learning Services (AML) and Azure resources for AML will focus on a gentle introduction to Azure Machine Learning services, Azure Subscription, Key Vaults, and services for storing data, like Azure Data Store, Azure SQL Database and other data sources. We will also create the Azure roles and memberships so that we can use them in the upcoming modules. Additionally we are going to get familiar with Azure CLI and Python SDK v1.
Module 2 - Create, and manage your workspace, data, and compute for your experiments will be focusing on creating, configuring and managing your machine learning workspace, attaching data from Azure data store (created in Module 1), registering datastores and creating datasets. Furthermore, we will be exploring and configuring compute assets for optimal training workload.
Module 3 - Explorative data analysis with Python and notebooks will dive into exploring approaches to data analysis, statistics and getting the most information and insights from your data. Using the instantiated workspace, attached with compute and datastores, we will be performing univariate, and multivariate statistics, exploring data using Python visualisation packages and taking advantage of notebooks.
Module 4 - Preparing the models, running experiments, and training the model will teach you how to configure a job run, configure the compute and consume data from a job. With the help of notebooks, we will evaluate a model, and train and track the model using MLflow. After the process will be completed, we will implement a training pipeline, learn how to pass data between the steps and also take a look into using custom components and component-based pipelines.
Module 5 - Using MLflow output, deploying, and retraining a model will explore the registered models in MLflow and how to use the relevant model, retrain and watch the model performance. We will also set both real-time and batch deployment and explore endpoints and how to use them using Azure ML CLI. The last part of this module will be focused on integrating the solution with Github and learning how to retrain the model with event-based triggers or scheduled triggers.
Module 6 - Building end-to-end solution will deliver the complete experience for an end-to-end solution. This module will wrap up the previous five modelsWe will deploy a model by using an online managed endpoint with the help of Azure ML CLI, register and track the model using MLflow, and create YAML for sweeping the model and instancing the inferring cluster for model consumption.
Module 7 - Using the Designer and Automated ML goes into more detail on the Designer and Automated ML options inside Azure Machine Learning. We will train a machine learning model, analyze it's performance and ultimately publish it so we can use it to score data from other services.
Learn how to use Azure Machine Learning service and be able to build a Machine Learning solution in a day from a scratch.
Data Scientists, Statisticians, Machine Learning Engineers
Data Analysts, BI Analysts, Big Data analysts, Data engineers, Data architects, Tech Leader, DevOps Engineers, and Business Leader.
Some background in Machine learning or statistics. Any additional knowledge of AML or EDA is a benefit for the workshop
- Working laptop with admin access (Win or Mac)
- Installed Visual Studio Code and Python Environment
- Conda environment and additional Python packages installed (Pytorch, ONNX, ...)
- Access to the internet
- Credentials and credit (free credit) for accessing the Azure portal
- SQL Server 2022 developer edition (optional)
All materials (Markdown, iPynb, Bicep, Pytorch, Py) and accompanying materials will be handed to attendees before the workshop. Material is prepared for self-paced learning.
- Microsoft Github
- Microsoft Learn
- Blogpost
- Book: https://github.com/PacktPublishing/Mastering-Azure-Machine-Learning-Second-Edition