Assignment - MLOps on Edge devices
Ayse Mutlu
Assignment 4 - MLOP
Submission:
Introduction to Azure IoT
Introduction to Azure IoT Hu
Introduction to Azure IoT Edge
Deploy a pre-built module to the Edge device
Train and package an Azure machine learning module for deployment to IoT Edge device
Introduction to MLOps for IoT Edge
Implement CI/CD for IoT Edge
Exercise questions
Assignment 4 - MLOps
You will use:
Assignment 4.1 to access to the assignment
Assignment 4.2 to submit the assignment
Assignment - MLOps
Assignment title Available link in Canvas Due date Submission type
Introduction to Azure IoT
Introduction to Azure IoT Hu
Introduction to Azure IoT Edge
Introduction to MLOps for IoT Edge https:
canvas.ox.ac.uk/courses/119867/assignments/335859 30 April 2022 Upload screenshot of the module badges in Canvas
Deploy a pre-built module to the Edge device
Train and package an Azure machine learning module for deployment to IoT Edge device
Implement CI/CD for IoT Edge
https:
canvas.ox.ac.uk/courses/119867/assignments/335859 30 April 2022
Upload output of the module from the command prompt in Canvas
Exercise questions https:
canvas.ox.ac.uk/courses/119867/assignments/335859 30 April 2022
Upload the answer for introduction modules and output for standard modules
Assignment - How to access
Assignment - How to access
Click on the module.
Assignment - How to access
Sign in to Learn page so that you can keep track of learning experience.
Assignment - How to submit
After completing the module, unlock achievement.
Assignment - How to submit
Go to your profile. Click on the “Achievements”. You will get the badges of completed modules.
Assignment - How to submit
Go to “Assignment 4.2. MLOps on Egde devices” and upload your submission files for learn modules and exercises.
Thank you!
Assignment 4 - MLOps
You will use:
Assignment 4.1 to access the assignment
Assignment 4.2 to submit the assignment
Submission:
Introduction to Azure IoT
Introduction to Azure IoT Hu
Introduction to Azure IoT Edge
Deploy a pre-built module to the Edge device
Train and package an Azure machine learning module for deployment to IoT Edge device
Introduction to MLOps for IoT Edge
Implement CI/CD for IoT Edge
Exercise questions
What to submit
For introduction modules, upload screenshot of the module badges
● Introduction to Azure IoT
● Introduction to Azure IoT Hu
● Introduction to Azure IoT Edge
● Introduction to MLOps for IoT Edge
For standard module, upload output of the module i.e. Output from the command prompt
● Deploy a pre-built module to the Edge device
● Train and package an Azure machine learning module for deployment to IoT Edge device
● Implement CI/CD for IoT Edge
Exercise questions: Please submit the answer for introduction modules and output for standard modules
Credits to use:
You should setup Azure credits based on the instructions
Assignment 1
Introduction to Azure IoT 1
Introduction to Azure IoT Hub 1
Introduction to Azure IoT Edge 1
Introduction to MLOps for IoT Edge 2
Deploy a pre-built module to the Edge device 2
Train and package an Azure machine learning module for deployment to IoT Edge device 3
Implement CI/CD for IoT Edge 3
3
Introduction to Azure IoT
Introduction to Azure IoT Hu
Introduction to Azure IoT Edge
Introduction to MLOps for IoT Edge
Consider the four scenarios below. Design a solution using IoT Hub, IoT Edge, Digital twin and MLOps.
Notes
a) Your solution should be in one or two paragraphs per scenario.
) Your solution should list the key criteria (benefits) of the Hub, Twin, Edge and MLOps.
c) In this context, the use of Digital twin includes both the shadow and the simulation feature.
Scenario 1
Rural healthcare
You are tasked with providing support to patients in a rural area. The patients are equipped with IoT devices (like blood sugar monitors, heart rate monitors etc) . Data is being captured by these devices. How will you use IoT Hub, Edge and Digital Twin to design a solution to cater for rural patients who may have intermittent connectivity?
Scenario 2
Accident and emergency
The Accident and emergency department of your hospital has now started to use IoT devices. Each patient is assigned an IoT device when they come to A and E. Data is being captured by these devices. How will you use IoT Hub, Edge and Digital Twin to design a solution to cater for patients who could be potentially at risk and need to be prioritized?
Scenario 3
Facility utilization
In a medical facility, all assets are tagged with IoT devices. This includes both the equipment (such as heart rate monitors) and the rooms (such as operating theaters). All personnel are also tagged with IoT devices. Thus, you get a real time view of the medical faculty.
How will you use IoT Hub, Edge and Digital Twin to design a solution to cater for the medical facility to cover both patient safety and reducing costs?
Scenario 4
Image classification in production
You are developing ML classification model and deploying it at the edge devices. You want to be able to retrain the model frequently (every couple of days) and you would like to automate this training. How will you use MLOps to design the solution?
Deploy a pre-built module to the Edge device
In this part, you will:
1. Read the generated data from the log file.
2. Preprocess the data.
3. Visualise the data using Histogram, Kernel, Density Plots, Time Variation,.
4. What insights can you gain from visualization?
Note: log file(log.log) which includes the generated data is in .zip file.
Train and package an Azure machine learning module for deployment to IoT Edge device
Cu
ently, the machine learning model is trained with a decision tree classifier. You will:
1. Change the baseline model to logistic regression and train the model.
2. Visualise the logistic regression model output.
3. What insights can you gain from visualization?
4. Compare the accuracy of the two models.
Implement CI/CD for IoT Edge
In this part, you will develop a custom IoT Edge Module and use Azure DevOps to build / deploy the module in an automated fashion using the concepts from the CICD module (Integration step in the Release Pipeline would not necessarily be required however it is great to know about it.)
MLOps on Edge devices
Submission - The following modules will be submitted.
· Summary of modules:
· Introduction to Azure IoT
· Introduction to Azure IoT Hu
· Introduction to Azure IoT Edge
· Deploy a pre-built module to the Edge device
· Train and package an Azure machine learning module for deployment to IoT Edge device
· Introduction to MLOps for IoT Edge
· Implement CI/CD for IoT Edge
Submission format:
Introduction module submissions will be made as a screenshot of the module badge:
· Introduction to Azure IoT
· Introduction to Azure IoT Hu
· Introduction to Azure IoT Edge
· Introduction to MLOps for IoT Edge
For standard modules, you need to show the screenshot of the output from command terminal/platform for the following modules:
· Deploy a pre-built module to the Edge device
· Train and package an Azure machine learning module for deployment to IoT Edge device
· Implement CI/CD for IoT Edge