Azure Machine learning Federated learning

Federated learning sampling introduction

This sample is to show how to run federated learning in azure machine learning from existing sample from documentation. AT the time of this tutorial, Federated learning is in public preview.

Prerequisites

Azure Account
Azure Machine Learning Service
clone the github repo — https://github.com/balakreshnan/azure-ml-federated-learning
Go to quick start — https://github.com/balakreshnan/azure-ml-federated-learning/blob/main/docs/quickstart.md

Federated learning Steps

Create a workspace
Create a compute instance
or local workstation is also fine
there is bicep code to create the workspace and compute instance

Code

az login --tenant <tenant-id>
az account set --name <subscription name>

create a reource group

az group create --name fltest --location eastus

now create the resource
Change the name fldemo to something else

az deployment group create --template-file ./mlops/bicep/open_sandbox_setup.bicep --resource-group fltest --parameters demoBaseName="fldemo22"

wait until the deployment is complete
now run the python code to setup the environment

python -m pip install -r ./examples/pipelines/fl_cross_silo_literal/requirements.txt

now update the config.json

{
    "subscription_id": "<subscription-id>",
    "resource_group": "<resource-group>",
    "workspace_name": "<workspace-name>"
}

now run the training code

python ./examples/pipelines/fl_cross_silo_literal/submit.py --example MNIST --submit

wait for the training to complete

Metric view

Steps review

Original article — Samples2022/FLintro.md at main · balakreshnan/Samples2022 (github.com)

Mlearning.ai Submission Suggestions

How to become a writer on Mlearning.ai

medium.com