MLOps-K3

This is a tutorial repo of Course 5 - Machine Learning in Production - MLOps organized by DataScienceWorld.Kan. Its' oriented target towards building an Azure DevOps CI/CD pipeline that comprises the steps as follows:

1. Setup infrastructure over Azure ML workspace
2. Connect to Azure ML workspace
3. Running register Dataset and Datastore
4. Training a machine learning model
5. Endpoint deployment of machine learning model
6. Test model performance.

Setup Infrastructure as Code via ARM template file

To manipulate datasets and train models you need to initialize an AzureML working space as a priority. The Lession 9 - Part II - Deploying infrastructure as code has provided a hands-on guideline for your to complete this.

Setup Azure-cli steps in CD

1. Get default workspace:

python aml-service/00-Workspace.py -config config/dev/config.json

2. Load datastore:

python aml-service/10-Datastore.py -config config/dev/config.json

3. Register Dataset:

python aml-service/22-TabularDataset.py -config config/dev/config.json

4. Initializing compute instances:

python aml-service/30-Compute.py -config config/dev/config.json

5. Register Environments:

python aml-service/40-Environment.py -config config/dev/config.json

6. Model training pipeline:

python aml-service/50-PipelineModelTraining.py -config config/dev/config.json -pipeline modeltraining

7. Model deployment Local service:

python aml-service/77-DeployToLocalService.py -config config/dev/config.json

8. Testing local deployment:

python aml-service/78-TestLocal.py -config config/dev/config.json