Intent

• A repo to expose some examples of 'data drift' in Azure Machine Learning

Steps

• First ensure you have a file called sub.env in the './scripts' folder with the following line: SUB_ID=<your subscription id>
• Then run the create-workspace-sprbac.sh shell script to create the AML workspace
  • This will also create two environment files: config.json and variables.env which will help with service principal authentication and be used in the authentication.py script.
  • As part of the create-workspace-sprbac.sh script, names are derived based upon a random choice combining the nouns.txt and the adjectives.txt file, implemented in the random_name.py script.
• In the './data-drift' folder, run the clusters.py script to create a cluster.
• The 'get-data' folder contains some scripts for pulling data from Azure Open Datasets.
• The 'seattle-weather-data' folder contains pre-downloaded files for the 2018-2020 Seattle NOAA Weather Data.
• To trigger the data drift monitor for the Seattle Weather, run seattle_weather_drift.py.
• To trigger the data drift monitor for the US County data, run us_county_drift.py.

Other

• If on Mac, brew install libomp, refer article
• While the clusters.py is more of a setup script, to allow for interaction with the variables.env file, have stored this under the data-drift folder.