Skip to content

Latest commit

 

History

History
29 lines (20 loc) · 1.37 KB

GettingStarted.md

File metadata and controls

29 lines (20 loc) · 1.37 KB

Getting Started

To setup your own MLOPs project in you azure subscription, follow these steps:

  1. To use the scripts on your local machine, add the azure ml workspace credentials in a config.json file in the root directory and very important (!) add it to the gitignore file, if it is not present already.

  2. Provide following Environment variables in ADO:

  • Mandatory
tenant id: service principal tenant id. Default name in code: 
principal id: service principal appId. Default name in code: AML_PRINCIPAL_ID
principal pass: service principal password. Default name in code: AML_PRINCIPAL_PASS
workspace name: workspace name of your test and/or prod (depending on your approach). Default name in code: AML_WORKSPACE_NAME
subscription id: azure subscription id containing your workspace. Default name in code: SUBSCRIPTION_ID
  • Optional
compute target name: if your are using different computes in your environments.
inference target: can be and ACI, AKS, VM for your inference.
AppInsight Instrumentation key: app insight key to use python logger.
datastore/dataset name: if you are using different data source during your CI/CD pipeline (though PROD data must be available for the data scientist)
  1. Create your own dataset in AML, then add the dataset name and other variables (model name, etc) to operation/configuration.yml