MachineLearningNotebooks
MachineLearningNotebooks copied to clipboard
Using Workspace Connections for Docker Authentication in Serverless Builds
Title
How to properly pass workspace connection credentials to Docker builds on AML serverless compute
Description
Problem Statement
I'm trying to set up a Docker build on Azure ML serverless compute that needs to authenticate to a private PyPI feed during the image build phase. The image needs to install Python packages from a private Azure DevOps Artifacts feed.
What I'm Trying to Achieve
- Create a Docker image that installs packages from a private PyPI mirror
- Use Azure DevOps Personal Access Token (PAT) for authentication (I am happy to use a different way if it is more sustainable and better)
- Run this on AML serverless compute (not a dedicated compute cluster)
- Securely pass credentials without hardcoding them