Get the PoC of Tekton pipeline for building a container of an AzureML model working on the PoC test cluster.

[devguyio]

User Story

• As a MLOps engineer I would like to take a trained model and use an automated pipeline to package it inside a container so that it can be deployed to the near edge. #2

@stefan-bergstein PoC code.

A/C

• Any software needed to be installed on the test cluster is identified as a prerequisites
• A small README of how to get successfully get that PoC running on the cluster

[piotrpdev]

Got Stefan's code working by following his README. You will need the OpenShift Pipelines Operator (which RHODS provides), a Quay repo, and an AzureML workspace with a MLFlow model (model is in Stefan's repo).

---

Note
As of writing this I haven't tested the final step in the README that describes how to deploy the model to the target system, since it isn't part of this issue.

If you plan to reproduce this, I would advise is the following:

Tips

• Make sure to give the Azure App the AzureML Data Scientist role. Here is a great guide on how to do this in case you don't know. Here is a quick tutorial on how to get the credential values for .azure-env.

  • 

• If running the code as-is, make sure the model is stored like this in AzureML:

  • 

• Change the target-namespace value in the test-mlflow-image pipeline to match your Quay username/namespace. I'm not very familiar with Tekton; you may find a better way than changing the file.

Note

Building the mlflow container does take a while.