Integrate Azure ML Studio with Azure DevOps repo

Pedro Gonzalez 6

Hello,

I am working with Azure Machine Learning Studio via GUI (upload dataset, run experiments, register ML models, register endpoints, develop code in notebooks, ...) and would like to have that code available in Azure DevOps repo in order to restore it if necessary.

Does anyone know how that integration between Azure ML Studio and Azure DevOps repo could be done?

Thanks,

romungi-MSFT 48,496 Reputation points Microsoft Employee

2022-10-21T11:42:50.543+00:00

@Pedro Gonzalez Did the below suggestion to use MLOps help to setup your CI/CD pipelines?

2 answers

Takahito Iwasa 4,851 Reputation points MVP

2022-10-17T20:52:43.673+00:00

Hi, @Pedro Gonzalez

I checked the documentation of Azure Machine Learning Studio, but it seems that there is no integration function with the version control system.

You can download the created model code by referring to the following questions.

https://learn.microsoft.com/en-us/answers/questions/355882/azure-ml-studio-and-gitazuredevops.html

By applying this to build a regular batch mechanism, you can periodically acquire the latest model on ML Studio with code and reflect it in Git management.
Please sign in to rate this answer.

0 comments No comments
Sign in to comment

Use comments to ask for clarification, additional information, or improvements to the question.
romungi-MSFT 48,496 Reputation points Microsoft Employee

2022-10-18T10:02:33.74+00:00

@Pedro Gonzalez You can use MLOps to setup Azure ML pipelines and experiments with Azure DevOps. A good working sample is available in the MLOpsPython repo. The setup should be straightforward where any changes in your pipeline code will trigger a job in your Azure ML workspace.

If an answer is helpful, please click on or upvote which might help other community members reading this thread.
Please sign in to rate this answer.
Pedro Gonzalez 6 Reputation points

2022-10-24T14:02:01.053+00:00

Hi @roungi-MSFT

Thank you for your response and the link provided!

I was reviewing the GitHub repo "Getting Started with MLOpsPython" and I saw that they explain how to build pipelines in order to create and deploy ML models.

Working with Azure Machine Learning Studio, I see that there are different options to create and deploy ML models, it can be done via GUI or it can be also done via pipelines or SDK.

In this case, my question is, if we choose to create and deploy ML models via GUI, is there some way to save that work in Azure DevOps Repo? Something similar to what there is already available in Azure Synapse.

In Azure Synapse, we can create the pipelines via GUI and there is an out of the box integration between Azure Synapse and Azure DevOps repo that sync that information from Synapse to Azure Repos.

Thanks,

Pedro Gonzalez 6 Reputation points

2022-10-24T14:09:54.35+00:00

For example, I have created several Notebooks in Azure Machine Learning Studio via GUI.

Can these Notebooks be saved to Azure DevOps Repo in an automatic way via some Git synchronization?

romungi-MSFT 48,496 Reputation points Microsoft Employee

2022-10-25T07:04:17.323+00:00

I think you can login to your compute instance terminal from the GUI and configure your Azure DevOps CLI to use your DevOps project and push the code to the repo. Please check this link to setup DevOps CLI.

Pedro Gonzalez 6 Reputation points

2022-10-26T08:26:55.32+00:00

You are absolutely right, the code of the notebooks can be found through the compute instance terminal from the GUI.

In this case, I am looking for something generic, about the Machine Learning Studio workspace, not only notebooks. For example, via GUI, we can build notebooks, ML models, endpoints, ....

Is there some way to have a sync between Azure Machine Learning Studio workspace and Git for all components in the workspace? Not just a few of them like pipelines or notebook. Something generic similar to what we have available for Azure Synapse

romungi-MSFT 48,496 Reputation points Microsoft Employee

2022-10-26T10:07:30.797+00:00

In this case the workspace and all its components that are deployed cannot be accessed through backend. You can maintain the configuration of your jobs in the repo and deploy using the SDK or CLI like in the case where the azureml-examples repo provides examples for the usage of the same.
Sign in to comment

Use comments to ask for clarification, additional information, or improvements to the question.

Share via

Integrate Azure ML Studio with Azure DevOps repo

2 answers

Your answer