Workspace advice for ML tasks

[DCELL]

Hello,

I need some advice/example use case for using ML in Fabric.

We have a Dev workspace containing artifacts for ingesting data, storing data, cleaning & feature engineering, model training & logging experiments, and finally the best model is selected from the experiment and saved as a new version of the model.

The problem is ML models are the only artifacts that can't be directly deployed using the pipeline. I know it's possible to save the models to the Dev workspace files, then copy them to the test / prod workspace files, then re-register the models from the files in to the model registry. However this process contains a lot of manual steps which introduces more room for error.

I was hoping that someone could point to a successful use case that I can reference or give advice on the best way to structure the workspaces for ML Ops.

Thanks

[DCELL]

I'll mark it as resolved. I guess we just have to wait for a Fabric update to address this.

[v-menakakota · Accepted solution]

Hi @DCELL ,
Thanks for reaching out to the Microsoft fabric community forum. 

Yes,you're absolutely right that Fabric currently doesn’t support deploying ML models via deployment pipelines, and that adds complexity when aligning with proper MLOps practices.

Tutorial: AutoML- train no-code classification models - Azure Machine Learning | Microsoft Learn

Automated ML in Fabric - Microsoft Fabric | Microsoft Learn

One of the user raised it in the Issues forum please go through the link:

Deployment Pipelines unsupported items - ML Model - Microsoft Fabric Community

Please go through the below solved link which may help you in resolving the issue:

Solved: Machine learning pipelines in Microsoft Fabric - Microsoft Fabric Community


I hope this information helps. Please do let us know if you have any further queries.

[v-menakakota]

Hi  @DCELL ,

May I ask if you have resolved this issue? If you have any issues please reach out to us.
Thank you.