Git Integration for ML Models & Experiments – Support the Git Ignore Idea (in general)

[nielsvdc]

As of August 27, 2025, we’ve noticed that support for Git integration now includes ML Models and Experiment items in Fabric. While I’m all for expanding Git support across Fabric, this particular change raises some concerns.

Many ML Models and experiments are temporary by nature and don’t need to be versioned or moved across development, test, and production environments. Including them by default in source control can clutter repositories and complicate workflows unnecessarily.

Surprisingly, there’s no blog post or update listing this change in the Supported items for Git integration, which makes it even more confusing.

To be able to exclude Fabric items from committing to Git, I encourage everyone to support the idea of a Git Ignore option for source control in Fabric.

[v-venuppu]

Hi @nielsvdc ,

As you said, please avoid committing these items. We will keep this thread open so that other users who are facing similar issue can upvote to the idea you have suggested.

Git Ignore option for Source Control - Microsoft Fabric Community

Thank you.

[v-venuppu]

Hi @nielsvdc ,

I wanted to check if you had the opportunity to review the information provided and resolve the issue..?Please let us know if you need any further assistance.We are happy to help.

Thank you.

[nielsvdc]

@v-venuppu

Currently, the only solution to this issue is to avoid committing these items. In our dev-test-prd setup, each environment contains over 500 models and experiments, which do not require git support at this time. Changing our process may be necessary for handling these items.

[v-venuppu]

Hi @nielsvdc ,

Thank you for reaching out to Microsoft Fabric Community.

Thank you @KevinChant @rohit1991 for the prompt response.

Thanks for clarifying. From what you have found, it seems that MLFlow doesn’t currently provide a way to move models or experiments between workspaces. Even if it might be possible in theory, there isn’t a straightforward solution available right now. That makes your point about needing a Git ignore option even more relevant - it would be a much simpler way to manage temporary models and experiments without adding extra workspaces.

[KevinChant]

Is branching out to separate workspaces to work with the ML Models and Experiments and option? That way you only eed to commit the Models needed back to the "main" workspace.

[nielsvdc]

Hi Kevin, we had a look before if it was possible with MLFlow to save or load models and experiment to or from other workspace and were unable to find a solution for this. So this might not be possible.
And, as we have dev-test-prd workspace for data science, we would like to avoid adding extra workspace to manage. I also think this would be a workaround for something that can be better managed with the ability to ignore specific items by type (folder extension) or item name.

[rohit1991]

Hi @nielsvdc 

I completely agree with your concern. ML Models and Experiments are usually temporary and don’t always need to be moved between dev, test, and production. Forcing them into Git by default can make the repository messy and harder to manage. Right now, the documentation doesn’t clearly mention this change, which only adds to the confusion for users. A lot of teams would benefit from having an option similar to a “.gitignore” so that they can decide what really needs to go into version control. This would give flexibility to exclude items like ML models or experiments that are not critical for long-term source control. It would also help keep Git cleaner, avoid unnecessary commits, and simplify workflows. Overall, giving users control to include or exclude items will make Git integration in Fabric much more useful and practical.