Is it wise to use a more recent MLFlow Python package version or is the DB Runtime compatibility matrix strict about MLFlow versions?

[fermin_vicente]

More concretely, should we fix the dependency version at MAJOR, MINOR or PATCH?

For example, MLFlow 1.30.0 is available and latest DBR 11.3 LTS is compatible with 1.29.0

My question comes from the fact that installing our own libraries that use MLFlow, dependency resolution might try to get the latest version if we don't properly pin it.

Thanks!

[Debayan]

As of now, DBR 11.3 LTS is compatible with MLFlow 1.29.0. (https://docs.databricks.com/release-notes/runtime/releases.html?&_ga=2.185224809.1729528441.16681138...)

[fermin_vicente]

Hi! thanks for the reply, although maybe you didn't notice that I linked to the same url, so we're aware of the matrix.

The question is, is it compatible solely with 1.29.0? We want to know which dependency should we use in all our projects that might be running against/in the platform:

• mlflow==1.29.0
• mlflow==1.29.*
• mlflow==1.*

Hope this is clearer! thanks