Deleting MLflow runs programmatically

[samini]

I would like to programmatically delete some MLflow runs based on a given run id. I am interested in the best practices on how to do this in Databricks workspaces.

I know that MLflow cli has gc command which seems quite useful since it also deletes associated artifacts with a run id. I tried to use that in a Databricks workspace but it gave me an error saying: 

mlflow.exceptions.MlflowException: This cli can only be used with a backend that allows hard-deleting runs

Does anyone have experience with running this command or other best practices deleting runs programmatically?

[Kaniz]

Hi @samini, The error message you're seeing indicates that the MLflow CLI garbage collection (gc) command is not supported by the backend store that's being used by your Databricks workspace.

[samini]

Hi, it does indeed suggest that. But, is it something configurable in Databricks workspaces? I tried to find a setting in the Admin Settings tab, but I couldn't find anything.

[mameli]

Do we have any updates on this? i'm using databricks and I have the same MlflowException