Hello! I am trying to figure out if setup_mlflow() (ray.air.integrations.mlflow.setup_mlflow) has the ability to specify an existing run_id? The use case of this is to resume MLFlow logging under an existing run_id in the event of a training job failing, server issue, etc. This behavior exists already in the MLflow python API link
I don’t see any references to this functionality in the source code or docs-wondering if this is a new issue or if it’s been encountered previously?
In looking through the code, looks like the run_name isn’t passed or handled as run_id anywhere. The current implementation is setup to handle run_name only, no way to pass run_id.
My goal is to be able to have potential multiple runs bundled under a single run_id to avoid the case where duplicate instances of the same run might be produced if only able to specify run_name, example of unwanted behavior below.
