What does it mean if an MLflow run is "UNFINISHED?"
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
09-15-2021 04:43 PM
The run has a model logged to it already, but the "Status" field says "UNFINISHED."
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
09-15-2021 04:45 PM
The MLflow run was probably created either (a) via notebook autologging or (b) via a call to `mlflow.start_run()`. With (a), when the notebook first logs something to MLflow, it starts a run. But if the notebook is still active and attached to a cluster, and nothing has ended the run, then the run is still active ("UNFINISHED").
You can end the run with `mlflow.end_run()`.
Alternatively, you can wrap your model fitting code in a "with" statement:
with mlflow.start_run():
# my ML code
This "with" syntax will automatically end the run once the ML code finishes.