Databricks ML Associate Practice Question 21: Set Experiment Location
Databricks Certified Machine Learning Associate · Test 3 · question 1 of 10
Three data scientists are working on the same churn project, each in their own notebook. Every run is landing in a separate notebook experiment, so nobody can sort all the runs together to find the best model. The team wants every run from all three notebooks recorded in one shared experiment.
What should they do?
Show answer and explanation
Answer:
By default a run is written to the notebook experiment tied to the notebook that created it, which is exactly the fragmentation described. mlflow.set_experiment() points the active session at a named workspace experiment instead, creating it if it does not already exist. Once each notebook sets the same path, every subsequent start_run() writes there, and all three scientists' runs appear in one list that can be sorted and compared on a shared metric.
Option A relies on a behaviour that does not exist. Notebook experiments are bound to individual notebooks, not to the folder containing them, so relocating the files changes nothing about where runs are recorded.
Option B keeps the runs in three separate experiments. Tags help you filter within an experiment, but the UI sorts and compares runs inside one experiment, so a cross-experiment ranking is still out of reach.
Option D confuses a display name with an identity. Two experiments with the same name are still two experiments with different IDs, and runs do not merge because their labels match.