Databricks Machine Learning Associate Question 159
Single answerYou are training a machine learning model on Databricks and want to track metrics such as loss, accuracy, and validation performance during training. To analyze the results after training, you plan to use the summary object from a Databricks MLflow run. Which of the following statements about using model summaries in Databricks is correct?
- A
The summary object only provides access to the final values of metrics logged during training.
- B
You can retrieve both the final and intermediate logged metrics from the summary object for a completed MLflow run.
- C
Summaries are automatically visualized in the Databricks MLflow UI without needing to log metrics explicitly.
- D
Summaries can display aggregated statistics like mean and standard deviation for logged metrics over multiple runs.
Show answer and explanation
Correct answer: B
Explanation
The summary object in Databricks MLflow is useful for retrieving logged metrics, including intermediate and final values, from a completed run. This allows users to analyze performance trends during training. However, summaries do not automatically generate aggregated statistics or appear in the UI without explicitly logged metrics.
- A. Incorrect.
The summary object provides access to both final and intermediate metrics, not just the final values, making this statement incorrect.
- B. Correct.
This is correct. The summary object allows you to retrieve both the final and intermediate metrics for a completed MLflow run, which can help you analyze performance trends.
- C. Incorrect.
While the MLflow UI visualizes metrics, summaries are not automatically generated unless metrics are explicitly logged during training.
- D. Incorrect.
Summaries do not inherently provide aggregated statistics like mean and standard deviation over multiple runs; this would require custom aggregation logic.