Databricks Machine Learning Associate Question 55
Single answerYou have trained a machine learning model using Databricks AutoML, which has generated multiple models and selected the best-performing one. How can you locate the source code for the best model to further inspect or customize it?
- A
Navigate to the AutoML experiment's notebook and locate the 'Generated Model' section.
- B
Open the MLflow experiment tracking UI and download the source code artifact from the best model's run.
- C
Inspect the AutoML pipeline configuration file for a direct link to the source code.
- D
Access the AutoML training dashboard and look for the 'Source Code' tab.
Show answer and explanation
Correct answer: B
Explanation
Databricks AutoML integrates with MLflow for experiment tracking. When AutoML produces multiple models, their details, metrics, and artifacts, including the source code for the best model, are logged into MLflow. To locate the source code, users can navigate to the MLflow UI, identify the best model's run, and download the source code from the 'Artifacts' section. This ensures users can inspect or customize the model as needed.
- A. Incorrect.
This option is incorrect because the AutoML experiment notebook does not directly contain the source code for the best model. It provides details of the experiment but not the specific model artifacts.
- B. Correct.
This option is correct because MLflow tracks all runs produced by AutoML, including the best model. The source code for the best model can be downloaded from the 'Artifacts' section of the best-performing run in the MLflow UI.
- C. Incorrect.
This option is incorrect because the AutoML pipeline configuration file does not provide access to the source code. Its primary purpose is to define the pipeline structure and parameters.
- D. Incorrect.
This option is incorrect because the AutoML dashboard in Databricks does not have a 'Source Code' tab. The source code is managed through MLflow artifacts.