Databricks-ML-Pro Model Development Practice Test 2 — 15 Questions

Question 1

A data scientist is using Databricks to train a model with a very large vocabulary, leading to high-dimensional sparse inputs. Which storage format for the feature table is most efficient for both storage and retrieval?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Delta Lake tables with Z-ordering on the primary key.. Delta tables are the standard for Databricks Feature Store, as they support efficient reads and writes, schema enforcement, and time travel. When dealing with high-dimensional sparse data, Delta's columnar storage format allows the engine to only read the necessary partitions and columns, significantly reducing I/O. This is critical for performance-sensitive feature engineering pipelines, as it allows rapid processing of large-scale datasets.

Answer

Parquet files on DBFS without using Delta Lake.

Answer

Storing the feature data as a series of JSON blobs in a single file.

Answer

Using a local SQLite database on the Spark driver node.

Question 2

Which TWO of the following are valid ways to register a model in the Databricks Model Registry?

Accepted Answer

Pass `registered_model_name='my_model'` to `mlflow.sklearn.log_model()`.. Registering models is a core step in the MLOps lifecycle. You can either register directly during the training flow (using log_model) or after the fact using the UI or the Python API. Understanding these methods is vital for automating the transition from development to staging and production, ensuring that all models undergo the necessary governance and versioning processes required by an enterprise Databricks environment.

Answer

Upload the model pickle file directly to the Model Registry's SQL database.

Answer

Manually copy the model code into the Databricks Model Registry UI.

Answer

Use the Spark `saveAsTable` command on the model object.

Question 3

When designing a custom MLflow model flavor for a proprietary framework, what is the most critical method to implement to ensure compatibility with `mlflow.pyfunc`?

Accepted Answer

The `load_model` method to enable loading the model artifact.. The `load_model` method is the entry point for the `pyfunc` flavor. It defines how the model is reconstructed from serialized files, which is necessary for serving. A well-implemented `load_model` method ensures that the model can be loaded by any deployment environment that supports the generic `pyfunc` interface, providing flexibility in model deployment across different platforms and runtimes.

Answer

The `train_model` method to support online learning.

Answer

The `register_model` method to support registry integration.

Answer

The `log_model` method to automatically record metrics.

Question 4

Refer to the exhibit. You are attempting to log a model to the MLflow Model Registry, but receive the serialization error shown. What is the cause of this error?

Accepted Answer

MLflow cannot resolve the absolute DBFS path during the serialization process.. The error occurs because the model object contains hardcoded references to absolute file paths that are not accessible on all worker nodes or serving endpoints. MLflow expects model artifacts to be self-contained or stored in a managed artifact repository. Referencing absolute paths creates a dependency on a specific storage mount that may not exist in the production environment.

Answer

The model file size exceeds the limit for the DBFS workspace.

Answer

The DBFS mount is currently in read-only mode during the training job.

Answer

The MLflow provider is not configured to use DBFS as the artifact store.

Question 5

Refer to the exhibit. You have multiple experimental runs in your MLflow tracking server. You want to retrieve the best-performing model based on accuracy. Which command should you use?

Accepted Answer

mlflow.search_runs(experiment_ids=['1'], order_by=['metrics.accuracy DESC']).iloc[0]. Using the MLflow client's search_runs method with an order_by parameter is the standard way to retrieve experiment results programmatically. This approach is superior to manual inspection, as it allows for automated pipeline selection of the top-performing model, ensuring that the model version promoted to the registry is objectively the most accurate according to the defined metrics.

Answer

mlflow.get_all_runs(experiment_id='1').sort_by('accuracy')

Answer

mlflow.list_experiments().get_best_run('accuracy')

Answer

mlflow.get_run(run_id='training_run_123').accuracy

Question 6

When training a model in Databricks, how should you handle sensitive data contained in your training set to ensure compliance?

Accepted Answer

Apply Spark SQL filtering to exclude PII columns before passing data to the model.. Data masking, encryption, and row-level security are fundamental to Databricks compliance. When training, ensuring that PII is removed or hashed before data enters the training pipeline prevents sensitive information from being persisted into model artifacts or logs. This protects the company from data leaks and ensures adherence to regulatory standards like GDPR or CCPA during the development lifecycle.

Answer

Log the sensitive raw features into MLflow tags for audit tracking.

Answer

Encrypt the entire training dataset using a public key during training.

Answer

Only use sensitive data for local testing; never use it in production.

Question 7

You are optimizing a model for inference latency on Databricks. Which strategy is most effective for reducing cold-start times?

Accepted Answer

Export the model to the ONNX format to minimize runtime overhead.. Inference cold-starts are primarily driven by the time required to pull model dependencies and load the model into memory. By utilizing optimized formats like ONNX or MLflow's native model flavors, and ensuring the environment image is pre-warmed or minimized, you reduce the initialization overhead. This is vital for real-time serving, where latency-sensitive applications require consistent response times under varying load conditions.

Answer

Increase the number of instances in the serving cluster.

Answer

Use a larger instance type for the model serving endpoint.

Answer

Disable authentication for the serving endpoint to speed up requests.

Question 8

You are using Databricks Feature Store to manage your features. Which TWO benefits does this provide for model development?

Accepted Answer

It prevents training-serving skew by using the same feature definitions.. The Feature Store enforces consistency between training and serving by centralizing feature logic. It eliminates training-serving skew, as the same code is used to compute features for training and real-time lookup. Additionally, it improves team collaboration by providing a discoverable catalog of pre-computed features, reducing redundant data engineering work and ensuring that models utilize standardized, high-quality data pipelines.

Answer

It automatically scales the compute cluster for arbitrary Python code.

Answer

It automatically retrains models when feature values change significantly.

Answer

It eliminates the need for any feature engineering by automatically learning features.

Question 9

You are deploying a model that requires a GPU for inference. How do you ensure the model is scheduled on an appropriate cluster?

Accepted Answer

Configure the Serving Endpoint to use a compute-optimized instance type.. Defining the cluster policy and GPU requirement in the model's deployment configuration ensures that Databricks assigns the model to nodes with physical GPU resources. This is a critical infrastructure requirement, as running GPU-dependent models on CPU nodes results in immediate execution failure. Correct configuration ensures that the serving environment provides the necessary hardware acceleration to meet performance and operational requirements for the model.

Answer

Write a custom wrapper that detects GPU availability at runtime.

Answer

Use a standard CPU cluster and virtualize the GPU via software emulation.

Answer

Rely on Spark to dynamically partition the model across multiple CPU nodes.

Question 10

During model development on Databricks, which TWO actions help manage model lifecycle and ensure compliance?

Accepted Answer

Register all candidate models in the MLflow Model Registry.. Using the Model Registry allows for central versioning and lifecycle stage management, ensuring that stakeholders can audit which model is in production and why. Integrating MLflow tracking provides an immutable history of training experiments. These practices together satisfy organizational compliance requirements by maintaining a clear, auditable trail from data selection and training parameters all the way through to deployment.

Answer

Hardcode the model's source S3 bucket path into the production notebook.

Answer

Delete old MLflow experiments to keep the tracking server clean.

Answer

Share the same MLflow tracking server across all teams and projects.

Question 11

You are training a model using MLflow on Databricks and notice that your model training runs are not logging parameters or metrics correctly to the experiment tracking server. What is the most likely cause if you are using an automated MLflow autologging configuration?

Accepted Answer

The mlflow.autolog() function was not called prior to initializing the model training code.. MLflow autologging in Databricks requires an active experiment to be set, or it defaults to the notebook's path. If the tracking URI or credentials are misconfigured, or if autologging is not explicitly enabled via mlflow.autolog(), the backend cannot intercept training hooks from common libraries like scikit-learn or XGBoost. Proper configuration ensures seamless experiment reproducibility and lineage, which is essential for professional model development and regulatory compliance within the Databricks environment.

Answer

The cluster is running in standard mode rather than high concurrency mode.

Answer

The experiment tracking server is unreachable due to a VPC network security policy.

Answer

The model training code is utilizing a custom-built ML library not supported by autologging.

Question 12

A data scientist is performing hyperparameter tuning using Hyperopt on Databricks. They want to ensure that each trial runs in parallel to reduce overall compute time. Which configuration is required to achieve this parallelism within an MLflow experiment?

Accepted Answer

Define a SparkTrials object and pass it to the fmin function.. Hyperopt requires a SparkTrials object to distribute the evaluation of hyperparameters across a Spark cluster. By wrapping the objective function in SparkTrials, Databricks orchestrates parallel execution of trials. This is crucial for efficient model development because it scales compute resources dynamically, allowing for faster convergence when searching large hyperparameter spaces, which is a common requirement in production-grade machine learning pipelines where time-to-market is critical.

Answer

Use the fmin function with a standard Trials object.

Answer

Implement a custom loop that manually spins up individual Spark jobs.

Answer

Increase the number of threads on the driver node in the cluster configuration.

Question 13

Which TWO of the following steps are considered best practices when performing feature engineering within a Databricks environment to ensure reproducibility and prevent data leakage?

Accepted Answer

Using Databricks Feature Store to log and version feature tables.. Preventing data leakage and ensuring feature consistency are paramount in machine learning. By using Feature Store to manage features, you ensure that the same logic is applied during training and inference. Time-traveling and avoiding the inclusion of future information in training features prevents the model from learning from data that would not be available at prediction time, ensuring the model generalizes well to new, unseen data in production.

Answer

Calculating aggregate statistics like means and variances on the entire dataset before splitting.

Answer

Hardcoding data transformation logic directly inside the model training notebook.

Answer

Dropping all rows with missing values to simplify the training dataset.

Question 14

Refer to the exhibit. You are reviewing the metadata of a completed MLflow run. You need to retrieve the actual model object that was serialized and saved during the training process using the provided run ID. Which MLflow API method is most appropriate for this task?

Accepted Answer

mlflow.pyfunc.load_model(f'runs:/{run_id}/models/model.pkl'). The mlfow.pyfunc.load_model method is the standard approach for loading a model saved in MLflow. This function loads the model as a generic Python function, which is compatible with any framework supported by MLflow. This abstraction is vital because it allows the model to be deployed consistently across different environments, regardless of the underlying library used to create the model, ensuring operational stability in production deployments.

Answer

mlflow.get_artifact(run_id, 'models/model.pkl')

Answer

mlflow.load_pickle(f'dbfs:/databricks/mlflow-tracking/{run_id}/model.pkl')

Answer

mlflow.sklearn.load_model(run_id)

Question 15

A machine learning engineer is training a PyTorch model on Databricks and wants to track hyperparameter sweeps using MLflow. Which approach ensures proper metric logging and artifact persistence across distributed worker nodes?

Accepted Answer

Wrap the training logic using MLflow autologging or explicit run contexts on the driver, ensuring worker metrics are logged via designated rank-zero processes.. Proper integration of MLflow with distributed PyTorch requires initializing the MLflow run on the driver node and safely broadcasting parameters while logging metrics globally. This guarantees that experiment tracking remains consistent, avoids race conditions, and correctly associates saved checkpoints with the master run ID during distributed training execution.

Answer

Call mlflow.log_metric inside every worker training loop independently without synchronizing run IDs.

Answer

Use MLflow model signatures exclusively on the driver node after gathering all model weights into local memory.

Answer

Store all hyperparameters locally in a JSON file on DBFS and disable the MLflow tracking URI during distributed execution.