Purchase Databricks : Databricks-Machine-Learning-Professional Exam Materials and then pass exam easily

A Machine Learning Engineer has deployed a fraud detection model in Databricks Model Serving to detect fraudulent transactions. The engineer wants to compare the model's predictions with the actual fraud classifications from the Fraud Ops team to monitor model performance. The Fraud Ops team uses a unique transaction_id to investigate fraudulent activity and persist their findings to a fraud_findings table. The engineer enabled inference tables on the endpoint, but they are not sure how to map the models' predictions to the Fraud Ops team's classifications. How can the engineer uniquely join the models' prediction to the fraud_findings table with the fewest code changes?

• A. Modify the model to include an additional input: transaction_id. Log, register and deploy the new model. In the model serving request body, add transaction_id as an additional input feature. Join the inference table with the fraud_findings table using transaction_id as the join key.
• B. Populate the client_request_id field with the transaction_id in the model serving request body.
  Join the inference table with the fraud_findings table using client_request_id (which contains the transaction_id) as the join key.
• C. Store databricks_request_id returned from each model serving request and persist it to the fraud_findings table. Join the inference table with the fraud_findings table using databricks_request_id as the join key.
• D. Join the inference table with the fraud_findings table using timestamp_ms as the join key.

Which of the following describes concept drift?

• A. Concept drift is when there is a change in the relationship between input variables and target variables
• B. Concept drift is when there is a change in the distribution of a target variable
• C. Concept drift is when there is a change in the distribution of the predicted target given by the model
• D. None of these describe Concept drift
• E. Concept drift is when there is a change in the distribution of an input variable

A machine learning engineer and data scientist are working together to convert a batch deployment to an always-on streaming deployment. The machine learning engineer has expressed that rigorous data tests must be put in place as a part of their conversion to account for potential changes in data formats. Which of the following describes why these types of data type tests and checks are particularly important for streaming deployments?

• A. All of these statements
• B. Because the streaming deployment is always on, there is a need to confirm that the deployment can autoscale
• C. Because the streaming deployment is always on, all types of data must be handled without producing an error
• D. Because the streaming deployment is always on, there is no practitioner to debug poor model performance
• E. None of these statements

Concept drift is when there is a change in which element?

• A. The relationship between input variables and target variables
• B. The distribution of the predicted target given by the model
• C. The distribution of an input variable
• D. The distribution of a target variable

A Machine Learning Engineer has a computer vision model in Databricks Model Serving that obscures sensitive data from images. Internal teams use the model throughout the work week when they request access to new files. Recently model users complained that the model takes much longer in the morning. This coincides with when people arrive at work and request files for the day. When the engineer reviews the endpoint health metrics, they see P50 model latency peaks around 9AM at 20 seconds. Request rate also peaks at 9AM at 15 requests/second. The GPU utilization is over 60%, GPU memory usage over 50%, and provisioned concurrency at 4 throughout the day. What can the engineer do to reduce user wait time when request rate peaks at 9AM each morning?

• A. The endpoint is constrained by its ability to handle simultaneous requests. Enable "scale_to_zero" on the endpoint to spin up additional endpoints quickly to handle the requests.
• B. The endpoint is constrained by the GPU. Scale the endpoint vertically by changing the endpoint workload_type from "GPU_SMALL" to "GPU_MEDIUM".
• C. The endpoint is constrained by its ability to handle simultaneous requests. Define a rate_1imit on the endpoint to spread out the requests over the course of the day.
• D. The endpoint is constrained by its ability to handle simultaneous requests. Scale the endpoint horizontally by editing the endpoint workload_size from "Small" to "Medium" to increase the model concurrency.