Databricks Certified Machine Learning Professional Practice Test 70
Question 690
Review this production model incident: An ML operator promotes a new model version but its automated rollback points to a model that was deleted from the endpoint configuration. Which failure risk follows most closely from the unresolved behavior?
- Rebuilding the old model under outage pressure extends an avoidable recovery window.
- The team cannot recover from a bad promotion without an emergency rebuild.
- Regulated traffic reaches an unapproved model version contrary to residency controls.
- A full immediate cutover could expose every booking to an untested model defect.
Explanation
Correct Answer
B:This consequence follows from the specific unresolved condition. The team cannot recover from a bad promotion without an emergency rebuild. The causal constraint is: An actionable rollback plan needs a retained deployable champion artifact and a tested restoration procedure.
Incorrect Answers
A:That direct consequence is directed at a different deployment strategies situation: A bank requires an immediate rollback after deployment and wants the prior serving environment intact until the new environment passes tests. The more immediate risk here is the team cannot recover from a bad promotion without an emergency rebuild.
C:That direct consequence is directed at a different deployment strategies situation: A model release requires all clients from a restricted region to keep the old version, but a simple percentage split sends some of those clients to the candidate. The more immediate risk here is the team cannot recover from a bad promotion without an emergency rebuild.
D:That direct consequence is directed at a different deployment strategies situation: A travel site serves 30,000 bookings per minute and wants to evaluate a new fraud model while limiting customer impact during its first hour. The more immediate risk here is the team cannot recover from a bad promotion without an emergency rebuild.
Question 691
The engineering team encounters the following issue: An automated retraining job triggers whenever a numerical feature mean shifts by 0.1, regardless of normal seasonal changes or sample size. What investigation would most effectively isolate this behavior?
- Compare champion and candidate per-segment false-negative rates on equivalent untouched evaluation cohorts.
- Compare candidate predictive metrics with model footprint, serving latency and expected concurrency.
- Compare baseline seasonality, sample sizes and repeated drift statistics before each retrain event.
- Compare holdout snapshot IDs, label maturity and metric computation versions in candidate MLflow runs.
Explanation
Correct Answer
C:This is the most discriminating check for the observed symptoms. Compare baseline seasonality, sample sizes and repeated drift statistics before each retrain event. The underlying technical issue is that drift-triggered retraining needs statistically and operationally justified thresholds, not arbitrary raw metric movement.
Incorrect Answers
A:That diagnostic evidence is directed at a different automated retraining situation: A new fraud classifier improves overall F1 but doubles false negatives for high-value payment accounts, and an automated policy chooses it on F1 alone. For the stated incident, the decisive observation is compare baseline seasonality, sample sizes and repeated drift statistics before each retrain event.
B:That diagnostic evidence is directed at a different automated retraining situation: An AutoML training job always chooses the model with the lowest training loss, even when the deployed service has a strict memory and P95 latency budget. For the stated incident, the decisive observation is compare baseline seasonality, sample sizes and repeated drift statistics before each retrain event.
D:That diagnostic evidence is directed at a different automated retraining situation: A retraining pipeline ranks candidate models using metrics calculated on different dataset versions, then selects the numerically highest result. For the stated incident, the decisive observation is compare baseline seasonality, sample sizes and repeated drift statistics before each retrain event.
Question 692
Review this production model incident: An ML engineer manually compares two logistic-regression regularization values on a single holdout but has no reusable SparkML parameter-search plan. What evidence would best distinguish the actual fault from neighboring ML concerns?
- Inspect the estimator ParamGridBuilder maps, evaluator metric and fold configuration used for the claimed search.
- Review predicted probabilities, a precision-recall threshold curve and analyst capacity at candidate cutoffs.
- Inspect the scoring job for collect and driver memory growth versus distributed transform stages.
- Check the chosen model library and training API for documented distributed support and GPU worker topology.
Explanation
Correct Answer
A:This is the most discriminating check for the observed symptoms. Inspect the estimator ParamGridBuilder maps, evaluator metric and fold configuration used for the claimed search. The underlying technical issue is that crossValidator and a ParamGridBuilder can evaluate estimator parameter maps across controlled folds when the validation design permits it.
Incorrect Answers
B:That diagnostic evidence is directed at a different using spark ml situation: A binary classifier ranks risky cases well but sends too many alerts at a default 0.5 decision threshold. For the stated incident, the decisive observation is inspect the estimator ParamGridBuilder maps, evaluator metric and fold configuration used for the claimed search.
C:That diagnostic evidence is directed at a different using spark ml situation: A logistics model must score 60 million daily routes already in Delta but the scoring script collects each route into Python before calling predict. For the stated incident, the decisive observation is inspect the estimator ParamGridBuilder maps, evaluator metric and fold configuration used for the claimed search.
D:That diagnostic evidence is directed at a different using spark ml situation: A media group has a specialized deep neural model that is not implemented as a SparkML estimator and assumes SparkML will automatically distribute its GPU training. For the stated incident, the decisive observation is inspect the estimator ParamGridBuilder maps, evaluator metric and fold configuration used for the claimed search.
Question 693
In an enterprise ML workflow, the reported issue is: A media group has a specialized deep neural model that is not implemented as a SparkML estimator and assumes SparkML will automatically distribute its GPU training. What technical distinction is essential to this investigation?
- Training-set fit does not establish out-of-sample predictive quality or cold-start behavior.
- SparkML distribution does not make every external estimator a natively distributed SparkML learner.
- VectorAssembler is a Transformer that creates a vector column, not a trainable predictor.
- Scoring should use preprocessing components trained and versioned with the exact estimator artifact.
Explanation
Correct Answer
B:The technical constraint is specific to the reported failure. SparkML distribution does not make every external estimator a natively distributed SparkML learner. To test it, check the chosen model library and training API for documented distributed support and GPU worker topology.
Incorrect Answers
A:That underlying principle is directed at a different using spark ml situation: A recommender experiment computes all evaluation metrics on training interactions and reports the result as generalization to new customers. The current incident instead turns on sparkML distribution does not make every external estimator a natively distributed SparkML learner.
C:That underlying principle is directed at a different using spark ml situation: A team wants to convert several numeric inputs into one features vector but treats VectorAssembler as a clustering Estimator that should output predictions. The current incident instead turns on sparkML distribution does not make every external estimator a natively distributed SparkML learner.
D:That underlying principle is directed at a different using spark ml situation: A batch scoring service loads a classifier version from last month while using today's new feature indexer, silently changing categorical positions. The current incident instead turns on sparkML distribution does not make every external estimator a natively distributed SparkML learner.
Question 694
Consider the following fictional Databricks production case: A team trains on curated development features but deploys the same job with credentials that accidentally read unapproved production source tables. Which result is the strongest acceptance criterion after remediation?
- An environment-mapping test resolves only approved datasets and rejects unauthorized paths.
- Production loads a pinned registered version and access audits identify its owner and version history.
- A replay from retained lineage recreates an example prediction within documented tolerance.
- Negative permission tests deny unapproved endpoint and registry changes while allowing permitted experiment logging.
Explanation
Correct Answer
A:This verification measures the required result rather than just process liveness. An environment-mapping test resolves only approved datasets and rejects unauthorized paths. It confirms whether the targeted action succeeded: parameterize approved environment resources and use narrowly scoped credentials per stage.
Incorrect Answers
B:That acceptance test is directed at a different model lifecycle management situation: An engineer stores model weights in a personal workspace folder and treats that path as a production artifact registry with audit history. Here the actual success criterion is an environment-mapping test resolves only approved datasets and rejects unauthorized paths.
C:That acceptance test is directed at a different model lifecycle management situation: An organization has reliable model serving but never archives feature definitions or experiment parameters and cannot reproduce predictions after drift. Here the actual success criterion is an environment-mapping test resolves only approved datasets and rejects unauthorized paths.
D:That acceptance test is directed at a different model lifecycle management situation: A platform team proposes using MLflow Tracking alone to enforce production endpoint authorization, ignoring Unity Catalog and service permissions. Here the actual success criterion is an environment-mapping test resolves only approved datasets and rejects unauthorized paths.
Question 695
A production ML review contains this situation: A model's inference table records output probabilities but omits ground-truth labels, yet reviewers expect immediate precision and recall metrics. Which statement best explains the failure mode?
- Categorical distribution drift needs category frequency or supported categorical distance measures, not only volume.
- Inference profiles are built for prediction monitoring when model ID, outputs, timestamps and optionally labels are recorded.
- Reference populations should match relevant seasonality and business context when interpreting drift.
- An inference profile can monitor inputs and predictions without labels, but true predictive performance requires observed outcomes.
Explanation
Correct Answer
D:The technical constraint is specific to the reported failure. An inference profile can monitor inputs and predictions without labels, but true predictive performance requires observed outcomes. To test it, check label availability, label delay and prediction-to-label matching IDs in the inference data.
Incorrect Answers
A:That underlying principle is directed at a different drift detection and lakehouse monitoring situation: A customer product-category feature acquires many new category values and analysts try to detect the change using only the overall row count. The current incident instead turns on an inference profile can monitor inputs and predictions without labels, but true predictive performance requires observed outcomes.
B:That underlying principle is directed at a different drift detection and lakehouse monitoring situation: A churn model logs requests, predictions, model versions and later ground-truth labels, but monitoring is configured as a generic feature snapshot. The current incident instead turns on an inference profile can monitor inputs and predictions without labels, but true predictive performance requires observed outcomes.
C:That underlying principle is directed at a different drift detection and lakehouse monitoring situation: An online feature monitor compares weekend shopping to weekday reference data and treats expected weekend behavior as a model failure. The current incident instead turns on an inference profile can monitor inputs and predictions without labels, but true predictive performance requires observed outcomes.
Question 696
Consider the following fictional Databricks production case: A new Spark developer calls transform on an unfitted StringIndexer instance and expects learned category labels to appear automatically. Which observation most directly tests the suspected failure boundary?
- Compare the original StringIndexer type with the fitted StringIndexerModel and inspect its label mapping.
- Check the chosen model library and training API for documented distributed support and GPU worker topology.
- Inspect the scoring job for collect and driver memory growth versus distributed transform stages.
- Trace when the indexer fit was called relative to the train/validation split and compare observed label inventories.
Explanation
Correct Answer
A:This is the most discriminating check for the observed symptoms. Compare the original StringIndexer type with the fitted StringIndexerModel and inspect its label mapping. The underlying technical issue is that stringIndexer is an Estimator that must fit its category mapping before its resulting model can transform records.
Incorrect Answers
B:That diagnostic evidence is directed at a different using spark ml situation: A media group has a specialized deep neural model that is not implemented as a SparkML estimator and assumes SparkML will automatically distribute its GPU training. For the stated incident, the decisive observation is compare the original StringIndexer type with the fitted StringIndexerModel and inspect its label mapping.
C:That diagnostic evidence is directed at a different using spark ml situation: A logistics model must score 60 million daily routes already in Delta but the scoring script collects each route into Python before calling predict. For the stated incident, the decisive observation is compare the original StringIndexer type with the fitted StringIndexerModel and inspect its label mapping.
D:That diagnostic evidence is directed at a different using spark ml situation: A credit-risk pipeline fits StringIndexer on the entire dataset before splitting train and holdout, and rare holdout labels influence its learned category mapping. For the stated incident, the decisive observation is compare the original StringIndexer type with the fitted StringIndexerModel and inspect its label mapping.
Question 697
In an enterprise ML workflow, the reported issue is: An engineer stores model weights in a personal workspace folder and treats that path as a production artifact registry with audit history. How should the engineer validate the correction?
- The same signed release revision and dependencies run in test and production with matching fixture output.
- Production loads a pinned registered version and access audits identify its owner and version history.
- An integration fixture confirms the promoted job reads the intended schema before traffic is moved.
- An environment-mapping test resolves only approved datasets and rejects unauthorized paths.
Explanation
Correct Answer
B:This verification measures the required result rather than just process liveness. Production loads a pinned registered version and access audits identify its owner and version history. It confirms whether the targeted action succeeded: use a supported governed model registry and controlled permissions for production model versions.
Incorrect Answers
A:That acceptance test is directed at a different model lifecycle management situation: A regulated company changes deployed Python notebook code directly in production to correct a model transform while its test environment runs a different untracked revision. Here the actual success criterion is production loads a pinned registered version and access audits identify its owner and version history.
C:That acceptance test is directed at a different model lifecycle management situation: A customer segmentation project promotes a scoring job but omits a database migration required by its new feature schema. Here the actual success criterion is production loads a pinned registered version and access audits identify its owner and version history.
D:That acceptance test is directed at a different model lifecycle management situation: A team trains on curated development features but deploys the same job with credentials that accidentally read unapproved production source tables. Here the actual success criterion is production loads a pinned registered version and access audits identify its owner and version history.
Question 698
In an enterprise ML workflow, the reported issue is: A custom PyFunc's predict method assumes every request contains a numeric feature that is optional under the documented inference schema. What likely operational or model-quality harm does this issue create?
- Automation falsely reports a completed rollout while clients still receive temporary failures.
- Clients throw parsing errors or accidentally interpret confidence as the prediction.
- The client amplifies throttling and creates a self-sustaining inference outage.
- A malformed request crashes the entire serving worker or produces an untraceable score.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. A malformed request crashes the entire serving worker or produces an untraceable score. The causal constraint is: Custom model input contracts should validate missing and malformed values before feature transformation.
Incorrect Answers
A:That direct consequence is directed at a different custom model serving situation: A Databricks SDK deployment script sends a model-version update but does not wait for the endpoint to reach a usable ready state before announcing success. The more immediate risk here is a malformed request crashes the entire serving worker or produces an untraceable score.
B:That direct consequence is directed at a different custom model serving situation: A serving client calls the REST endpoint expecting one number but the custom PyFunc returns a structured list of prediction and confidence fields. The more immediate risk here is a malformed request crashes the entire serving worker or produces an untraceable score.
C:That direct consequence is directed at a different custom model serving situation: A real-time client retries every 429 rate-limit error immediately in an unbounded loop and multiplies load on the serving endpoint. The more immediate risk here is a malformed request crashes the entire serving worker or produces an untraceable score.
Question 699
Consider the following fictional Databricks production case: A team logs custom MLflow model weights but forgets to include required Python package versions in the model environment. Which implementation is most appropriate for the stated incident?
- Record compatible dependencies with the PyFunc artifact and validate them in a clean runtime.
- Poll or otherwise verify documented readiness, then perform a permitted smoke prediction before marking the release ready.
- Format requests according to the supported MLflow Deployments client and registered model input schema.
- Update the caller to handle the supported structured response and validate required keys and types.
Explanation
Correct Answer
A:This corrects the relevant failure boundary instead of merely hiding the symptom. Record compatible dependencies with the PyFunc artifact and validate them in a clean runtime. The issue matters because the served model raises ImportError or changes semantics after deployment.
Incorrect Answers
B:That corrective action is directed at a different custom model serving situation: A Databricks SDK deployment script sends a model-version update but does not wait for the endpoint to reach a usable ready state before announcing success. The relevant corrective action here is record compatible dependencies with the PyFunc artifact and validate them in a clean runtime.
C:That corrective action is directed at a different custom model serving situation: A notebook uses the MLflow Deployments SDK to send a column-major payload to a custom model endpoint expecting a different documented dataframe schema. The relevant corrective action here is record compatible dependencies with the PyFunc artifact and validate them in a clean runtime.
D:That corrective action is directed at a different custom model serving situation: A serving client calls the REST endpoint expecting one number but the custom PyFunc returns a structured list of prediction and confidence fields. The relevant corrective action here is record compatible dependencies with the PyFunc artifact and validate them in a clean runtime.
Question 700
In an enterprise ML workflow, the reported issue is: A churn feature job succeeds and a classifier achieves acceptable F1, but the deployed endpoint returns HTTP 400 because request column names differ. What likely operational or model-quality harm does this issue create?
- Predictions drift between restarts because identical labels occupy different vector positions.
- Sensitive events leak into lower-trust test assets while making debugging harder.
- Different releases disagree on whether a model passes the same quality gate.
- A model that looks accurate offline cannot be called successfully by production clients.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. A model that looks accurate offline cannot be called successfully by production clients. The causal constraint is: A full ML integration test should traverse raw feature generation through training, registration, deployment and inference contracts.
Incorrect Answers
A:That direct consequence is directed at a different validation testing situation: A notebook's label encoder produces a different category order depending on an unordered Python set, but tests assert only output length. The more immediate risk here is a model that looks accurate offline cannot be called successfully by production clients.
B:That direct consequence is directed at a different validation testing situation: An engineer uses only production replay data to validate a code change and exposes customer events in an unrestricted test workspace. The more immediate risk here is a model that looks accurate offline cannot be called successfully by production clients.
C:That direct consequence is directed at a different validation testing situation: An engineering group duplicates a validation helper across six notebooks, and two copies enforce outdated threshold values. The more immediate risk here is a model that looks accurate offline cannot be called successfully by production clients.