Amazon AWS Certified Machine Learning Engineer - Associate MLA-C01 Practice Test 19
Question 180
While reproducing the alert in an isolated AWS ML integration account, the following condition is observed: A deployment pipeline is denied iam:PassRole for its approved SageMaker execution role, but the proposed fix allows passing every role in the account. The proposed change must preserve logging and least privilege. Which engineering response resolves the actual issue without masking the symptoms?
- Permit iam:PassRole only for the named execution role, constrained to the intended SageMaker service as appropriate.
- Grant only the needed S3 and KMS actions for the approved data prefix and ensure appropriate role trust.
- Tune justified thresholds and group actionable monitor violations by severity without suppressing critical data-integrity checks.
- Use supported managed Spot Training with durable checkpoints for interruptible phases and retain On-Demand capacity when deadlines demand it.
Explanation
Correct Answer
A:Permit iam:PassRole only for the named execution role, constrained to the intended SageMaker service as appropriate. This is the appropriate controlled remedy for iam passrole constraint for model deployment. PassRole authorizes delegation of role credentials to a service; wide permissions can enable privilege escalation. Verify the cause with: Inspect caller identity, pass-role resource ARN, service constraint, and denied event attributes. Then confirm: The pipeline deploys with the authorized role and cannot assign an unrelated privileged role. Reported signal: A deployment pipeline is denied iam:PassRole for its approved SageMaker execution role, but the proposed fix allows passing every role in the account. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
Incorrect Answers
B:Grant only the needed S3 and KMS actions for the approved data prefix and ensure appropriate role trust. This addresses a different issue involving sagemaker execution role s3 least privilege. It does not correct the stated iam passrole constraint for model deployment mechanism. The targeted response is: Permit iam:PassRole only for the named execution role, constrained to the intended SageMaker service as appropriate. Case evidence: A deployment pipeline is denied iam:PassRole for its approved SageMaker execution role, but the proposed fix allows passing every role in the account. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
C:Tune justified thresholds and group actionable monitor violations by severity without suppressing critical data-integrity checks. This addresses a different issue involving model monitor violation alert calibration. It does not correct the stated iam passrole constraint for model deployment mechanism. The targeted response is: Permit iam:PassRole only for the named execution role, constrained to the intended SageMaker service as appropriate. Case evidence: A deployment pipeline is denied iam:PassRole for its approved SageMaker execution role, but the proposed fix allows passing every role in the account. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
D:Use supported managed Spot Training with durable checkpoints for interruptible phases and retain On-Demand capacity when deadlines demand it. This addresses a different issue involving spot training cost versus interruption recovery. It does not correct the stated iam passrole constraint for model deployment mechanism. The targeted response is: Permit iam:PassRole only for the named execution role, constrained to the intended SageMaker service as appropriate. Case evidence: A deployment pipeline is denied iam:PassRole for its approved SageMaker execution role, but the proposed fix allows passing every role in the account. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
Question 181
During an operational load and release rehearsal for SageMaker-hosted inference, the following condition is observed: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Change control requires measurable endpoint or pipeline acceptance evidence. What targeted change should be implemented after verifying the stated cause?
- Grant the build role only required object and key operations for its specific artifact prefix and encryption key.
- Keep the blue environment available until green health gates pass and automate traffic rollback on failed alarms.
- Benchmark supported accelerated inference instances and choose GPU capacity only if latency and throughput gains justify cost.
- Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings.
Explanation
Correct Answer
D:Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. This is the appropriate controlled remedy for asynchronous endpoint for large payloads. Asynchronous inference decouples request submission from long-running processing and can work with larger payload contracts. Verify the cause with: Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency. Then confirm: Large inputs complete without client request timeout and clients receive tracked result locations or failure notifications. Reported signal: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
Incorrect Answers
A:Grant the build role only required object and key operations for its specific artifact prefix and encryption key. This addresses a different issue involving codebuild iam least-privilege artifact upload. It does not correct the stated asynchronous endpoint for large payloads mechanism. The targeted response is: Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. Case evidence: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
B:Keep the blue environment available until green health gates pass and automate traffic rollback on failed alarms. This addresses a different issue involving blue-green endpoint rollback after health failure. It does not correct the stated asynchronous endpoint for large payloads mechanism. The targeted response is: Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. Case evidence: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
C:Benchmark supported accelerated inference instances and choose GPU capacity only if latency and throughput gains justify cost. This addresses a different issue involving gpu versus cpu inference fit. It does not correct the stated asynchronous endpoint for large payloads mechanism. The targeted response is: Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. Case evidence: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
Question 182
While testing a second region for recoverable deployment of the same ML service, the following condition is observed: A pipeline pushes a newly trained model directly to production even though its evaluated recall has fallen below the business threshold. The deployment must be reproducible without manual console-only fixes. What targeted change should be implemented after verifying the stated cause?
- Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance.
- Require model evaluation and explicit approval gates before production deployment, blocking failed metric thresholds.
- Filter appropriate validated data-ready events and trigger retraining only after the required dataset manifest is complete.
- Use a compatible SageMaker multi-model endpoint with managed model loading and monitor cache/eviction behavior.
Explanation
Correct Answer
B:Require model evaluation and explicit approval gates before production deployment, blocking failed metric thresholds. This is the appropriate controlled remedy for codepipeline approval gate on model metrics. A CI/CD stage that records evaluation without enforcing its result is not a release control. Verify the cause with: Inspect pipeline stages, evaluation artifacts, test result propagation, registry approval state, and deployment dependencies. Then confirm: A deliberately poor candidate halts before production while a passing approved candidate can proceed. Reported signal: A pipeline pushes a newly trained model directly to production even though its evaluated recall has fallen below the business threshold. Operational setting: While testing a second region for recoverable deployment of the same ML service.
Incorrect Answers
A:Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance. This addresses a different issue involving sagemaker endpoint auto scaling metric selection. It does not correct the stated codepipeline approval gate on model metrics mechanism. The targeted response is: Require model evaluation and explicit approval gates before production deployment, blocking failed metric thresholds. Case evidence: A pipeline pushes a newly trained model directly to production even though its evaluated recall has fallen below the business threshold. Operational setting: While testing a second region for recoverable deployment of the same ML service.
C:Filter appropriate validated data-ready events and trigger retraining only after the required dataset manifest is complete. This addresses a different issue involving eventbridge conditional retraining trigger. It does not correct the stated codepipeline approval gate on model metrics mechanism. The targeted response is: Require model evaluation and explicit approval gates before production deployment, blocking failed metric thresholds. Case evidence: A pipeline pushes a newly trained model directly to production even though its evaluated recall has fallen below the business threshold. Operational setting: While testing a second region for recoverable deployment of the same ML service.
D:Use a compatible SageMaker multi-model endpoint with managed model loading and monitor cache/eviction behavior. This addresses a different issue involving multi-model endpoint hosting tradeoff. It does not correct the stated codepipeline approval gate on model metrics mechanism. The targeted response is: Require model evaluation and explicit approval gates before production deployment, blocking failed metric thresholds. Case evidence: A pipeline pushes a newly trained model directly to production even though its evaluated recall has fallen below the business threshold. Operational setting: While testing a second region for recoverable deployment of the same ML service.
Question 183
At the model-governance gate before promotion of a new ML artifact, the following condition is observed: A neural network continues training for many epochs after validation loss begins increasing while training loss keeps falling. The team needs a defensible technical decision rather than an untracked workaround. Select TWO: the appropriate diagnostic check and the targeted implementation change.
- Configure SageMaker Automatic Model Tuning with a bounded search space, objective metric, budget, and justified Bayesian or random strategy.
- Inspect tensor summaries, gradient magnitudes, learning-rate schedule, and training loss anomalies with supported debugging hooks.
- Inspect epoch-level training and validation curves, stopping criteria, and the best checkpoint location.
- Enable patience-based early stopping and restore the checkpoint with the best validation metric.
- Compare a more interpretable model or add tested SageMaker Clarify explainability to the decision process.
Explanation
Correct Answers
C:Inspect epoch-level training and validation curves, stopping criteria, and the best checkpoint location. This directly tests early stopping on validation plateau in the reported AWS ML workflow. More epochs are not always better; validation loss can worsen while training loss continues to decrease. It distinguishes the problem before choosing a change. The related targeted response is: Enable patience-based early stopping and restore the checkpoint with the best validation metric. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: A neural network continues training for many epochs after validation loss begins increasing while training loss keeps falling. Operational setting: At the model-governance gate before promotion of a new ML artifact.
D:Enable patience-based early stopping and restore the checkpoint with the best validation metric. This is the appropriate controlled remedy for early stopping on validation plateau. More epochs are not always better; validation loss can worsen while training loss continues to decrease. Verify the cause with: Inspect epoch-level training and validation curves, stopping criteria, and the best checkpoint location. Then confirm: Training stops before extended overfitting and the retained model corresponds to the best validated epoch. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: A neural network continues training for many epochs after validation loss begins increasing while training loss keeps falling. Operational setting: At the model-governance gate before promotion of a new ML artifact.
Incorrect Answers
A:Configure SageMaker Automatic Model Tuning with a bounded search space, objective metric, budget, and justified Bayesian or random strategy. This selection concerns hyperparameter bayesian tuning strategy rather than the required diagnostic-and-remedy pair for early stopping on validation plateau. Use evidence: Inspect epoch-level training and validation curves, stopping criteria, and the best checkpoint location. and the matching correction: Enable patience-based early stopping and restore the checkpoint with the best validation metric. Signal: A neural network continues training for many epochs after validation loss begins increasing while training loss keeps falling. Operational setting: At the model-governance gate before promotion of a new ML artifact.
B:Inspect tensor summaries, gradient magnitudes, learning-rate schedule, and training loss anomalies with supported debugging hooks. This selection concerns model debugger exploding gradient diagnosis rather than the required diagnostic-and-remedy pair for early stopping on validation plateau. Use evidence: Inspect epoch-level training and validation curves, stopping criteria, and the best checkpoint location. and the matching correction: Enable patience-based early stopping and restore the checkpoint with the best validation metric. Signal: A neural network continues training for many epochs after validation loss begins increasing while training loss keeps falling. Operational setting: At the model-governance gate before promotion of a new ML artifact.
E:Compare a more interpretable model or add tested SageMaker Clarify explainability to the decision process. This selection concerns interpretability for high-stakes prediction rather than the required diagnostic-and-remedy pair for early stopping on validation plateau. Use evidence: Inspect epoch-level training and validation curves, stopping criteria, and the best checkpoint location. and the matching correction: Enable patience-based early stopping and restore the checkpoint with the best validation metric. Signal: A neural network continues training for many epochs after validation loss begins increasing while training loss keeps falling. Operational setting: At the model-governance gate before promotion of a new ML artifact.
Question 184
While reproducing the alert in an isolated AWS ML integration account, the following condition is observed: A release pipeline downloads an ML model artifact from a mutable object key and deploys it without checking whether its bytes changed since approval. The proposed change must preserve logging and least privilege. What should the ML engineer inspect FIRST to distinguish this failure from unrelated issues?
- Inspect tuning max jobs and parallel jobs, usage forecasts, AWS Budgets thresholds, and quota settings.
- Inspect S3 object encryption, key policy, grant permissions, IAM role, and denied KMS decrypt events.
- Compare evaluated artifact hash, S3 object version, package provenance, and deployment manifest.
- Inspect resource tags, cost allocation tag activation, Cost Explorer grouping, and unallocated resource types.
Explanation
Correct Answer
C:Compare evaluated artifact hash, S3 object version, package provenance, and deployment manifest. This directly tests model artifact supply-chain integrity in the reported AWS ML workflow. Object names alone do not guarantee model provenance; supply-chain checks bind deployment to reviewed bytes. It distinguishes the problem before choosing a change. The related targeted response is: Pin an immutable approved artifact version or content digest and verify integrity before deployment. Reported signal: A release pipeline downloads an ML model artifact from a mutable object key and deploys it without checking whether its bytes changed since approval. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
Incorrect Answers
A:Inspect tuning max jobs and parallel jobs, usage forecasts, AWS Budgets thresholds, and quota settings. This is a possible diagnostic for aws budgets spend threshold for experimentation, but it does not directly test the model artifact supply-chain integrity failure reported here. Prioritize: Compare evaluated artifact hash, S3 object version, package provenance, and deployment manifest. The relevant case is: A release pipeline downloads an ML model artifact from a mutable object key and deploys it without checking whether its bytes changed since approval. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
B:Inspect S3 object encryption, key policy, grant permissions, IAM role, and denied KMS decrypt events. This is a possible diagnostic for s3 sse-kms cross-account model artifact access, but it does not directly test the model artifact supply-chain integrity failure reported here. Prioritize: Compare evaluated artifact hash, S3 object version, package provenance, and deployment manifest. The relevant case is: A release pipeline downloads an ML model artifact from a mutable object key and deploys it without checking whether its bytes changed since approval. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
D:Inspect resource tags, cost allocation tag activation, Cost Explorer grouping, and unallocated resource types. This is a possible diagnostic for tag-based ml training cost attribution, but it does not directly test the model artifact supply-chain integrity failure reported here. Prioritize: Compare evaluated artifact hash, S3 object version, package provenance, and deployment manifest. The relevant case is: A release pipeline downloads an ML model artifact from a mutable object key and deploys it without checking whether its bytes changed since approval. Operational setting: While reproducing the alert in an isolated AWS ML integration account.
Question 185
During an approved rollout of an AWS ML inference workload, the following condition is observed: An online classifier faces bursty demand and latency spikes because auto scaling tracks host CPU while requests spend most time waiting for accelerated inference. The team must meet its documented availability and rollback constraints. Which TWO steps best establish the evidence and correct the identified root cause? Select TWO.
- Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance.
- Set adequate minimum capacity, scaling limits, alarms, and request backpressure for the expected traffic ramp.
- Require model evaluation and explicit approval gates before production deployment, blocking failed metric thresholds.
- Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy.
- Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency.
Explanation
Correct Answers
A:Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance. This is the appropriate controlled remedy for sagemaker endpoint auto scaling metric selection. CPU percentage can be poorly correlated with accelerator-bound inference demand; scale on a meaningful signal. Verify the cause with: Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy. Then confirm: Capacity responds to peak request load and p95 latency improves without continual scale oscillation. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: An online classifier faces bursty demand and latency spikes because auto scaling tracks host CPU while requests spend most time waiting for accelerated inference. Operational setting: During an approved rollout of an AWS ML inference workload.
D:Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy. This directly tests sagemaker endpoint auto scaling metric selection in the reported AWS ML workflow. CPU percentage can be poorly correlated with accelerator-bound inference demand; scale on a meaningful signal. It distinguishes the problem before choosing a change. The related targeted response is: Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: An online classifier faces bursty demand and latency spikes because auto scaling tracks host CPU while requests spend most time waiting for accelerated inference. Operational setting: During an approved rollout of an AWS ML inference workload.
Incorrect Answers
B:Set adequate minimum capacity, scaling limits, alarms, and request backpressure for the expected traffic ramp. This selection concerns endpoint scale-to-zero cold capacity planning rather than the required diagnostic-and-remedy pair for sagemaker endpoint auto scaling metric selection. Use evidence: Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy. and the matching correction: Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance. Signal: An online classifier faces bursty demand and latency spikes because auto scaling tracks host CPU while requests spend most time waiting for accelerated inference. Operational setting: During an approved rollout of an AWS ML inference workload.
C:Require model evaluation and explicit approval gates before production deployment, blocking failed metric thresholds. This selection concerns codepipeline approval gate on model metrics rather than the required diagnostic-and-remedy pair for sagemaker endpoint auto scaling metric selection. Use evidence: Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy. and the matching correction: Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance. Signal: An online classifier faces bursty demand and latency spikes because auto scaling tracks host CPU while requests spend most time waiting for accelerated inference. Operational setting: During an approved rollout of an AWS ML inference workload.
E:Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency. This selection concerns asynchronous endpoint for large payloads rather than the required diagnostic-and-remedy pair for sagemaker endpoint auto scaling metric selection. Use evidence: Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy. and the matching correction: Configure a justified endpoint variant auto scaling policy using a workload-relevant supported metric such as invocations per instance. Signal: An online classifier faces bursty demand and latency spikes because auto scaling tracks host CPU while requests spend most time waiting for accelerated inference. Operational setting: During an approved rollout of an AWS ML inference workload.
Question 186
During an operational load and release rehearsal for SageMaker-hosted inference, the following condition is observed: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Change control requires measurable endpoint or pipeline acceptance evidence. Which TWO steps best establish the evidence and correct the identified root cause? Select TWO.
- Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy.
- Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings.
- Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency.
- Define a parameterized SageMaker Pipeline that orders steps and records source data, code and model artifacts.
- Use SageMaker Serverless Inference where the model and latency envelope fit, or justify provisioned capacity if it does not.
Explanation
Correct Answers
B:Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. This is the appropriate controlled remedy for asynchronous endpoint for large payloads. Asynchronous inference decouples request submission from long-running processing and can work with larger payload contracts. Verify the cause with: Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency. Then confirm: Large inputs complete without client request timeout and clients receive tracked result locations or failure notifications. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
C:Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency. This directly tests asynchronous endpoint for large payloads in the reported AWS ML workflow. Asynchronous inference decouples request submission from long-running processing and can work with larger payload contracts. It distinguishes the problem before choosing a change. The related targeted response is: Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
Incorrect Answers
A:Compare invocation count per instance, model latency, GPU utilization, and the behavior of current scaling policy. This selection concerns sagemaker endpoint auto scaling metric selection rather than the required diagnostic-and-remedy pair for asynchronous endpoint for large payloads. Use evidence: Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency. and the matching correction: Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. Signal: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
D:Define a parameterized SageMaker Pipeline that orders steps and records source data, code and model artifacts. This selection concerns sagemaker pipelines parameterized retraining rather than the required diagnostic-and-remedy pair for asynchronous endpoint for large payloads. Use evidence: Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency. and the matching correction: Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. Signal: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
E:Use SageMaker Serverless Inference where the model and latency envelope fit, or justify provisioned capacity if it does not. This selection concerns serverless inference for intermittent requests rather than the required diagnostic-and-remedy pair for asynchronous endpoint for large payloads. Use evidence: Inspect payload sizes, model processing duration, caller timeout limits, and request concurrency. and the matching correction: Use SageMaker asynchronous inference with S3 input/output and suitable queue and notification settings. Signal: A document embedding application has payloads too large for its synchronous inference path and some requests need minutes of processing. Operational setting: During an operational load and release rehearsal for SageMaker-hosted inference.
Question 187
During production evaluation of a SageMaker AI model candidate, the following condition is observed: A deep model begins producing NaN loss midway through training and the team increases epochs instead of checking gradient stability. The release is blocked until the underlying model behavior is understood. Which engineering response resolves the actual issue without masking the symptoms?
- Enable patience-based early stopping and restore the checkpoint with the best validation metric.
- Select a decision threshold using recall or cost-aware precision/recall targets and measure on the untouched holdout.
- Select a regression approach with suitable loss and evaluation metrics instead of forcing continuous amounts into unrelated classes.
- Use appropriate SageMaker Debugger or framework instrumentation; adjust learning rate or gradient clipping after confirming instability.
Explanation
Correct Answer
D:Use appropriate SageMaker Debugger or framework instrumentation; adjust learning rate or gradient clipping after confirming instability. This is the appropriate controlled remedy for model debugger exploding gradient diagnosis. Divergent gradients can cause nonfinite optimization; more epochs do not repair an unstable update rule. Verify the cause with: Inspect tensor summaries, gradient magnitudes, learning-rate schedule, and training loss anomalies with supported debugging hooks. Then confirm: Gradient norms stay within the chosen range and training loss remains finite under representative batches. Reported signal: A deep model begins producing NaN loss midway through training and the team increases epochs instead of checking gradient stability. Operational setting: During production evaluation of a SageMaker AI model candidate.
Incorrect Answers
A:Enable patience-based early stopping and restore the checkpoint with the best validation metric. This addresses a different issue involving early stopping on validation plateau. It does not correct the stated model debugger exploding gradient diagnosis mechanism. The targeted response is: Use appropriate SageMaker Debugger or framework instrumentation; adjust learning rate or gradient clipping after confirming instability. Case evidence: A deep model begins producing NaN loss midway through training and the team increases epochs instead of checking gradient stability. Operational setting: During production evaluation of a SageMaker AI model candidate.
B:Select a decision threshold using recall or cost-aware precision/recall targets and measure on the untouched holdout. This addresses a different issue involving precision versus recall fraud threshold. It does not correct the stated model debugger exploding gradient diagnosis mechanism. The targeted response is: Use appropriate SageMaker Debugger or framework instrumentation; adjust learning rate or gradient clipping after confirming instability. Case evidence: A deep model begins producing NaN loss midway through training and the team increases epochs instead of checking gradient stability. Operational setting: During production evaluation of a SageMaker AI model candidate.
C:Select a regression approach with suitable loss and evaluation metrics instead of forcing continuous amounts into unrelated classes. This addresses a different issue involving regression versus classification problem formulation. It does not correct the stated model debugger exploding gradient diagnosis mechanism. The targeted response is: Use appropriate SageMaker Debugger or framework instrumentation; adjust learning rate or gradient clipping after confirming instability. Case evidence: A deep model begins producing NaN loss midway through training and the team increases epochs instead of checking gradient stability. Operational setting: During production evaluation of a SageMaker AI model candidate.
Question 188
During a controlled replay of the data-preparation workflow in a separate ML staging account, the following condition is observed: A real-time feature consumer lags only for one tenant because producers use the same partition key for most high-volume events. The dataset contract must remain reproducible for a backfill. What postchange observation provides the strongest acceptance evidence?
- The transformed distribution is less skewed and inference uses identical guarded transformations.
- Application p95 latency returns to target while exported records meet a measured completeness and staleness objective.
- Write throttles and iterator age fall under burst load without losing required ordering for related records.
- Memory use stays within limits and previously unseen categories are processed without dimension mismatch.
Explanation
Correct Answer
C:Write throttles and iterator age fall under burst load without losing required ordering for related records. This is the acceptance evidence for the described kinesis data streams shard hot key scenario. Kinesis shard throughput is bounded; a hot partition key can overload one shard despite spare total stream capacity. The corrective implementation it validates is: Choose partition keys with enough cardinality or reshard the stream for the intended throughput, while preserving required per-key ordering. Reported signal: A real-time feature consumer lags only for one tenant because producers use the same partition key for most high-volume events. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
Incorrect Answers
A:The transformed distribution is less skewed and inference uses identical guarded transformations. This could validate skewed numerical feature log transformation, but it does not establish that kinesis data streams shard hot key is resolved for this workload. Confirm instead: Write throttles and iterator age fall under burst load without losing required ordering for related records. Observed problem: A real-time feature consumer lags only for one tenant because producers use the same partition key for most high-volume events. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
B:Application p95 latency returns to target while exported records meet a measured completeness and staleness objective. This could validate rds extraction overload into ml, but it does not establish that kinesis data streams shard hot key is resolved for this workload. Confirm instead: Write throttles and iterator age fall under burst load without losing required ordering for related records. Observed problem: A real-time feature consumer lags only for one tenant because producers use the same partition key for most high-volume events. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
D:Memory use stays within limits and previously unseen categories are processed without dimension mismatch. This could validate high-cardinality categorical encoding, but it does not establish that kinesis data streams shard hot key is resolved for this workload. Confirm instead: Write throttles and iterator age fall under burst load without losing required ordering for related records. Observed problem: A real-time feature consumer lags only for one tenant because producers use the same partition key for most high-volume events. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
Question 189
During a controlled replay of the data-preparation workflow in a separate ML staging account, the following condition is observed: Several distributed workers repeatedly read small shared checkpoints with POSIX file semantics, but the pipeline treats object-storage paths as a mutable shared filesystem. The dataset contract must remain reproducible for a backfill. Select TWO: the appropriate diagnostic check and the targeted implementation change.
- Place shared POSIX-style working files on suitable EFS or FSx storage and keep immutable large artifacts in S3.
- Inspect AWS Glue Data Quality rule evaluations, missing-key rates, source partitions, and prior accepted baselines.
- Inspect workload read/write semantics, shared access requirements, filesystem operations, and object-store request patterns.
- Revise instructions, route ambiguous samples for expert adjudication, and monitor label quality per class.
- Increase appropriate retention before long maintenance and persist a durable downstream historical copy for replay beyond retention.
Explanation
Correct Answers
A:Place shared POSIX-style working files on suitable EFS or FSx storage and keep immutable large artifacts in S3. This is the appropriate controlled remedy for s3 versus efs training storage choice. S3 object semantics differ from mounted shared filesystems; choose storage for workload access behavior. Verify the cause with: Inspect workload read/write semantics, shared access requirements, filesystem operations, and object-store request patterns. Then confirm: Concurrent workers observe consistent supported shared-file operations and training artifact delivery remains durable. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: Several distributed workers repeatedly read small shared checkpoints with POSIX file semantics, but the pipeline treats object-storage paths as a mutable shared filesystem. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
C:Inspect workload read/write semantics, shared access requirements, filesystem operations, and object-store request patterns. This directly tests s3 versus efs training storage choice in the reported AWS ML workflow. S3 object semantics differ from mounted shared filesystems; choose storage for workload access behavior. It distinguishes the problem before choosing a change. The related targeted response is: Place shared POSIX-style working files on suitable EFS or FSx storage and keep immutable large artifacts in S3. This is one part of a select-two decision; the matching evidence and corrective action are both required. Reported signal: Several distributed workers repeatedly read small shared checkpoints with POSIX file semantics, but the pipeline treats object-storage paths as a mutable shared filesystem. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
Incorrect Answers
B:Inspect AWS Glue Data Quality rule evaluations, missing-key rates, source partitions, and prior accepted baselines. This selection concerns glue data quality non-null contract rather than the required diagnostic-and-remedy pair for s3 versus efs training storage choice. Use evidence: Inspect workload read/write semantics, shared access requirements, filesystem operations, and object-store request patterns. and the matching correction: Place shared POSIX-style working files on suitable EFS or FSx storage and keep immutable large artifacts in S3. Signal: Several distributed workers repeatedly read small shared checkpoints with POSIX file semantics, but the pipeline treats object-storage paths as a mutable shared filesystem. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
D:Revise instructions, route ambiguous samples for expert adjudication, and monitor label quality per class. This selection concerns ground truth labeling agreement review rather than the required diagnostic-and-remedy pair for s3 versus efs training storage choice. Use evidence: Inspect workload read/write semantics, shared access requirements, filesystem operations, and object-store request patterns. and the matching correction: Place shared POSIX-style working files on suitable EFS or FSx storage and keep immutable large artifacts in S3. Signal: Several distributed workers repeatedly read small shared checkpoints with POSIX file semantics, but the pipeline treats object-storage paths as a mutable shared filesystem. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
E:Increase appropriate retention before long maintenance and persist a durable downstream historical copy for replay beyond retention. This selection concerns kinesis consumer retention gap rather than the required diagnostic-and-remedy pair for s3 versus efs training storage choice. Use evidence: Inspect workload read/write semantics, shared access requirements, filesystem operations, and object-store request patterns. and the matching correction: Place shared POSIX-style working files on suitable EFS or FSx storage and keep immutable large artifacts in S3. Signal: Several distributed workers repeatedly read small shared checkpoints with POSIX file semantics, but the pipeline treats object-storage paths as a mutable shared filesystem. Operational setting: During a controlled replay of the data-preparation workflow in a separate ML staging account.
Question 190
During an approved rollout of an AWS ML inference workload, the following condition is observed: An inference service needs a separate preprocessing transform before a predictor, but the team inserts inconsistent transformation code into many client applications. The team must meet its documented availability and rollback constraints. Which measured outcome would confirm that the correction worked as intended?
- The deployment loads artifacts privately and blocked public routes remain unused.
- The endpoint applies the same preprocessor to every client request and produces the expected model input schema.
- Large inputs complete without client request timeout and clients receive tracked result locations or failure notifications.
- Only the intended completed data batch launches one controlled training workflow, while unrelated objects do not.
Explanation
Correct Answer
B:The endpoint applies the same preprocessor to every client request and produces the expected model input schema. This is the acceptance evidence for the described multi-container inference pipeline responsibilities scenario. A managed inference pipeline chains preprocessing and prediction steps consistently, reducing client-side skew. The corrective implementation it validates is: Use a compatible SageMaker inference pipeline with ordered containers or a single packaged equivalent for shared preprocessing. Reported signal: An inference service needs a separate preprocessing transform before a predictor, but the team inserts inconsistent transformation code into many client applications. Operational setting: During an approved rollout of an AWS ML inference workload.
Incorrect Answers
A:The deployment loads artifacts privately and blocked public routes remain unused. This could validate sagemaker vpc endpoint access for private artifacts, but it does not establish that multi-container inference pipeline responsibilities is resolved for this workload. Confirm instead: The endpoint applies the same preprocessor to every client request and produces the expected model input schema. Observed problem: An inference service needs a separate preprocessing transform before a predictor, but the team inserts inconsistent transformation code into many client applications. Operational setting: During an approved rollout of an AWS ML inference workload.
C:Large inputs complete without client request timeout and clients receive tracked result locations or failure notifications. This could validate asynchronous endpoint for large payloads, but it does not establish that multi-container inference pipeline responsibilities is resolved for this workload. Confirm instead: The endpoint applies the same preprocessor to every client request and produces the expected model input schema. Observed problem: An inference service needs a separate preprocessing transform before a predictor, but the team inserts inconsistent transformation code into many client applications. Operational setting: During an approved rollout of an AWS ML inference workload.
D:Only the intended completed data batch launches one controlled training workflow, while unrelated objects do not. This could validate eventbridge conditional retraining trigger, but it does not establish that multi-container inference pipeline responsibilities is resolved for this workload. Confirm instead: The endpoint applies the same preprocessor to every client request and produces the expected model input schema. Observed problem: An inference service needs a separate preprocessing transform before a predictor, but the team inserts inconsistent transformation code into many client applications. Operational setting: During an approved rollout of an AWS ML inference workload.
