Google Professional Machine Learning Engineer Practice Test 90
Question 890
During a production ML inference release on Google Cloud, the following condition occurs: A multi-class model returns numeric class indices, but an API gateway uses a stale alphabetical mapping and mislabels predictions. The service must meet its agreed quality and latency requirements. What acceptance result verifies the intended behavior after remediation?
- The running endpoint resolves to a traceable evaluated package and cannot silently switch weights when an alias moves.
- Every controlled sample's returned label matches the model's trained class index with no misordered classes.
- A controlled failed rollout restores the prior model and returns errors below the service objective.
- Previously missed customer IDs retrieve the intended feature values without collapsing distinct leading-zero identifiers.
Explanation
Correct Answer
B:Every controlled sample's returned label matches the model's trained class index with no misordered classes. This is specific acceptance evidence for custom postprocessing label mapping drift. Index ordering is part of the model contract; separate stale label dictionaries can silently corrupt inference. It confirms the controlled correction: Ship the class vocabulary with the model artifact and version or validate the output adapter against it. This matters here because a multi-class model returns numeric class indices, but an API gateway uses a stale alphabetical mapping and….
Incorrect Answers
A:The running endpoint resolves to a traceable evaluated package and cannot silently switch weights when an alias moves. This result could demonstrate success for model registry version provenance but not for the stated custom postprocessing label mapping drift issue. The direct test is: Every controlled sample's returned label matches the model's trained class index with no misordered classes.
C:A controlled failed rollout restores the prior model and returns errors below the service objective. This result could demonstrate success for model traffic rollback on high error rate but not for the stated custom postprocessing label mapping drift issue. The direct test is: Every controlled sample's returned label matches the model's trained class index with no misordered classes.
D:Previously missed customer IDs retrieve the intended feature values without collapsing distinct leading-zero identifiers. This result could demonstrate success for feature online entity key mismatch but not for the stated custom postprocessing label mapping drift issue. The direct test is: Every controlled sample's returned label matches the model's trained class index with no misordered classes. This matters here because a multi-class model returns numeric class indices, but an API gateway uses a stale alphabetical mapping and….
Question 891
During a production ML inference release on Google Cloud, the following condition occurs: A deployment grants every project member broad model prediction permissions merely to make a limited partner integration work. The service must meet its agreed quality and latency requirements. Which underlying engineering principle explains why the original approach fails?
- A valid container image is not necessarily compatible with an inference service's HTTP startup and health contract.
- Online feature serving requires an up-to-date synchronized view; source-table changes need not appear instantly.
- Coercing identifier types can change keys, particularly when numeric-looking strings have meaningful leading zeros.
- Endpoint invoker permissions should follow least privilege, rather than expanding access for convenience.
Explanation
Correct Answer
D:Endpoint invoker permissions should follow least privilege, rather than expanding access for convenience. This explains the mechanism behind public endpoint iam oversharing and why the described failure is possible. A discriminating investigation is: Examine effective IAM bindings, invocation audit logs and the intended partner service account identity. The corresponding repair is: Restrict invoke access to the necessary principal and endpoint, using appropriate short-lived authentication.
Incorrect Answers
A:A valid container image is not necessarily compatible with an inference service's HTTP startup and health contract. This principle describes inference container health interface mismatch, which is not the cause of the observed public endpoint iam oversharing failure. The controlling explanation is: Endpoint invoker permissions should follow least privilege, rather than expanding access for convenience.
B:Online feature serving requires an up-to-date synchronized view; source-table changes need not appear instantly. This principle describes feature store freshness and online sync, which is not the cause of the observed public endpoint iam oversharing failure. The controlling explanation is: Endpoint invoker permissions should follow least privilege, rather than expanding access for convenience.
C:Coercing identifier types can change keys, particularly when numeric-looking strings have meaningful leading zeros. This principle describes feature online entity key mismatch, which is not the cause of the observed public endpoint iam oversharing failure. The controlling explanation is: Endpoint invoker permissions should follow least privilege, rather than expanding access for convenience.
Question 892
At a Google Cloud analytics team preparing a governed pilot, the following condition occurs: A reporting team exports millions of BigQuery rows to a laptop for daily scoring although a trained BigQuery ML model already exists in the dataset. The team must satisfy a reproducible low-code AI assessment. Which TWO selections represent the correct investigation and remediation together? Select TWO.
- Inspect the timestamp ordering, seasonal patterns, horizon requirement, and the BigQuery ML model_type and time-series options.
- Train an appropriately scaled BigQuery ML KMEANS model and assess cluster cohesion, size and business interpretability.
- Use ML.PREDICT on the appropriate BigQuery ML model within a controlled SQL batch job, writing results to governed tables.
- Evaluate Cloud Vision API for supported generic image detection and use custom training only for missing domain-specific classes.
- Review row volume, model location, permissions and the current SQL report dependency on the local scoring script.
Explanation
Correct Answers
C:Use ML.PREDICT on the appropriate BigQuery ML model within a controlled SQL batch job, writing results to governed tables. This is the targeted corrective response for bigquery ml sql batch prediction integration. In-database prediction avoids unnecessary bulk data transfer when model and features are already in BigQuery. Verify the diagnosis using: Review row volume, model location, permissions and the current SQL report dependency on the local scoring script. Confirm success using: The scheduled report processes the full authorized partition and stores traceable predictions without a local export. Here, the relevant situation is: the following condition occurs: A reporting team exports millions of BigQuery rows to a laptop for daily scoring….
E:Review row volume, model location, permissions and the current SQL report dependency on the local scoring script. This inspects evidence relevant to bigquery ml sql batch prediction integration. In-database prediction avoids unnecessary bulk data transfer when model and features are already in BigQuery. After identifying the cause, the matching correction is: Use ML.PREDICT on the appropriate BigQuery ML model within a controlled SQL batch job, writing results to governed tables. Here, the relevant situation is: the following condition occurs: A reporting team exports millions of BigQuery rows to a laptop for daily scoring….
Incorrect Answers
A:Inspect the timestamp ordering, seasonal patterns, horizon requirement, and the BigQuery ML model_type and time-series options. This is an action or diagnostic relevant to bigquery arima_plus forecast horizon, not one of the required two steps. In this scenario verify: Review row volume, model location, permissions and the current SQL report dependency on the local scoring script. Then correct: Use ML.PREDICT on the appropriate BigQuery ML model within a controlled SQL batch job, writing results to governed tables.
B:Train an appropriately scaled BigQuery ML KMEANS model and assess cluster cohesion, size and business interpretability. This is an action or diagnostic relevant to bigquery ml kmeans customer segmentation, not one of the required two steps. In this scenario verify: Review row volume, model location, permissions and the current SQL report dependency on the local scoring script. Then correct: Use ML.PREDICT on the appropriate BigQuery ML model within a controlled SQL batch job, writing results to governed tables.
D:Evaluate Cloud Vision API for supported generic image detection and use custom training only for missing domain-specific classes. This is an action or diagnostic relevant to vision api pretrained image labels versus custom detection, not one of the required two steps. In this scenario verify: Review row volume, model location, permissions and the current SQL report dependency on the local scoring script. Then correct: Use ML.PREDICT on the appropriate BigQuery ML model within a controlled SQL batch job, writing results to governed tables.
Question 893
While preparing a production training workflow on Gemini Enterprise Agent Platform, the following condition occurs: A customer-service model correctly follows general instructions but lacks newly updated internal warranty details, so developers schedule daily weight fine-tuning. Engineers need an auditable, resource-aware correction to the issue. Which evidence most directly establishes what is wrong with this AI or ML workflow?
- Determine whether failures arise from changing reference knowledge, stable behavior patterns or inadequate retrieval coverage.
- Inspect required output type, business loss, sample labels, and the configured model objective and metric.
- Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second.
- Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts.
Explanation
Correct Answer
A:Determine whether failures arise from changing reference knowledge, stable behavior patterns or inadequate retrieval coverage. This inspects evidence relevant to foundation model prompt versus fine-tuning. External reference updates need fresh context; changing foundation weights for every factual revision is inefficient. After identifying the cause, the matching correction is: Use grounded retrieval for frequently changing policy facts and reserve fine-tuning for measured durable behavior adaptation.
Incorrect Answers
B:Inspect required output type, business loss, sample labels, and the configured model objective and metric. This examines regression versus classification product choice rather than the specific mechanism reported here. It is not the decisive first check for foundation model prompt versus fine-tuning. The relevant evidence is: Determine whether failures arise from changing reference knowledge, stable behavior patterns or inadequate retrieval coverage.
C:Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second. This examines distributed gradient communication bottleneck rather than the specific mechanism reported here. It is not the decisive first check for foundation model prompt versus fine-tuning. The relevant evidence is: Determine whether failures arise from changing reference knowledge, stable behavior patterns or inadequate retrieval coverage.
D:Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts. This examines constrained generative json outputs rather than the specific mechanism reported here. It is not the decisive first check for foundation model prompt versus fine-tuning. The relevant evidence is: Determine whether failures arise from changing reference knowledge, stable behavior patterns or inadequate retrieval coverage.
Question 894
During an independent acceptance rehearsal of a customer-facing AI capability, the following condition occurs: An international retailer manually sends each product description to an English-only classifier although the business first needs translated content in several supported languages. The owners require measurable quality and controlled operating cost. Which TWO complementary choices jointly establish the evidence and appropriate correction? Select TWO.
- Evaluate Cloud Vision API for supported generic image detection and use custom training only for missing domain-specific classes.
- Evaluate Agent Platform AutoML tabular training against a simple baseline and select it if quality and operating cost justify the managed path.
- Use Cloud Translation API for supported text translation, then separately validate downstream classification against localized content.
- Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task.
- Inspect source languages, target locale requirements, formatting preservation and translation quality on a multilingual sample.
Explanation
Correct Answers
C:Use Cloud Translation API for supported text translation, then separately validate downstream classification against localized content. This is the targeted corrective response for translate api for multilingual product descriptions. Translation and classification are different tasks; an industry language API directly addresses supported cross-language conversion. Verify the diagnosis using: Inspect source languages, target locale requirements, formatting preservation and translation quality on a multilingual sample. Confirm success using: A bilingual reviewer confirms meaning and formatting for representative descriptions without retraining a translation model. This matters here because during an independent acceptance rehearsal of a customer-facing AI capability, the following condition occurs: An….
E:Inspect source languages, target locale requirements, formatting preservation and translation quality on a multilingual sample. This inspects evidence relevant to translate api for multilingual product descriptions. Translation and classification are different tasks; an industry language API directly addresses supported cross-language conversion. After identifying the cause, the matching correction is: Use Cloud Translation API for supported text translation, then separately validate downstream classification against localized content. This matters here because during an independent acceptance rehearsal of a customer-facing AI capability, the following condition occurs: An….
Incorrect Answers
A:Evaluate Cloud Vision API for supported generic image detection and use custom training only for missing domain-specific classes. This is an action or diagnostic relevant to vision api pretrained image labels versus custom detection, not one of the required two steps. In this scenario verify: Inspect source languages, target locale requirements, formatting preservation and translation quality on a multilingual sample. Then correct: Use Cloud Translation API for supported text translation, then separately validate downstream classification against localized content.
B:Evaluate Agent Platform AutoML tabular training against a simple baseline and select it if quality and operating cost justify the managed path. This is an action or diagnostic relevant to automl tabular supervised model selection, not one of the required two steps. In this scenario verify: Inspect source languages, target locale requirements, formatting preservation and translation quality on a multilingual sample. Then correct: Use Cloud Translation API for supported text translation, then separately validate downstream classification against localized content.
D:Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task. This is an action or diagnostic relevant to model garden task and license selection, not one of the required two steps. In this scenario verify: Inspect source languages, target locale requirements, formatting preservation and translation quality on a multilingual sample. Then correct: Use Cloud Translation API for supported text translation, then separately validate downstream classification against localized content.
Question 895
During a staged restoration test at a secondary serving deployment, the following condition occurs: A GPU inference server shows low CPU use but increasing queue length and slow responses under request bursts; the only scaling rule watches CPU. The team must prove an actual user-visible improvement without weakening access controls. Which targeted engineering change most directly addresses the failure?
- Set a justified warm minimum or scheduled capacity and balance initialization overhead against idle cost.
- Configure an appropriate request-concurrency or workload-driven scaling policy with resource and cooldown limits.
- Use stable cohort-level assignment and record model identity on requests and measured outcomes.
- Test a supported post-training quantization or optimization path and release only if target quality remains acceptable.
Explanation
Correct Answer
B:Configure an appropriate request-concurrency or workload-driven scaling policy with resource and cooldown limits. This is the targeted corrective response for throughput-based autoscaling versus cpu metric. CPU usage can be weakly related to accelerator or queue-bound inference workloads. Verify the diagnosis using: Compare request rate, queue depth, GPU utilization, per-instance concurrency and p95 response latency. Confirm success using: Peak requests trigger additional serving capacity and tail latency recovers without repeated scale oscillations. This matters here because during a staged restoration test at a secondary serving deployment, the following condition occurs: A GPU inference….
Incorrect Answers
A:Set a justified warm minimum or scheduled capacity and balance initialization overhead against idle cost. This action addresses autoscaling minimum capacity versus cold load, not the failure in the described throughput-based autoscaling versus cpu metric case. The targeted correction is: Configure an appropriate request-concurrency or workload-driven scaling policy with resource and cooldown limits.
C:Use stable cohort-level assignment and record model identity on requests and measured outcomes. This action addresses a/b assignment consistency for user journeys, not the failure in the described throughput-based autoscaling versus cpu metric case. The targeted correction is: Configure an appropriate request-concurrency or workload-driven scaling policy with resource and cooldown limits.
D:Test a supported post-training quantization or optimization path and release only if target quality remains acceptable. This action addresses inference optimization using quantization, not the failure in the described throughput-based autoscaling versus cpu metric case. The targeted correction is: Configure an appropriate request-concurrency or workload-driven scaling policy with resource and cooldown limits.
Question 896
During a postdeployment AI safety and reliability review, the following condition occurs: A deployed risk model receives income values far outside the training distribution but the dashboard checks only endpoint availability. The model owner must provide repeatable evidence rather than a cosmetic dashboard change. Which outcome most directly demonstrates a successful, measurable fix?
- Red-team test prompts trigger expected safeguards while legitimate requests continue to receive useful responses.
- A synthetic distribution shift triggers an attributable alert while an unchanged control sample stays within tolerance.
- Precision and recall recover on new labeled periods and monitoring distinguishes performance change from mere input shift.
- For controlled entities the online and offline features match at equivalent cutoffs and quality recovers.
Explanation
Correct Answer
B:A synthetic distribution shift triggers an attributable alert while an unchanged control sample stays within tolerance. This is specific acceptance evidence for input data distribution drift alarm. HTTP health is not evidence that incoming feature distributions still resemble training data. It confirms the controlled correction: Configure Agent Platform Model Monitoring on relevant features with representative baseline and actionable drift alerts. This matters here because a deployed risk model receives income values far outside the training distribution but the dashboard checks only….
Incorrect Answers
A:Red-team test prompts trigger expected safeguards while legitimate requests continue to receive useful responses. This result could demonstrate success for gen ai output safety policy enforcement but not for the stated input data distribution drift alarm issue. The direct test is: A synthetic distribution shift triggers an attributable alert while an unchanged control sample stays within tolerance. This matters here because a deployed risk model receives income values far outside the training distribution but the dashboard checks only….
C:Precision and recall recover on new labeled periods and monitoring distinguishes performance change from mere input shift. This result could demonstrate success for concept drift with stable feature statistics but not for the stated input data distribution drift alarm issue. The direct test is: A synthetic distribution shift triggers an attributable alert while an unchanged control sample stays within tolerance. This matters here because a deployed risk model receives income values far outside the training distribution but the dashboard checks only….
D:For controlled entities the online and offline features match at equivalent cutoffs and quality recovers. This result could demonstrate success for training-serving feature skew detection but not for the stated input data distribution drift alarm issue. The direct test is: A synthetic distribution shift triggers an attributable alert while an unchanged control sample stays within tolerance.
Question 897
While reproducing a training orchestration incident in isolated staging, the following condition occurs: A feature-building pipeline reuses a cached output after source data changed because the component's cache key does not incorporate dataset version. The team requires an idempotent pipeline behavior and traceable evidence. Which configuration or procedure is the most appropriate correction?
- Promote explicit immutable model and image versions with versioned release manifests and review gates.
- Tune component resource requests/limits and stream or partition embeddings rather than materializing every vector at once.
- Include immutable dataset versions and code revisions as explicit component inputs or disable inappropriate caching.
- Package the fitted preprocessing component with the model and apply the same validated transform in both paths.
Explanation
Correct Answer
C:Include immutable dataset versions and code revisions as explicit component inputs or disable inappropriate caching. This is the targeted corrective response for kubeflow component cache invalidation. Pipeline caches assume identical declared inputs; hidden external dependencies can make cached outputs stale. Verify the diagnosis using: Inspect component input parameters, artifact provenance, caching settings and source snapshot identifiers. Confirm success using: Changing the dataset version forces a rebuild while identical inputs may reuse validated cached results. This matters here because while reproducing a training orchestration incident in isolated staging, the following condition occurs: A….
Incorrect Answers
A:Promote explicit immutable model and image versions with versioned release manifests and review gates. This action addresses cloud build release version promotion, not the failure in the described kubeflow component cache invalidation case. The targeted correction is: Include immutable dataset versions and code revisions as explicit component inputs or disable inappropriate caching.
B:Tune component resource requests/limits and stream or partition embeddings rather than materializing every vector at once. This action addresses custom kfp component resource limits, not the failure in the described kubeflow component cache invalidation case. The targeted correction is: Include immutable dataset versions and code revisions as explicit component inputs or disable inappropriate caching.
D:Package the fitted preprocessing component with the model and apply the same validated transform in both paths. This action addresses training serving preprocessing artifact sharing, not the failure in the described kubeflow component cache invalidation case. The targeted correction is: Include immutable dataset versions and code revisions as explicit component inputs or disable inappropriate caching.
Question 898
During a regulated MLOps pipeline deployment, the following condition occurs: A production workflow starts model registration before validation and evaluation steps have finished, occasionally promoting untested weights. The workflow must be repeatable and protect the last approved production model. Which technical explanation best accounts for the reported behavior?
- Orchestration needs explicit data and control dependencies; notebook execution order is not a production DAG.
- Kubernetes resource limits constrain component runtime even when the model's training job itself has spare capacity.
- Pipeline caches assume identical declared inputs; hidden external dependencies can make cached outputs stale.
- Model retraining reproducibility requires stable inputs rather than unversioned reads of an evolving table.
Explanation
Correct Answer
A:Orchestration needs explicit data and control dependencies; notebook execution order is not a production DAG. This explains the mechanism behind pipeline component dependency order and why the described failure is possible. A discriminating investigation is: Inspect the pipeline DAG, dependency declarations, artifact references and execution ordering from recent runs. The corresponding repair is: Add explicit upstream validation and evaluation dependencies before registration and deployment components.
Incorrect Answers
B:Kubernetes resource limits constrain component runtime even when the model's training job itself has spare capacity. This principle describes custom kfp component resource limits, which is not the cause of the observed pipeline component dependency order failure. The controlling explanation is: Orchestration needs explicit data and control dependencies; notebook execution order is not a production DAG.
C:Pipeline caches assume identical declared inputs; hidden external dependencies can make cached outputs stale. This principle describes kubeflow component cache invalidation, which is not the cause of the observed pipeline component dependency order failure. The controlling explanation is: Orchestration needs explicit data and control dependencies; notebook execution order is not a production DAG.
D:Model retraining reproducibility requires stable inputs rather than unversioned reads of an evolving table. This principle describes data version reproducibility in ct jobs, which is not the cause of the observed pipeline component dependency order failure. The controlling explanation is: Orchestration needs explicit data and control dependencies; notebook execution order is not a production DAG.
Question 899
During an independent acceptance rehearsal of a customer-facing AI capability, the following condition occurs: A support assistant includes a full 200-page manual in every prompt despite users asking brief questions, producing excessive latency and cost. The owners require measurable quality and controlled operating cost. What should the engineer inspect before selecting a correction?
- Inspect prompt token usage, retrieval relevance, model context requirements, answer grounding and p95 latency across queries.
- Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task.
- Inspect label availability, feature types, tuning needs, governance requirements, and supported AutoML model interfaces.
- Inspect class definitions, sample error costs, throughput, p95 latency, pricing and published model capability constraints.
Explanation
Correct Answer
A:Inspect prompt token usage, retrieval relevance, model context requirements, answer grounding and p95 latency across queries. This inspects evidence relevant to gemini prompt context token budget. Unfocused context consumes tokens and can distract generation; targeted retrieval can reduce cost without discarding needed facts. After identifying the cause, the matching correction is: Retrieve only relevant passages with provenance and use a concise structured prompt with measured token limits.
Incorrect Answers
B:Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task. This examines model garden task and license selection rather than the specific mechanism reported here. It is not the decisive first check for gemini prompt context token budget. The relevant evidence is: Inspect prompt token usage, retrieval relevance, model context requirements, answer grounding and p95 latency across queries.
C:Inspect label availability, feature types, tuning needs, governance requirements, and supported AutoML model interfaces. This examines automl tabular supervised model selection rather than the specific mechanism reported here. It is not the decisive first check for gemini prompt context token budget. The relevant evidence is: Inspect prompt token usage, retrieval relevance, model context requirements, answer grounding and p95 latency across queries.
D:Inspect class definitions, sample error costs, throughput, p95 latency, pricing and published model capability constraints. This examines gemini model choice for latency versus complexity rather than the specific mechanism reported here. It is not the decisive first check for gemini prompt context token budget. The relevant evidence is: Inspect prompt token usage, retrieval relevance, model context requirements, answer grounding and p95 latency across queries.
Question 900
During model-training validation on a Google Cloud experimental project, the following condition occurs: An existing TensorFlow training pipeline has large dense matrix operations and engineers select a TPU solely from marketing without checking model operation support. The change should preserve a meaningful independent model-quality comparison. Which TWO complementary choices jointly establish the evidence and appropriate correction? Select TWO.
- Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second.
- Benchmark a suitable boosted-tree model against more complex candidates and retain complexity only when justified by held-out gains.
- Benchmark compatible TPU and GPU configurations and select hardware based on measured throughput and target budget.
- Request justified quota or select an available compatible accelerator and region without compromising data residency controls.
- Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost.
Explanation
Correct Answers
C:Benchmark compatible TPU and GPU configurations and select hardware based on measured throughput and target budget. This is the targeted corrective response for tpu versus gpu for tensorflow workload. TPU and GPU performance depends on operations, data pipeline and framework integration, not headline compute alone. Verify the diagnosis using: Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost. Confirm success using: The chosen accelerator meets throughput and convergence expectations without unsupported operations or unstable retracing. This matters here because during model-training validation on a Google Cloud experimental project, the following condition occurs: An existing….
E:Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost. This inspects evidence relevant to tpu versus gpu for tensorflow workload. TPU and GPU performance depends on operations, data pipeline and framework integration, not headline compute alone. After identifying the cause, the matching correction is: Benchmark compatible TPU and GPU configurations and select hardware based on measured throughput and target budget. This matters here because during model-training validation on a Google Cloud experimental project, the following condition occurs: An existing….
Incorrect Answers
A:Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second. This is an action or diagnostic relevant to distributed gradient communication bottleneck, not one of the required two steps. In this scenario verify: Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost. Then correct: Benchmark compatible TPU and GPU configurations and select hardware based on measured throughput and target budget.
B:Benchmark a suitable boosted-tree model against more complex candidates and retain complexity only when justified by held-out gains. This is an action or diagnostic relevant to deep model versus gradient boosted trees on tabular, not one of the required two steps. In this scenario verify: Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost. Then correct: Benchmark compatible TPU and GPU configurations and select hardware based on measured throughput and target budget.
D:Request justified quota or select an available compatible accelerator and region without compromising data residency controls. This is an action or diagnostic relevant to accelerator quota and regional availability, not one of the required two steps. In this scenario verify: Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost. Then correct: Benchmark compatible TPU and GPU configurations and select hardware based on measured throughput and target budget.