Inference Modes: 10 practice questions

Question 1 of 10

Which inference option is designed for offline processing of large datasets?

1. ABatch transform
2. BLow-latency real-time inference
3. CAsynchronous inference
4. DServerless inference

Question 2 of 10

Which option directly targets interactive workloads with low-latency requirements?

1. AServerless inference
2. BReal-time inference
3. CAsynchronous inference
4. DBatch transform

Question 3 of 10

Which inference option is intended for near-real-time requests with large inputs and longer preprocessing?

1. AServerless inference
2. BBatch transform
3. CAsynchronous inference
4. DLow-latency real-time inference

Question 4 of 10

What traffic pattern is the primary fit for serverless inference?

1. AContinuous offline processing of accumulated datasets
2. BIdle periods between intermittent traffic spurts
3. CSingle requests with payloads approaching one gigabyte
4. DSustained GPU requests through a multi-model endpoint

Question 5 of 10

Which pair of inference options can scale to zero according to the feature matrix?

1. AAsynchronous inference and serverless inference
2. BReal-time inference and batch transform
3. CReal-time inference and serverless inference
4. DBatch transform and asynchronous inference

Question 6 of 10

Which set of payload boundaries is correct?

1. AReal-time at most 100 MB; batch at most 4 MB; asynchronous below 6 MB; serverless at most 1 GB
2. BReal-time below 6 MB; batch at most 100 MB; asynchronous at most 1 GB; serverless at most 4 MB
3. CReal-time at most 4 MB; batch at most 1 GB; asynchronous at most 100 MB; serverless below 6 MB
4. DReal-time at most 1 GB; batch below 6 MB; asynchronous at most 4 MB; serverless at most 100 MB

Question 7 of 10

How do the documented request-duration boundaries compare?

1. AReal-time can run for days; serverless is under one hour; asynchronous and batch are under 60 seconds
2. BReal-time and batch are under one hour; asynchronous can run for days; serverless has no timeout
3. CReal-time is under one hour; batch is under 60 seconds; asynchronous and serverless can run for days
4. DReal-time and serverless are under 60 seconds; asynchronous is under one hour; batch can run for days

Question 8 of 10

An application sends one 700 MB request, preprocessing may take 40 minutes, and the result is needed sooner than an offline job would provide. Which mode fits?

1. ABatch transform, because it is the interactive mode for one long-running request
2. BReal-time inference, because it accepts up to 1 GB and requests under one hour
3. CServerless inference, because it accepts up to 1 GB and requests under one hour
4. DAsynchronous inference, because it accepts up to 1 GB and requests under one hour

Question 9 of 10

A prediction API receives 3 MB requests that finish in 20 seconds, but traffic arrives in brief spurts separated by long idle periods. The team wants capacity to reach zero between spurts. Which mode is the best fit?

1. ABatch transform, because short API requests are accumulated into a live endpoint
2. BReal-time inference, because it scales to zero whenever interactive traffic becomes idle
3. CServerless inference, because the requests fit its limits and the traffic has idle gaps
4. DAsynchronous inference, because requests under 4 MB require its one-hour timeout

Question 10 of 10

A team has an accumulated offline dataset, individual payloads can reach 80 MB, and processing may take several days. No interactive response is needed. Which mode should it choose?

1. ABatch transform, because it supports offline large datasets, payloads up to 100 MB, and jobs lasting days
2. BAsynchronous inference, because it is the offline dataset mode and each job can last several days
3. CReal-time inference, because 80 MB requests remain below its interactive payload boundary
4. DServerless inference, because long idle periods raise its payload limit to 100 MB