An AI practitioner is trying to choose an inference method (real-time / asynchronous / batch, and so on) suited to a use case. Which TWO of the following are appropriate considerations for choosing a method? (Choose TWO.)
Select two considerations for choosing the inference method.
Whether the training data has labels
Whether labels exist is a consideration for deciding the learning approach (supervised / unsupervised).
It does not relate to choosing which method to use for inference (real-time / asynchronous / batch) with a trained model, so it is incorrect.
The required response speed (immediate, or can it wait)
This is correct. Whether an immediate response is needed or the result can be received later (the acceptable latency) is an important consideration for choosing real-time / asynchronous / batch.
The setting value of the temperature parameter
Temperature is an inference parameter that adjusts the randomness of generative AI's output.
It is a setting at inference time, but it is not a consideration that decides the choice of the inference method itself (real-time / asynchronous / batch), so it is incorrect.
The size of the input data and the time it takes to process
This is correct. Whether each input is large and slow to process, or small and light, is an important consideration for choosing real-time / asynchronous / batch.
The number of epochs during training
The number of epochs is the setting for how many times the entire dataset is repeated during training.
It is a training setting, at a different stage from choosing the inference method, so it is incorrect.
The criteria for choosing an inference method are 'the response speed (immediate, or can it wait),' the 'size of the input data and processing time,' and 'whether traffic is steady or intermittent.' Immediate + steady → real-time, intermittent → serverless, large input + a few minutes OK → asynchronous, bulk processing → batch. Presence of labels (learning approach), temperature (output adjustment), and number of epochs (training setting) are considerations at a different stage from choosing the inference method.