AI-102 exam dumps

AI-102 practice question 150 of 493

Designing and Implementing a Microsoft Azure AI Solution. Professional level, Microsoft. Free question with the correct answer and a full explanation.

AI-102 Question 150

Select 3

You have trained a custom vision model using the Custom Vision service in Azure to classify images of different plant species. You now want to make this model available for use in a production environment. What steps should you take to publish the model?

  1. A

    Assign an iteration as the default iteration and publish it with a unique name and prediction resource.

  2. B

    Export the model as a TensorFlow file and deploy it to Azure Kubernetes Service (AKS).

  3. C

    Ensure the prediction resource is configured in the Custom Vision portal before publishing the model.

  4. D

    Create a new prediction endpoint in Azure Cognitive Services and bind it to the model after exporting the iteration.

  5. E

    Use the Custom Vision SDK or REST API to trigger the publishing of the model if not using the portal.

Show answer and explanation

Correct answers: A, C, E

Explanation

To publish a custom vision model, you need to assign an iteration as the default, configure a prediction resource in the Custom Vision portal, and either use the portal or the SDK/REST API to trigger the publishing process. Exporting the model or creating a separate prediction endpoint is not necessary unless you are deploying to an edge or offline environment.

  • A. Correct.

    Correct: Assigning an iteration as the default and publishing it with a unique name and prediction resource is required to make the model accessible via an endpoint.

  • B. Incorrect.

    Incorrect: Exporting the model as a TensorFlow file is an optional step for offline or edge deployment but is not required for publishing it via Azure's Custom Vision service.

  • C. Correct.

    Correct: The prediction resource must be configured in the Custom Vision portal to enable the publishing process.

  • D. Incorrect.

    Incorrect: While Azure Cognitive Services provides prediction endpoints, you do not explicitly create or bind a new prediction endpoint when publishing a model from the Custom Vision portal.

  • E. Correct.

    Correct: The Custom Vision SDK or REST API can be used as an alternative to the portal for publishing the model.

Timed practice exam

Take a AI-102 practice test under exam conditions

40 questions in 120 minutes, drawn from this bank, with a score report and a per-question review when you finish.

Start timed exam