Skip to Content

AI-102: How to Implement Azure Custom Speech CI/CD with GitHub Actions?

By: Author Alex Lim

Posted on

Categories Exam

Master Azure AI-102 exam DevOps requirements with our breakdown of 5 essential steps for Custom Speech pipelines using GitHub Actions. Learn service principal setup, WER optimization, and CI/CD workflow implementation for certification success – includes Azure Speech resource configuration templates.

Table of Contents

Question

Your organization, Verigon Inc., is developing a suite of innovative applications for custom speech to be used with common software engineering practices. You want to create a DevOps solution for custom speech using GitHub Actions.

Which five steps should you perform in order?

Unordered Choices:

  • When the model’s Word Error Rate (WER) improves, execute the CD workflow to create an endpoint.
  • Update training and testing data and test the data changes with a temporary development model.
  • Create a service principal for the CI/CD workflows for the GitHub Actions and necessary Azure resources.
  • Create a copy of the Speech DevOps template from the GitHub repository to your GitHub account.
  • Raise a pull request and use the GitHub Actions Cl workflow to train the models.

Answer

Correct Order:

  1. Create a copy of the Speech DevOps template from the GitHub repository to your GitHub account.
  2. Create a service principal for the CI/CD workflows for the GitHub Actions and necessary Azure resources.
  3. Update training and testing data and test the data changes with a temporary development model.
  4. Raise a pull request and use the GitHub Actions CI workflow to train the models.
  5. When the model’s Word Error Rate (WER) improves, execute the CD workflow to create an endpoint.

Explanation

You would create a copy of the Speech DevOps template from the GitHub repository to your GitHub account.

You would create a service principal for the continuous integration/continuous delivery (CI/CD) workflows for the GitHub Actions and necessary Azure resources.

You would then update training and testing data, test the data changes with a temporary development model, and raise a pull request to review the data changes.

Once the pull request has retrieved the updated training data, you would use the GitHub Actions CI workflow to train the models.

Test the accuracy of the model’s Word Error Rate (WER), and store the test results in an Azure Blob.

When the Word Error Rate (WER) improves, execute the CD workflow to create an endpoint. The WER should be within 5-10% to ensure good quality. If the WER is 20%, the rate is acceptable but might require more training.

You would move the AI Voice Cloning template to your GitHub account. This template is used for speaker verification and cloning the voice of an individual. You cannot move a template from GitHub, only copy one.

Microsoft Azure AI Engineer Associate AI-102 certification exam practice question and answer (Q&A) dump with detail explanation and reference available free, helpful to pass the Microsoft Azure AI Engineer Associate AI-102 exam and earn Microsoft Azure AI Engineer Associate AI-102 certification.