Continue Deployment Document Intelligence with Github

Question

Hello Community,

I'm exploring ways to integrate Git&GitHub for managing trained models in Azure AI Document Intelligence. I understand that Azure AI Document Intelligence Studio can train a model, let's call it "ModelName," which "MyApp" utilizes for text extraction.

For CI/CD, I plan to track "MyApp" using Git&GitHub, creating a YML file for its deployment. My question is: Is there a method to version control Azure AI Document Intelligence Projects and Trained Models using Git&GitHub? This would enable multiple developers to collaborate on "MyApp" and deploy the latest stable version of "ModelName."

Additionally, Azure AI Language has a 'UseAllTrainingData' setting set to true. Does Azure AI Document Intelligence offer a similar feature and achieve my goal?

Answer 1

Thanks for the question, Azure AI Document Intelligence does not directly support version control of trained models using Git&GitHub.

you can achieve a similar effect by following these steps:

1. Model Serialization: After training a model in Azure AI Document Intelligence, you can serialize the model and save it as a file.
2. Store in Repository: You can then store this file in your GitHub repository.
3. Pull and Deserialize: During the deployment process, you can pull the latest version of the model file from the repository, deserialize it, and load it into your application.

This way, you can keep track of different versions of your model, and multiple developers can collaborate on the project.