How to enable neural "neural" build model in Document Intelligence Container?

Question

I'm running a local instance of Azure Document Intelligence (Form Recognizer) Container. Accessing the local URL (http://localhost:5001), I'm able to open the Form Recognizer studio and select a custom model. So, I can see this screen:

User's image

After that, I can label my files. When I'm ready to trigger the train, just the "template" Build Mode option is available. I can train using it, but my main goal is to run the neural one.

User's image

Reading the documentation it is possible to see that the version should be 3.0.

All my images defined in the docker-compose.yml are set to 3.0 (layout, custom-template, and studio). I also checked the build request, it was set for the "2022-08-31" API version.

The Region could also be an issue, but I checked the Region of my resource, and it seems to be compatible with the neural Build Mode.

User's image

Does anyone know if it is possible to train the custom neural extraction model using the local container? What do you think I should do to make it work?

Accepted Answer

Hello @Diego Lima Pereira , Thanks for using Microsoft Q&A Platform.

I believe this is a known behavior. Custom Neural model is not supported in containers. Unfortunately, we don't have any ETA to share with you at this time. These are the supported containers for the document models: User's image

I hope this helps.

Regards,

Vasavi

-Please kindly accept the answer and vote 'yes' if you feel helpful to support the community, thanks.

Share via

How to enable neural "neural" build model in Document Intelligence Container?

0 additional answers