search

microsoft / OCR-Form-Tools

A set of tools to use in Microsoft Azure Form Recognizer and OCR services.
MIT License
509 stars 171 forks source link

Bug - Compose Model - Multiple custom trained models #935

Closed prajwal-ainapur closed 2 years ago

prajwal-ainapur commented 3 years ago

In the compose model feature, I tried to compose 7-8 custom trained models, unfortunately no model is being generated nor any error message being shown. I tried to compose 5 models instead and it generated a compose model successfully. Same behavior was observed on FOTT website as well as docker image. Is there any limit on the number of models that could be composed together? If yes, How to go about it as my application require a minimum of 7 models to be composed together.

To Reproduce Steps to reproduce the behavior:

  1. Go to 'Compose feature on FOTT'
  2. Select any '7 custom trained models'
  3. Click on Compose button on top
  4. A loading page appears stating 'model is composing'. Post that, no new model would be created.

Expected behavior After the loading page, a new compose model should have been created and it should have appeared in the list of models.

Desktop (please complete the following information):

prajwal-ainapur commented 2 years ago

Update:

Approached the dev team for form recognizer through mail. The issue seems that under F0 tier only 5 models can be composed whereas in S0 any number of models can be composed together. They seems to have missed to highlight this difference on the documentation.