Hello,
I've been training a document extraction AI Skill in one region and then deployed the AI Skill to a different region. Interestingly, I'm finding that the document extraction in the first region performs better than the document extraction in the newer region.
Do AI skills have to be retrained in each region, like when you deploy from TEST to PROD?
Thanks!
Discussion posts and replies are publicly visible
AFAIK the trained model is part of the deployment. Could it be that the documents are slightly different?
I see. I've ran more tests using the same document on both models, and I'm still getting a difference in results. Though now, sometimes one model correctly maps a certain field better than another. Perhaps I'm just observing performance variance after the document extraction model is run?
can you run a test case if you get different results if you have two totally seperate trained AI models?
The issue here is, that with each reconciliation step, the model is modified and trained. As soon as you deploy a model to another environment and perform test in different ways, the models will diverge more and more.