Hello all,
I have created a structured Document Processing Model with 7 collections. Each collection has around 20 documents. Usually, it used to take 15 to 12 mins to train. But yesterday I just added one document in one of the collections and retrained. But retraining is taking more than 24 hours. Retraining is not completed yet. How to resolve this? Is there any option to stop the retraining? Kindly suggest.
Hi @CedrickB
Thank you for the recommendations! Will follow them accordingly.
I have a doubt here if we add 50 documents per collection, How come it would make the model smaller?
This can happen if your model is too large.
Here are some recommendations to make it smaller:
Add up to 50 documents per collections
Add 2-3 samples per layout type (It is not necessary to train similar layouts from different vendors, same terms, same kind of field positioning)
Training process
1. Train an model adding 10 collections
2. Test various vendors with this model
3. Restart at 1 with vendors not properly captured in 2
Hi @samiak ,
This issue got resolved after 28 hours of retraining with status "Needs Attention." When I checked the model, it is not having the last trained version instead it had a training document that caused the error. But that particular document was trained in the early stage. It should have thrown error earlier. But not sure why this happened. Any comments on this?
Hello,
Could you open a support ticket please. Our support team will unblock you.
We are working on improving this experience.
Thanks,
Samia