Document Manager pipeline has been running one of my jobs for over 48 hours. It is “Training” and only has 35 documents with about 15-60 lines per document. Why is it taking so long? We tried creating a new one yesterday and that one is still running as well.
Hi! Thanks so much for the reply.
1.The skill is the UiPath Document Understanding for Invoices Skill that has been retrained with additional fields
2. We have not tried enabling GPU yet
GPU really helped! Thank you!
I have 2 more questions, not sure if you know the answers but…
- What happens if I don’t press “eval set” on an import but use it for an eval run?
- Can I grab fields on an invoice multiple times? Example, I have an invoice and need to use a service period multiple times but it’s only listed once on the bill. Can I only use it on one “line”?
- You cannot use same dataset for train and eval…you need to mark dataset as eval to use it
- Why you want to multiply the same field rather extract once and use anymany times as you need
Cheers
Thank you!
For extracting once and using many times, how would we do that? We have meters and service periods on bills that need to be used next to each charge. How I see it right now is if meter and service period is only used once on the bill, I can only use it once in document manager. If we can use it multiple times without having much to do manually, that would be great.
So in the attached, we would need that meter # to attach to every single line on this bill as this is how our client prefers it. But since the meter # is only listed in one spot, I can only use it on one line (in the pic, line 1). So once this data transfers to the front end, as it stands now, we would have to manually add that meter # to all the additional lines. I’m curious if there’s a way to attach that meter # to each line (basic charge, electric energy, electric cons. program, etc) even though that meter # is only used once on the bill.
ok thank you! That’s what we were thinking but wanted to see if there ways a way around it. Last question…I think. Is there a way to do less than 10 images for eval? I get 10 images/documents to train the bot what to look for but you would think you could import as many as one to “test” it with the eval.
Its advised to doe a 80 w0 ratio…but you can do with one as well…eval is not a must…as the name suggests it just evaluates or gives the results based on the trained model to check the results before opening it in studio
Cheers