Will using the validation station re-train the model?


Just curious, will my usage of the validation station automatically retrain the model when we use it on-prem? Or is that a setting we can change?

Also do you have a timeline for adding multiple languages and more models?



Hi @sonnymeyer,

Training is not possible for now, the models are available as is. We are continuously improving them and adding more fields - please let me know what fields you need and we will try to add them. Beyond that, enabling you to train your own models on premises on arbitrary documents is a capability we are planning for Q1 2020.

As far as languages are concerned, our first priority is to add Japanese. What are the main languages that you need? Also what other document types do you have in mind?



:smile: I am looking forward to the Japanese language! Thank you!


Sorry for the late reply. Being Danish I of course would like the Danish language to appear at some point, however I know that we are small compared to many other languages.

In terms of documents I can’t say for sure right now, but it could be a wide range of laboratory equipment analysis reports or calibration reports of these.


1 Like

Hi Alex,

It´s a great feature.

I need adding the follow fields:
- VAT Vendor.
- Material Code in ítems.

It’s possible for a new realese?

Also, will the restriction of having 2 pages document maximum be improved?


1 Like

Hi @JoseLuis,

Customers that we have talked to use the Vendor name and/or Vendor address to look up the vendor in their database, where they also have the VAT. So this has not been a priority. Also, Material Code sounds like something very specific to your usecase, so this would only be possible by training our ML model on your data.

We are planning to offer this capability of training ML models on premises, in Q1 next year.


Hi @alexcabuz,

As the first quarter of 2020 is passed. Have UIPath now offer training ML models on premises or not?

Hello @AhmadAshfaq,

We have :slight_smile: Please reach out to your UiPath contact and ask about the on-prem Document Understanding options.

Hi Ioana_Gligan

Are you able to provide an example of how to retrain model after validation station?
I’ve gone through DU training but was not able to find any examples.

hello @Annie_Wen,

(sorry for the late reply!)

You will need to

  1. give your ML Extractor an Alias (in the Configure Extractors wizard, under the extractor name in its column)
  2. add a Train Extractors Scope after the Validation Station, with a ML Extractor Trainer (with the same alias as the one above) in it.
    This will collect data in a folder of your choice.
  3. when you have data collected (enough), zip your folder and upload it to Data manager
  4. curate your data in Data Manager
  5. export data from Data Manager for training
  6. train your model.

DM is available in preview in AI Fabric, Cloud or On-Prem - if you have an Enterprise or Enterprise Trial account, please ask your UiPath representative about how to get access to the preview version of DM.

Hope this helps,


I am still unsure about the online retraining of the model in the validation station. Will the model be retrained whenever a user submits a new document to the deployed model, without a need to manually label it in the Data Manager? The Validation Station kind of looks like the Data Manager. Is it possible to enable the online training of the model?

Kind regards,