Can we Export ML model somehow which is already trained?

[Tejaskumar_Darji]

Hi,

For multi environment based deployment. Suppose i have trained a model in Dev environment and later lets say we want to move to Prod environment. So is there any option that we can Export trained ML model to other environment?

Or do we have to train model entirely again in new environment?

Thanks.

------------------------------
Tejaskumar Darji
------------------------------

[KrishnaA]

Hello Tejaskumar,

Please refer to the below help link for details.

https://bpdocs.blueprism.com/decipher/user-guide/machine-learning.htm?Highlight=export

Best regards,

------------------------------
Krishna A
Blue Prism
------------------------------

[Tejaskumar_Darji]

Hello,

Thanks Krishna for direct link. I checked that documents. So there is option to Import ML Model but there is no option to export existing ML model. And in documentation "how to import" is mentioned but "how to export" is not mentioned.

Thanks.

------------------------------
Tejaskumar Darji
------------------------------

[PatrickAucoin]

From the Decipher FAQs:

How do I move the Machine learning training data from one install to another?
We expect to have the ability to save the machine learning data in upcoming releases of Decipher IDP. In the meantime, the easiest way is to create a backup of the database used and restore it on the production system. If this does not work for the client, then use the following solution:

1. Create a batch of the type that uses the ML model.
2. When the batch is in data capture in the windows temp folder there should be a DataCaptureDir folder.
3. In this folder or in some of its sub folders there should be a file with the name of the GUID corresponding to the ID of the model in the database. This is the needed file which later can be imported on other systems.


------------------------------
Patrick Aucoin
Senior Product Consultant
Blue Prism
------------------------------

[Tejaskumar_Darji]

Thank you Patrick for the comprehensive answer.

Regards,
Tejas

------------------------------
Tejaskumar Darji
------------------------------