topic Re: Capture model not training, and 0 available documents for training in Product Forum

Capture model not training, and 0 available documents for training

US0TRBM — Fri, 19 Jul 2024 19:44:33 GMT

We are using Decipher IDP version 2.33 and are trying to train a machine learning model.

We have created the capture model, enabled machine learning on the document type and have assigned it the respective learning model and have updated the model.

To make sure that the capture model is working we set the default value to 50 and began verifying documents, but the capture model still won't update.

Instead, no matter what we try the values don't change from

What are we doing wrong, or not doing so that the machine won't start training?

Re: Capture model not training, and 0 available documents for training

Ben.Lyons1 — Wed, 24 Jul 2024 07:40:10 GMT

Hi,

Have you checked out this topic in the troubleshooting pages?

First thing I would check is machine learning has been enabled in the config file.

Thanks

Re: Capture model not training, and 0 available documents for training

US0TRBM — Wed, 24 Jul 2024 13:28:15 GMT

Yes, we have tried all of the troubleshooting steps and have enabled machine learning in the config file

Re: Capture model not training, and 0 available documents for training

Ben.Lyons1 — Thu, 25 Jul 2024 07:25:02 GMT

Ok, please can you provide screenshots of the Document Type configuration?
Also, what is the Retention Period set to in the overall configuration page?

Re: Capture model not training, and 0 available documents for training

Tejaskumar_Darji — Thu, 25 Jul 2024 08:10:27 GMT

I had a similar observation and could never make these training settings work although it shows a count of available docs to train. The terminology here is confusing.

Re: Capture model not training, and 0 available documents for training

Ben.Lyons1 — Thu, 25 Jul 2024 09:07:31 GMT

Worth checking as well.

The document count is based on those which have been fully processed, this includes the export stage. In v2.2 and older this is automatic as the message is pushed to Rabbit MQ, in 2.3 this is now a manual step. So for a document to count towards the ML training it needs to be marked as complete by the respective VBO action.

Re: Capture model not training, and 0 available documents for training

Yaman__Neameh — Mon, 09 Sep 2024 14:41:23 GMT

Hello Ben, Yes that's right, the data will be trained after marking the Batch as completed using the VBO action. Thx