I’m facing an issue with my ML Skills, they aren’t extracting any values when I trained them.
Here are the steps I’ve made :
I’ve used an Out-of-the-box Package (9.0).
I’ve created a DataLabeling in which I’ve used a pre-labeling for invoices, but I’ve deleted most of the fields to keep those I’m interested in : Name, Date, Total and Invoice-no. I have 13 invoices in my DataSet.
After this, I’ve trained the package 9.0, which gives me a 9.1 package. The problem is that my Skill isn’t returning any result in my program.
I’ve tried to do it with or without GPU, no difference.
Good point :
My program seems to work because I have results with 9.0 Skill OR when using an endpoint for ML Extractor.
So the problem must be located in the ML Skill or in the DataLabeling ?
Hello, can you please share more information on the error you see and how you are configuring the extraction call inside your workflow?
A couple areas for you to look at and to try:
Make sure you have deployed this v9.1 package to a ML Skill (e.g., let’s call this NewInvoice91)
Make sure your Studio is connected to the Orchestrator instance where this still is running
Try using the “GetCapabilities” feature inside the extractor configuration to make sure you can connect to your “NewInvoice91” skill and also get back the correct model schema
In the document classification, make sure the workflow is able to correctly classify the document type to be of an “invoice” type
In the extractor configuration, make sure you have mapped the appropriate model schema fields to your invoice taxonomy fields
Hi All,
I am facing same issue with the ML skill returning no value/incorrect value or very few values (just 2 header fields are extracted correctly). I have done all these steps (ML extractor - configuration - get capabilities) etc… Refreshed many times. Still it looks like during extraction, it is not taking my ML skill. But it is taking a default ML skill for Invoice.