Training classifiers and extractors from multiple processes


I’d like to build a project split into multiple processes, where each process interacts with the same classifiers and extractors.

What I would like to achieve, is to be able to train the shared classifiers and extractors from each of the processes. So if Process 1 performs the training, this strengthens the same classifiers and extractors that are used by process 2, and vice versa.

I imagine to achieve this, we need to have a shared taxonomy file and shared learning files for classifiers.

But what about extractors? There are no learning files for those. Does it mean that the training gets saved via the endpoint and is stored remotely against the API key? Or is there a file stored somewhere that needs to be shared similar to classifiers?

Hi @totoro84

When using UiPath’s Document Understanding framework with classifiers and extractors, the training data for classifiers is typically stored in learning files, while the training for extractors is performed through the API endpoint and is stored remotely.

For classifiers, you can have shared learning files that multiple processes can access and use for training. These learning files contain the training data and are usually stored in a shared location or version control system. Each process can read and update these learning files, allowing them to contribute to the training of the classifiers.

On the other hand, extractors do not have learning files like classifiers. When training an extractor, the training data is sent to the Document Understanding API endpoint, and the training happens remotely. The API processes the training data and updates the machine learning model associated with the extractor. This means that the training for extractors is stored remotely against the API key, and there is no file that needs to be shared between processes.

So, to achieve the shared training of classifiers and extractors across multiple processes, you would need to ensure that the processes can access and update the shared learning files for classifiers, while the training for extractors happens remotely through the API endpoiint

Hope this helps


thank you for the exhaustive answer!

1 Like

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.