I have some questions regarding AI-Fabric and DU, can anyone help with this?
What is the main purpose of GPU in AI-Fabric? Is that possible to train and deploy my document understanding model without GPU?
How to reset the data manager for the new project?
While extracting the data from the file, is that possible to limit the search scope?
Use case: I have a single PDF file which has “commercial Invoice”, “Delivery Note” and “Tax Invoice”. I need to extract details only from the “commercial invoice” page or based on the the classification token bot need to extract the data.
Using GPU is more like adding extra CPU power to your computer. So what the GPU does is, it improves the efficiency of the AI models. This affects training, deployment, and usage of the models. You can do the same thing without GPU as well.
If you want to train the AI models based on the documents you process, you can use the set of new activities that came in with the latest Preview release 20.8, to send the data to AI Fabric, and to train the models.
So if you have multiple pages, you can use the Classifiers to identify which page your data is available in, and use some PDF activities, to remove other pages if needed if the document is too big.
Once you have identified the page, you can use the Machine Learning extractor to use the AI models you have in the AI Fabric, and do the extraction.