BynePDF is a simple wrapper for our SOTA model Byne-LayoutLM. This model enables the retrieval of pages from visually rich documents. With this librar

Search code, repositories, users, issues, pull requests...

submited by

Style Pass

2024-09-25 16:00:13

BynePDF is a simple wrapper for our SOTA model Byne-LayoutLM. This model enables the retrieval of pages from visually rich documents. With this library, you can implement document-understanding systems and create visual or agentic RAG pipelines with industry-leading performance for analysing documents like pitch decks, company reports or scientific papers.

LayoutLM-Byne: This model is used for document and query embedding. It's pre-trained on a large corpus of document images and is specifically designed for document understanding tasks. In BynePDF, this model is non-substitutable and is always used for embedding.

MiniCPM: This model is used for question answering. It takes a page and query as input and generates an answer. The path to this model can be customized during initialization.