byaldi
byaldi copied to clipboard
Dependency related modifications
Hello, The dependency on pdf2image and poppler-utils gave me some headaches because it is not possible to install poppler-utils in my work environment. So, I implemented custom classes that does the same thing as pdf2image.convert_from_path() function but using PyMuPDF. I also inherited RAGMultiModalModel and ColPaliModel classes and modified the certain methods to make them work with PyMuPDF-based new custom classes. As a result, it is still working in the same way but without the dependency on poppler-utils.
If it is welcomed and suitable, I'd like to contribute this implementation, but I don't know much about licensing issues.