RuntimeError: Conversion failed with status: ConversionStatus.FAILURE

[AdityaMannu1709]

Using the below code, I am getting this error on a PDF.

def docling_markdown(pdf_path):

    # Docling Configs
    pipeline_options = PipelineOptions(do_table_structure=True, do_ocr=True)
    pipeline_options.table_structure_options.mode = TableFormerMode.ACCURATE 
    pipeline_options.table_structure_options.do_cell_matching = False  

    doc_converter = DocumentConverter(
        pipeline_options=pipeline_options,
    )

    source = pdf_path  # PDF path or URL
    result = doc_converter.convert_single(source)
    print(result.render_as_markdown()) 

Error traceback -

• document_converter.py -> _process_document -> for assembled_page in assembled_pages:
• page_assemble_model.py -> call -> for page in page_batch:
• base_model_pipeline.py -> apply -> yield from page_batch
• table_structure_model.py -> call -> text_piece = page._backend.get_text_in_rect(the_bbox)
• docling_parse_backend.py -> get_text_in_rect -> overlap_frac = cell_bbox.intersection_area_with(bbox) / cell_bbox.area()
• ZeroDivisionError: float division by zero

It happened on 1 PDF out of many that i tested.

[staar]

@AdityaMannu1709 Could you share the PDF with us? We are trying to build now a good evaluation set (hopefully soon open-source), and it would be great to include it in the test set.

[AdityaMannu1709]

Actually that PDF has some personal information but I found that if I am excluding this below pipeline option then it works smoothly on that PDF and proper output is coming -

pipeline_options.table_structure_options.do_cell_matching = False