PDFDocument object has no attribute 'render' and pdfium library destroyed (despite transformers 4.38.2) -- what next?

[MarHai]

Hello there,

I'm trying to get nougat up and running within a Docker environment. My Dockerfile looks as follows:

FROM nvidia/cuda:13.0.2-cudnn-devel-ubuntu24.04

WORKDIR /usr/src/app

RUN apt-get update && apt-get install -y git python3-full
RUN python3 -m venv nougatpy/

RUN nougatpy/bin/pip install transformers==4.38.2 requests git+https://github.com/facebookresearch/nougat

COPY parse_pdfs.sh nougatpy/
RUN chmod +x nougatpy/parse_pdfs.sh

EXPOSE 7950

CMD ["bash", "nougatpy/parse_pdfs.sh"]

The parse_pdfs.sh file checks for files in a directory and then runs them through nougat using:

nougat "$FILE" -o "$OUTPUT_DIR" -m 0.1.0-base --full-precision --recompute --batchsize 8

However, the result always looks as follows:

/usr/src/app/nougatpy/lib/python3.12/site-packages/torch/functional.py:505: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /pytorch/aten/src/ATen/native/TensorShape.cpp:4317.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
  0%|          | 0/3 [00:00<?, ?it/s]ERROR:root:'PdfDocument' object has no attribute 'render'
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
WARNING:root:Image not found
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
WARNING:root:Image not found
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
ERROR:root:list index out of range
 67%|██████▋   | 2/3 [00:00<00:00, 307.34it/s]
Traceback (most recent call last):
  File "/usr/src/app/nougatpy/bin/nougat", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/usr/src/app/nougatpy/lib/python3.12/site-packages/predict.py", line 166, in main
    for i, (sample, is_last_page) in enumerate(tqdm(dataloader)):
  File "/usr/src/app/nougatpy/lib/python3.12/site-packages/tqdm/std.py", line 1181, in __iter__
    for obj in iterable:
  File "/usr/src/app/nougatpy/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 732, in __next__
    data = self._next_data()
           ^^^^^^^^^^^^^^^^^
  File "/usr/src/app/nougatpy/lib/python3.12/site-packages/torch/utils/data/dataloader.py", line 788, in _next_data
    data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/src/app/nougatpy/lib/python3.12/site-packages/torch/utils/data/_utils/fetch.py", line 55, in fetch
    return self.collate_fn(data)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/src/app/nougatpy/lib/python3.12/site-packages/nougat/utils/dataset.py", line 114, in ignore_none_collate
    _batch[-1] = (_batch[-1][0], name)
                  ~~~~~~^^^^
IndexError: list index out of range
-> Cannot close object; pdfium library is destroyed. This may cause a memory leak.

I've seen the downgrade discussion on transformers (and I did; see the Dockerfile), but this does not help at all.

Any ideas on how to proceed?

Thanks
Mario

[Aniruddha521]

@MarHai ,
you may try this #265