stuck/not working

[gitpushmainforce]

I tried to launch this on both Windows and WSL2, either have same behavior

Launch command
docling hello.pdf

Output:
2025-10-27 19:37:52,460 - INFO - Loading plugin 'docling_defaults' 2025-10-27 19:37:52,462 - INFO - Registered ocr engines: ['auto', 'easyocr', 'ocrmac', 'rapidocr', 'tesserocr', 'tesseract'] 2025-10-27 19:37:52,479 - INFO - paths: [PosixPath('/tmp/tmpecnqgfyb/hello.pdf'), PosixPath('/tmp/tmpecnqgfyb/hello.pdf')] 2025-10-27 19:37:52,481 - INFO - detected formats: [<InputFormat.PDF: 'pdf'>] 2025-10-27 19:37:52,501 - INFO - Going to convert document batch... 2025-10-27 19:37:52,501 - INFO - Initializing pipeline for StandardPdfPipeline with options hash be940d8e09a8a29d0e083636610e060b 2025-10-27 19:37:52,508 - INFO - Loading plugin 'docling_defaults' 2025-10-27 19:37:52,508 - INFO - Registered picture descriptions: ['vlm', 'api'] 2025-10-27 19:37:52,776 - INFO - Accelerator device: 'cpu' [INFO] 2025-10-27 19:37:52,805 [RapidOCR] base.py:22: Using engine_name: onnxruntime [INFO] 2025-10-27 19:37:52,812 [RapidOCR] download_file.py:60: File exists and is valid: /home/pudge/.local/lib/python3.10/site-packages/rapidocr/models/ch_PP-OCRv4_det_infer.onnx [INFO] 2025-10-27 19:37:52,813 [RapidOCR] main.py:53: Using /home/pudge/.local/lib/python3.10/site-packages/rapidocr/models/ch_PP-OCRv4_det_infer.onnx [INFO] 2025-10-27 19:37:52,922 [RapidOCR] base.py:22: Using engine_name: onnxruntime [INFO] 2025-10-27 19:37:52,924 [RapidOCR] download_file.py:60: File exists and is valid: /home/pudge/.local/lib/python3.10/site-packages/rapidocr/models/ch_ppocr_mobile_v2.0_cls_infer.onnx [INFO] 2025-10-27 19:37:52,924 [RapidOCR] main.py:53: Using /home/pudge/.local/lib/python3.10/site-packages/rapidocr/models/ch_ppocr_mobile_v2.0_cls_infer.onnx [INFO] 2025-10-27 19:37:52,975 [RapidOCR] base.py:22: Using engine_name: onnxruntime [INFO] 2025-10-27 19:37:52,988 [RapidOCR] download_file.py:60: File exists and is valid: /home/pudge/.local/lib/python3.10/site-packages/rapidocr/models/ch_PP-OCRv4_rec_infer.onnx [INFO] 2025-10-27 19:37:52,988 [RapidOCR] main.py:53: Using /home/pudge/.local/lib/python3.10/site-packages/rapidocr/models/ch_PP-OCRv4_rec_infer.onnx 2025-10-27 19:37:53,112 - INFO - Auto OCR model selected rapidocr with onnxruntime. 2025-10-27 19:37:53,119 - INFO - Accelerator device: 'cpu'

After that binary just keeps running, but nothing happens, its just stuck. Also there is no any CPU usage increase during running. Tried both -v and -vv flags, but output is just the same. Tried multiple different PDF's (simple ones too), none worked.

[dosubot[bot]]

Hi @gitpushmainforce! I'm Dosu and I’m helping the docling team.

This kind of hang (no CPU usage, no progress after "Accelerator device: 'cpu'") is a known issue and can be caused by a few things in Docling:

• The default PDF backend sometimes blocks indefinitely on certain PDFs or environments. Switching to the pypdfium2 backend often resolves this. Try:

  docling --pdf-backend pypdfium2 hello.pdf
  

  This backend is faster and more robust for many PDFs, especially if the default backend hangs or is slow [source].

• Native dependency mismatches (like ONNX Runtime or PyTorch installed for the wrong architecture) can cause silent hangs, especially in WSL2 or mixed-arch setups. Make sure all dependencies are for your platform (not ARM64 on x86, or vice versa) [source].

• Some OCR engines or models can hang if their dependencies are missing or incompatible. Try switching to a different OCR engine, for example:

  docling --ocr-engine easyocr hello.pdf
  

  or

  docling --ocr-engine tesseract hello.pdf
  

  This can help isolate if the issue is with RapidOCR or the pipeline [source].

• If the process still hangs, interrupt it with Ctrl+C and check the stack trace. This can help pinpoint where it's stuck (e.g., PDF parsing, OCR, or model loading) [source].

• Upgrading to the latest Docling version is recommended, as recent releases have improved timeout handling and better error reporting for these kinds of hangs [source].

Let me know if switching the PDF backend or OCR engine helps, or if you get a stack trace after interrupting the process.

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.

[gitpushmainforce]

Changing pdf backend worked, thanks!