Does this project need to download a model from huggingface?

[neverlatetolearn0]

Bug

When I run the code, I encounter the following problem
requests.exceptions.SSLError: (MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /api/models/ds4sd/docling-models/revision/v2.1.0 (Caused by SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1000)')))"), '(Request ID: 17d20384-a725-4353-8425-f27cb0dc2ff6)')

Does anyone else have the same problem

[Leflak]

Got the same problem, tried to update everything but does not work.

[tanaha2002]

same to me, do you know how to fix it?

[Leflak]

Was initially using personnal project to launch docling and got that error then executed the main page script which somehow solved my problem, hope that will help you too:

from docling.document_converter import DocumentConverter

source = "https://arxiv.org/pdf/2408.09869"  # document per local path or URL
converter = DocumentConverter()
result = converter.convert(source)
print(result.document.export_to_markdown())  # output: "## Docling Technical Report[...]"

[tanaha2002]

Was initially using personnal project to launch docling and got that error then executed the main page script which somehow solved my problem, hope that will help you too:

from docling.document_converter import DocumentConverter

source = "https://arxiv.org/pdf/2408.09869"  # document per local path or URL
converter = DocumentConverter()
result = converter.convert(source)
print(result.document.export_to_markdown())  # output: "## Docling Technical Report[...]"

There's still an error, but I found an alternative solution. You guys can try running it on Google Colab. In my case, installing it on my computer didn’t work, but it worked on Colab. And yes, it downloads some weights from Hugging Face. Somehow, my network blocks Hugging Face, which caused the problem.

[Leflak]

Requires transformers but maybe launch the simple python script (picked from the "Use this model" button from https://huggingface.co/ds4sd/docling-models) and then retry with my previous answer:

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("ds4sd/docling-models")

[trinanjan12]

Yes, There is Layout model and the other one is Table detection model from hugging face
#Example code to only run layout model
from docling_ibm_models.layoutmodel.layout_predictor import LayoutPredictor
test_layout_predictor = LayoutPredictor("/home/your_name/.cache/huggingface/hub/models--ds4sd--docling-models/snapshots/36bebf56681740529abd09f5473a93a69373fbf0/model_artifacts/layout/")

import pypdfium2 as pdfium
test_pdoc = pdfium.PdfDocument("./tests/data/2305.03393v1-pg9.pdf")

#layout prediction for one page
test_layout_op = test_layout_predictor.predict(test_pdoc[0].render().to_pil())
print(list(test_layout_op))

[jdbranham]

I was also getting models errors with 2.14.0

Saw this in the log -

  File "/Users/###/###/.venv/lib/python3.12/site-packages/docling_ibm_models/layoutmodel/layout_predictor.py", line 98, in __init__
    raise FileNotFoundError("Missing ONNX file: {}".format(self._onnx_fn))
FileNotFoundError: Missing ONNX file: /Users/###/.cache/huggingface/hub/models--ds4sd--docling-models/snapshots/36bebf56681740529abd09f5473a93a69373fbf0/model_artifacts/layout/beehive_v0.0.5/model.pt

I'd cleared the hugging face cache but it didn't seem to fix the issue.

Downgrading to 2.13.0 seemed to work, then I went back to 2.14.0 and it works without error now. 😓

https://ds4sd.github.io/docling/usage/#setting-up-a-documentconverter

There's a section in the docs that might be helpful, but these imports don't appear to be valid -

from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import PdfPipelineOptions
from docling.document_converter import DocumentConverter, PdfFormatOption
from docling.pipeline.standard_pdf_pipeline import StandardPdfPipeline

# # to explicitly prefetch:
# artifacts_path = StandardPdfPipeline.download_models_hf()

artifacts_path = "/local/path/to/artifacts"

pipeline_options = PdfPipelineOptions(artifacts_path=artifacts_path)
doc_converter = DocumentConverter(
    format_options={
        InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options)
    }
)