fix: reuse easyocr models from docling cache dir #743
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
smoya
commented
May 15, 2025
| assert "Use cases include providing chatbot" in chunks_str | ||
|
|
||
| # electromagnetic_radiation.docx | ||
| assert "All forms of EMR travel at the speed of light in a vacuum" in chunks_str |
There was a problem hiding this comment.
We can remove this test and save some KB but I added it in order to prove no extra models were being downloaded during the parsing.
Askir
reviewed
May 15, 2025
Comment on lines
+121
to
+127
| basic_pipeline_options = PdfPipelineOptions( | ||
| do_ocr=False, # we do not want to do OCR in PDF (yet) | ||
| artifacts_path=self.cache_dir if os.path.isdir(self.cache_dir) else None, | ||
| ) # pyright: ignore[reportCallIssue] | ||
|
|
||
| with_ocr_pipeline_options = basic_pipeline_options | ||
| with_ocr_pipeline_options.do_ocr = True |
There was a problem hiding this comment.
should we actually enable ocr in pdfs? I thought it was enabled tbh
There was a problem hiding this comment.
We decided to leave this option disabled due to the fact it makes PDF parsing process way longer and it is not very clear the user intent is to run ocr on the images inside the PDF.
I think we should make it configurable from the parser when creating a vectorizer. Either keep it disabled by default or enabled, still can be a discussion we want to retake.
JamesGuthrie
approved these changes
May 15, 2025
JamesGuthrie
approved these changes
May 15, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR configures docling parser for reusing the EasyOCR models downloaded (included in the docker image) when running OCR in image files.
Without this fix, models were being downloaded again (duplicated) in the
~/.EasyOCR/modeldir whenever docling needed to run OCR on image files (note that we don't enable OCR in PDF files).Models downloaded by docling:
Models being duplicated by EasyOCR prior to this fix:
❯ ls -la ~/.EasyOCR/model .rw-r--r--@ 83M smoya 15 May 12:23 craft_mlt_25k.pth .rw-r--r--@ 15M smoya 15 May 12:23 latin_g2.pthWondering if we would need to also include models for other languages in advance, as I can see we only download english and latin.