Repository navigation
Benchmark liteparse with its markdown output - #1144
Closed
logan-markewich wants to merge 1 commit into
Closed
logan-markewich wants to merge 1 commit into
logan-markewich wants to merge 1 commit into
Conversation
liteparse_infer.py scored page.text, liteparse's spatial plain-text layout, which carries no table or heading markup, so liteparse scored 0% on tables and layout. Score its markdown output instead (output_format="markdown", page.markdown), the same kind of output marker is scored on. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
|
CLA Assistant Lite bot All contributors have signed the CLA ✍️ ✅ |
Contributor
Author
|
I have read the CLA Document and I hereby sign the CLA |
Member
|
Thanks @logan-markewich! Merged into dev and master. |
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
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The existing harness for liteparse in this repo used text outputs, which will fail quite badly when benchmarks expect markdown.
This PR switches the output to markdown and rescores. A sister PR in datalab-to/omni_parse_bench#80 adds the LiteParse harness to OmniParseBench.
olmOCR-bench:
OmniParseBench:
--no-ocr)Honestly the timings are likely wrong since my machine won't match yours, so you'll probably want to recalculate these for consistency. I've scoped this PR to harness-only changes for now.
Tesseract is pretty meh, so id honestly go with the no-cpu results only