feat(libcommon): add duckdb-based leak detection prototype (#2994) #3202
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
…ce#2994) This PR adds a prototype module for detecting data leaks and duplicates between dataset splits, as suggested in issue huggingface#2994 and inspired by the [LLE blog post](https://huggingface.co/blog/lbourdois/lle). ### 🔍 What it does: - Introduces `detect_leaks(dataset_name, subset=None)` in `libs/libcommon/src/libcommon/leak_detection.py` - Uses DuckDB to efficiently compute: - Leaks from train → test - Leaks from validation → test - Duplicates within train/validation/test splits - Supports datasets with `tokens`, `text`, and optionally `ner_tags` ### 🛠️ Example usage: ```bash python libs/libcommon/src/libcommon/leak_detection.py Will output stats for conll2003 like: yaml Copy Edit train_test_leaks: 78 validation_test_leaks: 25 train_dup: 1350 validation_dup: 180 test_dup: 269 📌 Notes: This is a standalone prototype and not yet wired into jobs or UI. Future steps may include integrating this into dataset QA dashboards or automated Hub checks. Closes huggingface#2994
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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 adds a prototype module for detecting data leaks and duplicates between dataset splits, as suggested in issue #2994 and inspired by the LLE blog post.
🔍 What it does:
detect_leaks(dataset_name, subset=None)inlibs/libcommon/src/libcommon/leak_detection.pytokens,text, and optionallyner_tags🛠️ Example usage:
Will output stats for
conll2003like:📌 Notes:
This is a standalone prototype, not yet wired into backend jobs or UI components.
No local test coverage is added at this point, as it's an early isolated utility.
Future steps may include:
Closes #2994