Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Tool for comparing dolma data after preprocessing. #86

Merged
merged 1 commit into from
Jul 31, 2024
Merged

Conversation

blester125
Copy link
Collaborator

This adds a script I have been using to spot check my text cleaning implementations. Point it at the aligned dolma shards from two different versions (raw vs v0, etc.) and you can use the buttons to move through example or jump to a specific example. Note: If documents are removed between the versions, this tool will be unaligned, could be fixed in the future by moving though the example id's instead of a simple numerical index.

The data is shown in independently scrollable boxes to help align sections when there are heavy changes.

Example Screenshot

image

Copy link
Collaborator

@craffel craffel left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice! Maybe a stretch goal, but perhaps it would be cool to do some kind of highlighted diff between the two verisons?

licensed_pile/compare_data.py Outdated Show resolved Hide resolved
Data is read (and cached), then it can be navigated with next/prev to
move long as if scrolling though the file on disk. On can also jump to a
specific page via id or title (both which support type-ahead search). It
also allows for jumping to a random page for spot checking. Once an
example is selected, the two versions are shown to the user side-by-side
@blester125 blester125 merged commit 9a4d292 into main Jul 31, 2024
2 checks passed
@blester125 blester125 deleted the feat/compare branch July 31, 2024 17:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants