Skip to content

Conversation

@rvashurin
Copy link
Collaborator

  • Adds stripped_input column to most of the datasets
  • Improves memory footprint of dataset publishing by switching to generator-based iteration
  • Adds validation script to prevent regressions in datasets
  • Adds publishing script allowing for batch-publishing of all datasets to specified namespace

- QA: `Q: <question>\nA:`
- Multiple choice: `Q: <question>\nA. <choice0>\nB. <choice1>\n...\nA:`
- Translation: `Original:\n<source>\nTranslation:`
- Summarization: `Original:\n<source>\nSummary:`
Copy link
Collaborator

Choose a reason for hiding this comment

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

Should we keep this structure? It seems that all methods using stripped_input expect raw input, without prefixes like Original:, Q:, etc

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

I'm not sure if it's best to mark parts of the task (input/output) for the NLI task. I guess for summarization it makes sense to mark output as a summary, since otherwise it could just be thought of as another sentence from the text. I'll check how it works with and without these markers and report back

x.append(prompt.format(text=q) if len(prompt) else q)
y.append("")
return x, y
s.append(qa_stripped(q))
Copy link
Collaborator

@ArtemVazh ArtemVazh Aug 19, 2025

Choose a reason for hiding this comment

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

I’m not sure QA is the best format for this dataset

@rvashurin rvashurin merged commit c70e378 into main Jan 10, 2026
1 check passed
@rvashurin rvashurin deleted the raw_input_column branch January 10, 2026 18:47
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.

4 participants