huggingface / huggingface/datasets

Fix: Canonical 'multi_news' dataset is broken and should be updated to a Parquet version

Open
#7,746 1 comment 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
22k
Forks
3.4k
Avg merge
5d 7h
Merged PRs (30d)
17

Description

Hi,

The canonical `multi_news` dataset is currently broken and fails to load. This is because it points to the [alexfabri/multi_news](https://huggingface.co/datasets/alexfabbri/multi_news) repository, which contains a legacy loading script (`multi_news.py`) that requires the now-removed `trust_remote_code` parameter.

The original maintainer's GitHub and Hugging Face repositories appear to be inactive, so a community-led fix is needed.

I have created a working fix by converting the dataset to the modern Parquet format, which does not require a loading script. The fixed version is available here and loads correctly:

**[Awesome075/multi_news_parquet](https://huggingface.co/datasets/Awesome075/multi_news_parquet)**

Could the maintainers please guide me or themselves update the official `multi_news` dataset to use this working Parquet version? This would involve updating the canonical pointer for "multi_news" to resolve to the new repository.

This action would fix the dataset for all users and ensure its continued availability.

Thank you!

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.