Description
Feature request
Add support for subset_name
as an alias for config_name
in the datasets library and related tools (such as loading scripts, documentation, and metadata).
Motivation
The Hugging Face Hub dataset viewer displays a column named "Subset", which refers to what is currently technically called config_name in the datasets library. This inconsistency has caused confusion for many users, especially those unfamiliar with the internal terminology.
I have repeatedly received questions from users trying to understand what "config" means, and why it doesn’t match what they see as "subset" on the Hub. Renaming everything to subset_name
might be too disruptive, but introducing subset_name as a clear alias for config_name could significantly improve user experience without breaking backward compatibility.
This change would:
- Align terminology across the Hub UI and datasets codebase
- Reduce user confusion, especially for newcomers
- Make documentation and examples more intuitive