huggingface / huggingface/datasets
Make creating/editing dataset cards easier, by editing on site and dumping info from test command.
- Dominant language
- Python
- Stars
- 22k
- Forks
- 3.4k
- Avg merge
- 5d 7h
- Merged PRs (30d)
- 17
Description
**Is your feature request related to a problem? Please describe.**
Creating and editing dataset cards should be but not that easy
- If other else know Some information I don't know (bias of dataset, dataset curation, supported dataset, ...), he/she should know the description on hf.co comes from README.md under github huggingface/datasets/datasets/the dataset, and willing to make a pr to add or fix information.
- Many information is also saved in `dataset_info.json` (citaion, description), but still need to write it down to README.md again.
- Contributor need to pip install and start a local server just for tagging the dataset's size. And contributor may be creating the dataset on lab's server, which can't open a browser.
- if any one proposes a new tag, it doesn't show in the list that another creator see. (a stackoverflow way may be ideal)
- dataset card generator web app doesn't generate the necessary subsecion `Contributions` for us.
**Describe the solution you'd like**
- Everyone (or at least the author/contributor) can edit the description, information, tags of the dataset, on hf.co website. Just like wikipedia+stackoverflow
- We can infer the actual data size, citation, data instance, ... from `dataset_info.json` and `dataset.arrow` via `dataset-cli test`
Contributor guide
Assessment
This issue has not been assessed yet.