huggingface / huggingface/datasets
Only-data dataset loaded unexpectedly as validation split
Open
bug
- Dominant language
- Python
- Stars
- 22k
- Forks
- 3.4k
- Avg merge
- 5d 7h
- Merged PRs (30d)
- 17
Description
## Describe the bug
As reported by @thomasw21 and @lhoestq, a dataset containing only a data file whose name matches the pattern `*dev*` will be returned as VALIDATION split, even if this is not the desired behavior, e.g. a file named `datosdevision.jsonl.gz`.
Contributor guide
Assessment
This issue has not been assessed yet.