huggingface / huggingface/datasets
Support multiprocessing for download/extract nestedly
Open
enhancement
- Dominant language
- Python
- Stars
- 22k
- Forks
- 3.4k
- Avg merge
- 5d 7h
- Merged PRs (30d)
- 17
Description
### Feature request
Current multiprocessing for download/extract is not done nestedly. For example, when processing SlimPajama, there is only 3 processes (for train/test/val), while there are many files inside these 3 folders
```
Downloading data files #0: 0%| | 0/1 [00:00
Contributor guide
Assessment
This issue has not been assessed yet.