EveryVoiceTTS / EveryVoiceTTS/EveryVoice
Choose reasonable validation split
Open
enhancement
help wanted
- Dominant language
- Python
- Stars
- 45
- Forks
- 4
- Avg merge
- 1d 8h
- Merged PRs (30d)
- 14
Description
I think in some cases a default of 90% training 10% validation is not appropriate. Maybe we could discuss a better heuristic for this, with special consideration to the low-resource case.
Contributor guide
Assessment
This issue has not been assessed yet.