EveryVoiceTTS / EveryVoiceTTS/EveryVoice
Only display warning about missing character during training once
- Dominant language
- Python
- Stars
- 45
- Forks
- 4
- Avg merge
- 1d 8h
- Merged PRs (30d)
- 14
Description
### Description & Motivation
From @joanise:
"I’m getting a ton of warnings like this:
WARNING - Token '£' has no mapping in the pretrained StyleTTS2 symbol table and will be silently dropped.
The same warnings appear for 7, 8,ü, 4, [, (, é, - It makes sense to have these warnings, but they’re also printed again for the same characters later, it’s not just once to problematic character. That’s annoying and would be worth fixing."
The problem is that the warning is already de-duplicated in StyleTTS2 within each dataloader process, but because there are multiple dataloader processes, the warnings get duplicated.
### Pitch
_No response_
### Alternatives
_No response_
### Additional context
_No response_
Contributor guide
Research direction
Start by tracing where the missing-character warning is emitted during StyleTTS2 training and how the dataloader processes are created. Reproduce training with multiple dataloader processes, then verify that each problematic character produces only one warning across the run rather than one per process.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 48/100