Support StyleTTS2 OOD data requirements

[roedoejet]

PR Goal?

Figure out how to handle OOD data properly. This is data that is used for the WavLM based loss used in StyleTTS2. It intentionally uses text (from the same language) to synthesize that is unseen during training, to make the synthesizer more robust to out of domain utterances. This PR:

• moves the default English data to HuggingFace: https://huggingface.co/datasets/everyvoice/StyleTTS2-English-OOD
• adds ood to the preprocessor (exports a per-language psv file)
• adds ood steps to the wizard
• allows the user to provide a HF repo link, a local file, or use the validation set

Fixes?

Fixes part of #686

Feedback sought?

This is a big one, so mostly looking for feedback on the UX in the wizard for handling OOD data. Is it explained OK? How intuitive is it? Try configuring + preprocessing + training if you can.

Priority?

high. this is one of the last things for StyleTTS2, and probably the second-to-last major change.

Tests added?

I added a few.

How to test?

Configure a new project, add some OOD data, preprocess, train.

Confidence?

medium-high

Version change?

already changed for StyleTTS2

Related PRs?

EveryVoiceTTS/StyleTTS2#13

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
everyvoice/wizard/basic.py	64% smaller
everyvoice/tests/test_cli.py	55% smaller
everyvoice/tests/data/relative/config/everyvoice-text-to-wav.yaml	49% smaller
mkdocs.yml	40% smaller
everyvoice/tests/test_custom_g2p.py	39% smaller
everyvoice/tests/test_wizard.py	23% smaller
everyvoice/cli.py	9% smaller
everyvoice/tests/test_preprocessing.py	6% smaller
everyvoice/preprocessor/preprocessor.py	1% smaller
Contributing.md	Unsupported file format
README.md	Unsupported file format
docs/guides/custom.md	Unsupported file format
docs/guides/fastspeech2.md	Unsupported file format
docs/guides/styletts2.md	Unsupported file format
everyvoice/.schema/everyvoice-text-to-wav-0.5.json	0% smaller
everyvoice/base_cli/helpers.py	0% smaller
everyvoice/model/e2e/StyleTTS2_lightning	0% smaller
everyvoice/tests/data/mixed-cleaners-resume	Unsupported file format
everyvoice/wizard/__init__.py	0% smaller

[github-actions]

CLI load time: 0:00.28
Pull Request HEAD: 9b3edf619bc132c64553d9c61c33711eeadfb635

Imports that take more than 0.1 s:
import time: self [us] | cumulative | imported package
import time:       648 |     101588 |   typer
import time:      3603 |     101215 |             loguru
import time:      2554 |     150482 |           everyvoice.utils
import time:       849 |     151330 |         everyvoice.model.feature_prediction.FastSpeech2_lightning.fs2.cli.benchmark
import time:       531 |     159821 |       everyvoice.model.feature_prediction.FastSpeech2_lightning.fs2.cli.cli
import time:       321 |     160141 |     everyvoice.model.feature_prediction.FastSpeech2_lightning.fs2.cli
import time:      1309 |     161450 |   everyvoice.model.feature_prediction.FastSpeech2_lightning.fs2.cli.check_data
import time:      8340 |     407152 | everyvoice.cli
import time:      5385 |     132153 |   rich.markdown
import time:      4420 |     221490 | typer.rich_utils

[codecov]

Codecov Report

❌ Patch coverage is 78.32168% with 62 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.41%. Comparing base (f7b2d4d) to head (9b3edf6).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
everyvoice/wizard/basic.py	81.30%	31 Missing and 12 partials ⚠️
everyvoice/cli.py	38.09%	12 Missing and 1 partial ⚠️
everyvoice/preprocessor/preprocessor.py	82.35%	3 Missing and 3 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #816      +/-   ##
==========================================
- Coverage   84.99%   84.41%   -0.58%     
==========================================
  Files          46       46              
  Lines        4219     4403     +184     
  Branches      632      656      +24     
==========================================
+ Hits         3586     3717     +131     
- Misses        494      533      +39     
- Partials      139      153      +14     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[joanise]

Just my comments so far, I'll keep reviewing next week.

I know some of my comments are on text that you didn't just write, but rather only moved, but I noticed things we could improve anyway.

And I know, so far only on docs, nothing on the code yet. Next week...