Fix Config_v1 Configs and Random Scales Settings In Train.py for NSF-HiFiGAN

fishaudio / fish-diffusion

An easy to understand TTS / SVS / SVC framework

https://diff.fish.audio

MIT License

635 stars 81 forks source link

Closed catalystfrank closed 1 year ago

catalystfrank commented 1 year ago

Addition scales for SingGAN like mel/stft loss. Fix Config_v1 kernel sizes, will update config_v2 for later experiments.

leng-yue commented 1 year ago

Closed PR because @catalystfrank has a newer and better version :)