search

p0p4k / vits2_pytorch

unofficial vits2-TTS implementation in pytorch
https://arxiv.org/abs/2307.16430
MIT License
465 stars 81 forks source link

KeyError when training the Chinese dataset #53

Closed KevinWang676 closed 10 months ago

KevinWang676 commented 10 months ago

Hi, does anyone know how to fix the KeyError: ',' when training the Chinese dataset? I'm using chinese_cleaners defined here. Thank you!

image

p0p4k commented 10 months ago

fixed in latest patch. unavailable tokens will just be skipped.

KevinWang676 commented 10 months ago

Hi, thank you so much for the fix! However, you may need to change the variable to cleaned_text in line 40 and define sequence = [] in line 38. I fix the typos in my folk.

Also, I wonder if missing the symbols like the comma , would affect my training process and results. Thank you!

KevinWang676 commented 10 months ago

Hi @p0p4k, I wonder if missing some symbols like the comma , would affect my training process and results. Thank you!

p0p4k commented 10 months ago

You can add missing punctuations in the punctuations list and try training with that first.