search

yl4579 / StarGANv2-VC

StarGANv2-VC: A Diverse, Unsupervised, Non-parallel Framework for Natural-Sounding Voice Conversion
MIT License
479 stars 107 forks source link

training issue #22

Closed Kristopher-Chen closed 2 years ago

Kristopher-Chen commented 2 years ago

https://github.com/yl4579/StarGANv2-VC/blob/ed20538971a03d699351a349a3631767333baeb7/trainer.py#L205

When training with the mapping network, the style encoder is backward; shouldn't it be moved after the loss of the style encoder is calculated?

yl4579 commented 2 years ago

14