modelscope/FunCodec

Discriminator loss?

a897456 opened this issue · 5 comments

image
As far as I know, should the loss in the figure above be the generator loss? What does the discriminator loss look like?

The discriminator is adversarially trained to maximize the equation (5) in Page 3.
For implement details, please refer to https://github.com/alibaba-damo-academy/FunCodec/blob/master/funcodec/models/codec_basic.py#L631-L653

The discriminator is adversarially trained to maximize the equation (5) in Page 3. For implement details, please refer to https://github.com/alibaba-damo-academy/FunCodec/blob/master/funcodec/models/codec_basic.py#L631-L653

https://github.com/alibaba-damo-academy/FunCodec?tab=readme-ov-file#batch-inference
How can I get this folder: input_wav.scp?

6ada5568ba2c6ef3a8ffc6e8b5d0e8d
I am using Windows, so run.sh keeps reporting errors and does not get ·input_wav.scp·. So is there a good way to run it on Windows?

This repo has be only tested on linux, I haven't try it on windows. The sh script is written for Linux.

ok,thanks