The Issue of training loss

[wangtao201919]

First of all, thank you very much to the author for providing this amazing model. I am very interested in NeuCoSVC, so I want to use this model to train my own dataset locally. I followed the data processing process exactly as described in the Readme. However, during training, the “generator_loss” suddenly increased after 100k iterations. Therefore, I would like to ask about the changes in each loss during the training process. Thank you very much!

[jerry1331]

In the default configuration, the discriminator is added to training at 100k step, which will cause the generator_loss to increase.