Whether kl loss is negative will affect the model convergence?

jaywalnut310 / vits

VITS: Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech

MIT License

6.73k stars 1.24k forks source link

Open 980202006 opened 1 year ago

980202006 commented 1 year ago

I noticed that my kl loss is already less than 0.01, and sometimes it suddenly becomes negative. Does this affect the convergence of the model?