Why use `randn_like` when sampling from latent distribution?

jaywalnut310 / vits

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

https://jaywalnut310.github.io/vits-demo/index.html

MIT License

6.91k stars 1.27k forks source link

Why use `randn_like` when sampling from latent distribution? #165

Open bejaeger opened 1 year ago

bejaeger commented 1 year ago

I am wondering why we don't use gaussian sample but rather sample from the interval [0, 1) when sampling from the latents [1]. Is there a particular theoretical advantage? Or did this prove more stable empirically?

[1] https://github.com/jaywalnut310/vits/blob/2e561ba58618d021b5b8323d3765880f7e0ecfdb/models.py#L520