Use predicted noise as base when taking steps in latent diffusion - Githubissues

samedii / perceptor

Modular image generation library

Other

5 stars 1 forks source link

Use predicted noise as base when taking steps in latent diffusion #40

Closed samedii closed 2 years ago

samedii commented 2 years ago

We get numerical errors when alpha/sigma is close to zero otherwise

Additionally

Refactor and rename eps to "predicted noise"
Refactor to use "prediction" class as output