Might be a typo on formula for diffusion step using alpha
hangjoni opened this issue · 5 comments
On the formula for q(xt | x0) on text between cell 9 and 10 of notebook 01_introduction_to_diffusers, currently the formula is stated as :
q(xt | x0) ~ N(sqrt(1-alpha)*x0, sqrt(1-alpha)*I)
=> This might be a typo, according to the paper it should be:
q(xt | x0) ~ N(sqrt(1-alpha)*x0, (1-alpha)*I)
Sorry for the bad typing
I believe the typo may be in the paper! Elsewhere (eg in the derivation in this video there is a square root there.
@hangjoni @johnowhitaker actually it's the same thing :
Source:Understanding Diffusion Models: A Unified Perspective by calvin luo(Page-11)
Ah right, got my notation mixed and was confused between variance and standard deviation. I'll tweak the notebook to correct and add a better explanation, and see if I've propagated my incorrect version anywhere else!
Thank you for explaining!
Fixed