diffusion-DDIM-pytorch icon indicating copy to clipboard operation
diffusion-DDIM-pytorch copied to clipboard

Published 20 hours ago • verified publisher icon Alokia

About Eta value

Open railgun122 opened this issue 1 year ago • 2 comments
trafficstars

Good Afternoon,

Firstly, Thank you for your great effort on implementing DDIM,

I have seen a lot of DDPM implementation, but less on DDIM implementation. I have learned alot on the current code

One question raises me on trying to learn your code. Since in the code it is written eta = 0, isnt in that case, sigma_t is also a 0 while calculating getting x(t-1) is sampling proecss?

Appreciate your work again.

railgun122 avatar Jul 04 '24 07:07 railgun122

Good Afternoon,

Firstly, Thank you for your great effort on implementing DDIM,

I have seen a lot of DDPM implementation, but less on DDIM implementation. I have learned alot on the current code

One question raises me on trying to learn your code. Since in the code it is written eta = 0, isnt in that case, sigma_t is also a 0 while calculating getting x(t-1) is sampling proecss?

Appreciate your work again.

Sorry, it's been so long, I can't remember many details. You can read this blogs, which inspired me greatly.

  • https://kexue.fm/archives/9119
  • https://kexue.fm/archives/9152
  • https://kexue.fm/archives/9164
  • https://kexue.fm/archives/9181

I'm sure you can gain a lot of insight.

Alokia avatar Jul 07 '24 08:07 Alokia

Thank you for your references

It help me alot in understanding of ddim

Cheers

railgun122 avatar Jul 10 '24 07:07 railgun122