icon-lab/FDB

Mismatch between code and paper

Closed this issue · 1 comments

Thansk for you nice job!

I got a question about the code. In the defination of DiffusionBridge, I can't find any usage about Gaussian noise, but in the main paper, both forward and reverse process contain Gaussian noise.

Thanks a lot and look forward to your answer.

We are currently testing many variations to optimize our method. The current version does not contain Gaussian noise but we will soon release a comprehensive version fixing this issue.