Skip to content

c_skip and c_out in Inference #23

Description

@w86763777

Hi, thank you for your great work!

I have a question regarding the following lines:

c_skip = self.sigma_data ** 2 / ((sigma - self.sigma_min) ** 2 + self.sigma_data ** 2)
c_out = (sigma - self.sigma_min) * self.sigma_data / (sigma ** 2 + self.sigma_data ** 2).sqrt()

It seems that c_skip and c_out used during inference differ from those used in the training phase. I think this behavior is not present in EDM. Is there any particular intuition or motivation behind this change?

Thanks again for your time and looking forward to your insights!
Please kindly correct me if I have misunderstood anything.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions