Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Question about GVBD and GVBG #2

Open
JalinWang opened this issue Oct 20, 2022 · 0 comments
Open

Question about GVBD and GVBG #2

JalinWang opened this issue Oct 20, 2022 · 0 comments

Comments

@JalinWang
Copy link

JalinWang commented Oct 20, 2022

    def get_alignment_loss(self, name, logits, x, align_logits):
        discriminator = getattr(self, name + "_discriminator")
        gvbg = getattr(self, name + "_gvbg")
        coeff = calc_coeff(self.iteration)
        bridge = gvbg(x)
        if self.opt.gvbg_weight > 0 and align_logits:
            logits = logits - bridge
        setattr(self, "loss_%s_gvbg" % (name), loss_gvbg)
        setattr(self, "loss_%s_gvbd" % (name), loss_gvbd)

Hi~I'm very confused by these two linear layers. They seem to be a logit-related trick, but I find no description in the paper/supp.
I would appreciate it if any explanation could be given on what they are used for.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant