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

Might be a bug during fade in #20

Open
kunrenzhilu opened this issue May 18, 2018 · 0 comments
Open

Might be a bug during fade in #20

kunrenzhilu opened this issue May 18, 2018 · 0 comments

Comments

@kunrenzhilu
Copy link

kunrenzhilu commented May 18, 2018

This line

        for it in range(from_it, total_it):
            if phase == 'stabilize':
                cur_level = R
            else:
                cur_level = R + total_it/float(from_it)
            cur_resol = 2 ** int(np.ceil(cur_level+1))

is problematic.

That's because, when fade in, the from_it = (train_kimg)//batch_size, total_it = (train_kimg+total_kimg)//batch_size so the according to your code, cur_level will always be R+1.99. But it should be something progressively increasing from [R, R+1].

It should be

        for it in range(from_it, total_it):
            if phase == 'stabilize':
                cur_level = R
            else:
                cur_level = R + (it - from_it) / float(total_it - from_it)
            cur_resol = 2 ** int(np.ceil(cur_level+1))

Please correct me if my understanding is wrong.

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