Skip to content

Conversation

@Johnson-yue
Copy link

I 'am not sure this is right or false!! please check it , thank you

I 'am not sure this is right or false!! please check it , thank you
# generate random epsilon
epsilon = th.rand((batch_size, 1, 1, 1)).to(fake_samps.device)
# generate random epsilon
epsilon = th.rand((batch_size, 1, 1, 1)).to(fake_samp.device)
Copy link
Owner

@akanimax akanimax Sep 6, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is a tiny issue here. The epsilon would be different for different real_samp and fake_samp, which is not intended; and perhaps also incorrect. Could you add the epsilon as another parameter to this _interp function and then select an epsilon only once outside? Basically the epsilon should be same for interpolating real and fake samples at all scales.
Rest everything looks good. 👍. Thanks.

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ok, I will try it

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@Johnson-yue @akanimax any result of this? This PR is still open?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi, I try to adding gp in MSG-GAN, but, it is confusion, the multi-output need gradient for D ,so it can be compute

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

Successfully merging this pull request may close these issues.

3 participants