You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Since the d_loss = fake-real+gp, how to decide the value of LAMBDA to get a reasonable gradient penalty.
I mean, if fake=-10, real=10, is gp=10 appropriate?or the gp should be smaller(e.g. gp=5)?
i have the same question about it, did you solve this problem? at the start of my train, the wass loss is very small, about 0.01, so how should i set the lambda?
Penalty coefficient All experiments in this paper use λ = 10, which we found to work well across a variety of architectures and datasets ranging from toy tasks to large ImageNet CNNs.
Activity
hangexuexi commentedon Nov 29, 2022
i have the same question about it, did you solve this problem? at the start of my train, the wass loss is very small, about 0.01, so how should i set the lambda?
cjblocker commentedon Aug 7, 2023
The original paper suggests sticking with 10