RAGAN Loss

Search…

RAGAN Loss

Adapted from https://arxiv.org/abs/1807.00734​
Source: /losses/ragan_loss.py​
1
# wasserstein type
2
cr = torch.mean(d_real,0)
3
cf = torch.mean(d_fake,0)
4
d_loss = -(d_real-cf) + (d_fake-cr)
5
g_loss = -(d_fake-cr)
Copied!
examples
Configurations: /losses/ragan_loss/​
1
{
2
  "class": "function:hypergan.losses.ragan_loss.RaganLoss",
3
  "type": "hinge"
4
}
Copied!
options
attribute
description
type
type
least_squares,hinge,wasserstein or standard.  Defaults to standard
string (optional)
rgan
rgan does not average over batch.  Defaults to false
boolean (optional)
labels
[a,b,c].  Defaults to [-1,1,1].  Only used in least_squares type
array of floats (optional)

QP Loss

Realness Loss

Last modified 1yr ago

Copy link

Contents

examples

options