def test_metric(self):
with self.test_session():
gan = mock_gan()
loss = SupervisedLoss(gan, loss_config)
d_loss, g_loss = loss.create()
metrics = loss.metrics
self.assertTrue(metrics['d_class_loss'] != None)
def test_create(self):
with self.test_session():
gan = mock_gan()
loss = SupervisedLoss(gan, loss_config)
d_loss, g_loss = loss.create()
d_shape = gan.ops.shape(d_loss)
self.assertEqual(d_shape, [1])
self.assertEqual(g_loss, None)
def add_supervised_loss(self):
if self.args.classloss:
print("[discriminator] Class loss is on. Semi-supervised learning mode activated.")
supervised_loss = SupervisedLoss(self.gan, self.gan.config.loss)
self.gan.loss = MultiComponent(components=[supervised_loss, self.gan.loss], combine='add')
supervised_loss.create()
#EWW
else:
print("[discriminator] Class loss is off. Unsupervised learning mode activated.")
def test_config(self):
with self.test_session():
loss = SupervisedLoss(mock_gan(), loss_config)
self.assertTrue(loss.config.test)