def train(self, x=None, z=None, y=None):
if x is None:
return self.sess.run([self.G_optim, self.G_loss],
feed_dict={self.Z: z})[1]
else:
x = processing.img_preprocessing(x)
return self.sess.run([self.D_optim, self.D_loss],
feed_dict={
self.X: x,
self.Z: z
})[1]
def train(self, y= None, x= None, z=None):
if x is None:
return self.sess.run([self.G_optim, self.G_loss], feed_dict = {
self.Z: z,
self.y: y,
self.X: np.zeros(shape=[y.shape[0], self.input_shape[0]])
})[1]
else:
x = processing.img_preprocessing(x)
return self.sess.run([self.D_optim, self.D_loss], feed_dict = {
self.X: x,
self.Z: z,
self.y: y,
})[1]
def train(self, x= None, z=None, y=None):
if x is None:
disc, cont = self.sess.run([self.Q_disc_loss, self.Q_cont_loss], feed_dict = {
self.Z: z[:, :self.noise_dim],
self.c_dis: y,
self.c_cont: z[:,self.noise_dim:]
})
#print('Q_loss 1. discrete:{} 2. continuous: {}'.format(disc,cont))
return self.sess.run([self.G_optim, self.G_loss], feed_dict = {
self.Z: z[:,:self.noise_dim],
self.c_dis: y,
self.c_cont: z[:,self.noise_dim:]
})[1]
else:
x = processing.img_preprocessing(x)
return self.sess.run([self.D_optim, self.D_loss], feed_dict = {
self.X: x,
self.Z: z[:, :self.noise_dim],
self.c_dis: y,
self.c_cont: z[:,self.noise_dim:]
})[1]