def get_loss(self, x, z):
feed = {}
feed[tbn('x:0')] = x
feed[tbn('z:0')] = z
feed[tbn('is_training:0')] = False
return self.sess.run([self.loss_G, self.loss_Dreal, self.loss_Dfake], feed_dict=feed)
def sample(self, z):
feed = {}
feed[tbn('is_training:0')] = False
feed[tbn('z:0')] = z
out = self.sess.run(tbn('samples:0'), feed_dict=feed)
return out
def train(self):
"""Take a training step with batches from each domain."""
self.iteration += 1
feed = {
tbn('lr:0'): self.args.learning_rate,
tbn('is_training:0'): True
}
self.sess.run([obn('train_op_D')], feed_dict=feed)
self.sess.run([obn('train_op_G')], feed_dict=feed)
def train(self, x=None, z=None, learning_rate=.001):
feed = {}
if x:
feed[tbn('x:0')] = x
if z:
feed[tbn('z:0')] = z
feed[tbn('learning_rate:0')] = learning_rate
feed[tbn('is_training:0')] = True
self.sess.run(self.update_op_D, feed_dict=feed)
self.sess.run(self.update_op_G, feed_dict=feed)
def get_loss(self, xb1, xb2):
"""Return all of the loss values for the given input."""
feed = {
tbn('xb1:0'): xb1,
tbn('xb2:0'): xb2,
tbn('is_training:0'): False
}
losses = self.sess.run(tf.get_collection('losses'), feed_dict=feed)
lstring = ' '.join(['{:.3f}'.format(loss) for loss in losses])
return lstring
def get_layer(self, xb1, xb2, name):
"""Get a layer of the network by name for the entire datasets given in xb1 and xb2."""
tensor_name = "{}:0".format(name)
tensor = tbn(tensor_name)
feed = {
tbn('xb1:0'): xb1,
tbn('xb2:0'): xb2,
tbn('is_training:0'): False
}
layer = self.sess.run(tensor, feed_dict=feed)
return layer