def generator(z):
# because up to now we can not derive bias_add's higher order derivative in tensorflow,
# so I use vanilla implementation of FC instead of a FC layer in tensorflow.contrib.layers
# the following conv case is out of the same reason
weights = slim.model_variable(
'fn_weights', shape=(FLAGS.z_dim, 4 * 4 * 512), initializer=ly.xavier_initializer())
bias = slim.model_variable(
'fn_bias', shape=(4 * 4 * 512, ), initializer=tf.zeros_initializer)
train = tf.nn.relu(ly.batch_norm(fully_connected(z, weights, bias)))
train = tf.reshape(train, (-1, 4, 4, 512))
train = ly.conv2d_transpose(train, 256, 3, stride=2,
activation_fn=tf.nn.relu, normalizer_fn=ly.batch_norm, padding='SAME')
train = ly.conv2d_transpose(train, 128, 3, stride=2,
activation_fn=tf.nn.relu, normalizer_fn=ly.batch_norm, padding='SAME')
train = ly.conv2d_transpose(train, 64, 3, stride=2,
activation_fn=tf.nn.relu, normalizer_fn=ly.batch_norm, padding='SAME')
train = ly.conv2d_transpose(train, 1, 3, stride=1,
activation_fn=None, padding='SAME', biases_initializer=None)
bias = slim.model_variable('bias', shape=(
1, ), initializer=tf.zeros_initializer)
train += bias
train = tf.nn.tanh(train)
return train
python类model_variable()的实例源码
def generator(z):
weights = slim.model_variable(
'fn_weights', shape=(FLAGS.z_dim, 4 * 4 * 512), initializer=ly.xavier_initializer())
bias = slim.model_variable(
'fn_bias', shape=(4 * 4 * 512, ), initializer=tf.zeros_initializer)
train = tf.nn.relu(fully_connected(z, weights, bias))
train = tf.reshape(train, (-1, 4, 4, 512))
train = ly.conv2d_transpose(train, 256, 3, stride=2,
activation_fn=tf.nn.relu, normalizer_fn=ly.batch_norm, padding='SAME', weights_initializer=tf.random_normal_initializer(0, 0.02))
train = ly.conv2d_transpose(train, 128, 3, stride=2,
activation_fn=tf.nn.relu, normalizer_fn=ly.batch_norm, padding='SAME', weights_initializer=tf.random_normal_initializer(0, 0.02))
train = ly.conv2d_transpose(train, 64, 3, stride=2,
activation_fn=tf.nn.relu, normalizer_fn=ly.batch_norm, padding='SAME', weights_initializer=tf.random_normal_initializer(0, 0.02))
train = ly.conv2d_transpose(train, 1, 3, stride=1,
activation_fn=None, padding='SAME', weights_initializer=tf.random_normal_initializer(0, 0.02), biases_initializer=None)
bias = slim.model_variable('bias', shape=(
1, ), initializer=tf.zeros_initializer)
train += bias
train = tf.nn.tanh(train)
return train
def discriminator(img, name, target):
size = 64
with tf.variable_scope(name):
# img = ly.conv2d(img, num_outputs=size, kernel_size=3,
# stride=2, activation_fn=None, biases_initializer=None)
# bias = slim.model_variable('conv_bias', shape=(
# size, ), initializer=tf.zeros_initializer)
# img += bias
# img = lrelu(img)
img = ly.conv2d(img, num_outputs=size, kernel_size=3,
stride=2, activation_fn=lrelu, normalizer_fn=ly.batch_norm)
img = ly.conv2d(img, num_outputs=size * 2, kernel_size=3,
stride=2, activation_fn=lrelu, normalizer_fn=ly.batch_norm)
img = ly.conv2d(img, num_outputs=size * 4, kernel_size=3,
stride=2, activation_fn=lrelu, normalizer_fn=ly.batch_norm)
img = tf.reshape(img, (2 * batch_size, -1))
weights = slim.model_variable('weights', shape=[img.get_shape().as_list()[-1], 1],
initializer=ly.xavier_initializer())
bias = slim.model_variable('bias', shape=(
1,), initializer=tf.zeros_initializer)
logit = fully_connected(img, weights, bias)
fake_logit = logit[:FLAGS.batch_size]
true_logit = logit[FLAGS.batch_size:]
d_loss_fake = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
fake_logit, tf.zeros_like(fake_logit)))
d_loss_true = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
true_logit, tf.ones_like(true_logit)))
f = tf.reduce_mean(d_loss_fake + d_loss_true)
return f, logit, d_loss_true, d_loss_fake
