def sample(self, z, flat=True):
z_reshaped = []
num_samples = -1 #z.get_shape().as_list()[0]
x = tf.reshape(z[:, :self.z_shape_list[-1][0]*
self.z_shape_list[-1][1]*
self.z_shape_list[-1][2]],
[num_samples] + list(self.z_shape_list[-1]))
z = z[:, self.z_shape_list[-1][0]*self.z_shape_list[-1][1]*self.z_shape_list[-1][2]:]
for l in reversed(range(1, self.L)):
for k in reversed(range(0, self.K)):
x = self.blocks[l][k](x, inverse=True)
x = ops.unsqueeze2d(x)
x_concat = tf.reshape(z[:, :self.z_shape_list[l-1][0]*
self.z_shape_list[l-1][1]*
self.z_shape_list[l-1][2]],
[num_samples] + list(self.z_shape_list[l-1]))
z = z[:, self.z_shape_list[l-1][0]*
self.z_shape_list[l-1][1]*
self.z_shape_list[l-1][2]:]
x = tf.concat([x_concat, x], axis=-1)
l = 0
for k in reversed(range(0, self.K)):
x = self.blocks[l][k](x, inverse=True)
x_reconstructed = ops.unsqueeze2d(x)
return x_reconstructed
def split2d_reverse(z, eps, hps=None, name=None):
with tf.variable_scope(name, reuse=tf.AUTO_REUSE):
z1 = unsqueeze2d(z)
pz = split2d_prior(z1, hps=hps)
z2 = pz.sample(eps=eps)
z = tf.concat([z1, z2], 3)
return z
def split2d_reverse(z, eps, hps=None, name=None):
with tf.variable_scope(name, reuse=tf.AUTO_REUSE):
z1 = unsqueeze2d(z)
pz = split2d_prior(z1, hps=hps)
z2 = pz.sample(eps=eps)
z = tf.concat([z1, z2], 3)
return z
def f_decode(self, y, eps):
with tf.variable_scope('model', reuse=tf.AUTO_REUSE):
y_onehot = tf.cast(tf.one_hot(y, self.cfg.n_y, 1, 0), 'float32')
_, sample, _ = prior("prior", y_onehot, self.cfg)
z = sample(eps=eps[-1])
z = self.decoder(z, eps=eps[:-1])
z = ops.unsqueeze2d(z, 2) # 8x8x12 -> 16x16x3
x = self.postprocess(z)
return x
def sample(self, y, temp=1.0, eps=None, post_process=True):
if eps is None:
eps = [None]*self.cfg.n_levels
with tf.variable_scope('model', reuse=tf.AUTO_REUSE):
y_onehot = tf.cast(tf.one_hot(y, self.cfg.n_y, 1, 0), 'float32')
_, sample, _ = prior("prior", y_onehot, self.cfg)
z = sample(temp=temp)
x = self.decoder(z, eps)
x = ops.unsqueeze2d(x, 2) # 8x8x12 -> 16x16x3
if post_process:
x = self.postprocess(x)
return x
def decode(self, labels=None, condition=None, epsilon=None):
"""
参数列表:
标签: Class label, could be none
条件参数: 2D or 4D tensor, condition for dynamic linear transformation
epsilon: None or list. If specified, it should be a list with `num_levels` elements
返回值:
x: 4D tensor, generated samples
"""
with tf.variable_scope("prior", reuse=tf.AUTO_REUSE):
if labels is None:
y_onehot = self.y
elif len(labels.shape) == 1:
y_onehot = tf.one_hot(labels,
depth=self.num_classes,
dtype=tf.float32)
elif len(labels.shape) == 2:
y_onehot = labels
_, sample, get_eps = prior(y_onehot, self.hps)
if epsilon is not None:
eps = epsilon if len(
epsilon) == self.hps.num_levels else [None] * (
self.hps.num_levels - 1) + epsilon
else:
eps = [None] * self.hps.num_levels
z = sample(eps=eps[-1])
objective = tf.zeros(tf.shape(z)[0])
if self.hps.conditioning and condition is None:
condition = y_onehot
# with tf.variable_scope("cond_preprocess", reuse=tf.AUTO_REUSE):
# condition = tf.layers.dense(condition, units=10, use_bias=False)
z, objective = codec(z,
cond=condition,
hps=self.hps,
reverse=True,
objective=objective,
eps=eps[:-1],
reuse=tf.AUTO_REUSE)
with tf.variable_scope("postprocess"):
x = unsqueeze2d(z)
x = tf.clip_by_value(
tf.floor((x + .5) * self.num_bins) * (256. / self.num_bins), 0,
255)
self.gen_x = tf.cast(x, 'uint8')
return self.gen_x
def split2d_reverse(name, z, eps, eps_std, cfg):
with tf.variable_scope(name):
z1 = ops.unsqueeze2d(z)
pz = split2d_prior(z1, cfg)
if eps is not None:
# Already sampled eps
z2 = pz.sample_eps(eps)
elif eps_std is not None:
# Sample with given eps_std
z2 = pz.sample_eps(pz.eps * tf.reshape(eps_std, [-1, 1, 1, 1]))
else:
# Sample normally
z2 = pz.sample(1.0)
z = tf.concat([z1, z2], 3)
return z
def decode(self, labels=None, condition=None, epsilon=None):
"""
Args:
labels: Class label, could be none
condition: 2D or 4D tensor, condition for dynamic linear transformation
epsilon: None or list. If specified, it should be a list with `num_levels` elements
Returns:
x: 4D tensor, generated samples
"""
with tf.variable_scope("prior", reuse=tf.AUTO_REUSE):
if labels is None:
y_onehot = self.y
elif len(labels.shape) == 1:
y_onehot = tf.one_hot(labels, depth=self.num_classes, dtype=tf.float32)
elif len(labels.shape) == 2:
y_onehot = labels
_, sample, get_eps = prior(y_onehot, self.hps)
if epsilon is not None:
eps = epsilon if len(epsilon) == self.hps.num_levels else [None] * (self.hps.num_levels-1) + epsilon
else:
eps = [None] * self.hps.num_levels
z = sample(eps=eps[-1])
objective = tf.zeros(tf.shape(z)[0])
if self.hps.conditioning and condition is None:
condition = y_onehot
# with tf.variable_scope("cond_preprocess", reuse=tf.AUTO_REUSE):
# condition = tf.layers.dense(condition, units=10, use_bias=False)
z, objective = codec(z, cond=condition, hps=self.hps, reverse=True,
objective=objective, eps=eps[:-1], reuse=tf.AUTO_REUSE)
with tf.variable_scope("postprocess"):
x = unsqueeze2d(z)
x = tf.clip_by_value(tf.floor((x+.5)*self.num_bins)*(256./self.num_bins), 0, 255)
self.gen_x = tf.cast(x, 'uint8')
return self.gen_x