def _build(self, incoming, loss_config, encoder_fn, decoder_fn, *args,
**kwargs):
self._build_dependencies()
losses = None
loss = None
if Modes.GENERATE == self.mode:
results = self.decode(incoming=incoming, decoder_fn=decoder_fn)
elif Modes.ENCODE == self.mode:
encoded = self.encode(incoming=incoming, encoder_fn=encoder_fn)
results = self.z_mean(encoded)
else:
encoded = self.encode(incoming=incoming, encoder_fn=encoder_fn)
z_mean = self.z_mean(encoded)
z_log_sigma = self.z_log_sigma(encoded)
shape = self._get_decoder_shape(incoming)
eps = tf.random_normal(shape=shape,
mean=self.mean,
stddev=self.stddev,
dtype=tf.float32,
name='eps')
z = tf.add(z_mean, tf.multiply(tf.sqrt(tf.exp(z_log_sigma)), eps))
results = self.decode(incoming=z, decoder_fn=decoder_fn)
losses, loss = self._build_loss(incoming,
results,
loss_config,
z_mean=z_mean,
z_log_sigma=z_log_sigma)
return BridgeSpec(results=results, losses=losses, loss=loss)
def _build(self, incoming, loss_config, encoder_fn, decoder_fn, *args, **kwargs):
losses, loss = None, None
if Modes.GENERATE == self.mode:
results = self.decode(incoming=incoming, decoder_fn=decoder_fn)
elif Modes.ENCODE == self.mode:
results = self.encode(incoming=incoming, encoder_fn=encoder_fn)
else:
x = self.encode(incoming=incoming, encoder_fn=encoder_fn)
results = self.decode(incoming=x, decoder_fn=decoder_fn)
if not Modes.is_infer(self.mode):
losses, loss = self._build_loss(incoming, results, loss_config)
return BridgeSpec(results=results, losses=losses, loss=loss)
def _build(self, features, labels, loss, encoder_fn, decoder_fn, *args,
**kwargs):
losses, loss = None, None
if Modes.GENERATE == self.mode:
results = self.decode(incoming=features,
features=features,
labels=labels,
decoder_fn=decoder_fn)
elif Modes.ENCODE == self.mode:
results = self.encode(features=features,
labels=labels,
encoder_fn=encoder_fn)
else:
x = self.encode(features=features,
labels=labels,
encoder_fn=encoder_fn)
results = self.decode(features=x,
labels=labels,
decoder_fn=decoder_fn)
if not Modes.is_infer(self.mode):
losses, loss = self._build_loss(results, features, labels,
loss)
return BridgeSpec(results=results, losses=losses, loss=loss)