def composite_layer(inputs, mask, hparams, for_output=False):
"""Composite layer."""
x = inputs
# Applies ravanbakhsh on top of each other.
if hparams.composite_layer_type == "ravanbakhsh":
for layer in xrange(hparams.layers_per_layer):
with tf.variable_scope(".%d" % layer):
x = common_layers.ravanbakhsh_set_layer(
hparams.hidden_size,
x,
mask=mask,
sequential=for_output,
dropout=hparams.relu_dropout)
# Transforms elements to get a context, and then uses this in a final layer.
elif hparams.composite_layer_type == "reembedding":
# Transform elements n times and then pool.
for layer in xrange(hparams.layers_per_layer):
with tf.variable_scope("sub_layer_%d" % layer):
x = common_layers.linear_set_layer(
hparams.hidden_size, x, dropout=hparams.relu_dropout)
if for_output:
context = common_layers.running_global_pool_1d(x)
else:
context = common_layers.global_pool_1d(x, mask=mask)
# Final layer.
x = common_layers.linear_set_layer(hparams.hidden_size,
x,
context=context,
dropout=hparams.relu_dropout)
return x
def testLinearSetLayer(self):
x1 = np.random.rand(5, 4, 11)
cont = np.random.rand(5, 13)
x1_ = tf.Variable(x1, dtype=tf.float32)
cont_ = tf.Variable(cont, dtype=tf.float32)
simple_ff = common_layers.linear_set_layer(32, x1_)
cont_ff = common_layers.linear_set_layer(32, x1_, context=cont_)
self.evaluate(tf.global_variables_initializer())
actual = self.evaluate([simple_ff, cont_ff])
self.assertEqual(actual[0].shape, (5, 4, 32))
self.assertEqual(actual[1].shape, (5, 4, 32))
def testLinearSetLayer(self):
x1 = np.random.rand(5, 4, 11)
cont = np.random.rand(5, 13)
x1_ = tf.Variable(x1, dtype=tf.float32)
cont_ = tf.Variable(cont, dtype=tf.float32)
simple_ff = common_layers.linear_set_layer(32, x1_)
cont_ff = common_layers.linear_set_layer(32, x1_, context=cont_)
self.evaluate(tf.global_variables_initializer())
actual = self.evaluate([simple_ff, cont_ff])
self.assertEqual(actual[0].shape, (5, 4, 32))
self.assertEqual(actual[1].shape, (5, 4, 32))
