def testPool(self):
x = np.random.rand(5, 8, 1, 11)
y = common_layers.pool(tf.constant(x, dtype=tf.float32), (2, 2), "AVG",
"SAME")
self.evaluate(tf.global_variables_initializer())
res = self.evaluate(y)
self.assertEqual(res.shape, (5, 8, 1, 11))
def testPool(self):
x = np.random.rand(5, 8, 1, 11)
y = common_layers.pool(
tf.constant(x, dtype=tf.float32), (2, 2), "AVG", "SAME")
self.evaluate(tf.global_variables_initializer())
res = self.evaluate(y)
self.assertEqual(res.shape, (5, 8, 1, 11))
def testPool(self):
x = np.random.rand(5, 8, 1, 11)
with self.test_session() as session:
y = common_layers.pool(tf.constant(x, dtype=tf.float32), (2, 2),
"AVG", "SAME")
session.run(tf.global_variables_initializer())
res = session.run(y)
self.assertEqual(res.shape, (5, 8, 1, 11))
def xnet_resblock(x, filters, res_relu, name):
with tf.variable_scope(name):
y = common_layers.separable_conv_block(
x,
filters, [((1, 1), (3, 3)), ((1, 1), (3, 3))],
first_relu=True,
padding="SAME",
force2d=True,
name="sep_conv_block")
y = common_layers.pool(y, (3, 3), "MAX", "SAME", strides=(2, 2))
return y + common_layers.conv_block(
x,
filters, [((1, 1), (1, 1))],
padding="SAME",
strides=(2, 2),
first_relu=res_relu,
force2d=True,
name="res_conv0")
def xnet_resblock(x, filters, res_relu, name):
with tf.variable_scope(name):
y = common_layers.separable_conv_block(
x,
filters, [((1, 1), (3, 3)), ((1, 1), (3, 3))],
first_relu=True,
padding="SAME",
force2d=True,
name="sep_conv_block")
y = common_layers.pool(y, (3, 3), "MAX", "SAME", strides=(2, 2))
return y + common_layers.conv_block(
x,
filters, [((1, 1), (1, 1))],
padding="SAME",
strides=(2, 2),
first_relu=res_relu,
force2d=True,
name="res_conv0")
def xnet_resblock(x, filters, res_relu, name):
with tf.variable_scope(name):
# We only stride along the length dimension to preserve the spectral
# bins (which are tiny in dimensionality relative to length)
y = common_layers.separable_conv_block(
x,
filters, [((1, 1), (3, 3)), ((1, 1), (3, 3))],
first_relu=True,
padding="SAME",
force2d=True,
name="sep_conv_block")
y = common_layers.pool(y, (3, 3), "MAX", "SAME", strides=(2, 1))
return y + common_layers.conv_block(
x,
filters, [((1, 1), (1, 1))],
padding="SAME",
strides=(2, 1),
first_relu=res_relu,
force2d=True,
name="res_conv0")
def xnet_resblock(x, filters, res_relu, name):
with tf.variable_scope(name):
# Typically audio samples are >100k samples in length and have a width
# of 2 or 4. Mono audio has a single channel while stereo has 2.
y = common_layers.separable_conv_block(
x,
filters, [((1, 1), (3, 3)), ((1, 1), (3, 3))],
first_relu=True,
padding="SAME",
force2d=True,
name="sep_conv_block")
y = common_layers.pool(y, (3, 3), "MAX", "SAME", strides=(2, 2))
return y + common_layers.conv_block(
x,
filters, [((1, 1), (1, 1))],
padding="SAME",
strides=(2, 2),
first_relu=res_relu,
force2d=True,
name="res_conv0")
def xnet_resblock(x, filters, res_relu, name):
with tf.variable_scope(name):
# We only stride along the length dimension to preserve the spectral
# bins (which are tiny in dimensionality relative to length)
y = common_layers.separable_conv_block(
x,
filters, [((1, 1), (3, 3)), ((1, 1), (3, 3))],
first_relu=True,
padding="SAME",
force2d=True,
name="sep_conv_block")
y = common_layers.pool(y, (3, 3), "MAX", "SAME", strides=(2, 1))
return y + common_layers.conv_block(
x,
filters, [((1, 1), (1, 1))],
padding="SAME",
strides=(2, 1),
first_relu=res_relu,
force2d=True,
name="res_conv0")
def xnet_resblock(x, filters, res_relu, name):
with tf.variable_scope(name):
# Typically audio samples are >100k samples in length and have a width
# of 2 or 4. Mono audio has a single channel while stereo has 2.
y = common_layers.separable_conv_block(
x,
filters, [((1, 1), (3, 3)), ((1, 1), (3, 3))],
first_relu=True,
padding="SAME",
force2d=True,
name="sep_conv_block")
y = common_layers.pool(y, (3, 3), "MAX", "SAME", strides=(2, 2))
return y + common_layers.conv_block(
x,
filters, [((1, 1), (1, 1))],
padding="SAME",
strides=(2, 2),
first_relu=res_relu,
force2d=True,
name="res_conv0")