First step of migrating layers to new API.
Change: 121435753
This commit is contained in:
A. Unique TensorFlower
TensorFlower Gardener
parent
ec0b1393c7
commit
0393436023
@ -22,6 +22,7 @@ import numpy as np
|
||||
import tensorflow as tf
|
||||
|
||||
|
||||
# TODO(b/28426988): Add separate tests for non-legacy versions.
|
||||
class FullyConnectedTest(tf.test.TestCase):
|
||||
|
||||
def setUp(self):
|
||||
@ -41,8 +42,9 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
assert not tf.get_collection(tf.GraphKeys.SUMMARIES)
|
||||
|
||||
def _fully_connected_basic_use(self, x, num_output_units, expected_shape):
|
||||
output = tf.contrib.layers.fully_connected(x, num_output_units,
|
||||
activation_fn=tf.nn.relu)
|
||||
output = tf.contrib.layers.legacy_fully_connected(x,
|
||||
num_output_units,
|
||||
activation_fn=tf.nn.relu)
|
||||
|
||||
with tf.Session() as sess:
|
||||
with self.assertRaises(tf.errors.FailedPreconditionError):
|
||||
@ -71,7 +73,7 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
self.input_3_dim, last_dim, [2, 4, last_dim])
|
||||
|
||||
def test_relu_layer_basic_use(self):
|
||||
output = tf.contrib.layers.relu(self.input, 8)
|
||||
output = tf.contrib.layers.legacy_relu(self.input, 8)
|
||||
|
||||
with tf.Session() as sess:
|
||||
with self.assertRaises(tf.errors.FailedPreconditionError):
|
||||
@ -90,7 +92,7 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
len(tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)))
|
||||
|
||||
def test_relu6_layer_basic_use(self):
|
||||
output = tf.contrib.layers.relu6(self.input, 8)
|
||||
output = tf.contrib.layers.legacy_relu6(self.input, 8)
|
||||
|
||||
with tf.Session() as sess:
|
||||
with self.assertRaises(tf.errors.FailedPreconditionError):
|
||||
@ -112,11 +114,11 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
|
||||
def test_variable_reuse_with_scope(self):
|
||||
with tf.variable_scope('test') as vs:
|
||||
output1 = tf.contrib.layers.relu(self.input, 8)
|
||||
output2 = tf.contrib.layers.relu(self.input, 8)
|
||||
output1 = tf.contrib.layers.legacy_relu(self.input, 8)
|
||||
output2 = tf.contrib.layers.legacy_relu(self.input, 8)
|
||||
|
||||
with tf.variable_scope(vs, reuse=True):
|
||||
output3 = tf.contrib.layers.relu(self.input, 8)
|
||||
output3 = tf.contrib.layers.legacy_relu(self.input, 8)
|
||||
|
||||
with tf.Session() as sess:
|
||||
tf.initialize_all_variables().run()
|
||||
@ -127,7 +129,7 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
|
||||
def test_variable_reuse_with_template(self):
|
||||
tmpl1 = tf.make_template('test',
|
||||
tf.contrib.layers.fully_connected,
|
||||
tf.contrib.layers.legacy_fully_connected,
|
||||
num_output_units=8)
|
||||
output1 = tmpl1(self.input)
|
||||
output2 = tmpl1(self.input)
|
||||
@ -138,9 +140,11 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
self.assertAllClose(out_value1, out_value2)
|
||||
|
||||
def _custom_initializers(self, x, num_output_units, expected_outputs):
|
||||
output = tf.contrib.layers.relu(x, num_output_units,
|
||||
weight_init=tf.constant_initializer(2.0),
|
||||
bias_init=tf.constant_initializer(1.0))
|
||||
output = tf.contrib.layers.legacy_relu(
|
||||
x,
|
||||
num_output_units,
|
||||
weight_init=tf.constant_initializer(2.0),
|
||||
bias_init=tf.constant_initializer(1.0))
|
||||
|
||||
with tf.Session() as sess:
|
||||
tf.initialize_all_variables().run()
|
||||
@ -165,10 +169,11 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
[49.6, 49.6]]])
|
||||
|
||||
def test_custom_collections(self):
|
||||
tf.contrib.layers.relu(self.input, 2,
|
||||
weight_collections=['unbiased'],
|
||||
bias_collections=['biased'],
|
||||
output_collections=['output'])
|
||||
tf.contrib.layers.legacy_relu(self.input,
|
||||
2,
|
||||
weight_collections=['unbiased'],
|
||||
bias_collections=['biased'],
|
||||
output_collections=['output'])
|
||||
|
||||
self.assertEquals(1, len(tf.get_collection('unbiased')))
|
||||
self.assertEquals(1, len(tf.get_collection('biased')))
|
||||
@ -176,9 +181,10 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
self.assertEquals(2, len(tf.get_collection(tf.GraphKeys.VARIABLES)))
|
||||
|
||||
def test_all_custom_collections(self):
|
||||
tf.contrib.layers.relu(self.input, 2,
|
||||
weight_collections=['unbiased', 'all'],
|
||||
bias_collections=['biased', 'all'])
|
||||
tf.contrib.layers.legacy_relu(self.input,
|
||||
2,
|
||||
weight_collections=['unbiased', 'all'],
|
||||
bias_collections=['biased', 'all'])
|
||||
|
||||
self.assertEquals(1, len(tf.get_collection('unbiased')))
|
||||
self.assertEquals(1, len(tf.get_collection('biased')))
|
||||
@ -186,16 +192,16 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
tf.get_collection('all'))
|
||||
|
||||
def test_no_bias(self):
|
||||
tf.contrib.layers.relu(self.input, 2, bias_init=None)
|
||||
tf.contrib.layers.legacy_relu(self.input, 2, bias_init=None)
|
||||
self.assertEqual(1, len(tf.get_collection(tf.GraphKeys.VARIABLES)))
|
||||
|
||||
def test_no_activation(self):
|
||||
y = tf.contrib.layers.fully_connected(self.input, 2)
|
||||
y = tf.contrib.layers.legacy_fully_connected(self.input, 2)
|
||||
self.assertEquals(2, len(tf.get_collection(tf.GraphKeys.VARIABLES)))
|
||||
self.assertEquals('BiasAdd', y.op.type)
|
||||
|
||||
def test_no_activation_no_bias(self):
|
||||
y = tf.contrib.layers.fully_connected(self.input, 2, bias_init=None)
|
||||
y = tf.contrib.layers.legacy_fully_connected(self.input, 2, bias_init=None)
|
||||
self.assertEquals(1, len(tf.get_collection(tf.GraphKeys.VARIABLES)))
|
||||
self.assertEquals('MatMul', y.op.type)
|
||||
|
||||
@ -206,7 +212,9 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
cnt[0] += 1
|
||||
return tensor
|
||||
|
||||
tf.contrib.layers.fully_connected(self.input, 2, weight_regularizer=test_fn)
|
||||
tf.contrib.layers.legacy_fully_connected(self.input,
|
||||
2,
|
||||
weight_regularizer=test_fn)
|
||||
|
||||
self.assertEqual([tensor],
|
||||
tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES))
|
||||
@ -219,10 +227,12 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
cnt[0] += 1
|
||||
return tensor
|
||||
|
||||
tf.contrib.layers.fully_connected(self.input, 2,
|
||||
weight_regularizer=test_fn)
|
||||
tf.contrib.layers.fully_connected(self.input, 2,
|
||||
weight_regularizer=test_fn)
|
||||
tf.contrib.layers.legacy_fully_connected(self.input,
|
||||
2,
|
||||
weight_regularizer=test_fn)
|
||||
tf.contrib.layers.legacy_fully_connected(self.input,
|
||||
2,
|
||||
weight_regularizer=test_fn)
|
||||
|
||||
self.assertEqual([tensor, tensor],
|
||||
tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES))
|
||||
@ -236,12 +246,14 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
return tensor
|
||||
|
||||
with tf.variable_scope('test') as vs:
|
||||
tf.contrib.layers.fully_connected(self.input, 2,
|
||||
weight_regularizer=test_fn)
|
||||
tf.contrib.layers.legacy_fully_connected(self.input,
|
||||
2,
|
||||
weight_regularizer=test_fn)
|
||||
|
||||
with tf.variable_scope(vs, reuse=True):
|
||||
tf.contrib.layers.fully_connected(self.input, 2,
|
||||
weight_regularizer=test_fn)
|
||||
tf.contrib.layers.legacy_fully_connected(self.input,
|
||||
2,
|
||||
weight_regularizer=test_fn)
|
||||
|
||||
self.assertEqual([tensor],
|
||||
tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES))
|
||||
@ -254,7 +266,9 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
with self.test_session():
|
||||
x = tf.constant([[]], shape=[0, 3])
|
||||
self.assertEqual(0, tf.size(x).eval())
|
||||
y = tf.contrib.layers.fully_connected(x, 2, activation_fn=tf.nn.softmax)
|
||||
y = tf.contrib.layers.legacy_fully_connected(x,
|
||||
2,
|
||||
activation_fn=tf.nn.softmax)
|
||||
tf.initialize_all_variables().run()
|
||||
expected_y = np.array([]).reshape(0, 2)
|
||||
np.testing.assert_array_equal(expected_y, y.eval())
|
||||
@ -262,7 +276,7 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
def test_shapes_variable_first_dim(self):
|
||||
# first dimension is not known statically.
|
||||
x = tf.placeholder(tf.float32, shape=[None, 4, 3])
|
||||
y = tf.contrib.layers.fully_connected(x, 1)
|
||||
y = tf.contrib.layers.legacy_fully_connected(x, 1)
|
||||
# in the output we still only know the 2nd and 3rd dimensions statically.
|
||||
self.assertEquals(y.get_shape().as_list(), [None, 4, 1])
|
||||
with self.test_session() as sess:
|
||||
@ -280,7 +294,7 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
|
||||
def _unknown_dim_invalid_input(self, last_dim):
|
||||
x = tf.placeholder(tf.float32, shape=[3, last_dim])
|
||||
tf.contrib.layers.fully_connected(x, 2, activation_fn=None)
|
||||
tf.contrib.layers.legacy_fully_connected(x, 2, activation_fn=None)
|
||||
|
||||
def test_known_dim_valid_input(self):
|
||||
self._unknown_dim_invalid_input(last_dim=3)
|
||||
@ -295,7 +309,9 @@ class FullyConnectedTest(tf.test.TestCase):
|
||||
with self.assertRaisesRegexp(ValueError,
|
||||
'rank of x must be at least 2 not: 1'):
|
||||
x = tf.constant([[]], shape=[0])
|
||||
tf.contrib.layers.fully_connected(x, 2, activation_fn=tf.nn.softmax)
|
||||
tf.contrib.layers.legacy_fully_connected(x,
|
||||
2,
|
||||
activation_fn=tf.nn.softmax)
|
||||
|
||||
|
||||
class Convolution2dTest(tf.test.TestCase):
|
||||
@ -308,8 +324,9 @@ class Convolution2dTest(tf.test.TestCase):
|
||||
assert not tf.get_collection(tf.GraphKeys.SUMMARIES)
|
||||
|
||||
def test_basic_use(self):
|
||||
output = tf.contrib.layers.convolution2d(self.input, 8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
output = tf.contrib.layers.legacy_convolution2d(self.input,
|
||||
8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
|
||||
with tf.Session() as sess:
|
||||
with self.assertRaises(tf.errors.FailedPreconditionError):
|
||||
@ -328,17 +345,17 @@ class Convolution2dTest(tf.test.TestCase):
|
||||
|
||||
def test_variable_reuse_with_scope(self):
|
||||
with tf.variable_scope('test') as vs:
|
||||
output1 = tf.contrib.layers.convolution2d(self.input,
|
||||
8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
output2 = tf.contrib.layers.convolution2d(self.input,
|
||||
8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
output1 = tf.contrib.layers.legacy_convolution2d(self.input,
|
||||
8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
output2 = tf.contrib.layers.legacy_convolution2d(self.input,
|
||||
8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
|
||||
with tf.variable_scope(vs, reuse=True):
|
||||
output3 = tf.contrib.layers.convolution2d(self.input,
|
||||
8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
output3 = tf.contrib.layers.legacy_convolution2d(self.input,
|
||||
8, (3, 3),
|
||||
activation_fn=tf.nn.relu)
|
||||
|
||||
with tf.Session() as sess:
|
||||
tf.initialize_all_variables().run()
|
||||
@ -349,7 +366,7 @@ class Convolution2dTest(tf.test.TestCase):
|
||||
|
||||
def test_variable_reuse_with_template(self):
|
||||
tmpl1 = tf.make_template('test',
|
||||
tf.contrib.layers.convolution2d,
|
||||
tf.contrib.layers.legacy_convolution2d,
|
||||
kernel_size=(3, 3),
|
||||
num_output_channels=8)
|
||||
output1 = tmpl1(self.input)
|
||||
@ -361,10 +378,9 @@ class Convolution2dTest(tf.test.TestCase):
|
||||
self.assertAllClose(out_value1, out_value2)
|
||||
|
||||
def test_custom_initializers(self):
|
||||
output = tf.contrib.layers.convolution2d(
|
||||
output = tf.contrib.layers.legacy_convolution2d(
|
||||
self.input,
|
||||
2,
|
||||
(3, 3),
|
||||
2, (3, 3),
|
||||
activation_fn=tf.nn.relu,
|
||||
weight_init=tf.constant_initializer(2.0),
|
||||
bias_init=tf.constant_initializer(1.0),
|
||||
@ -378,21 +394,22 @@ class Convolution2dTest(tf.test.TestCase):
|
||||
np.array([[[[1261., 1261.]]], [[[3853., 3853.]]]]), out_value)
|
||||
|
||||
def test_custom_collections(self):
|
||||
tf.contrib.layers.convolution2d(self.input,
|
||||
2, (3, 3),
|
||||
activation_fn=tf.nn.relu,
|
||||
weight_collections=['unbiased'],
|
||||
bias_collections=['biased'])
|
||||
tf.contrib.layers.legacy_convolution2d(self.input,
|
||||
2, (3, 3),
|
||||
activation_fn=tf.nn.relu,
|
||||
weight_collections=['unbiased'],
|
||||
bias_collections=['biased'])
|
||||
|
||||
self.assertEquals(1, len(tf.get_collection('unbiased')))
|
||||
self.assertEquals(1, len(tf.get_collection('biased')))
|
||||
|
||||
def test_all_custom_collections(self):
|
||||
tf.contrib.layers.convolution2d(self.input,
|
||||
2, (3, 3),
|
||||
activation_fn=tf.nn.relu,
|
||||
weight_collections=['unbiased', 'all'],
|
||||
bias_collections=['biased', 'all'])
|
||||
tf.contrib.layers.legacy_convolution2d(
|
||||
self.input,
|
||||
2, (3, 3),
|
||||
activation_fn=tf.nn.relu,
|
||||
weight_collections=['unbiased', 'all'],
|
||||
bias_collections=['biased', 'all'])
|
||||
|
||||
self.assertEquals(1, len(tf.get_collection('unbiased')))
|
||||
self.assertEquals(1, len(tf.get_collection('biased')))
|
||||
@ -407,15 +424,18 @@ class Convolution2dTest(tf.test.TestCase):
|
||||
cnt[0] += 1
|
||||
return tensor
|
||||
|
||||
tf.contrib.layers.convolution2d(self.input, 2, (3, 3),
|
||||
weight_regularizer=test_fn)
|
||||
tf.contrib.layers.legacy_convolution2d(self.input,
|
||||
2, (3, 3),
|
||||
weight_regularizer=test_fn)
|
||||
|
||||
self.assertEqual([tensor],
|
||||
tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES))
|
||||
self.assertEqual(1, cnt[0])
|
||||
|
||||
def test_no_bias(self):
|
||||
tf.contrib.layers.convolution2d(self.input, 2, (3, 3), bias_init=None)
|
||||
tf.contrib.layers.legacy_convolution2d(self.input,
|
||||
2, (3, 3),
|
||||
bias_init=None)
|
||||
self.assertEqual(1,
|
||||
len(tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES)))
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user