experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Make core layers tests run in graph and eager mode.

Browse Source 
PiperOrigin-RevId: 225231668
This commit is contained in:
Francois Chollet 2018-12-12 12:23:40 -08:00 committed by TensorFlower Gardener
parent 16069bf874
commit 250ab666f6
2 changed files with 191 additions and 206 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorflow/python/keras/BUILD
View File

@ -401,7 +401,7 @@ py_test(
name = "core_test",
size = "medium",
srcs = ["layers/core_test.py"],
shard_count = 2,
shard_count = 3,
srcs_version = "PY2AND3",
deps = [
":keras",

395
tensorflow/python/keras/layers/core_test.py
View File

@ -22,43 +22,36 @@ import numpy as np
from tensorflow.python import keras
from tensorflow.python.eager import context
from tensorflow.python.framework import test_util as tf_test_util
from tensorflow.python.keras import keras_parameterized
from tensorflow.python.keras import testing_utils
from tensorflow.python.ops import math_ops
from tensorflow.python.platform import test
class CoreLayersTest(test.TestCase):
def test_masking(self):
with self.cached_session():
testing_utils.layer_test(
keras.layers.Masking, kwargs={}, input_shape=(3, 2, 3))
@keras_parameterized.run_all_keras_modes
class DropoutLayersTest(keras_parameterized.TestCase):
def test_dropout(self):
with self.cached_session():
testing_utils.layer_test(
keras.layers.Dropout, kwargs={'rate': 0.5}, input_shape=(3, 2))
testing_utils.layer_test(
keras.layers.Dropout, kwargs={'rate': 0.5}, input_shape=(3, 2))
with self.cached_session():
testing_utils.layer_test(
keras.layers.Dropout,
kwargs={'rate': 0.5,
'noise_shape': [3, 1]},
input_shape=(3, 2))
testing_utils.layer_test(
keras.layers.Dropout,
kwargs={'rate': 0.5,
'noise_shape': [3, 1]},
input_shape=(3, 2))
# https://github.com/tensorflow/tensorflow/issues/14819
with self.cached_session():
dropout = keras.layers.Dropout(0.5)
self.assertEqual(True, dropout.supports_masking)
def test_dropout_supports_masking(self):
dropout = keras.layers.Dropout(0.5)
self.assertEqual(True, dropout.supports_masking)
@tf_test_util.run_in_graph_and_eager_modes
def test_spatial_dropout(self):
def test_spatial_dropout_1d(self):
testing_utils.layer_test(
keras.layers.SpatialDropout1D,
kwargs={'rate': 0.5},
input_shape=(2, 3, 4))
def test_spatial_dropout_2d(self):
testing_utils.layer_test(
keras.layers.SpatialDropout2D,
kwargs={'rate': 0.5},
@ -69,6 +62,7 @@ class CoreLayersTest(test.TestCase):
kwargs={'rate': 0.5, 'data_format': 'channels_first'},
input_shape=(2, 3, 4, 5))
def test_spatial_dropout_3d(self):
testing_utils.layer_test(
keras.layers.SpatialDropout3D,
kwargs={'rate': 0.5},
@ -79,94 +73,9 @@ class CoreLayersTest(test.TestCase):
kwargs={'rate': 0.5, 'data_format': 'channels_first'},
input_shape=(2, 3, 4, 4, 5))
@tf_test_util.run_in_graph_and_eager_modes
def test_activation(self):
# with string argument
testing_utils.layer_test(
keras.layers.Activation,
kwargs={'activation': 'relu'},
input_shape=(3, 2))
# with function argument
testing_utils.layer_test(
keras.layers.Activation,
kwargs={'activation': keras.backend.relu},
input_shape=(3, 2))
@tf_test_util.run_in_graph_and_eager_modes
def test_reshape(self):
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (8, 1)},
input_shape=(3, 2, 4))
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (-1, 1)},
input_shape=(3, 2, 4))
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (1, -1)},
input_shape=(3, 2, 4))
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (-1, 1)},
input_shape=(None, None, 2))
@tf_test_util.run_in_graph_and_eager_modes
def test_permute(self):
testing_utils.layer_test(
keras.layers.Permute, kwargs={'dims': (2, 1)}, input_shape=(3, 2, 4))
@tf_test_util.run_in_graph_and_eager_modes
def test_permute_errors_on_invalid_starting_dims_index(self):
with self.assertRaisesRegexp(ValueError, r'Invalid permutation .*dims.*'):
testing_utils.layer_test(
keras.layers.Permute,
kwargs={'dims': (0, 1, 2)}, input_shape=(3, 2, 4))
@tf_test_util.run_in_graph_and_eager_modes
def test_permute_errors_on_invalid_set_of_dims_indices(self):
with self.assertRaisesRegexp(ValueError, r'Invalid permutation .*dims.*'):
testing_utils.layer_test(
keras.layers.Permute,
kwargs={'dims': (1, 4, 2)}, input_shape=(3, 2, 4))
@tf_test_util.run_in_graph_and_eager_modes
def test_flatten(self):
testing_utils.layer_test(
keras.layers.Flatten, kwargs={}, input_shape=(3, 2, 4))
# Test channels_first
inputs = np.random.random((10, 3, 5, 5)).astype('float32')
outputs = testing_utils.layer_test(
keras.layers.Flatten,
kwargs={'data_format': 'channels_first'},
input_data=inputs)
target_outputs = np.reshape(
np.transpose(inputs, (0, 2, 3, 1)), (-1, 5 * 5 * 3))
self.assertAllClose(outputs, target_outputs)
@tf_test_util.run_in_graph_and_eager_modes
def test_flatten_scalar_channels(self):
testing_utils.layer_test(
keras.layers.Flatten, kwargs={}, input_shape=(3,))
# Test channels_first
inputs = np.random.random((10,)).astype('float32')
outputs = testing_utils.layer_test(
keras.layers.Flatten,
kwargs={'data_format': 'channels_first'},
input_data=inputs)
target_outputs = np.expand_dims(inputs, -1)
self.assertAllClose(outputs, target_outputs)
@tf_test_util.run_in_graph_and_eager_modes
def test_repeat_vector(self):
testing_utils.layer_test(
keras.layers.RepeatVector, kwargs={'n': 3}, input_shape=(3, 2))
@keras_parameterized.run_all_keras_modes
class LambdaLayerTest(keras_parameterized.TestCase):
def test_lambda(self):
testing_utils.layer_test(
@ -202,7 +111,6 @@ class CoreLayersTest(test.TestCase):
config = ld.get_config()
ld = keras.layers.Lambda.from_config(config)
@tf_test_util.run_in_graph_and_eager_modes
def test_lambda_multiple_inputs(self):
ld = keras.layers.Lambda(lambda x: x[0], output_shape=lambda x: x[0])
x1 = np.ones([3, 2], np.float32)
@ -210,7 +118,158 @@ class CoreLayersTest(test.TestCase):
out = ld([x1, x2])
self.assertAllEqual(out.shape, [3, 2])
@tf_test_util.run_in_graph_and_eager_modes
def test_lambda_output_shape(self):
l = keras.layers.Lambda(lambda x: x + 1, output_shape=(1, 1))
l(keras.backend.variable(np.ones((1, 1))))
self.assertEqual((1, 1), l.get_config()['output_shape'])
def test_lambda_output_shape_function(self):
def get_output_shape(input_shape):
return 1 * input_shape
l = keras.layers.Lambda(lambda x: x + 1, output_shape=get_output_shape)
l(keras.backend.variable(np.ones((1, 1))))
self.assertEqual('lambda', l.get_config()['output_shape_type'])
def test_lambda_output_shape_autocalculate_multiple_inputs(self):
def lambda_fn(x):
return math_ops.matmul(x[0], x[1])
l = keras.layers.Lambda(lambda_fn)
output_shape = l.compute_output_shape([(10, 10), (10, 20)])
self.assertAllEqual((10, 20), output_shape)
def test_lambda_output_shape_list_multiple_outputs(self):
def lambda_fn(x):
return x
l = keras.layers.Lambda(lambda_fn, output_shape=[(10,), (20,)])
output_shape = l.compute_output_shape([(10, 10), (10, 20)])
self.assertAllEqual([(10, 10), (10, 20)], output_shape)
def test_lambda_output_shape_tuple_with_none(self):
def lambda_fn(x):
return x
l = keras.layers.Lambda(lambda_fn, output_shape=(None, 10))
output_shape = l.compute_output_shape((5, 10, 20))
self.assertAllEqual([5, None, 10], output_shape.as_list())
def test_lambda_output_shape_function_multiple_outputs(self):
def lambda_fn(x):
return x
def output_shape_fn(input_shape):
return input_shape
l = keras.layers.Lambda(lambda_fn, output_shape=output_shape_fn)
output_shape = l.compute_output_shape([(10, 10), (10, 20)])
self.assertAllEqual([(10, 10), (10, 20)], output_shape)
def test_lambda_config_serialization(self):
# Test serialization with output_shape and output_shape_type
layer = keras.layers.Lambda(lambda x: x + 1, output_shape=(1, 1))
layer(keras.backend.variable(np.ones((1, 1))))
config = layer.get_config()
layer = keras.layers.deserialize({
'class_name': 'Lambda',
'config': config
})
layer = keras.layers.Lambda.from_config(config)
@keras_parameterized.run_all_keras_modes
class CoreLayersTest(keras_parameterized.TestCase):
def test_masking(self):
testing_utils.layer_test(
keras.layers.Masking, kwargs={}, input_shape=(3, 2, 3))
def test_activation(self):
# with string argument
testing_utils.layer_test(
keras.layers.Activation,
kwargs={'activation': 'relu'},
input_shape=(3, 2))
# with function argument
testing_utils.layer_test(
keras.layers.Activation,
kwargs={'activation': keras.backend.relu},
input_shape=(3, 2))
def test_reshape(self):
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (8, 1)},
input_shape=(3, 2, 4))
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (-1, 1)},
input_shape=(3, 2, 4))
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (1, -1)},
input_shape=(3, 2, 4))
testing_utils.layer_test(
keras.layers.Reshape,
kwargs={'target_shape': (-1, 1)},
input_shape=(None, None, 2))
def test_permute(self):
testing_utils.layer_test(
keras.layers.Permute, kwargs={'dims': (2, 1)}, input_shape=(3, 2, 4))
def test_permute_errors_on_invalid_starting_dims_index(self):
with self.assertRaisesRegexp(ValueError, r'Invalid permutation .*dims.*'):
testing_utils.layer_test(
keras.layers.Permute,
kwargs={'dims': (0, 1, 2)}, input_shape=(3, 2, 4))
def test_permute_errors_on_invalid_set_of_dims_indices(self):
with self.assertRaisesRegexp(ValueError, r'Invalid permutation .*dims.*'):
testing_utils.layer_test(
keras.layers.Permute,
kwargs={'dims': (1, 4, 2)}, input_shape=(3, 2, 4))
def test_flatten(self):
testing_utils.layer_test(
keras.layers.Flatten, kwargs={}, input_shape=(3, 2, 4))
# Test channels_first
inputs = np.random.random((10, 3, 5, 5)).astype('float32')
outputs = testing_utils.layer_test(
keras.layers.Flatten,
kwargs={'data_format': 'channels_first'},
input_data=inputs)
target_outputs = np.reshape(
np.transpose(inputs, (0, 2, 3, 1)), (-1, 5 * 5 * 3))
self.assertAllClose(outputs, target_outputs)
def test_flatten_scalar_channels(self):
testing_utils.layer_test(
keras.layers.Flatten, kwargs={}, input_shape=(3,))
# Test channels_first
inputs = np.random.random((10,)).astype('float32')
outputs = testing_utils.layer_test(
keras.layers.Flatten,
kwargs={'data_format': 'channels_first'},
input_data=inputs)
target_outputs = np.expand_dims(inputs, -1)
self.assertAllClose(outputs, target_outputs)
def test_repeat_vector(self):
testing_utils.layer_test(
keras.layers.RepeatVector, kwargs={'n': 3}, input_shape=(3, 2))
def test_dense(self):
testing_utils.layer_test(
keras.layers.Dense, kwargs={'units': 3}, input_shape=(3, 2))
@ -225,105 +284,31 @@ class CoreLayersTest(test.TestCase):
keras.layers.Dense, kwargs={'units': 3}, input_shape=(3, 4, 5, 2))
def test_dense_regularization(self):
with self.cached_session():
layer = keras.layers.Dense(
3,
kernel_regularizer=keras.regularizers.l1(0.01),
bias_regularizer='l1',
activity_regularizer='l2',
name='dense_reg')
layer(keras.backend.variable(np.ones((2, 4))))
self.assertEqual(3, len(layer.losses))
layer = keras.layers.Dense(
3,
kernel_regularizer=keras.regularizers.l1(0.01),
bias_regularizer='l1',
activity_regularizer='l2',
name='dense_reg')
layer(keras.backend.variable(np.ones((2, 4))))
self.assertEqual(3, len(layer.losses))
def test_dense_constraints(self):
with self.cached_session():
k_constraint = keras.constraints.max_norm(0.01)
b_constraint = keras.constraints.max_norm(0.01)
layer = keras.layers.Dense(
3, kernel_constraint=k_constraint, bias_constraint=b_constraint)
layer(keras.backend.variable(np.ones((2, 4))))
self.assertEqual(layer.kernel.constraint, k_constraint)
self.assertEqual(layer.bias.constraint, b_constraint)
k_constraint = keras.constraints.max_norm(0.01)
b_constraint = keras.constraints.max_norm(0.01)
layer = keras.layers.Dense(
3, kernel_constraint=k_constraint, bias_constraint=b_constraint)
layer(keras.backend.variable(np.ones((2, 4))))
self.assertEqual(layer.kernel.constraint, k_constraint)
self.assertEqual(layer.bias.constraint, b_constraint)
def test_activity_regularization(self):
with self.cached_session():
layer = keras.layers.ActivityRegularization(l1=0.1)
layer(keras.backend.variable(np.ones((2, 4))))
self.assertEqual(1, len(layer.losses))
_ = layer.get_config()
layer = keras.layers.ActivityRegularization(l1=0.1)
layer(keras.backend.variable(np.ones((2, 4))))
self.assertEqual(1, len(layer.losses))
config = layer.get_config()
self.assertEqual(config.pop('l1'), 0.1)
def test_lambda_output_shape(self):
with self.cached_session():
l = keras.layers.Lambda(lambda x: x + 1, output_shape=(1, 1))
l(keras.backend.variable(np.ones((1, 1))))
self.assertEqual((1, 1), l.get_config()['output_shape'])
def test_lambda_output_shape_function(self):
def get_output_shape(input_shape):
return 1 * input_shape
with self.cached_session():
l = keras.layers.Lambda(lambda x: x + 1, output_shape=get_output_shape)
l(keras.backend.variable(np.ones((1, 1))))
self.assertEqual('lambda', l.get_config()['output_shape_type'])
@tf_test_util.run_in_graph_and_eager_modes
def test_lambda_output_shape_autocalculate_multiple_inputs(self):
def lambda_fn(x):
return math_ops.matmul(x[0], x[1])
l = keras.layers.Lambda(lambda_fn)
output_shape = l.compute_output_shape([(10, 10), (10, 20)])
self.assertAllEqual((10, 20), output_shape)
@tf_test_util.run_in_graph_and_eager_modes
def test_lambda_output_shape_list_multiple_outputs(self):
def lambda_fn(x):
return x
l = keras.layers.Lambda(lambda_fn, output_shape=[(10,), (20,)])
output_shape = l.compute_output_shape([(10, 10), (10, 20)])
self.assertAllEqual([(10, 10), (10, 20)], output_shape)
@tf_test_util.run_in_graph_and_eager_modes
def test_lambda_output_shape_tuple_with_none(self):
def lambda_fn(x):
return x
l = keras.layers.Lambda(lambda_fn, output_shape=(None, 10))
output_shape = l.compute_output_shape((5, 10, 20))
self.assertAllEqual([5, None, 10], output_shape.as_list())
@tf_test_util.run_in_graph_and_eager_modes
def test_lambda_output_shape_function_multiple_outputs(self):
def lambda_fn(x):
return x
def output_shape_fn(input_shape):
return input_shape
l = keras.layers.Lambda(lambda_fn, output_shape=output_shape_fn)
output_shape = l.compute_output_shape([(10, 10), (10, 20)])
self.assertAllEqual([(10, 10), (10, 20)], output_shape)
def test_lambda_config_serialization(self):
with self.cached_session():
# test serialization with output_shape and output_shape_type
layer = keras.layers.Lambda(lambda x: x + 1, output_shape=(1, 1))
layer(keras.backend.variable(np.ones((1, 1))))
config = layer.get_config()
layer = keras.layers.deserialize({
'class_name': 'Lambda',
'config': config
})
layer = keras.layers.Lambda.from_config(config)
@tf_test_util.run_in_graph_and_eager_modes
def test_numpy_inputs(self):
if context.executing_eagerly():
layer = keras.layers.RepeatVector(2)