def test_does_not_support_dict_columns(self):
with self.assertRaisesRegexp(
ValueError,
'Expected feature_columns to be iterable, found dict.'):
df.DenseFeatures(feature_columns={'a': fc.numeric_column('a')})(
features={
'a': [[0]]
})
def test_raises_if_duplicate_name(self):
with self.assertRaisesRegexp(
ValueError, 'Duplicate feature column name found for columns'):
df.DenseFeatures(feature_columns=[
fc.numeric_column('a'),
fc.numeric_column('a')
])(features={
'a': [[0]]
})
def test_bare_column(self):
with ops.Graph().as_default():
features = features = {'a': [0.]}
net = df.DenseFeatures(fc.numeric_column('a'))(features)
self.evaluate(variables_lib.global_variables_initializer())
self.evaluate(lookup_ops.tables_initializer())
self.assertAllClose([[0.]], self.evaluate(net))
def test_raises_if_shape_mismatch(self):
price = fc.numeric_column('price', shape=2)
with ops.Graph().as_default():
features = {'price': [[1.], [5.]]}
with self.assertRaisesRegexp(
Exception,
r'Cannot reshape a tensor with 2 elements to shape \[2,2\]'
):
df.DenseFeatures([price])(features)
def test_column_generator(self):
with ops.Graph().as_default():
features = features = {'a': [0.], 'b': [1.]}
columns = (fc.numeric_column(key) for key in features)
net = df.DenseFeatures(columns)(features)
self.evaluate(variables_lib.global_variables_initializer())
self.evaluate(lookup_ops.tables_initializer())
self.assertAllClose([[0., 1.]], self.evaluate(net))
def test_reshaping(self):
price = fc.numeric_column('price', shape=[1, 2])
with ops.Graph().as_default():
features = {'price': [[[1., 2.]], [[5., 6.]]]}
net = df.DenseFeatures([price])(features)
self.evaluate(variables_lib.global_variables_initializer())
self.evaluate(lookup_ops.tables_initializer())
self.assertAllClose([[1., 2.], [5., 6.]], self.evaluate(net))
def test_with_1d_sparse_tensor(self):
embedding_values = (
(1., 2., 3., 4., 5.), # id 0
(6., 7., 8., 9., 10.), # id 1
(11., 12., 13., 14., 15.) # id 2
)
def _initializer(shape, dtype, partition_info=None):
del shape, dtype, partition_info
return embedding_values
# price has 1 dimension in dense_features
price = fc.numeric_column('price')
# one_hot_body_style has 3 dims in dense_features.
body_style = fc.categorical_column_with_vocabulary_list(
'body-style', vocabulary_list=['hardtop', 'wagon', 'sedan'])
one_hot_body_style = fc.indicator_column(body_style)
# embedded_body_style has 5 dims in dense_features.
country = fc.categorical_column_with_vocabulary_list(
'country', vocabulary_list=['US', 'JP', 'CA'])
embedded_country = fc.embedding_column(country,
dimension=5,
initializer=_initializer)
# Provides 1-dim tensor and dense tensor.
features = {
'price':
constant_op.constant([
11.,
12.,
]),
'body-style':
sparse_tensor.SparseTensor(indices=((0, ), (1, )),
values=('sedan', 'hardtop'),
dense_shape=(2, )),
# This is dense tensor for the categorical_column.
'country':
constant_op.constant(['CA', 'US']),
}
self.assertEqual(1, features['price'].shape.ndims)
self.assertEqual(1, features['body-style'].dense_shape.get_shape()[0])
self.assertEqual(1, features['country'].shape.ndims)
net = df.DenseFeatures([price, one_hot_body_style,
embedded_country])(features)
self.assertEqual(1 + 3 + 5, net.shape[1])
with _initialized_session() as sess:
# Each row is formed by concatenating `embedded_body_style`,
# `one_hot_body_style`, and `price` in order.
self.assertAllEqual([[0., 0., 1., 11., 12., 13., 14., 15., 11.],
[1., 0., 0., 1., 2., 3., 4., 5., 12.]],
sess.run(net))
def test_fails_for_categorical_column(self):
animal = fc.categorical_column_with_identity('animal', num_buckets=4)
with ops.Graph().as_default():
features = {
'animal':
sparse_tensor.SparseTensor(indices=[[0, 0], [0, 1]],
values=[1, 2],
dense_shape=[1, 2])
}
with self.assertRaisesRegexp(Exception, 'must be a .*DenseColumn'):
df.DenseFeatures([animal])(features)
def test_multi_column(self):
price1 = fc.numeric_column('price1', shape=2)
price2 = fc.numeric_column('price2')
with ops.Graph().as_default():
features = {'price1': [[1., 2.], [5., 6.]], 'price2': [[3.], [4.]]}
net = df.DenseFeatures([price1, price2])(features)
self.evaluate(variables_lib.global_variables_initializer())
self.evaluate(lookup_ops.tables_initializer())
self.assertAllClose([[1., 2., 3.], [5., 6., 4.]],
self.evaluate(net))
def test_dense_feature_with_training_arg(self):
price1 = fc.numeric_column('price1', shape=2)
price2 = fc.numeric_column('price2')
# Monkey patch the second numeric column to simulate a column that has
# different behavior by mode.
def training_aware_get_dense_tensor(transformation_cache,
state_manager,
training=None):
return transformation_cache.get(price2,
state_manager,
training=training)
def training_aware_transform_feature(transformation_cache,
state_manager,
training=None):
input_tensor = transformation_cache.get(price2.key,
state_manager,
training=training)
if training:
return input_tensor * 10.0
else:
return input_tensor * 20.0
price2.get_dense_tensor = training_aware_get_dense_tensor
price2.transform_feature = training_aware_transform_feature
with ops.Graph().as_default():
features = {'price1': [[1., 2.], [5., 6.]], 'price2': [[3.], [4.]]}
train_mode = df.DenseFeatures([price1, price2])(features,
training=True)
predict_mode = df.DenseFeatures([price1, price2])(features,
training=False)
self.evaluate(variables_lib.global_variables_initializer())
self.evaluate(lookup_ops.tables_initializer())
self.assertAllClose([[1., 2., 30.], [5., 6., 40.]],
self.evaluate(train_mode))
self.assertAllClose([[1., 2., 60.], [5., 6., 80.]],
self.evaluate(predict_mode))
def test_static_batch_size_mismatch(self):
price1 = fc.numeric_column('price1')
price2 = fc.numeric_column('price2')
with ops.Graph().as_default():
features = {
'price1': [[1.], [5.], [7.]], # batchsize = 3
'price2': [[3.], [4.]] # batchsize = 2
}
with self.assertRaisesRegexp(
ValueError,
r'Batch size \(first dimension\) of each feature must be same.'
): # pylint: disable=anomalous-backslash-in-string
df.DenseFeatures([price1, price2])(features)
def test_with_rank_0_feature(self):
# price has 1 dimension in dense_features
price = fc.numeric_column('price')
features = {
'price': constant_op.constant(0),
}
self.assertEqual(0, features['price'].shape.ndims)
# Static rank 0 should fail
with self.assertRaisesRegexp(ValueError,
'Feature .* cannot have rank 0'):
df.DenseFeatures([price])(features)
# Dynamic rank 0 should fail
features = {
'price': array_ops.placeholder(dtypes.float32),
}
net = df.DenseFeatures([price])(features)
self.assertEqual(1, net.shape[1])
with _initialized_session() as sess:
with self.assertRaisesOpError('Feature .* cannot have rank 0'):
sess.run(net, feed_dict={features['price']: np.array(1)})
def test_multiple_layers_with_same_embedding_column(self):
some_sparse_column = fc.categorical_column_with_hash_bucket(
'sparse_feature', hash_bucket_size=5)
some_embedding_column = fc.embedding_column(some_sparse_column,
dimension=10)
with ops.Graph().as_default():
features = {
'sparse_feature': [['a'], ['x']],
}
all_cols = [some_embedding_column]
df.DenseFeatures(all_cols)(features)
df.DenseFeatures(all_cols)(features)
# Make sure that 2 variables get created in this case.
self.assertEqual(
2, len(ops.get_collection(ops.GraphKeys.GLOBAL_VARIABLES)))
expected_var_names = [
'dense_features/sparse_feature_embedding/embedding_weights:0',
'dense_features_1/sparse_feature_embedding/embedding_weights:0'
]
self.assertItemsEqual(expected_var_names, [
v.name
for v in ops.get_collection(ops.GraphKeys.GLOBAL_VARIABLES)
])
def test_runtime_batch_size_mismatch(self):
price1 = fc.numeric_column('price1')
price2 = fc.numeric_column('price2')
with ops.Graph().as_default():
features = {
'price1':
array_ops.placeholder(dtype=dtypes.int64), # batchsize = 3
'price2': [[3.], [4.]] # batchsize = 2
}
net = df.DenseFeatures([price1, price2])(features)
with _initialized_session() as sess:
with self.assertRaisesRegexp(
errors.OpError, 'Dimensions of inputs should match'):
sess.run(
net,
feed_dict={features['price1']: [[1.], [5.], [7.]]})
def test_feature_column_dense_features_gradient(self):
with context.eager_mode():
sparse_input = sparse_tensor.SparseTensor(indices=((0, 0), (1, 0),
(2, 0)),
values=(0, 1, 2),
dense_shape=(3, 3))
# Create feature columns (categorical and embedding).
categorical_column = fc.categorical_column_with_identity(
key='a', num_buckets=3)
embedding_dimension = 2
def _embedding_column_initializer(shape,
dtype,
partition_info=None):
del shape # unused
del dtype # unused
del partition_info # unused
embedding_values = (
(1, 0), # id 0
(0, 1), # id 1
(1, 1)) # id 2
return embedding_values
embedding_column = fc.embedding_column(
categorical_column,
dimension=embedding_dimension,
initializer=_embedding_column_initializer)
dense_features = df.DenseFeatures([embedding_column])
features = {'a': sparse_input}
def scale_matrix():
matrix = dense_features(features)
return 2 * matrix
# Sanity check: Verify that scale_matrix returns the correct output.
self.assertAllEqual([[2, 0], [0, 2], [2, 2]], scale_matrix())
# Check that the returned gradient is correct.
grad_function = backprop.implicit_grad(scale_matrix)
grads_and_vars = grad_function()
indexed_slice = grads_and_vars[0][0]
gradient = grads_and_vars[0][0].values
self.assertAllEqual([0, 1, 2], indexed_slice.indices)
self.assertAllEqual([[2, 2], [2, 2], [2, 2]], gradient)
def test_reuses_variables(self):
with context.eager_mode():
sparse_input = sparse_tensor.SparseTensor(indices=((0, 0), (1, 0),
(2, 0)),
values=(0, 1, 2),
dense_shape=(3, 3))
# Create feature columns (categorical and embedding).
categorical_column = fc.categorical_column_with_identity(
key='a', num_buckets=3)
embedding_dimension = 2
def _embedding_column_initializer(shape,
dtype,
partition_info=None):
del shape # unused
del dtype # unused
del partition_info # unused
embedding_values = (
(1, 0), # id 0
(0, 1), # id 1
(1, 1)) # id 2
return embedding_values
embedding_column = fc.embedding_column(
categorical_column,
dimension=embedding_dimension,
initializer=_embedding_column_initializer)
dense_features = df.DenseFeatures([embedding_column])
features = {'a': sparse_input}
inputs = dense_features(features)
variables = dense_features.variables
# Sanity check: test that the inputs are correct.
self.assertAllEqual([[1, 0], [0, 1], [1, 1]], inputs)
# Check that only one variable was created.
self.assertEqual(1, len(variables))
# Check that invoking dense_features on the same features does not create
# additional variables
_ = dense_features(features)
self.assertEqual(1, len(variables))
self.assertEqual(variables[0], dense_features.variables[0])
def test_runtime_batch_size_matches(self):
price1 = fc.numeric_column('price1')
price2 = fc.numeric_column('price2')
with ops.Graph().as_default():
features = {
'price1':
array_ops.placeholder(dtype=dtypes.int64), # batchsize = 2
'price2':
array_ops.placeholder(dtype=dtypes.int64), # batchsize = 2
}
net = df.DenseFeatures([price1, price2])(features)
with _initialized_session() as sess:
sess.run(net,
feed_dict={
features['price1']: [[1.], [5.]],
features['price2']: [[1.], [5.]],
})
def test_cols_to_output_tensors(self):
price1 = fc.numeric_column('price1', shape=2)
price2 = fc.numeric_column('price2')
with ops.Graph().as_default():
cols_dict = {}
features = {'price1': [[1., 2.], [5., 6.]], 'price2': [[3.], [4.]]}
dense_features = df.DenseFeatures([price1, price2])
net = dense_features(features, cols_dict)
self.evaluate(variables_lib.global_variables_initializer())
self.evaluate(lookup_ops.tables_initializer())
self.assertAllClose([[1., 2.], [5., 6.]],
self.evaluate(cols_dict[price1]))
self.assertAllClose([[3.], [4.]], self.evaluate(cols_dict[price2]))
self.assertAllClose([[1., 2., 3.], [5., 6., 4.]],
self.evaluate(net))
def DISABLED_test_train_with_dense_features_v2(self):
feature_dict = {
'sex': np.int64([1, 1, 1, 1, 0]),
'cp': np.int64([0, 3, 3, 2, 1]),
'slope': np.int64([3, 2, 0, 3, 1]),
}
label = np.int64([0, 1, 0, 0, 0])
train_input_fn = numpy_io.numpy_input_fn(x=feature_dict,
y=label,
num_epochs=1,
shuffle=False)
feature_columns = list()
input_features = dict()
for feature_name, data_array in feature_dict.items():
feature_columns.append(
tf.feature_column.embedding_column(
tf.feature_column.categorical_column_with_identity(
key=feature_name,
num_buckets=np.size(np.unique(data_array))),
dimension=3))
input_features[feature_name] = keras.layers.Input(
name=feature_name,
shape=(np.size(np.unique(data_array)), ),
dtype=tf.dtypes.int64)
df = dense_features_v2.DenseFeatures(feature_columns)
x = df(input_features)
x = keras.layers.Dense(16, activation='relu')(x)
logits = keras.layers.Dense(1, activation='linear')(x)
model = keras.Model(inputs=input_features, outputs=logits)
model.compile(optimizer='rmsprop',
loss='binary_crossentropy',
metrics=['accuracy'])
estimator_model = keras_lib.model_to_estimator(keras_model=model)
estimator_model.train(input_fn=train_input_fn, steps=5)
# We assert that we find the embedding_weights variables in the dependencies
# for the DenseFeatures layer.
dependency_names = [x.name for x in df._checkpoint_dependencies]
self.assertNotIn('embedding_weights', dependency_names)
self.assertIn('cp_embedding/embedding_weights', dependency_names)
self.assertIn('sex_embedding/embedding_weights', dependency_names)
self.assertIn('slope_embedding/embedding_weights', dependency_names)
def test_compute_output_shape(self):
price1 = fc.numeric_column('price1', shape=2)
price2 = fc.numeric_column('price2', shape=4)
with ops.Graph().as_default():
features = {
'price1': [[1., 2.], [5., 6.]],
'price2': [[3., 4., 5., 6.], [7., 8., 9., 10.]]
}
dense_features = df.DenseFeatures([price1, price2])
self.assertEqual((None, 6),
dense_features.compute_output_shape((None, )))
net = dense_features(features)
self.evaluate(variables_lib.global_variables_initializer())
self.evaluate(lookup_ops.tables_initializer())
self.assertAllClose(
[[1., 2., 3., 4., 5., 6.], [5., 6., 7., 8., 9., 10.]],
self.evaluate(net))
def test_linear_model_with_feature_column(self):
vocab_list = ['alpha', 'beta', 'gamma']
vocab_val = [0.4, 0.6, 0.9]
data = np.random.choice(vocab_list, size=256)
y = np.zeros_like(data, dtype=np.float32)
for vocab, val in zip(vocab_list, vocab_val):
indices = np.where(data == vocab)
y[indices] = val + np.random.uniform(
low=-0.01, high=0.01, size=indices[0].shape)
cat_column = feature_column_v2.categorical_column_with_vocabulary_list(
key='symbol', vocabulary_list=vocab_list)
ind_column = feature_column_v2.indicator_column(cat_column)
dense_feature_layer = dense_features_v2.DenseFeatures([ind_column])
linear_model = linear.LinearModel(
use_bias=False, kernel_initializer='zeros')
combined = keras.Sequential([dense_feature_layer, linear_model])
opt = gradient_descent.SGD(learning_rate=0.1)
combined.compile(opt, 'mse', [])
combined.fit(x={'symbol': data}, y=y, batch_size=32, epochs=10)
self.assertAllClose([[0.4], [0.6], [0.9]],
combined.layers[1].dense_layers[0].kernel.numpy(),
atol=0.01)
def test_wide_deep_model_with_single_feature_column(self):
vocab_list = ['alpha', 'beta', 'gamma']
vocab_val = [0.4, 0.6, 0.9]
data = np.random.choice(vocab_list, size=256)
y = np.zeros_like(data, dtype=np.float32)
for vocab, val in zip(vocab_list, vocab_val):
indices = np.where(data == vocab)
y[indices] = val + np.random.uniform(
low=-0.01, high=0.01, size=indices[0].shape)
cat_column = fc.categorical_column_with_vocabulary_list(
key='symbol', vocabulary_list=vocab_list)
ind_column = fc.indicator_column(cat_column)
dense_feature_layer = dense_features_v2.DenseFeatures([ind_column])
linear_model = linear.LinearModel(use_bias=False,
kernel_initializer='zeros')
dnn_model = sequential.Sequential([core.Dense(units=1)])
wide_deep_model = wide_deep.WideDeepModel(linear_model, dnn_model)
combined = sequential.Sequential(
[dense_feature_layer, wide_deep_model])
opt = gradient_descent.SGD(learning_rate=0.1)
combined.compile(opt,
'mse', [],
run_eagerly=testing_utils.should_run_eagerly())
combined.fit(x={'symbol': data}, y=y, batch_size=32, epochs=10)
def __init__(self,
units,
hidden_units,
feature_columns,
activation_fn,
dropout,
batch_norm,
name=None,
**kwargs):
super(_DNNModelV2, self).__init__(name=name, **kwargs)
with ops.name_scope(
'input_from_feature_columns') as input_feature_column_scope:
layer_name = input_feature_column_scope + 'input_layer'
if feature_column_lib.is_feature_column_v2(feature_columns):
self._input_layer = dense_features_v2.DenseFeatures(
feature_columns=feature_columns, name=layer_name)
else:
raise ValueError(
'Received a feature column from TensorFlow v1, but this is a '
'TensorFlow v2 Estimator. Please either use v2 feature columns '
'(accessible via tf.feature_column.* in TF 2.x) with this '
'Estimator, or switch to a v1 Estimator for use with v1 feature '
'columns (accessible via tf.compat.v1.estimator.* and '
'tf.compat.v1.feature_column.*, respectively.')
self._dropout = dropout
self._batch_norm = batch_norm
self._hidden_layers = []
self._dropout_layers = []
self._batch_norm_layers = []
self._hidden_layer_scope_names = []
for layer_id, num_hidden_units in enumerate(hidden_units):
with ops.name_scope('hiddenlayer_%d' % layer_id) as hidden_layer_scope:
# Get scope name without the trailing slash.
hidden_shared_name = _name_from_scope_name(hidden_layer_scope)
hidden_layer = keras_core.Dense(
units=num_hidden_units,
activation=activation_fn,
kernel_initializer=tf.compat.v1.glorot_uniform_initializer(),
name=hidden_shared_name)
self._hidden_layer_scope_names.append(hidden_shared_name)
self._hidden_layers.append(hidden_layer)
if self._dropout is not None:
dropout_layer = keras_core.Dropout(rate=self._dropout)
self._dropout_layers.append(dropout_layer)
if self._batch_norm:
batch_norm_name = hidden_shared_name + '/batchnorm_%d' % layer_id
# TODO(scottzhu): Change back to use BatchNormalization when the
# cleanup is done.
batch_norm_layer = keras_norm.BatchNormalizationBase(
# The default momentum 0.99 actually crashes on certain
# problem, so here we use 0.999, which is the default of
# tf.contrib.layers.batch_norm.
momentum=0.999,
trainable=True,
name=batch_norm_name)
self._batch_norm_layers.append(batch_norm_layer)
with ops.name_scope('logits') as logits_scope:
logits_shared_name = _name_from_scope_name(logits_scope)
self._logits_layer = keras_core.Dense(
units=units,
activation=None,
kernel_initializer=tf.compat.v1.glorot_uniform_initializer(),
name=logits_shared_name)
self._logits_scope_name = logits_shared_name
def test_retrieving_input(self):
features = {'a': [0.]}
dense_features = df.DenseFeatures(fc.numeric_column('a'))
inputs = self.evaluate(dense_features(features))
self.assertAllClose([[0.]], inputs)
def test_raises_if_empty_feature_columns(self):
with self.assertRaisesRegexp(ValueError,
'feature_columns must not be empty'):
df.DenseFeatures(feature_columns=[])(features={})
