[Py] Added elu operator to Python API. (#3236)

* Added elu operator to Python API.

* Added missing file.

* Specified elu function description.

* Expand docstring

* [Py] Added test with scalar for elu operator.

* Bugfix

*  [Py] Changed input type in elu test.

* Update test_ops_binary.py

* [Py] Syntax bugfix.

* [Py] Added elu operator to list in documentation.

doc/sphinx/source/python_api/_autosummary/ngraph.ops.rst

@@ -32,6 +32,7 @@ ngraph.ops
       cosh
       divide
       dot
+      elu
       equal
       exp
       floor

python/ngraph/__init__.py

@@ -45,6 +45,7 @@
 from ngraph.ops import cosh
 from ngraph.ops import divide
 from ngraph.ops import dot
+from ngraph.ops import elu
 from ngraph.ops import equal
 from ngraph.ops import exp
 from ngraph.ops import floor

python/ngraph/impl/op/__init__.py

@@ -69,6 +69,7 @@
 from _pyngraph.op import Cosh
 from _pyngraph.op import Divide
 from _pyngraph.op import Dot
+from _pyngraph.op import Elu
 from _pyngraph.op import Equal
 from _pyngraph.op import Exp
 from _pyngraph.op import Floor

python/ngraph/ops.py

@@ -22,7 +22,7 @@
 
 from ngraph.impl.op import Abs, Acos, Add, And, Asin, ArgMax, ArgMin, Atan, AvgPool, \
     BatchNormTraining, BatchNormInference, Broadcast, Ceiling, Concat, Constant, Convert, \
-    Convolution, ConvolutionBackpropData, Cos, Cosh, Divide, Dot, Equal, Exp, Floor, \
+    Convolution, ConvolutionBackpropData, Cos, Cosh, Divide, Dot, Elu, Equal, Exp, Floor, \
     GetOutputElement, Greater, GreaterEq, Less, LessEq, Log, LRN, Max, Maximum, MaxPool, \
     Min, Minimum, Multiply, Negative, Not, NotEqual, OneHot, Or, Pad, Parameter, Product, \
     Power, Relu, ReplaceSlice, Reshape, Reverse, Select, Sign, Sin, Sinh, Slice, Softmax, \
@@ -35,7 +35,7 @@
 from ngraph.utils.input_validation import assert_list_of_ints
 from ngraph.utils.reduction import get_reduction_axes
 from ngraph.utils.types import NumericType, NumericData, TensorShape, make_constant_node, \
-    NodeInput, ScalarData
+    NodeInput, ScalarData, as_node
 from ngraph.utils.types import get_element_type
 
 
@@ -60,6 +60,24 @@ def constant(value, dtype=None, name=None):  # type: (NumericData, NumericType,
     return make_constant_node(value, dtype)
 
 
+@nameable_op
+def elu(data, alpha, name=None):  # type: (NodeInput, NodeInput, str) -> Node
+    """Perform Exponential Linear Unit operation element-wise on data from input node.
+
+    Computes exponential linear: alpha * (exp(data) - 1) if < 0, data otherwise.
+
+    For more information refer to:
+    `Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
+    <http://arxiv.org/abs/1511.07289>`_
+
+    :param data: Input tensor. One of: input node, array or scalar.
+    :param alpha: Multiplier for negative values. One of: input node or scalar value.
+    :param name: Optional output node name.
+    :return: The new node performing an ELU operation on its input data element-wise.
+    """
+    return Elu(as_node(data), as_node(alpha))
+
+
 # Unary ops
 @unary_op
 def absolute(node, name=None):  # type: (NodeInput, str) -> Node

python/pyngraph/ops/elu.cpp

@@ -0,0 +1,30 @@
+//*****************************************************************************
+// Copyright 2017-2019 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//*****************************************************************************
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+
+#include "ngraph/op/fused/elu.hpp"
+#include "pyngraph/ops/elu.hpp"
+
+namespace py = pybind11;
+
+void regclass_pyngraph_op_Elu(py::module m)
+{
+    py::class_<ngraph::op::Elu, std::shared_ptr<ngraph::op::Elu>, ngraph::op::Op> elu(m, "Elu");
+    elu.doc() = "ngraph.impl.op.Elu wraps ngraph::op::Elu";
+    elu.def(py::init<const std::shared_ptr<ngraph::Node>&, const std::shared_ptr<ngraph::Node>&>());
+}

python/pyngraph/ops/elu.hpp

@@ -0,0 +1,23 @@
+//*****************************************************************************
+// Copyright 2017-2019 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//*****************************************************************************
+
+#pragma once
+
+#include <pybind11/pybind11.h>
+
+namespace py = pybind11;
+
+void regclass_pyngraph_op_Elu(py::module m);

python/pyngraph/ops/fused/elu.cpp

@@ -0,0 +1,30 @@
+//*****************************************************************************
+// Copyright 2017-2019 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//*****************************************************************************
+
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+
+#include "ngraph/op/fused/elu.hpp"
+#include "pyngraph/ops/fused/elu.hpp"
+
+namespace py = pybind11;
+
+void regclass_pyngraph_op_Elu(py::module m)
+{
+    py::class_<ngraph::op::Elu, std::shared_ptr<ngraph::op::Elu>, ngraph::op::Op> elu(m, "Elu");
+    elu.doc() = "ngraph.impl.op.Elu wraps ngraph::op::Elu";
+    elu.def(py::init<const std::shared_ptr<ngraph::Node>&, const std::shared_ptr<ngraph::Node>&>());
+}

python/pyngraph/ops/regmodule_pyngraph_op.cpp

@@ -49,6 +49,7 @@ void regmodule_pyngraph_op(py::module m_op)
     regclass_pyngraph_op_Cosh(m_op);
     regclass_pyngraph_op_Divide(m_op);
     regclass_pyngraph_op_Dot(m_op);
+    regclass_pyngraph_op_Elu(m_op);
     regclass_pyngraph_op_Equal(m_op);
     regclass_pyngraph_op_Exp(m_op);
     regclass_pyngraph_op_Floor(m_op);

python/pyngraph/ops/regmodule_pyngraph_op.hpp

@@ -39,6 +39,7 @@
 #include "pyngraph/ops/cosh.hpp"
 #include "pyngraph/ops/divide.hpp"
 #include "pyngraph/ops/dot.hpp"
+#include "pyngraph/ops/elu.hpp"
 #include "pyngraph/ops/equal.hpp"
 #include "pyngraph/ops/exp.hpp"
 #include "pyngraph/ops/floor.hpp"

python/setup.py

@@ -179,6 +179,7 @@ def cpp_flag(compiler):
     'pyngraph/ops/ceiling.cpp',
     'pyngraph/ops/divide.cpp',
     'pyngraph/ops/dot.cpp',
+    'pyngraph/ops/elu.cpp',
     'pyngraph/ops/equal.cpp',
     'pyngraph/ops/exp.cpp',
     'pyngraph/ops/floor.cpp',

python/test/ngraph/test_ops_fused.py

@@ -0,0 +1,69 @@
+# ******************************************************************************
+# Copyright 2017-2019 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ******************************************************************************
+import numpy as np
+
+import ngraph as ng
+from test.ngraph.util import get_runtime
+
+
+def test_elu_operator():
+    runtime = get_runtime()
+
+    data_shape = [2, 2]
+    alpha_shape = [2]
+    parameter_data = ng.parameter(data_shape, name='Data', dtype=np.float32)
+    parameter_alpha = ng.parameter(alpha_shape, name='Alpha', dtype=np.float32)
+
+    model = ng.elu(parameter_data, parameter_alpha)
+    computation = runtime.computation(model, parameter_data, parameter_alpha)
+
+    value_data = np.array([[-5, 1], [-2, 3]], dtype=np.float32)
+    value_alpha = np.array([3, 3], dtype=np.float32)
+
+    result = computation(value_data, value_alpha)
+    expected = np.array([[-2.9797862, 1.], [-2.5939941, 3.]], dtype=np.float32)
+    assert np.allclose(result, expected)
+
+
+def test_elu_operator_with_scalar_and_array():
+    runtime = get_runtime()
+
+    data_value = np.array([[-5, 1], [-2, 3]], dtype=np.float32)
+    alpha_value = np.float32(3)
+
+    model = ng.elu(data_value, alpha_value)
+    computation = runtime.computation(model)
+
+    result = computation()
+    expected = np.array([[-2.9797862, 1.], [-2.5939941, 3.]], dtype=np.float32)
+    assert np.allclose(result, expected)
+
+
+def test_elu_operator_with_scalar():
+    runtime = get_runtime()
+
+    data_value = np.array([[-5, 1], [-2, 3]], dtype=np.float32)
+    alpha_value = np.float32(3)
+
+    data_shape = [2, 2]
+    parameter_data = ng.parameter(data_shape, name='Data', dtype=np.float32)
+
+    model = ng.elu(parameter_data, alpha_value)
+    computation = runtime.computation(model, parameter_data)
+
+    result = computation(data_value)
+    expected = np.array([[-2.9797862, 1.], [-2.5939941, 3.]], dtype=np.float32)
+    assert np.allclose(result, expected)

src/ngraph/op/fused/elu.cpp

@@ -39,7 +39,7 @@ NodeVector op::Elu::decompose_op() const
     auto data = get_argument(0);
     auto alpha_node = get_argument(1);
 
-    alpha_node = ngraph::op::make_broadcast_node(alpha_node, data->get_shape());
+    alpha_node = ngraph::op::numpy_style_broadcast(alpha_node, data->get_shape());
 
     shared_ptr<ngraph::Node> zero_node =
         builder::make_constant(data->get_element_type(), data->get_shape(), 0);