add hardshrink kernel

tensorflow_addons/activations/BUILD

@@ -7,6 +7,7 @@ py_library(
     srcs = [
         "__init__.py",
         "gelu.py",
+        "hardshrink.py",
         "sparsemax.py",
     ],
     data = [
@@ -31,7 +32,7 @@ py_test(
 
 py_test(
     name = "gelu_test",
-    size = "large",
+    size = "medium",
     srcs = [
         "gelu_test.py",
     ],
@@ -41,3 +42,16 @@ py_test(
         ":activations",
     ],
 )
+
+py_test(
+    name = "hardshrink_test",
+    size = "medium",
+    srcs = [
+        "hardshrink_test.py",
+    ],
+    main = "hardshrink_test.py",
+    srcs_version = "PY2AND3",
+    deps = [
+        ":activations",
+    ],
+)

tensorflow_addons/activations/README.md

@@ -4,12 +4,14 @@
 | Submodule | Maintainers               | Contact Info                             |
 |:----------|:--------------------------|:-----------------------------------------|
 | gelu      | @AakashKumarNain @WindQAQ | [email protected] [email protected] |
+| hardshrink| @WindQAQ                  | [email protected]
 | sparsemax | @AndreasMadsen            | [email protected]                |
 
 ## Contents
 | Submodule | Activation | Reference                        |
 |:----------|:-----------|:---------------------------------|
 | gelu      | gelu       | https://arxiv.org/abs/1606.08415 |
+| hardshrink| hardshrink |                                  |
 | sparsemax | Sparsemax  | https://arxiv.org/abs/1602.02068 |
 
 

tensorflow_addons/activations/__init__.py

@@ -19,4 +19,5 @@
 from __future__ import print_function
 
 from tensorflow_addons.activations.gelu import gelu
+from tensorflow_addons.activations.hardshrink import hardshrink
 from tensorflow_addons.activations.sparsemax import sparsemax

tensorflow_addons/activations/hardshrink.py

@@ -0,0 +1,52 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# 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.
+# ==============================================================================
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import tensorflow as tf
+from tensorflow_addons.utils import keras_utils
+from tensorflow_addons.utils.resource_loader import get_path_to_datafile
+
+_activation_ops_so = tf.load_op_library(
+    get_path_to_datafile("custom_ops/activations/_activation_ops.so"))
+
+
+@keras_utils.register_keras_custom_object
+@tf.function
+def hardshrink(x, lower=-1.0, upper=1.0):
+    """Hard shrink function.
+
+    Computes hard shrink function:
+    `x if x < lower or x > upper else 0`.
+
+    Args:
+        x: A `Tensor`. Must be one of the following types:
+            `float16`, `float32`, `float64`.
+        lower: `float`, lower bound for setting values to zeros.
+        upper: `float`, upper bound for setting values to zeros.
+    Returns:
+        A `Tensor`. Has the same type as `x`.
+    """
+    x = tf.convert_to_tensor(x)
+    return _activation_ops_so.hardshrink(x, lower, upper)
+
+
+@tf.RegisterGradient("Hardshrink")
+def _hardshrink_grad(op, grad):
+    return _activation_ops_so.hardshrink_grad(grad, op.inputs[0],
+                                              op.get_attr("lower"),
+                                              op.get_attr("upper"))

tensorflow_addons/activations/hardshrink_test.py

@@ -0,0 +1,99 @@
+# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+#
+# 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.
+# ==============================================================================
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from absl.testing import parameterized
+
+import numpy as np
+import tensorflow as tf
+from tensorflow_addons.activations import hardshrink
+from tensorflow_addons.utils import test_utils
+
+
+def _ref_hardshrink(x, lower=-1.0, upper=1.0):
+    x = tf.convert_to_tensor(x)
+    return tf.where(tf.math.logical_or(x < lower, x > upper), x, 0.0)
+
+
+@test_utils.run_all_in_graph_and_eager_modes
+class HardshrinkTest(tf.test.TestCase, parameterized.TestCase):
+    def test_invalid(self):
+        with self.assertRaisesOpError(
+                "lower must be less than or equal to upper."):  # pylint: disable=bad-continuation
+            y = hardshrink(tf.ones(shape=(1, 2, 3)), lower=2.0, upper=-2.0)
+            self.evaluate(y)
+
+    @parameterized.named_parameters(("float16", np.float16),
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_hardshrink(self, dtype):
+        x = (np.random.rand(2, 3, 4) * 2.0 - 1.0).astype(dtype)
+        self.assertAllCloseAccordingToType(hardshrink(x), _ref_hardshrink(x))
+        self.assertAllCloseAccordingToType(
+            hardshrink(x, -2.0, 2.0), _ref_hardshrink(x, -2.0, 2.0))
+
+    @parameterized.named_parameters(("float16", np.float16),
+                                    ("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_gradients(self, dtype):
+        x = tf.constant([-1.5, -0.5, 0.5, 1.5], dtype=dtype)
+
+        with tf.GradientTape(persistent=True) as tape:
+            tape.watch(x)
+            y_ref = _ref_hardshrink(x)
+            y = hardshrink(x)
+        grad_ref = tape.gradient(y_ref, x)
+        grad = tape.gradient(y, x)
+        self.assertAllCloseAccordingToType(grad, grad_ref)
+
+    @parameterized.named_parameters(("float32", np.float32),
+                                    ("float64", np.float64))
+    def test_theoretical_gradients(self, dtype):
+        # Only test theoretical gradients for float32 and float64
+        # because of the instability of float16 while computing jacobian
+        x = tf.constant([-1.5, -0.5, 0.5, 1.5], dtype=dtype)
+
+        theoretical, numerical = tf.test.compute_gradient(
+            lambda x: hardshrink(x), [x])
+        self.assertAllCloseAccordingToType(theoretical, numerical, atol=1e-4)
+
+    def test_unknown_shape(self):
+        fn = hardshrink.get_concrete_function(
+            tf.TensorSpec(shape=None, dtype=tf.float32))
+
+        for shape in [(1,), (1, 2), (1, 2, 3), (1, 2, 3, 4)]:
+            x = tf.ones(shape=shape, dtype=tf.float32)
+            self.assertAllClose(fn(x), hardshrink(x))
+
+    def test_serialization(self):
+        ref_fn = hardshrink
+        config = tf.keras.activations.serialize(ref_fn)
+        fn = tf.keras.activations.deserialize(config)
+        self.assertEqual(fn, ref_fn)
+
+    def test_serialization_with_layers(self):
+        layer = tf.keras.layers.Dense(3, activation=hardshrink)
+        config = tf.keras.layers.serialize(layer)
+        deserialized_layer = tf.keras.layers.deserialize(config)
+        self.assertEqual(deserialized_layer.__class__.__name__,
+                         layer.__class__.__name__)
+        self.assertEqual(deserialized_layer.activation.__name__, "hardshrink")
+
+
+if __name__ == "__main__":
+    tf.test.main()

tensorflow_addons/custom_ops/activations/BUILD

@@ -27,12 +27,37 @@ cc_library(
     alwayslink = 1,
 )
 
+cc_library(
+    name = "hardshrink_op_gpu",
+    srcs = [
+        "cc/kernels/hardshrink_op.h",
+        "cc/kernels/hardshrink_op_gpu.cu.cc",
+    ],
+    copts = if_cuda_is_configured([
+        "-DGOOGLE_CUDA=1",
+        "-x cuda",
+        "-nvcc_options=relaxed-constexpr",
+        "-nvcc_options=ftz=true",
+    ]),
+    deps = [
+        "@local_config_tf//:libtensorflow_framework",
+        "@local_config_tf//:tf_header_lib",
+    ] + if_cuda_is_configured([
+        "@local_config_cuda//cuda:cuda_libs",
+        "@local_config_cuda//cuda:cuda_headers",
+    ]),
+    alwayslink = 1,
+)
+
 cc_binary(
     name = "_activation_ops.so",
     srcs = [
         "cc/kernels/gelu_op.cc",
         "cc/kernels/gelu_op.h",
+        "cc/kernels/hardshrink_op.cc",
+        "cc/kernels/hardshrink_op.h",
         "cc/ops/gelu_op.cc",
+        "cc/ops/hardshrink_op.cc",
     ],
     copts = [
         "-pthread",
@@ -43,5 +68,8 @@ cc_binary(
     deps = [
         "@local_config_tf//:libtensorflow_framework",
         "@local_config_tf//:tf_header_lib",
-    ] + if_cuda_is_configured([":gelu_op_gpu"]),
+    ] + if_cuda_is_configured([
+        ":gelu_op_gpu",
+        ":hardshrink_op_gpu",
+    ]),
 )

tensorflow_addons/custom_ops/activations/cc/kernels/hardshrink_op.cc

@@ -0,0 +1,77 @@
+/* Copyright 2019 The TensorFlow Authors. All Rights Reserved.
+
+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.
+==============================================================================*/
+
+#define EIGEN_USE_THREADS
+
+#include "tensorflow_addons/custom_ops/activations/cc/kernels/hardshrink_op.h"
+#include "tensorflow/core/framework/op_kernel.h"
+#include "tensorflow/core/framework/register_types.h"
+#include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
+
+namespace tensorflow {
+
+using CPUDevice = Eigen::ThreadPoolDevice;
+
+#define REGISTER_HARDSHRINK_KERNELS(type)                                  \
+  REGISTER_KERNEL_BUILDER(                                                 \
+      Name("Hardshrink").Device(DEVICE_CPU).TypeConstraint<type>("T"),     \
+      HardshrinkOp<CPUDevice, type>);                                      \
+  REGISTER_KERNEL_BUILDER(                                                 \
+      Name("HardshrinkGrad").Device(DEVICE_CPU).TypeConstraint<type>("T"), \
+      HardshrinkGradOp<CPUDevice, type>);
+
+// Hardshrink only makes sense with floating points.
+TF_CALL_GPU_NUMBER_TYPES(REGISTER_HARDSHRINK_KERNELS);
+#undef REGISTER_HARDSHRINK_KERNELS
+
+#if GOOGLE_CUDA
+
+using GPUDevice = Eigen::GpuDevice;
+
+// Forward declarations of the functor specializations for GPU.
+namespace functor {
+#define DECLARE_GPU_SPEC(T)                                                  \
+  template <>                                                                \
+  void Hardshrink<GPUDevice, T>::operator()(                                 \
+      const GPUDevice& d, typename TTypes<T>::ConstTensor features, T lower, \
+      T upper, typename TTypes<T>::Tensor activations);                      \
+  extern template struct Hardshrink<GPUDevice, T>;                           \
+                                                                             \
+  template <>                                                                \
+  void HardshrinkGrad<GPUDevice, T>::operator()(                             \
+      const GPUDevice& d, typename TTypes<T>::ConstTensor gradients,         \
+      typename TTypes<T>::ConstTensor features, T lower, T upper,            \
+      typename TTypes<T>::Tensor backprops);                                 \
+  extern template struct HardshrinkGrad<GPUDevice, T>;
+
+TF_CALL_GPU_NUMBER_TYPES(DECLARE_GPU_SPEC);
+#undef DECLARE_GPU_SPEC
+}  // namespace functor
+
+// Registration of the GPU implementations.
+#define REGISTER_HARDSHRINK_GPU_KERNELS(type)                              \
+  REGISTER_KERNEL_BUILDER(                                                 \
+      Name("Hardshrink").Device(DEVICE_GPU).TypeConstraint<type>("T"),     \
+      HardshrinkOp<GPUDevice, type>);                                      \
+  REGISTER_KERNEL_BUILDER(                                                 \
+      Name("HardshrinkGrad").Device(DEVICE_GPU).TypeConstraint<type>("T"), \
+      HardshrinkGradOp<GPUDevice, type>);
+
+TF_CALL_GPU_NUMBER_TYPES(REGISTER_HARDSHRINK_GPU_KERNELS);
+#undef REGISTER_HARDSHRINK_GPU_KERNELS
+
+#endif  // GOOGLE_CUDA
+
+}  // namespace tensorflow