diff --git a/paddle/fluid/operators/reverse_op.cc b/paddle/fluid/operators/reverse_op.cc
new file mode 100644
index 0000000000000000000000000000000000000000..58ca8da7ac0696546929465dd2a703c796db5a88
--- /dev/null
+++ b/paddle/fluid/operators/reverse_op.cc
@@ -0,0 +1,100 @@
+// Copyright (c) 2018 PaddlePaddle 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.
+
+#include "paddle/fluid/operators/reverse_op.h"
+#include <vector>
+
+namespace paddle {
+namespace operators {
+
+class ReverseOp : public framework::OperatorWithKernel {
+ public:
+  using framework::OperatorWithKernel::OperatorWithKernel;
+
+  void InferShape(framework::InferShapeContext* ctx) const override {
+    PADDLE_ENFORCE(ctx->HasInput("X"), "Input(X) should not be null");
+    PADDLE_ENFORCE(ctx->HasOutput("Out"), "Output(Out) should not be null");
+    const auto& x_dims = ctx->GetInputDim("X");
+    const auto& axis = ctx->Attrs().Get<std::vector<int>>("axis");
+    PADDLE_ENFORCE(!axis.empty(), "'axis' can not be empty.");
+    for (int a : axis) {
+      PADDLE_ENFORCE_LT(a, x_dims.size(),
+                        "The axis must be less than input tensor's rank.");
+    }
+    ctx->SetOutputDim("Out", x_dims);
+  }
+};
+
+class ReverseOpMaker : public framework::OpProtoAndCheckerMaker {
+ public:
+  void Make() override {
+    AddInput("X", "The LoDTensor to be flipped.");
+    AddOutput("Out", "The LoDTensor after flipping.");
+    AddAttr<std::vector<int>>(
+        "axis", "The axises that along which order of elements is reversed.");
+    AddComment(R"DOC(
+      Reverse Operator.
+
+      Reverse the order of elements in the input LoDTensor along given axises.
+
+      Case 1:
+        Given
+            X = [[1, 2, 3, 4, 5]
+                 [6, 7, 8, 9, 10]
+                 [11, 12, 13, 14, 15]],
+        and
+            axis = [0],
+        we get:
+            Out = [[11, 12, 13, 14, 15]
+                   [6, 7, 8, 9, 10]
+                   [1, 2, 3, 4, 5]].
+        
+      Case 2:
+        Given
+            X = [[[1, 2, 3, 4]
+                  [5, 6, 7, 8]]
+                 [[9, 10, 11, 12]
+                  [13, 14, 15, 16]]],
+        and
+            axis = [0, 2],
+        we get:
+            Out = [[[12, 11, 10, 9]
+                    [16, 15, 14, 13]]
+                   [[4, 3, 2, 1]
+                    [8, 7, 6, 5]]],
+    )DOC");
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+REGISTER_OPERATOR(reverse, ops::ReverseOp, ops::ReverseOpMaker,
+                  paddle::framework::DefaultGradOpDescMaker<true>);
+REGISTER_OPERATOR(reverse_grad, ops::ReverseOp);
+REGISTER_OP_CPU_KERNEL(
+    reverse, ops::ReverseKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, int8_t>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, int64_t>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, bool>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, float>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, double>)
+REGISTER_OP_CPU_KERNEL(
+    reverse_grad, ops::ReverseKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, int8_t>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, int64_t>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, bool>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, float>,
+    ops::ReverseKernel<paddle::platform::CPUDeviceContext, double>)
diff --git a/paddle/fluid/operators/reverse_op.cu b/paddle/fluid/operators/reverse_op.cu
new file mode 100644
index 0000000000000000000000000000000000000000..820be6b580803e5a974d68027de0b4072fdfac4b
--- /dev/null
+++ b/paddle/fluid/operators/reverse_op.cu
@@ -0,0 +1,31 @@
+// Copyright (c) 2018 PaddlePaddle 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.
+
+#include "paddle/fluid/operators/reverse_op.h"
+
+namespace ops = paddle::operators;
+REGISTER_OP_CUDA_KERNEL(
+    reverse, ops::ReverseKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, int8_t>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, bool>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, double>)
+REGISTER_OP_CUDA_KERNEL(
+    reverse_grad, ops::ReverseKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, int8_t>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, int64_t>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, bool>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::ReverseKernel<paddle::platform::CUDADeviceContext, double>)
diff --git a/paddle/fluid/operators/reverse_op.h b/paddle/fluid/operators/reverse_op.h
new file mode 100644
index 0000000000000000000000000000000000000000..2af4268dfacc333b124d812f9d666b6cde2416e8
--- /dev/null
+++ b/paddle/fluid/operators/reverse_op.h
@@ -0,0 +1,85 @@
+// Copyright (c) 2018 PaddlePaddle 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.
+
+#pragma once
+#include <vector>
+#include "paddle/fluid/framework/eigen.h"
+#include "paddle/fluid/framework/op_registry.h"
+
+namespace paddle {
+namespace operators {
+template <typename DeviceContext, typename T, int Rank>
+struct ReverseFunctor {
+  void operator()(const DeviceContext& context, const framework::LoDTensor& in,
+                  framework::LoDTensor* out, const std::vector<int>& axis) {
+    Eigen::array<bool, Rank> reverse_axis;
+    reverse_axis.fill(false);
+    for (int a : axis) {
+      reverse_axis[a] = true;
+    }
+
+    auto in_eigen = framework::EigenTensor<T, Rank>::From(in);
+    auto out_eigen = framework::EigenTensor<T, Rank>::From(*out);
+    auto* dev = context.eigen_device();
+
+    out_eigen.device(*dev) = in_eigen.reverse(reverse_axis);
+  }
+};
+
+template <typename DeviceContext, typename T>
+class ReverseKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& context) const override {
+    auto* x = context.Input<framework::LoDTensor>("X");
+    auto* out = context.Output<framework::LoDTensor>("Out");
+    out->mutable_data<T>(context.GetPlace());
+    const auto& axis = context.Attr<std::vector<int>>("axis");
+    int rank = x->dims().size();
+    auto& dev_ctx = context.template device_context<DeviceContext>();
+
+    switch (rank) {
+      case 1:
+        ReverseFunctor<DeviceContext, T, 1> functor1;
+        functor1(dev_ctx, *x, out, axis);
+        break;
+      case 2:
+        ReverseFunctor<DeviceContext, T, 2> functor2;
+        functor2(dev_ctx, *x, out, axis);
+        break;
+      case 3:
+        ReverseFunctor<DeviceContext, T, 3> functor3;
+        functor3(dev_ctx, *x, out, axis);
+        break;
+      case 4:
+        ReverseFunctor<DeviceContext, T, 4> functor4;
+        functor4(dev_ctx, *x, out, axis);
+        break;
+      case 5:
+        ReverseFunctor<DeviceContext, T, 5> functor5;
+        functor5(dev_ctx, *x, out, axis);
+        break;
+      case 6:
+        ReverseFunctor<DeviceContext, T, 6> functor6;
+        functor6(dev_ctx, *x, out, axis);
+        break;
+      default:
+        PADDLE_THROW(
+            "Reserve operator doesn't supports tensors whose ranks are greater "
+            "than 6.");
+    }
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle