Add sequence slice operator

paddle/operators/sub_sequence_op.cc

@@ -0,0 +1,99 @@
+/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
+
+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/operators/sub_sequence_op.h"
+
+namespace paddle {
+namespace operators {
+
+class SubSequenceOp : public framework::OperatorWithKernel {
+ public:
+  using framework::OperatorWithKernel::OperatorWithKernel;
+
+  void InferShape(framework::InferShapeContext* ctx) const override {
+    PADDLE_ENFORCE(ctx->HasInput("X"),
+                   "Input(X) of SubSequenceOp should not be null.");
+    PADDLE_ENFORCE(ctx->HasOutput("Out"),
+                   "Output(Out) of SubSequenceOp should not be null.");
+    auto input_dims = ctx->GetInputDim("X");
+
+    auto offsets = ctx->Attrs().Get<std::vector<int>>("offset");
+    auto sizes = ctx->Attrs().Get<std::vector<int>>("size");
+
+    auto dim_0 = 0;
+    for (size_t i = 0; i < sizes.size(); ++i) {
+      dim_0 += sizes[i];
+    }
+
+    framework::DDim out_dims = input_dims;
+    out_dims[0] = dim_0;
+    ctx->SetOutputDim("Out", out_dims);
+  }
+};
+
+class SubSequenceGradOp : public framework::OperatorWithKernel {
+ public:
+  using framework::OperatorWithKernel::OperatorWithKernel;
+
+  void InferShape(framework::InferShapeContext* ctx) const override {
+    PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("Out")),
+                   "The gradient of Out should not be null.");
+    PADDLE_ENFORCE(ctx->HasOutputs(framework::GradVarName("X")),
+                   "The gradient of X should not be null.");
+    ctx->SetOutputsDim(framework::GradVarName("X"), ctx->GetInputsDim("X"));
+  }
+};
+
+class SubSequenceOpMaker : public framework::OpProtoAndCheckerMaker {
+ public:
+  SubSequenceOpMaker(framework::OpProto* proto,
+                        framework::OpAttrChecker* op_checker)
+      : OpProtoAndCheckerMaker(proto, op_checker) {
+    AddInput("X", "(LoDTensor), "
+            "the variable-length input of SubSequenceOp");
+    AddAttr<std::vector<int>>(
+        "offset",
+        "A list<int> to describes offset for sub sequence item.");
+    AddAttr<std::vector<int>>(
+        "size",
+        "A list<int> to describes size for sub sequence item.");
+    AddOutput("Out",
+              "(Tensor), Variable-length output of "
+              "sequence_concat Op.");
+    AddComment(R"DOC(
+Sub Sequence operator
+
+The operator crop a subsequence from given sequence with given start offset and subsequence size.
+It only supports sequence (LoD Tensor with level number is 1).
+- Case:
+    LoD(x) = {{0, 3, 6, 10}}; Dims(x0) = (10, 3, 2)
+    offset = (0, 1, 1); size = (2, 1, 2)
+    LoD(Out) = {{0, 2, 3, 5}}; Dims(Out) = (5,3,2)
+NOTE: The length of the input, offset and size should be the same. The offset start from 0.
+    )DOC");
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+REGISTER_OP(sub_sequence, ops::SubSequenceOp, ops::SubSequenceOpMaker,
+            sub_sequence_grad, ops::SubSequenceGradOp);
+REGISTER_OP_CPU_KERNEL(
+    sub_sequence,
+    ops::SubSequenceOpKernel<paddle::platform::CPUPlace, float>);
+REGISTER_OP_CPU_KERNEL(
+    sub_sequence_grad,
+    ops::SubSequenceGradOpKernel<paddle::platform::CPUPlace, float>);

paddle/operators/sub_sequence_op.cu

@@ -0,0 +1,25 @@
+/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
+
+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_GPU
+
+#include "paddle/operators/sub_sequence_op.h"
+
+namespace ops = paddle::operators;
+REGISTER_OP_GPU_KERNEL(
+    sub_sequence,
+    ops::SubSequenceOpKernel<paddle::platform::GPUPlace, float>);
+REGISTER_OP_GPU_KERNEL(
+    sub_sequence_grad,
+    ops::SubSequenceGradOpKernel<paddle::platform::GPUPlace, float>);

paddle/operators/sub_sequence_op.h

@@ -0,0 +1,156 @@
+/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
+
+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 "paddle/framework/eigen.h"
+#include "paddle/framework/op_registry.h"
+#include "paddle/operators/strided_memcpy.h"
+
+namespace paddle {
+namespace operators {
+
+using Tensor = framework::Tensor;
+using LoDTensor = framework::LoDTensor;
+using LoD = framework::LoD;
+
+template <typename T>
+LoD subsequenceLoD(const T* in, const std::vector<int> offsets,
+                   const std::vector<int> sizes) {
+  auto out_lod = in->lod();
+  size_t lod_offset = 0;
+
+  auto n = in->lod()[0].size() - 1;
+  out_lod[0][0] = 0;
+  for (size_t i = 0; i < n; ++i) {
+    lod_offset += sizes[i];
+    out_lod[0][i+1] = lod_offset;
+  }
+  return out_lod;
+}
+
+template <typename Place, typename T>
+class SubSequenceOpKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* in = ctx.Input<LoDTensor>("X");
+    std::vector<int> offsets = ctx.Attr<std::vector<int>>("offset");
+    std::vector<int> sizes = ctx.Attr<std::vector<int>>("size");
+    auto* out = ctx.Output<LoDTensor>("Out");
+
+    auto offset_len = offsets.size();
+    auto size_len = sizes.size();
+
+    auto lod = in->lod();
+    auto n = lod[0].size() - 1;
+
+    PADDLE_ENFORCE_EQ(lod.size(), 1UL, "Only support one level sequence now.");
+    PADDLE_ENFORCE_EQ(n, offset_len,
+                      "The length of input and offset should be the same")
+    PADDLE_ENFORCE_EQ(n, size_len,
+                      "The length of input and size should be the same")
+
+    for (size_t i = 0; i < n; ++i) {
+      auto offset = offsets[i];
+      auto size = sizes[i];
+      PADDLE_ENFORCE_LT(lod[0][i] + offset + size, lod[0][i + 1],
+                        "The target tensor's length overflow")
+    }
+
+    out->mutable_data<T>(ctx.GetPlace());
+    auto out_lod = subsequenceLoD(in, offsets, sizes);
+    out->set_lod(out_lod);
+
+    auto in_stride = framework::stride(in->dims());
+    auto out_stride = framework::stride(out->dims());
+
+    size_t out_offset = 0;
+    for (size_t i = 0; i < n; ++i) {
+      auto offset = offsets[i];
+      auto size = sizes[i];
+
+      Tensor in_t = in->Slice(static_cast<int>(lod[0][i] + offset),
+                               static_cast<int>(lod[0][i] + offset + size));
+
+      StridedMemcpy<T>(ctx.device_context(), in_t.data<T>(),
+                       in_stride, in_t.dims(), out_stride,
+                       out->data<T>() + out_offset);
+      out_offset += size * in_stride[0];
+    }
+  }
+};
+
+template <typename Place, typename T>
+class SubSequenceGradOpKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* in = ctx.Input<LoDTensor>("X");
+    std::vector<int> offsets = ctx.Attr<std::vector<int>>("offset");
+    std::vector<int> sizes = ctx.Attr<std::vector<int>>("size");
+    auto* out_grad =
+        ctx.Input<framework::LoDTensor>(framework::GradVarName("Out"));
+    auto* x_grad =
+        ctx.Output<framework::LoDTensor>(framework::GradVarName("X"));
+
+    auto offset_len = offsets.size();
+    auto size_len = sizes.size();
+
+    auto lod = in->lod();
+    auto n = lod[0].size() - 1;
+
+    // check input data format
+    PADDLE_ENFORCE_EQ(lod.size(), 1UL, "Only support one level sequence now.");
+    PADDLE_ENFORCE_EQ(n, offset_len,
+                      "The length of input and offset should be the same")
+    PADDLE_ENFORCE_EQ(n, size_len,
+                      "The length of input and size should be the same")
+
+    for (size_t i = 0; i < n; ++i) {
+      auto offset = offsets[i];
+      auto size = sizes[i];
+      PADDLE_ENFORCE_LT(lod[0][i] + offset + size, lod[0][i + 1],
+                        "The target tensor's length overflow")
+    }
+
+    auto out_lod = subsequenceLoD(in, offsets, sizes);
+
+    x_grad->set_lod(lod);
+    x_grad->mutable_data<T>(ctx.GetPlace());
+    auto temp = framework::EigenVector<T>::Flatten(*x_grad);
+    temp.device(ctx.GetEigenDevice<Place>()) = temp.constant(static_cast<T>(0));
+
+    auto out_grad_stride = framework::stride(out_grad->dims());
+
+    for (size_t i = 0; i < out_lod[0].size() - 1; ++i) {
+      Tensor out_grad_t =
+          out_grad->Slice(static_cast<int>(out_lod[0][i]),
+                          static_cast<int>(out_lod[0][i + 1]));
+      auto out_grad_stride = framework::stride(out_grad_t.dims());
+
+      auto x_grad_stride = framework::stride(x_grad->dims());
+
+      auto offset = offsets[i];
+      auto size = sizes[i];
+
+      Tensor x_grad_t = x_grad->Slice(static_cast<int>(lod[0][i] + offset),
+                         static_cast<int>(lod[0][i] + offset + size));
+
+      StridedMemcpy<T>(ctx.device_context(), out_grad_t.data<T>(),
+                       out_grad_stride, out_grad_t.dims(), x_grad_stride,
+                       x_grad_t.data<T>());
+    }
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle

python/paddle/v2/framework/tests/test_sub_sequence_op.py

@@ -0,0 +1,40 @@
+import unittest
+import numpy as np
+import sys
+from op_test import OpTest
+
+class TestSubSequenceOp(OpTest):
+    def set_data(self):
+        # only supprot one level LoD
+        x = np.random.random((100, 3, 2)).astype('float32')
+        lod = [[0, 20, 40, 60, 80, 100]]
+        offsets = np.array([1, 2, 3, 4, 5]).flatten()
+        sizes = np.array([10, 8, 6, 4, 2]).flatten()
+
+        self.inputs = {'X': (x, lod)}
+        self.attrs = {'offset': offsets, 'size': sizes}
+        outs = []
+        out_lod = [[0]]
+        out_lod_offset = 0
+        for i in range(len(offsets)):
+            sub_x = x[lod[0][i] + offsets[i]: lod[0]
+                      [i] + offsets[i] + sizes[i], :]
+            outs.append(sub_x)
+            out_lod_offset = out_lod_offset + len(sub_x)
+            out_lod[0].append(out_lod_offset)
+
+        outs = np.concatenate(outs, axis=0)
+        self.outputs = {'Out': outs}
+
+    def setUp(self):
+        self.op_type = "sub_sequence"
+        self.set_data()
+
+    def test_check_output(self):
+        self.check_output()
+
+    def test_check_grad(self):
+        self.check_grad(['X'], 'Out')
+
+if __name__ == '__main__':
+    unittest.main()