Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into develop

doc/howto/cross_compiling/cross_compiling_for_ios_cn.md

+# 构建iOS平台上的PaddlePaddle库
+交叉编译iOS平台上适用的PaddlePaddle库，需要在MacOS系统上进行。本文的将介绍在MacOS上，从源码交叉编译iOS平台上适用的PaddlePaddle库。
+
+## 准备交叉编译环境
+Apple官方为iOS开发提供了完整的交叉编译工具和集成开发环境，用户从App Store下载安装Xcode即可。也可自行前往官网下载，[Xcode](https://developer.apple.com/cn/xcode/)。安装完成之后，可在命令行执行`xcodebuild -version`，判断是否安装成功。
+
+```bash
+$ xcodebuild -version
+Xcode 9.0
+Build version 9A235
+```
+
+## 配置交叉编译参数
+
+PaddlePaddle为交叉编译提供了工具链配置文档[cmake/cross_compiling/ios.cmake](https://github.com/PaddlePaddle/Paddle/blob/develop/cmake/cross_compiling/ios.cmake)，以提供一些默认的编译器和编译参数配置。
+
+交叉编译iOS版本的PaddlePaddle库时，有一些必须配置的参数：
+
+- `CMAKE_SYSTEM_NAME`，CMake编译的目标平台，必须设置为`iOS`。在设置`CMAKE_SYSTEM_NAME=iOS`后，PaddlePaddle的CMake系统会自动编译所有的第三方依赖库，并且强制设置一些PaddlePaddle参数的值（`WITH_C_API=ON`、`WITH_GPU=OFF`、`WITH_AVX=OFF`、`WITH_PYTHON=OFF`、`WITH_RDMA=OFF`）。
+- `WITH_C_API`，是否编译C-API预测库，必须设置为ON。在iOS平台上只支持使用C-API来预测。
+- `WITH_SWIG_PY`，必须设置为ON。在iOS平台上不支持通过swig调用来训练或者预测。
+
+iOS平台可选配置参数：
+
+- `IOS_PLATFORM`，可设置为`OS/SIMULATOR`，默认值为`OS`。
+  - `OS`，构建目标为`arm`架构的iPhone或者iPad等物理设备。
+  - `SIMULATOR`，构建目标为`x86`架构的模拟器平台。
+- `IOS_ARCH`，目标架构。针对不同的`IOS_PLATFORM`，可设置的目标架构如下表所示：
+
+   | IOS_PLATFORM | IOS_ARCH             |
+   |--------------|----------------------|
+   |   OS         | armv7, armv7s, arm64 (默认) |
+   | SIMULATOR    | i386, x86_64 (默认)         |   
+
+- `IOS_DEPLOYMENT_TARGET`，最小的iOS部署版本，默认值为`7.0`。
+- `IOS_ENABLE_BITCODE`，是否使能[Bitcode](https://developer.apple.com/library/content/documentation/IDEs/Conceptual/AppDistributionGuide/AppThinning/AppThinning.html#//apple_ref/doc/uid/TP40012582-CH35-SW3)，可设置`ON/OFF`，默认值为`ON`。
+- `IOS_USE_VECLIB_FOR_BLAS`，是否使用[vecLib](https://developer.apple.com/documentation/accelerate/veclib)框架进行BLAS矩阵计算，可设置`ON/OFF`，默认值为`OFF`。
+- `IOS_DEVELOPMENT_ROOT`，`Developer`目录，可显式指定为`/path/to/platform/Developer`。若未显式指定，PaddlePaddle将会根据`IOS_PLATFORM`自动选择`Xcode`对应`platform`的`Developer`目录。
+- `IOS_SDK_ROOT`，所使用`SDK`的根目录，可显式指定为`/path/to/platform/Developer/SDKs/SDK`。若未显式指定，PaddlePaddle将会自动选择`IOS_DEVELOPMENT_ROOT`目录下最新的`SDK`版本。
+
+其他配置参数：
+
+- `USE_EIGEN_FOR_BLAS`，是否使用Eigen库进行矩阵计算，在`IOS_USE_VECLIB_FOR_BLAS=OFF`时有效。可设置`ON/OFF`，默认值为`OFF`。
+- `HOST_C/CXX_COMPILER`，宿主机的C/C++编译器。默认值为环境变量`CC/CXX`的值；若环境变量`CC/CXX`未设置，则使用`cc/c++`编译器。
+
+常用的cmake配置如下：
+
+```bash
+cmake -DCMAKE_SYSTEM_NAME=iOS \
+      -DIOS_PLATFORM=OS \
+      -DIOS_ARCH="arm64" \
+      -DIOS_ENABLE_BITCODE=ON \
+      -DIOS_USE_VECLIB_FOR_BLAS=ON \
+      -DCMAKE_INSTALL_PREFIX=your/path/to/install \
+      -DWITH_C_API=ON \
+      -DWITH_TESTING=OFF \
+      -DWITH_SWIG_PY=OFF \
+      ..
+```
+
+```bash
+cmake -DCMAKE_SYSTEM_NAME=iOS \
+      -DIOS_PLATFORM=SIMULATOR \
+      -DIOS_ARCH="x86_64" \
+      -DIOS_USE_VECLIB_FOR_BLAS=ON \
+      -DCMAKE_INSTALL_PREFIX=your/path/to/install \
+      -DWITH_C_API=ON \
+      -DWITH_TESTING=OFF \
+      -DWITH_SWIG_PY=OFF \
+      ..
+```
+
+用户还可根据自己的需求设置其他编译参数。比如希望最小化生成库的大小，可以设置`CMAKE_BUILD_TYPE`为`MinSizeRel`；若希望得到最快的执行速度，则可设置`CMAKE_BUILD_TYPE`为`Release`。亦可以通过手动设置`CMAKE_C/CXX_FLAGS`来影响PaddlePaddle的编译过程。
+
+**性能TIPS**，为了达到最快的计算速度，在CMake参数配置上，有以下建议：
+
+- 设置`CMAKE_BUILD_TYPE`为`Release`
+- 设置`IOS_USE_VECLIB_FOR_BLAS=ON`，调用`vecLib`框架提供的BLAS函数进行矩阵计算。
+
+## 编译和安装
+
+CMake配置完成后，执行以下命令，PaddlePaddle将自动下载和编译所有第三方依赖库、编译和安装PaddlePaddle预测库。
+
+```
+$ make
+$ make install
+```
+
+注意：如果你曾在源码目录下编译过其他平台的PaddlePaddle库，请先使用`rm -rf`命令删除`third_party`目录和`build`目录，以确保所有的第三方依赖库和PaddlePaddle代码都是针对新的CMake配置重新编译的。
+
+执行完安装命令后，`your/path/to/install`目录中会包含以下内容：
+
+- `include`目录，其中包含所有C-API的头文件
+- `lib`目录，其中包含PaddlePaddle的C-API静态库
+- `third_party`目录，其中包含所依赖的所有第三方库
+
+注意，不同架构的PaddlePaddle库建议安装到不同的目录下，然后使用`lipo`工具将多个静态库合并成一个支持多个架构的fat库。
+
+自此，PaddlePaddle库已经安装完成，用户可将合成的fat库用于深度学习相关的iOS App中，调用方法见C-API文档。

doc/howto/cross_compiling/cross_compiling_for_raspberry_cn.md

@@ -59,4 +59,4 @@ make install
 
 注意：如果你曾经在源码目录下编译过其他平台的PaddlePaddle库，请先使用`rm -rf`命令删除`third_party`目录和`build`目录，以确保所有的第三方依赖库和PaddlePaddle代码都是针对新的CMake配置重新编译的。
 
-执行完安装命令后，，`your/path/to/install`目录中会包含`include`和`lib`目录，其中`include`中包含C-API的头文件，`lib`中包含一个Raspberry Pi版本的库。
+执行完安装命令后，`your/path/to/install`目录中会包含`include`和`lib`目录，其中`include`中包含C-API的头文件，`lib`中包含一个Raspberry Pi版本的库。

doc/howto/cross_compiling/cross_compiling_for_raspberry_en.md

@@ -44,7 +44,7 @@ cmake -DCMAKE_SYSTEM_NAME=RPi \
       ..
 ```
 
-To build the inference library, please set the argument WITH_API to ON: `WITH_C_API=ON`.
+To build the inference library, please set the argument WITH\_C\_API to ON: `WITH_C_API=ON`.
 
 You can add more arguments. For example, to minimize the size of the generated inference library, you may use `CMAKE_BUILD_TYPE=MinSizeRel`. For performance optimization, you may use `CMAKE_BUILD_TYPE=Release`.
 

paddle/framework/lod_tensor_test.cu

@@ -36,8 +36,8 @@ TEST(LoDTensor, LoDInGPU) {
   lod_tensor.mutable_data<float>(place);
 
   lod_tensor.set_lod(src_lod);
-  CHECK_EQ(lod_tensor.lod_element(0, 2).first, 4UL);
-  CHECK_EQ(lod_tensor.lod_element(0, 4).first, 8UL);
+  EXPECT_EQ(lod_tensor.lod_element(0, 2).first, 4UL);
+  EXPECT_EQ(lod_tensor.lod_element(0, 4).first, 8UL);
 
   auto lod = lod_tensor.lod();
 
@@ -45,6 +45,6 @@ TEST(LoDTensor, LoDInGPU) {
   cudaDeviceSynchronize();
 
   for (size_t i = 0; i < src_lod[0].size(); ++i) {
-    CHECK_EQ(lod[0].data()[i], src_lod[0].data()[i] * 2);
+    EXPECT_EQ(lod[0].data()[i], src_lod[0].data()[i] * 2);
   }
-}
\ No newline at end of file
+}

paddle/framework/operator.cc

@@ -37,32 +37,32 @@ ExecutionContext::GetEigenDevice<platform::GPUPlace, Eigen::GpuDevice>() const {
 std::string OperatorBase::Input(const std::string& name) const {
   auto& ins = Inputs(name);
   PADDLE_ENFORCE_LE(ins.size(), 1UL,
-                    "Op %s input %s should contain only one variable", type_,
-                    name);
+                    "Operator %s's input %s should contain only one variable.",
+                    type_, name);
   return ins.empty() ? kEmptyVarName : ins[0];
 }
 
 const std::vector<std::string>& OperatorBase::Inputs(
     const std::string& name) const {
   auto it = inputs_.find(name);
-  PADDLE_ENFORCE(it != inputs_.end(), "Op %s do not have input %s", type_,
-                 name);
+  PADDLE_ENFORCE(it != inputs_.end(), "Operator %s does not have the input %s.",
+                 type_, name);
   return it->second;
 }
 
 std::string OperatorBase::Output(const std::string& name) const {
   auto& outs = Outputs(name);
   PADDLE_ENFORCE_LE(outs.size(), 1UL,
-                    "Op %s output %s should contain only one variable", type_,
-                    name);
+                    "Operator %s's output %s should contain only one variable.",
+                    type_, name);
   return outs.empty() ? kEmptyVarName : outs[0];
 }
 
 const std::vector<std::string>& OperatorBase::Outputs(
     const std::string& name) const {
   auto it = outputs_.find(name);
-  PADDLE_ENFORCE(it != outputs_.end(), "Op %s does not have output called %s",
-                 type_, name);
+  PADDLE_ENFORCE(it != outputs_.end(),
+                 "Operator %s does not have an output called %s.", type_, name);
   return it->second;
 }
 

paddle/framework/operator.h

@@ -427,7 +427,8 @@ class OperatorWithKernel : public OperatorBase {
             int tmp = static_cast<int>(ToDataType(t->type()));
             VLOG(3) << "Input " << ipt_name << " with data_type " << tmp;
             PADDLE_ENFORCE(tmp == data_type || data_type == -1,
-                           "DataType of Paddle Op %s must be same.", Type());
+                           "DataType of Paddle Op %s must be the same.",
+                           Type());
             data_type = tmp;
           }
         }

paddle/framework/tensor.h

@@ -118,10 +118,12 @@ class Tensor {
                              const platform::DeviceContext& ctx);
 
   /**
-   * @brief   Return the slice of the tensor.
+   * @brief  Return a sub-tensor of the given tensor.
    *
-   * @param[in] begin_idx   The begin index of the slice.
-   * @param[in] end_idx     The end index of the slice.
+   * @param[in] begin_idx   The index of the start row(inclusive) to slice.
+   *                        The index number begins from 0.
+   * @param[in] end_idx     The index of the end row(exclusive) to slice.
+   *                        The index number begins from 0.
    */
   inline Tensor Slice(const int& begin_idx, const int& end_idx) const;
 

paddle/framework/tensor_impl.h

@@ -112,9 +112,10 @@ inline void* Tensor::mutable_data(platform::Place place, std::type_index type) {
   if (holder_ != nullptr) {
     holder_->set_type(type);
   }
-  PADDLE_ENFORCE_GT(numel(), 0,
-                    "Tensor's numel must be larger than zero to call "
-                    "Tensor::mutable_data. Call Tensor::set_dim first.");
+  PADDLE_ENFORCE_GT(
+      numel(), 0,
+      "When calling this method, the Tensor's numel must be larger than zero. "
+      "Please check Tensor::Resize has been called first.");
   int64_t size = numel() * SizeOfType(type);
   /* some versions of boost::variant don't have operator!= */
   if (holder_ == nullptr || !(holder_->place() == place) ||
@@ -229,10 +230,12 @@ inline void Tensor::CopyFromVector(const std::vector<T>& src,
 
 inline Tensor Tensor::Slice(const int& begin_idx, const int& end_idx) const {
   check_memory_size();
-  PADDLE_ENFORCE_GE(begin_idx, 0, "Slice begin index is less than zero.");
-  PADDLE_ENFORCE_LE(end_idx, dims_[0], "Slice end index is out of bound.");
-  PADDLE_ENFORCE_LT(begin_idx, end_idx,
-                    "Begin index must be less than end index.");
+  PADDLE_ENFORCE_GE(begin_idx, 0,
+                    "The start row index must be greater than 0.");
+  PADDLE_ENFORCE_LE(end_idx, dims_[0], "The end row index is out of bound.");
+  PADDLE_ENFORCE_LT(
+      begin_idx, end_idx,
+      "The start row index must be lesser than the end row index.");
 
   if (dims_[0] == 1) {
     return *this;

paddle/gserver/layers/CRFLayer.cpp

@@ -101,8 +101,10 @@ void CRFLayer::backward(const UpdateCallback& callback) {
                               : real(1.0f);
     instanceWeight *= coeff_;
 
-    MatrixPtr grad = output.grad->subRowMatrix(starts[i], starts[i + 1]);
-    grad->add(*crfs_[i].getXGrad(), real(1.0f), instanceWeight);
+    if (output.grad) {
+      MatrixPtr grad = output.grad->subRowMatrix(starts[i], starts[i + 1]);
+      grad->add(*crfs_[i].getXGrad(), real(1.0f), instanceWeight);
+    }
     if (needWGrad) {
       weight_->getWGrad()->add(
           *crfs_[i].getWGrad(), real(1.0f), instanceWeight);

paddle/gserver/layers/LinearChainCRF.cpp

@@ -102,7 +102,6 @@ real LinearChainCRF::forward(real* x, int* s, int length) {
 }
 
 void LinearChainCRF::backward(real* x, int* s, int length, bool needWGrad) {
-  MatrixPtr matX = Matrix::create(x, length, numClasses_);
   Matrix::resizeOrCreate(matGrad_, length, numClasses_);
   Matrix::resizeOrCreate(beta_, length, numClasses_);
   real* b = b_->getData();

paddle/gserver/layers/SequenceReshapeLayer.cpp

@@ -70,11 +70,23 @@ void SequenceReshapeLayer::forward(PassType passType) {
   size_t outDim = getSize();
 
   size_t numSequences = input.getNumSequences();
-  auto startPositions = input.sequenceStartPositions->getVector(false);
-  const int* starts = startPositions->getData();
 
-  CHECK_EQ(starts[numSequences], input.getBatchSize());
-  CHECK_EQ(numSequences, startPositions->getSize() - 1);
+  // by default, we assume each instance as a sequence
+  IVectorPtr seqStarts;
+  IVector::resizeOrCreate(seqStarts, input.getBatchSize() + 1, false);
+  int* startsData = seqStarts->getData();
+  for (int i = 0; i < input.getBatchSize() + 1; i++) {
+    startsData[i] = i;
+  }
+  const int* starts = startsData;
+
+  // if there is sequence, then use start positions
+  if (input.sequenceStartPositions) {
+    auto startPositions = input.sequenceStartPositions->getVector(false);
+    starts = startPositions->getData();
+    CHECK_EQ(starts[numSequences], input.getBatchSize());
+    CHECK_EQ(numSequences, startPositions->getSize() - 1);
+  }
 
   for (size_t seqID = 0; seqID < numSequences; seqID++) {
     size_t inNumIns = starts[seqID + 1] - starts[seqID];

paddle/memory/detail/system_allocator.cc

@@ -41,7 +41,16 @@ void* CPUAllocator::Alloc(size_t& index, size_t size) {
 
   index = 0;  // unlock memory
 
-  void* p = malloc(size);
+  void* p;
+
+#ifdef PADDLE_USE_MKLDNN
+  // refer to https://github.com/01org/mkl-dnn/blob/master/include/mkldnn.hpp
+  // memory alignment
+  PADDLE_ENFORCE_EQ(posix_memalign(&p, 4096ul, size), 0);
+#else
+  PADDLE_ENFORCE_EQ(posix_memalign(&p, 32ul, size), 0);
+#endif
+  PADDLE_ENFORCE(p, "Fail to allocate CPU memory: size = %d .", size);
 
   if (p != nullptr) {
     if (FLAGS_use_pinned_memory) {

paddle/operators/cross_entropy_op.cc

@@ -28,8 +28,9 @@ class CrossEntropyOp : public framework::OperatorWithKernel {
 
     auto x_dims = ctx->GetInputDim("X");
     auto label_dims = ctx->GetInputDim("Label");
-    PADDLE_ENFORCE_EQ(x_dims.size(), 2, "Input(X)'s rank should be 2.");
-    PADDLE_ENFORCE_EQ(label_dims.size(), 2, "Input(Label)'s rank should be 2.");
+    PADDLE_ENFORCE_EQ(x_dims.size(), 2UL, "Input(X)'s rank should be 2.");
+    PADDLE_ENFORCE_EQ(label_dims.size(), 2UL,
+                      "Input(Label)'s rank should be 2.");
     PADDLE_ENFORCE_EQ(x_dims[0], label_dims[0],
                       "The 1st dimension of Input(X) and Input(Label) should "
                       "be equal.");
@@ -38,8 +39,8 @@ class CrossEntropyOp : public framework::OperatorWithKernel {
                         "If Attr(soft_label) == true, the 2nd dimension of "
                         "Input(X) and Input(Label) should be equal.");
     } else {
-      PADDLE_ENFORCE_EQ(label_dims[1], 1,
-                        "If Attr(soft_label) == false, the 2nd dimension of "
+      PADDLE_ENFORCE_EQ(label_dims[1], 1UL,
+                        "If Attr(softLabel) == false, the 2nd dimension of "
                         "Input(Label) should be 1.");
     }
 
@@ -48,7 +49,8 @@ class CrossEntropyOp : public framework::OperatorWithKernel {
   }
 
  protected:
-  // CrossEntropy's data type just determined by "X"
+  // Explicitly set that data type of the output of the cross_entropy operator
+  // is determined by its input "X".
   framework::DataType IndicateDataType(
       const framework::ExecutionContext& ctx) const override {
     return framework::ToDataType(ctx.Input<Tensor>("X")->type());

paddle/operators/linear_chain_crf_op.cc

+/* 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/linear_chain_crf_op.h"
+
+namespace paddle {
+namespace operators {
+
+class LinearChainCRFOpMaker : public framework::OpProtoAndCheckerMaker {
+ public:
+  LinearChainCRFOpMaker(framework::OpProto* proto,
+                        framework::OpAttrChecker* op_checker)
+      : OpProtoAndCheckerMaker(proto, op_checker) {
+    AddInput(
+        "Emission",
+        "(LoDTensor, default: LoDTensor<float>). "
+        "The unscaled emission weight matrix for the linear chain CRF. "
+        "This input is a LoDTensor with shape [N x D] where N is the size of "
+        "the mini-batch and D is the total tag number.");
+    AddInput(
+        "Transition",
+        "(Tensor, default: Tensor<float>). A Tensor with shape [(D + 2) x D]. "
+        "The learnable parameter for the linear_chain_crf operator. "
+        "See more details in the operator's comments.");
+    AddInput(
+        "Label",
+        "(LoDTensor, default: LoDTensor<int>). The ground truth which is a 2-D "
+        "LoDTensor with shape [N x 1], where N is the total element number in "
+        "a mini-batch.");
+    AddOutput(
+        "Alpha",
+        "Tensor, default: Tensor<float>. The forward vectors for the entire "
+        "batch. A two dimensional tensor with shape [N x D], "
+        "denoted as \f$\alpha\f$. \f$\alpha$\f is a memo table used to "
+        "calculate the normalization factor in CRF. \f$\alpha[k, v]$\f stores "
+        "the unnormalized probabilites of all possible unfinished sequences of "
+        "tags that end at position \f$k$\f with tag \f$v$\f. For each \f$k$\f, "
+        "\f$\alpha[k, v]$\f is a vector of length \f$D$\f with a component for "
+        "each tag value \f$v$\f. This vector is called a forward vecotr and "
+        "will also be used in backward computations.")
+        .AsIntermediate();
+    AddOutput("EmissionExps",
+              "The exponentials of Input(Emission). This is an intermediate "
+              "computational result in forward computation, and will be reused "
+              "in backward computation.")
+        .AsIntermediate();
+    AddOutput("TransitionExps",
+              "The exponentials of Input(Transition). This is an intermediate "
+              "computational result in forward computation, and will be reused "
+              "in backward computation.")
+        .AsIntermediate();
+    AddOutput(
+        "LogLikelihood",
+        "(Tensor, default: Tensor<float>). The logarithm of the conditional "
+        "likelihood of each training sample in a mini-batch. This is a 2-D "
+        "tensor with shape [S x 1], where S is the sequence number in a "
+        "mini-batch. Note: S is equal to the sequence number in a mini-batch. "
+        "The output is no longer a LoDTensor.");
+    AddComment(R"DOC(
+Conditional Random Field defines an undirected probabilistic graph with nodes
+denoting random variables and edges denoting dependencies between these
+variables. CRF learns the conditional probability \f$P(Y|X)\f$, where
+\f$X = (x_1, x_2, ... , x_n)\f$ are structured inputs and
+\f$Y = (y_1, y_2, ... , y_n)\f$ are labels for the inputs.
+
+Linear chain CRF is a special case of CRF that is useful for sequence labeling
+task. Sequence labeling tasks do not assume a lot of conditional
+independences among inputs. The only constraint they impose is that the input
+and output must be linear sequences. Thus, the graph of such a CRF is a simple
+chain or a line, which results in the linear chain CRF.
+
+This operator implements the Forward-Backward algorithm for the linear chain
+CRF. Please see http://www.cs.columbia.edu/~mcollins/fb.pdf and
+http://cseweb.ucsd.edu/~elkan/250Bwinter2012/loglinearCRFs.pdf for reference.
+
+Equation:
+
+- Denote Input(Emission) to this operator as \f$x\f$ here.
+- The first D values of Input(Transition) to this operator are for starting
+weights, denoted as \f$a\f$ here.
+- The next D values of Input(Transition) of this operator are for ending
+weights, denoted as \f$b\f$ here.
+- The remaning values of Input(Transition) are for transition weights,
+denoted as \f$w\f$ here.
+- Denote Input(Label) as \f$s\f$ here.
+
+The probability of a sequence \f$s\f$ of length \f$L\f$ is defined as:
+\f$P(s) = (1/Z) exp(a_{s_1} + b_{s_L}
+                 + \sum_{l=1}^L x_{s_l}
+                 + \sum_{l=2}^L w_{s_{l-1},s_l})\f$
+where \f$Z\f$ is a normalization value so that the sum of \f$P(s)\f$ over
+all possible sequences is \f$1\f$, and \f$x\f$ is the emission feature weight
+to the linear chain CRF.
+
+Finaly, the linear chain CRF operator outputs the logarithm of the conditional
+likelihood of each training sample in a mini-batch.
+
+NOTE:
+1. The feature function for a CRF is made up of the emission features and the
+transition features. The emission feature weights are NOT computed in
+this operator. They MUST be computed first before this operator is called.
+
+2. Because this operator performs global normalization over all possible
+sequences internally, it expects UNSCALED emission feature weights.
+Please do not call this op with the emission feature being output of any
+nonlinear activation.
+
+3. The 2nd dimension of Input(Emission) MUST be equal to the tag number.
+
+)DOC");
+  }
+};
+
+class LinearChainCRFOp : public framework::OperatorWithKernel {
+ public:
+  using framework::OperatorWithKernel::OperatorWithKernel;
+
+  void InferShape(framework::InferShapeContext* ctx) const override {
+    PADDLE_ENFORCE(ctx->HasInput("Emission"),
+                   "Input(Emission) should be not null.");
+    PADDLE_ENFORCE(ctx->HasInput("Transition"),
+                   "Input(Transition) should be not null.");
+    PADDLE_ENFORCE(ctx->HasInput("Label"), "Input(Label) should be not null.");
+
+    PADDLE_ENFORCE(ctx->HasOutput("Alpha"),
+                   "Output(Alpha) should be not null.");
+    PADDLE_ENFORCE(ctx->HasOutput("EmissionExps"),
+                   "Output(EmissionExps) should be not null.");
+    PADDLE_ENFORCE(ctx->HasOutput("TransitionExps"),
+                   "Output(TransitionExps) should be not null.");
+    PADDLE_ENFORCE(ctx->HasOutput("LogLikelihood"),
+                   "Output(LogLikelihood) should be not null.");
+
+    auto emission_dims = ctx->GetInputDim("Emission");
+    PADDLE_ENFORCE_EQ(emission_dims.size(), 2UL,
+                      "The Input(Emission) should be a 2-D tensor.");
+    PADDLE_ENFORCE(emission_dims[0], "An empty mini-batch is not allowed.");
+
+    auto transition_dims = ctx->GetInputDim("Transition");
+    PADDLE_ENFORCE_EQ(transition_dims.size(), 2UL,
+                      "The Input(Transition) should be a 2-D tensor.");
+    PADDLE_ENFORCE_EQ(
+        transition_dims[0] - 2, transition_dims[1],
+        "An invalid dimension for the Input(Transition), which should "
+        "be a 2-D tensor with shape [(D + 2) x D].");
+    PADDLE_ENFORCE_EQ(
+        emission_dims[1], transition_dims[1],
+        "The 2nd dimension of the Input(Emission) and the Input(Transition) "
+        "should be equal to the tag number.");
+
+    auto label_dims = ctx->GetInputDim("Label");
+    PADDLE_ENFORCE(label_dims.size() == 2UL && label_dims[1] == 1UL,
+                   "The Input(Label) should be a 2-D tensor with the 2nd "
+                   "dimensions fixed to 1.");
+    PADDLE_ENFORCE_EQ(
+        emission_dims[0], label_dims[0],
+        "The height of Input(Emission) and the height of Input(Label) "
+        "should be the same.");
+
+    ctx->SetOutputDim("Alpha", emission_dims);
+    ctx->SetOutputDim("EmissionExps", emission_dims);
+    ctx->SetOutputDim("TransitionExps", transition_dims);
+    // TODO(caoying) This is tricky. The 1st dimension of Output(LogLikelihood)
+    // is the sequence number in a mini-batch. The dimension set here should be
+    // resized to its correct size in the function Compute. Fix this once we can
+    // get LoD information in the InferShape interface.
+    ctx->SetOutputDim("LogLikelihood", {emission_dims[0], 1});
+  }
+
+ protected:
+  // Explicitly set that the data type of output of the linear_chain_crf
+  // operator is determined by its input "Emission".
+  framework::DataType IndicateDataType(
+      const framework::ExecutionContext& ctx) const override {
+    return framework::ToDataType(ctx.Input<LoDTensor>("Emission")->type());
+  }
+};
+
+class LinearChainCRFGradOp : public framework::OperatorWithKernel {
+ public:
+  using framework::OperatorWithKernel::OperatorWithKernel;
+
+  void InferShape(framework::InferShapeContext* ctx) const override {
+    PADDLE_ENFORCE(ctx->HasInput("EmissionExps"),
+                   "Input(EmissionExps) should be not null.");
+    PADDLE_ENFORCE(ctx->HasInput("TransitionExps"),
+                   "Input(TransitionExps) should be not null.");
+    PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("LogLikelihood")),
+                   "Input(LogLikelihood@GRAD) shoudl be not null.");
+
+    auto emission_exps_dims = ctx->GetInputDim("EmissionExps");
+    PADDLE_ENFORCE_EQ(emission_exps_dims.size(), 2UL,
+                      "The Input(EmissionExps) should be a 2-D tensor.");
+    PADDLE_ENFORCE(emission_exps_dims[0],
+                   "An empty mini-batch is not allowed.");
+
+    auto transition_exps_dims = ctx->GetInputDim("TransitionExps");
+    PADDLE_ENFORCE_EQ(transition_exps_dims.size(), 2UL,
+                      "The Input(TransitionExps) should be a 2-D tensor.");
+    PADDLE_ENFORCE_EQ(
+        transition_exps_dims[0] - 2, transition_exps_dims[1],
+        "An invalid dimension for the Input(TransitionExps), which should "
+        "be a 2-D tensor with shape [(D + 2) x D].");
+    PADDLE_ENFORCE_EQ(
+        emission_exps_dims[1], transition_exps_dims[1],
+        "The 2nd dimension of the Input(EmissionExps) and the "
+        "Input(TransitionExps) should be equal to the tag number.");
+
+    auto label_dims = ctx->GetInputDim("Label");
+    PADDLE_ENFORCE(label_dims.size() == 2UL && label_dims[1] == 1UL,
+                   "The Input(Label) should be a 2-D tensor with the 2nd "
+                   "dimensions fixed to 1.");
+    PADDLE_ENFORCE_EQ(
+        emission_exps_dims[0], label_dims[0],
+        "The height of Input(EmissionExps) and the height of Input(Label) "
+        "should be the same.");
+
+    if (ctx->HasOutput(framework::GradVarName("Emission"))) {
+      ctx->SetOutputDim(framework::GradVarName("Emission"), emission_exps_dims);
+    }
+    if (ctx->HasOutput(framework::GradVarName("Transition"))) {
+      ctx->SetOutputDim(framework::GradVarName("Transition"),
+                        transition_exps_dims);
+    }
+  }
+
+ protected:
+  // Explicitly set that the data type of output of the linear_chain_crf_grad
+  // operator is determined by its input: gradients of LogLikelihood.
+  framework::DataType IndicateDataType(
+      const framework::ExecutionContext& ctx) const override {
+    return framework::ToDataType(
+        ctx.Input<LoDTensor>(framework::GradVarName("LogLikelihood"))->type());
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+REGISTER_OP(linear_chain_crf, ops::LinearChainCRFOp, ops::LinearChainCRFOpMaker,
+            linear_chain_crf_grad, ops::LinearChainCRFGradOp);
+REGISTER_OP_CPU_KERNEL(
+    linear_chain_crf,
+    ops::LinearChainCRFOpKernel<paddle::platform::CPUPlace, float>,
+    ops::LinearChainCRFOpKernel<paddle::platform::CPUPlace, double>);
+REGISTER_OP_CPU_KERNEL(
+    linear_chain_crf_grad,
+    ops::LinearChainCRFGradOpKernel<paddle::platform::CPUPlace, float>,
+    ops::LinearChainCRFGradOpKernel<paddle::platform::CPUPlace, double>);

paddle/operators/linear_chain_crf_op.cu

+/* 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/linear_chain_crf_op.h"
+
+namespace ops = paddle::operators;
+
+REGISTER_OP_GPU_KERNEL(
+    linear_chain_crf,
+    ops::LinearChainCRFOpKernel<paddle::platform::GPUPlace, float>,
+    ops::LinearChainCRFOpKernel<paddle::platform::GPUPlace, double>);
+REGISTER_OP_GPU_KERNEL(
+    linear_chain_crf_grad,
+    ops::LinearChainCRFGradOpKernel<paddle::platform::GPUPlace, float>,
+    ops::LinearChainCRFGradOpKernel<paddle::platform::GPUPlace, double>);

paddle/operators/linear_chain_crf_op.h

+/* 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/math/math_function.h"
+
+namespace paddle {
+namespace operators {
+
+template <typename T>
+static inline T NormalizeL1(T* x, size_t len) {
+  T sum = 0.;
+  for (size_t i = 0; i < len; ++i) sum += x[i];
+  // (This comment is from the old LinearChainCRFLayer.)
+  // Right now, we just bet that sum won't be zero. If this really happens, we
+  // will figure out what should be done then.
+  PADDLE_ENFORCE(sum,
+                 "The unnormalized probabilities of all possible unfinished "
+                 "sequences must be greater than 0.");
+  T s = 1. / sum;
+  for (size_t i = 0; i < len; ++i) x[i] *= s;
+  return sum;
+}
+
+template <typename T>
+struct ScalarMul {
+  explicit ScalarMul(const T& scalar) : scalar(scalar) {}
+  T operator()(const T& val) const { return val * scalar; }
+
+  T scalar;
+};
+
+using framework::LoDTensor;
+using framework::LoD;
+using framework::Tensor;
+template <typename T, int MajorType = Eigen::RowMajor,
+          typename IndexType = Eigen::DenseIndex>
+using EigenMatrix = framework::EigenMatrix<T, MajorType, IndexType>;
+
+template <typename Place, typename T>
+class LinearChainCRFOpKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    // TODO(caoying) The checks related to LoD information should be
+    // moved into InferShape once after the InferShape is refactored.
+    PADDLE_ENFORCE_EQ(ctx.Input<LoDTensor>("Emission")->NumLevels(), 1UL,
+                      "The Input(Emission) should be a sequence.");
+    PADDLE_ENFORCE_EQ(ctx.Input<LoDTensor>("Label")->NumLevels(), 1UL,
+                      "The Input(Label) should be a sequence.");
+    auto in_lod = ctx.Input<LoDTensor>("Label")->lod();
+    PADDLE_ENFORCE(in_lod.size(), "Input(Label) must be a sequence.");
+    const size_t level = 0;
+    const size_t seq_num = in_lod[level].size() - 1;
+
+    // These local variables hold the inputs and outputs, garanteeing them on
+    // CPU memory, to provide a consistent reference.
+    // TODO(caoying) Fix this by moving all these local variables into the
+    // class's data members once we can profile the whole training process.
+    LoDTensor* emission_weights = nullptr;
+    LoDTensor emission_weight_tensor;
+    Tensor* transition_weights = nullptr;
+    Tensor transition_weight_tensor;
+    LoDTensor* label = nullptr;
+    LoDTensor label_tensor;
+
+    Tensor* emission_exps = nullptr;
+    Tensor emission_exps_tensor;
+    Tensor* transition_exps = nullptr;
+    Tensor transition_exps_tensor;
+    Tensor* alpha = nullptr;
+    Tensor alpha_tensor;
+    Tensor* ll = nullptr;
+    Tensor ll_tensor;
+
+    if (platform::is_gpu_place(ctx.GetPlace())) {
+      emission_weights = &emission_weight_tensor;
+      transition_weights = &transition_weight_tensor;
+      label = &label_tensor;
+
+      CopyInputsToCpuMemory(
+          ctx.device_context(), *ctx.Input<LoDTensor>("Emission"),
+          *ctx.Input<Tensor>("Transition"), *ctx.Input<LoDTensor>("Label"),
+          emission_weights, transition_weights, label);
+
+      emission_exps = &emission_exps_tensor;
+      emission_exps->Resize(emission_weights->dims());
+
+      transition_exps = &transition_exps_tensor;
+      transition_exps->Resize(transition_weights->dims());
+
+      alpha = &alpha_tensor;
+      alpha->Resize(ctx.Output<Tensor>("Alpha")->dims());
+
+      ll = &ll_tensor;
+    } else {
+      emission_weights =
+          const_cast<LoDTensor*>(ctx.Input<LoDTensor>("Emission"));
+      transition_weights = const_cast<Tensor*>(ctx.Input<Tensor>("Transition"));
+      label = const_cast<LoDTensor*>(ctx.Input<LoDTensor>("Label"));
+
+      emission_exps = ctx.Output<Tensor>("EmissionExps");
+      transition_exps = ctx.Output<Tensor>("TransitionExps");
+      alpha = ctx.Output<Tensor>("Alpha");
+      ll = ctx.Output<Tensor>("LogLikelihood");
+    }
+
+    // Because the computation codes only runs on CPU, here the memory for all
+    // the outputs is FIXED to be allocated on the CPU memory.
+    emission_exps->mutable_data<T>(platform::CPUPlace());
+    transition_exps->mutable_data<T>(platform::CPUPlace());
+    alpha->mutable_data<T>(platform::CPUPlace());
+
+    // Resize the output tensor to its correct dimension.
+    ll->Resize({static_cast<int>(seq_num), 1});
+    ll->mutable_data<T>(platform::CPUPlace());
+
+    // Now, all the inputs and outputs should be on the CPU memory.
+    auto emission_dims = emission_weights->dims();
+    const size_t batch_size = emission_dims[0];
+    const size_t tag_num = emission_dims[1];
+
+    Tensor emission_row_max;
+    emission_row_max.mutable_data<T>(
+        framework::make_ddim({static_cast<int>(batch_size), 1}),
+        platform::CPUPlace());
+
+    auto place = ctx.GetEigenDevice<platform::CPUPlace>();
+    auto x = EigenMatrix<T>::From(*emission_weights);
+    auto x_row_max = EigenMatrix<T>::From(emission_row_max);
+    x_row_max.device(place) =
+        x.maximum(Eigen::DSizes<int, 1>(1))
+            .reshape(Eigen::DSizes<int, 2>(int(batch_size), 1));
+
+    auto x_exps = EigenMatrix<T>::From(*emission_exps);
+    x_exps.device(place) =
+        (x - x_row_max.broadcast(Eigen::DSizes<int, 2>(1, tag_num))).exp();
+
+    auto w = EigenMatrix<T>::From(*transition_weights);
+    auto w_exps = EigenMatrix<T>::From(*transition_exps);
+    w_exps.device(place) = w.exp();
+
+    T* log_likelihood = ll->data<T>();
+    for (size_t i = 0; i < seq_num; ++i) {
+      int start_pos = static_cast<int>(in_lod[level][i]);
+      int end_pos = static_cast<int>(in_lod[level][i + 1]);
+      if (end_pos == start_pos) {
+        // If an empty input sequence is given, pad 0 for its cost.
+        log_likelihood[i] = 0.;
+        continue;
+      }
+
+      const Tensor one_seq = emission_weights->Slice(start_pos, end_pos);
+      Tensor one_seq_row_max = emission_row_max.Slice(start_pos, end_pos);
+      Tensor one_seq_exps = emission_exps->Slice(start_pos, end_pos);
+      const Tensor one_seq_label = label->Slice(start_pos, end_pos);
+      Tensor one_seq_alpha = alpha->Slice(start_pos, end_pos);
+
+      log_likelihood[i] = ForwardOneSequence(
+          one_seq, one_seq_row_max, one_seq_exps, *transition_weights,
+          *transition_exps, one_seq_label, &one_seq_alpha);
+    }
+
+    if (platform::is_gpu_place(ctx.GetPlace())) {
+      CopyOutputsToGpuMemory(
+          ctx.device_context(), *emission_exps, *transition_exps, *alpha, *ll,
+          ctx.Output<Tensor>("EmissionExps"),
+          ctx.Output<Tensor>("TransitionExps"), ctx.Output<Tensor>("Alpha"),
+          ctx.Output<Tensor>("LogLikelihood"));
+    }
+  };
+
+ private:
+  void CopyInputsToCpuMemory(const platform::DeviceContext& ctx,
+                             const LoDTensor& emission_weights_src,
+                             const Tensor& transition_weights_src,
+                             const LoDTensor& label_src,
+                             LoDTensor* emission_weights_dst,
+                             Tensor* transition_weights_dst,
+                             LoDTensor* label_dst) const {
+    // Copy the inputs from GPU memory to CPU memory if this operators runs on
+    // GPU device.
+    auto copyLoDTensor = [](const platform::DeviceContext& ctx,
+                            const LoDTensor& src, LoDTensor* dst) {
+      dst->mutable_data<T>(src.dims(), platform::CPUPlace());
+      dst->CopyFrom(src, platform::CPUPlace(), ctx);
+    };
+
+    copyLoDTensor(ctx, emission_weights_src, emission_weights_dst);
+    copyLoDTensor(ctx, label_src, label_dst);
+
+    transition_weights_dst->mutable_data<T>(transition_weights_src.dims(),
+                                            platform::CPUPlace());
+    transition_weights_dst->CopyFrom(transition_weights_src,
+                                     platform::CPUPlace(), ctx);
+  }
+
+  void CopyOutputsToGpuMemory(const platform::DeviceContext& ctx,
+                              const Tensor& emission_exps_src,
+                              const Tensor& transition_exps_src,
+                              const Tensor& alpha_src, const Tensor& ll_src,
+                              Tensor* emission_exps_dst,
+                              Tensor* transition_exps_dst, Tensor* alpha_dst,
+                              Tensor* ll_dst) const {
+    // Copy the forward results from CPU memory to GPU memory if this
+    // operators runs on GPU device.
+    auto copyTensor = [](const platform::DeviceContext& ctx, const Tensor& src,
+                         Tensor* dst) {
+      dst->mutable_data<T>(platform::GPUPlace());
+      dst->CopyFrom(src, platform::GPUPlace(), ctx);
+    };
+    copyTensor(ctx, emission_exps_src, emission_exps_dst);
+    copyTensor(ctx, transition_exps_src, transition_exps_dst);
+    copyTensor(ctx, alpha_src, alpha_dst);
+    copyTensor(ctx, ll_src, ll_dst);
+  }
+
+  T ForwardOneSequence(const Tensor& emission, const Tensor& emission_row_max,
+                       const Tensor& emission_exps, const Tensor& trans_weights,
+                       const Tensor& trans_weight_exps, const Tensor& label,
+                       Tensor* alpha) const {
+    const T* x = emission.data<T>();
+    const T* x_row_max = emission_row_max.data<T>();
+    const T* x_exps = emission_exps.data<T>();
+    const T* w = trans_weights.data<T>();
+    const T* w_exps = trans_weight_exps.data<T>();
+    T* alpha_value = alpha->data<T>();
+
+    auto x_dims = emission.dims();
+    const size_t seq_length = x_dims[0];
+    const size_t tag_num = x_dims[1];
+    // The 1st row of w are transition weights for start mask.
+    // The 2nd row of w are transition weights for end mask.
+    // Transition weights between other tags begin from the 3rd row of w.
+    const size_t state_trans_base_idx = 2;
+
+    for (size_t i = 0; i < tag_num; ++i) {
+      alpha_value[i] = w_exps[i] * x_exps[i];
+    }
+    T ll = -x_row_max[0] - std::log(NormalizeL1<T>(alpha_value, tag_num));
+
+    for (size_t k = 1; k < seq_length; ++k) {
+      for (size_t i = 0; i < tag_num; ++i) {
+        T sum = 0.;
+        for (size_t j = 0; j < tag_num; ++j) {
+          sum += alpha_value[(k - 1) * tag_num + j] *  // (*)
+                 w_exps[(j + state_trans_base_idx) * tag_num + i];
+        }
+        alpha_value[k * tag_num + i] = x_exps[k * tag_num + i] * sum;
+      }
+      // NormalizeL1 is to avoid underflow or overflow at (*).
+      ll -= x_row_max[k] +
+            std::log(NormalizeL1<T>(alpha_value + k * tag_num, tag_num));
+    }
+    T sum = 0.;
+    for (size_t i = 0; i < tag_num; ++i) {
+      sum += alpha_value[(seq_length - 1) * tag_num + i] * w_exps[tag_num + i];
+    }
+    ll -= std::log(sum);
+    // Now ll is equal to -log(Z).
+
+    const int* lbl = label.data<int>();
+    PADDLE_ENFORCE_LT(
+        *std::max_element(lbl, lbl + seq_length), tag_num,
+        "An invalid tag label that execesses the largest tag number.");
+
+    // Calculate the nominator part, which depends on the label sequence.
+    ll += w[lbl[0]] /*start transition*/ + x[lbl[0]] +
+          w[tag_num + lbl[seq_length - 1]] /*end transition*/;
+    for (size_t k = 1; k < seq_length; ++k) {
+      ll += x[k * tag_num + lbl[k]] +
+            w[(lbl[k - 1] + state_trans_base_idx) * tag_num + lbl[k]];
+    }
+    return -ll;
+  }
+};
+
+template <typename Place, typename T>
+class LinearChainCRFGradOpKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    const size_t level = 0;  // currently, only support sequence.
+    auto lod = ctx.Input<LoDTensor>("Label")->lod();
+    PADDLE_ENFORCE(lod.size(), "Input(Label) must be a sequence.");
+
+    // These local variables hold the inputs and outputs, garanteeing them on
+    // CPU memory, to provide a consistent reference.
+    // TODO(caoying) Fix this by moving all these local variables into the
+    // class's data members once we can profile the training process, or
+    // implementing a real GPU kernel for CRF.
+    Tensor* label = nullptr;
+    Tensor label_tensor;
+    Tensor* emission_exps = nullptr;
+    Tensor emission_exps_tensor;
+    Tensor* transition_exps = nullptr;
+    Tensor transition_exps_tensor;
+    Tensor* alpha = nullptr;
+    Tensor alpha_tensor;
+    Tensor ll_grad_tensor;
+    T* ll_grad = nullptr;
+
+    Tensor* emission_grad = nullptr;
+    Tensor emission_grad_tensor;
+    Tensor* transition_grad = nullptr;
+    Tensor transition_grad_tensor;
+
+    if (platform::is_gpu_place(ctx.GetPlace())) {
+      label = &label_tensor;
+      emission_exps = &emission_exps_tensor;
+      transition_exps = &transition_exps_tensor;
+      alpha = &alpha_tensor;
+      CopyInputsToCpuMemory(
+          ctx.device_context(), *ctx.Input<LoDTensor>("Label"),
+          *ctx.Input<Tensor>("EmissionExps"),
+          *ctx.Input<Tensor>("TransitionExps"), *ctx.Input<Tensor>("Alpha"),
+          *ctx.Input<Tensor>(framework::GradVarName("LogLikelihood")), label,
+          emission_exps, transition_exps, alpha, &ll_grad_tensor);
+      ll_grad = ll_grad_tensor.data<T>();
+
+      if (ctx.Output<Tensor>(framework::GradVarName("Emission"))) {
+        emission_grad = &emission_grad_tensor;
+        emission_grad->Resize(emission_exps->dims());
+      }
+
+      if (ctx.Output<Tensor>(framework::GradVarName("Transition"))) {
+        transition_grad = &transition_grad_tensor;
+        transition_grad->Resize(transition_exps->dims());
+      }
+    } else {
+      label = const_cast<LoDTensor*>(ctx.Input<LoDTensor>("Label"));
+      emission_exps = const_cast<Tensor*>(ctx.Input<Tensor>("EmissionExps"));
+      transition_exps =
+          const_cast<Tensor*>(ctx.Input<Tensor>("TransitionExps"));
+      alpha = const_cast<Tensor*>(ctx.Input<Tensor>("Alpha"));
+      ll_grad = const_cast<Tensor*>(
+                    ctx.Input<Tensor>(framework::GradVarName("LogLikelihood")))
+                    ->data<T>();
+
+      emission_grad = ctx.Output<Tensor>(framework::GradVarName("Emission"));
+      transition_grad =
+          ctx.Output<Tensor>(framework::GradVarName("Transition"));
+    }
+
+    // TODO(caoying) Fix this constraint. When the Input(Emission) is from the
+    // data reader operator, it can have no gradients.
+    PADDLE_ENFORCE(emission_grad, "Output(Emission@Grad) should not be null.");
+    emission_grad->mutable_data<T>(platform::CPUPlace());
+    if (transition_grad) {
+      transition_grad->mutable_data<T>(platform::CPUPlace());
+      math::SetConstant<platform::CPUPlace, T>()(ctx.device_context(),
+                                                 transition_grad, 0.);
+    }
+    // Now, all the inputs and outputs should be on the CPU memory.
+
+    auto emission_dims = emission_exps->dims();
+    // Beta is the memo table used in dynamic programming to calculate the
+    // backwark vectors. For a backward vector i (the i-th row of beta), it
+    // captures the unnormalized probabilities of partial sequences starting
+    // at position i.
+    Tensor beta;
+    beta.mutable_data<T>(emission_dims, platform::CPUPlace());
+
+    for (size_t i = 0; i < lod[level].size() - 1; ++i) {
+      int start_pos = static_cast<int>(lod[level][i]);
+      int end_pos = static_cast<int>(lod[level][i + 1]);
+      if (end_pos == start_pos) continue;
+
+      const Tensor one_seq_emission_exps =
+          emission_exps->Slice(start_pos, end_pos);
+      const Tensor one_seq_label = label->Slice(start_pos, end_pos);
+      const Tensor one_seq_alpha = alpha->Slice(start_pos, end_pos);
+      Tensor one_seq_beta = beta.Slice(start_pos, end_pos);
+      Tensor one_seq_emission_grad = emission_grad->Slice(start_pos, end_pos);
+
+      BackwardOneSequence(ctx.device_context(), ll_grad[i],
+                          one_seq_emission_exps, *transition_exps,
+                          one_seq_alpha, one_seq_label, &one_seq_beta,
+                          transition_grad, &one_seq_emission_grad);
+    }
+
+    if (platform::is_gpu_place(ctx.GetPlace())) {
+      CopyOutputsToGpuMemory(
+          ctx.device_context(), emission_grad, transition_grad,
+          ctx.Output<Tensor>(framework::GradVarName("Emission")),
+          ctx.Output<Tensor>(framework::GradVarName("Transition")));
+    }
+  };
+
+ private:
+  void CopyInputsToCpuMemory(const platform::DeviceContext& ctx,
+                             const LoDTensor& label_src,
+                             const Tensor& emission_exps_src,
+                             const Tensor& transition_exps_src,
+                             const Tensor& alpha_src, const Tensor& ll_grad_src,
+                             Tensor* label_dst, Tensor* emission_exps_dst,
+                             Tensor* transition_exps_dst, Tensor* alpha_dst,
+                             Tensor* ll_grad_dst) const {
+    // Copy the inputs from GPU memory to CPU memory when this operators runs on
+    // GPU device.
+    label_dst->mutable_data<T>(label_src.dims(), platform::CPUPlace());
+    label_dst->CopyFrom(label_src, platform::CPUPlace(), ctx);
+
+    auto copyTensor = [](const platform::DeviceContext& ctx, const Tensor& src,
+                         Tensor* dst) {
+      dst->mutable_data<T>(src.dims(), platform::CPUPlace());
+      dst->CopyFrom(src, platform::CPUPlace(), ctx);
+    };
+    copyTensor(ctx, emission_exps_src, emission_exps_dst);
+    copyTensor(ctx, transition_exps_src, transition_exps_dst);
+    copyTensor(ctx, alpha_src, alpha_dst);
+    copyTensor(ctx, ll_grad_src, ll_grad_dst);
+  }
+
+  void CopyOutputsToGpuMemory(const platform::DeviceContext& ctx,
+                              const Tensor* emission_grad_src,
+                              const Tensor* transition_grad_src,
+                              Tensor* emission_grad_dst,
+                              Tensor* transition_grad_dst) const {
+    // Copy the backward results from CPU memory to GPU
+    // memory if this operators runs on GPU device.
+    auto copyTensor = [](const platform::DeviceContext& ctx, const Tensor* src,
+                         Tensor* dst) {
+      if (src && dst) {
+        dst->mutable_data<T>(platform::GPUPlace());
+        dst->CopyFrom(*src, platform::GPUPlace(), ctx);
+      }
+    };
+    copyTensor(ctx, emission_grad_src, emission_grad_dst);
+    copyTensor(ctx, transition_grad_src, transition_grad_dst);
+  }
+
+  void BackwardOneSequence(const platform::DeviceContext& ctx, const T ll_grad,
+                           const Tensor& emission_exps,
+                           const Tensor& transition_exps, const Tensor& alpha,
+                           const Tensor& label, Tensor* beta,
+                           Tensor* transition_grad,
+                           Tensor* emission_grad) const {
+    const T* w_exps = transition_exps.data<T>();
+    const T* x_exps = emission_exps.data<T>();
+    const int* label_value = label.data<int>();
+    T* beta_value = beta->data<T>();
+
+    auto x_dims = emission_exps.dims();
+    const size_t seq_length = x_dims[0];
+    const size_t tag_num = x_dims[1];
+    const size_t state_trans_base_idx = 2;
+
+    // Calculate the backward vectors: beta.
+    // First, calculate the initialition state.
+    for (size_t i = 0; i < tag_num; ++i) {
+      beta_value[(seq_length - 1) * tag_num + i] = w_exps[tag_num + i];
+    }
+    NormalizeL1<T>(beta_value + (seq_length - 1) * tag_num, tag_num);
+    for (int k = static_cast<int>(seq_length) - 2; k >= 0; --k) {
+      for (size_t i = 0; i < tag_num; ++i) {
+        T sum = 0.;
+        for (size_t j = 0; j < tag_num; ++j) {
+          sum += w_exps[(i + state_trans_base_idx) * tag_num + j] *  // (**)
+                 x_exps[(k + 1) * tag_num + j] *
+                 beta_value[(k + 1) * tag_num + j];
+        }
+        beta_value[k * tag_num + i] = sum;
+      }
+      // NormalizeL1 is to avoid underflow or overflow at (**).
+      NormalizeL1<T>(beta_value + k * tag_num, tag_num);
+    }
+
+    auto x_grad_mat = EigenMatrix<T>::From(*emission_grad);
+    auto alpha_mat = EigenMatrix<T>::From(alpha);
+    auto beta_mat = EigenMatrix<T>::From(*beta);
+
+    auto* place = ctx.GetEigenDevice<platform::CPUPlace>();
+    auto prob = alpha_mat * beta_mat;
+    auto row_sum = prob.sum(Eigen::DSizes<int, 1>(1))
+                       .reshape(Eigen::DSizes<int, 2>(seq_length, 1))
+                       .broadcast(Eigen::DSizes<int, 2>(1, tag_num));
+    x_grad_mat.device(*place) =
+        (prob / row_sum).unaryExpr(ScalarMul<T>(ll_grad));
+
+    for (size_t k = 0; k < seq_length; ++k) {
+      x_grad_mat(k, label_value[k]) -= static_cast<T>(ll_grad);
+    }
+
+    if (transition_grad) {
+      T* trans_grad = transition_grad->data<T>();
+      for (size_t k = 0; k < tag_num; ++k) {
+        // Do not multiply by the output gradient here, because x_grad_mat has
+        // alrealy done this.
+        trans_grad[k] += x_grad_mat(/*from start state*/ 0, k);
+        trans_grad[tag_num + k] +=
+            x_grad_mat(/*to end state*/ seq_length - 1, k);
+      }
+
+      auto x_exps_mat = EigenMatrix<T>::From(emission_exps);
+
+      // TODO(caoying): Fix this to avoid using this local variable if we can
+      // profile the training process.
+      Tensor tmp;
+      tmp.mutable_data<T>(beta->dims(), platform::CPUPlace());
+      auto tmp_mat = EigenMatrix<T>::From(tmp);
+      auto prob = beta_mat * x_exps_mat;
+      auto row_sum = prob.sum(Eigen::DSizes<int, 1>(1))
+                         .reshape(Eigen::DSizes<int, 2>(seq_length, 1))
+                         .broadcast(Eigen::DSizes<int, 2>(1, tag_num));
+      tmp_mat.device(*place) = prob / row_sum;
+
+      for (size_t k = 1; k < seq_length; ++k) {
+        T sum = 0.;
+        for (size_t i = 0; i < tag_num; ++i) {
+          for (size_t j = 0; j < tag_num; ++j) {
+            sum += w_exps[(i + state_trans_base_idx) * tag_num + j] *  // (**)
+                   alpha_mat(k - 1, i) * tmp_mat(k, j);
+          }
+        }
+        sum = 1. / sum;
+        for (size_t i = 0; i < tag_num; ++i) {
+          for (size_t j = 0; j < tag_num; ++j) {
+            trans_grad[(i + state_trans_base_idx) * tag_num + j] +=
+                sum * w_exps[(i + state_trans_base_idx) * tag_num + j] *
+                alpha_mat(k - 1, i) * tmp_mat(k, j) * ll_grad;
+          }
+        }
+        trans_grad[(label_value[k - 1] + state_trans_base_idx) * tag_num +
+                   label_value[k]] -= static_cast<T>(ll_grad);
+      }
+    }
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle

paddle/operators/lstm_op.cc

@@ -21,7 +21,6 @@ class LSTMOp : public framework::OperatorWithKernel {
  public:
   using framework::OperatorWithKernel::OperatorWithKernel;
 
- protected:
   void InferShape(framework::InferShapeContext* ctx) const override {
     PADDLE_ENFORCE(ctx->HasInput("Input"),
                    "Input(Input) of LSTM should not be null.");
@@ -29,9 +28,13 @@ class LSTMOp : public framework::OperatorWithKernel {
                    "Output(Hidden) of LSTM should not be null.");
     PADDLE_ENFORCE(ctx->HasOutput("Cell"),
                    "Output(Cell) of LSTM should not be null.");
+    PADDLE_ENFORCE(ctx->HasOutput("BatchGate"),
+                   "Output(BatchGate) of LSTM should not be null.");
+    PADDLE_ENFORCE(ctx->HasOutput("BatchCellPreAct"),
+                   "Output(BatchGate) of LSTM should not be null.");
 
-    auto x_dims = ctx->GetInputDim("Input");
-    PADDLE_ENFORCE_EQ(x_dims.size(), 2, "Input(X)'s rank must be 2.");
+    auto in_dims = ctx->GetInputDim("Input");
+    PADDLE_ENFORCE_EQ(in_dims.size(), 2, "Input(X)'s rank must be 2.");
 
     if (ctx->HasInput("H0")) {
       PADDLE_ENFORCE(ctx->HasInput("C0"),
@@ -44,7 +47,7 @@ class LSTMOp : public framework::OperatorWithKernel {
                      "should be the same.");
     }
 
-    int frame_size = x_dims[1] / 4;
+    int frame_size = in_dims[1] / 4;
     auto w_dims = ctx->GetInputDim("Weight");
     PADDLE_ENFORCE_EQ(w_dims.size(), 2,
                       "The rank of Input(Weight) should be 2.");
@@ -71,12 +74,21 @@ class LSTMOp : public framework::OperatorWithKernel {
                         "4 * %d if disable peepholes connection",
                         frame_size);
     }
-    ctx->SetOutputDim("Hidden", {x_dims[0], frame_size});
-    ctx->SetOutputDim("Cell", {x_dims[0], frame_size});
-    ctx->SetOutputDim("BatchGate", x_dims);
+    framework::DDim out_dims({in_dims[0], frame_size});
+    ctx->SetOutputDim("Hidden", out_dims);
+    ctx->SetOutputDim("Cell", out_dims);
+    ctx->SetOutputDim("BatchGate", in_dims);
+    ctx->SetOutputDim("BatchCellPreAct", out_dims);
     ctx->ShareLoD("Input", "Hidden");
     ctx->ShareLoD("Input", "Cell");
   }
+
+ protected:
+  framework::DataType IndicateDataType(
+      const framework::ExecutionContext& ctx) const override {
+    return framework::ToDataType(
+        ctx.Input<framework::LoDTensor>("Input")->type());
+  }
 };
 
 class LSTMOpMaker : public framework::OpProtoAndCheckerMaker {
@@ -86,16 +98,18 @@ class LSTMOpMaker : public framework::OpProtoAndCheckerMaker {
     AddInput("Input",
              "(LoDTensor) the first input is a LodTensor, which support "
              "variable-time length input sequence. The underlying tensor in "
-             "this LoDTensor is a matrix with shape (T X 4D), where, T is the "
+             "this LoDTensor is a matrix with shape (T X 4D), where T is the "
              "total time steps in this mini-batch, D is the hidden size.");
     AddInput("H0",
              "(Tensor, optional) the initial hidden state is an optional "
              "input. This is a tensor with shape (N x D), where N is the "
-             "batch size, D is the hidden size.");
+             "batch size, D is the hidden size.")
+        .AsDispensable();
     AddInput("C0",
              "(Tensor, optional) the initial cell state is an optional "
              "input. This is a tensor with shape (N x D), where N is the "
-             "batch size. `H0` and `C0` can be NULL but only at the same time");
+             "batch size. `H0` and `C0` can be NULL but only at the same time")
+        .AsDispensable();
     AddInput("Weight",
              "(Tensor) the learnable hidden-hidden weights."
              " - The shape is (D x 4D), where D is the hidden size. "
@@ -109,22 +123,27 @@ class LSTMOpMaker : public framework::OpProtoAndCheckerMaker {
              " - Bias = {b_c, b_i, b_f, b_o}."
              "2. `usePeepholes = True` "
              " - The shape is (1 x 7D). "
-             " - Bias = {b_c, b_i, b_f, b_o, W_ic, W_fc, W_oc}.");
+             " - Bias = {b_c, b_i, b_f, b_o, W_ic, W_fc, W_oc}.")
+        .AsDispensable();
+    AddOutput("Hidden",
+              "(LoDTensor) the hidden state of LSTM operator. "
+              "The shape is (T x D), and lod is the same with the `Input`.");
+    AddOutput("Cell",
+              "(LoDTensor) the cell state of LSTM operator. "
+              "The shape is (T x D), and lod is the same with the `Input`.");
     AddOutput("BatchGate",
               "(LoDTensor) This LoDTensor contains input gate, forget gate "
               "and output gate after the nonlinear computation. This "
               "LoDTensor has the same shape with the reorganized input, which "
-              "was also be called batch input. The LoD size is 2. The first "
+              "is also be called batch input. The LoD size is 2. The first "
               "LoD is the batch offsets and the second LoD contains the "
               "indexes, which denote the position of reorganized sequence "
               "in the raw input.")
         .AsIntermediate();
-    AddOutput("Hidden",
-              "(LoDTensor) the hidden state lod tensor of LSTM operator. "
-              "The shape and lod is the same with the `Input`.");
-    AddOutput("Cell",
-              "(LoDTensor) the cell state lod tensor of LSTM operator. "
-              "The shape and lod is the same with the `Input`.");
+    AddOutput("BatchCellPreAct",
+              "(LoDTensor) This LoDTensor is got in the forward and used "
+              "in the backward.")
+        .AsIntermediate();
     AddAttr<bool>("usePeepholes",
                   "(bool, defalut: True) "
                   "whether to enable diagonal/peephole connections.")
@@ -202,15 +221,37 @@ class LSTMGradOp : public framework::OperatorWithKernel {
  public:
   using framework::OperatorWithKernel::OperatorWithKernel;
 
- protected:
   void InferShape(framework::InferShapeContext* ctx) const override {
-    PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("Hidden")),
-                   "Input(Hidden@GRAD) should not be null");
-    PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("Cell")),
-                   "Input(Cell@GRAD) should not be null");
-    ctx->SetOutputDim(framework::GradVarName("Weight"),
-                      ctx->GetInputDim("Weight"));
-    ctx->SetOutputDim(framework::GradVarName("Bias"), ctx->GetInputDim("Bias"));
+    PADDLE_ENFORCE(ctx->HasInput("Input"),
+                   "Input(Input) of LSTM should not be null.");
+    PADDLE_ENFORCE(ctx->HasInput("Hidden"),
+                   "Input(Hidden) of LSTM should not be null.");
+    PADDLE_ENFORCE(ctx->HasInput("Cell"),
+                   "Input(Cell) of LSTM should not be null.");
+
+    PADDLE_ENFORCE(ctx->HasInput("BatchGate"),
+                   "Input(BatchGate) of LSTM should not be null.");
+    PADDLE_ENFORCE(ctx->HasInput("BatchCellPreAct"),
+                   "Input(BatchGate) of LSTM should not be null.");
+
+    auto in_g_name = framework::GradVarName("Input");
+    if (ctx->HasOutput(in_g_name))
+      ctx->SetOutputDim(in_g_name, ctx->GetInputDim("Input"));
+
+    auto w_g_name = framework::GradVarName("Weight");
+    if (ctx->HasOutput(w_g_name))
+      ctx->SetOutputDim(w_g_name, ctx->GetInputDim("Weight"));
+
+    auto b_g_name = framework::GradVarName("Bias");
+    if (ctx->HasOutput(b_g_name))
+      ctx->SetOutputDim(b_g_name, ctx->GetInputDim("Bias"));
+  }
+
+ protected:
+  framework::DataType IndicateDataType(
+      const framework::ExecutionContext& ctx) const override {
+    return framework::ToDataType(
+        ctx.Input<framework::LoDTensor>("Input")->type());
   }
 };
 

paddle/operators/lstm_op.h

@@ -21,8 +21,9 @@ limitations under the License. */
 namespace paddle {
 namespace operators {
 
-using framework::LoDTensor;
-using framework::Tensor;
+using LoDTensor = framework::LoDTensor;
+using Tensor = framework::Tensor;
+
 template <typename T, int MajorType = Eigen::RowMajor,
           typename IndexType = Eigen::DenseIndex>
 using EigenMatrix = framework::EigenMatrix<T, MajorType, IndexType>;
@@ -31,15 +32,15 @@ template <typename Place, typename T>
 class LSTMKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& ctx) const override {
-    auto* input = ctx.Input<framework::LoDTensor>("Input");
-    auto* weight = ctx.Input<framework::Tensor>("Weight");
-    auto* bias = ctx.Input<framework::Tensor>("Bias");
+    auto* input = ctx.Input<LoDTensor>("Input");
+    auto* weight = ctx.Input<Tensor>("Weight");
+    auto* bias = ctx.Input<Tensor>("Bias");
 
-    auto* batch_gate = ctx.Output<framework::LoDTensor>("BatchGate");
+    auto* batch_gate = ctx.Output<LoDTensor>("BatchGate");
     batch_gate->mutable_data<T>(ctx.GetPlace());
-    auto* hidden_out = ctx.Output<framework::LoDTensor>("Hidden");
+    auto* hidden_out = ctx.Output<LoDTensor>("Hidden");
     hidden_out->mutable_data<T>(ctx.GetPlace());
-    auto* cell_out = ctx.Output<framework::LoDTensor>("Cell");
+    auto* cell_out = ctx.Output<LoDTensor>("Cell");
     cell_out->mutable_data<T>(ctx.GetPlace());
 
     // Now the function ShareLoD in InferShape is not implemented.
@@ -49,7 +50,8 @@ class LSTMKernel : public framework::OpKernel<T> {
 
     bool is_reverse = ctx.Attr<bool>("isReverse");
     math::LoDTensor2BatchFunctor<Place, T> to_batch;
-    to_batch(ctx.device_context(), *input, *batch_gate, is_reverse);
+    auto& device_ctx = ctx.device_context();
+    to_batch(device_ctx, *input, *batch_gate, true, is_reverse);
 
     auto in_dims = input->dims();
     int frame_size = static_cast<int>(in_dims[1] / 4);
@@ -69,17 +71,26 @@ class LSTMKernel : public framework::OpKernel<T> {
     }
 
     math::LstmMetaValue<T> lstm_value;
-    T* bias_data = const_cast<T*>(bias->data<T>());
-    // the code style in LstmMetaValue will be updated later.
-    lstm_value.checkIg = bias_data + 4 * frame_size;
-    lstm_value.checkFg = lstm_value.checkIg + frame_size;
-    lstm_value.checkOg = lstm_value.checkFg + frame_size;
+    if (bias) {
+      T* bias_data = const_cast<T*>(bias->data<T>());
+      // the code style in LstmMetaValue will be updated later.
+
+      lstm_value.checkIg = bias_data + 4 * frame_size;
+      lstm_value.checkFg = lstm_value.checkIg + frame_size;
+      lstm_value.checkOg = lstm_value.checkFg + frame_size;
+    } else {
+      lstm_value.checkIg = nullptr;
+      lstm_value.checkFg = nullptr;
+      lstm_value.checkOg = nullptr;
+    }
     lstm_value.prevStateValue = nullptr;
 
-    framework::LoDTensor batch_out, batch_cell, batch_cell_pre_act;
-    batch_out.mutable_data<T>(dims, ctx.GetPlace());
+    // Use the local variable as here.
+    LoDTensor batch_hidden, batch_cell;
+    auto* batch_cell_pre_act = ctx.Output<LoDTensor>("BatchCellPreAct");
+    batch_hidden.mutable_data<T>(dims, ctx.GetPlace());
     batch_cell.mutable_data<T>(dims, ctx.GetPlace());
-    batch_cell_pre_act.mutable_data<T>(dims, ctx.GetPlace());
+    batch_cell_pre_act->mutable_data<T>(dims, ctx.GetPlace());
 
     auto batch_starts = batch_gate->lod()[0];
     size_t num_batch = batch_starts.size() - 1;
@@ -92,18 +103,18 @@ class LSTMKernel : public framework::OpKernel<T> {
       int bend = static_cast<int>(batch_starts[n + 1]);
 
       Tensor gate_t = batch_gate->Slice(bstart, bend);
-      Tensor out_t = batch_out.Slice(bstart, bend);
+      Tensor out_t = batch_hidden.Slice(bstart, bend);
       Tensor cell_t = batch_cell.Slice(bstart, bend);
-      Tensor cell_pre_act_t = batch_cell_pre_act.Slice(bstart, bend);
+      Tensor cell_pre_act_t = batch_cell_pre_act->Slice(bstart, bend);
 
       int cur_batch_size = bend - bstart;
 
       if (n != 0) {
         int pre_h_start = static_cast<int>(batch_starts[n - 1]);
         int pre_h_end = pre_h_start + cur_batch_size;
-        auto pre_hidden_t = batch_out.Slice(pre_h_start, pre_h_end);
-        math::matmul<Place, T>(ctx.device_context(), pre_hidden_t, false,
-                               *weight, false, static_cast<T>(1.0), &gate_t,
+        auto pre_hidden_t = batch_hidden.Slice(pre_h_start, pre_h_end);
+        math::matmul<Place, T>(device_ctx, pre_hidden_t, false, *weight, false,
+                               static_cast<T>(1.0), &gate_t,
                                static_cast<T>(1.0));
       }
       // else if : FIXME support the initial hidden and cell
@@ -112,27 +123,186 @@ class LSTMKernel : public framework::OpKernel<T> {
       lstm_value.outputValue = out_t.data<T>();
       lstm_value.stateValue = cell_t.data<T>();
       lstm_value.stateActiveValue = cell_pre_act_t.data<T>();
-      math::LstmUnitFunctor<Place, T>::compute(ctx.device_context(), lstm_value,
+      math::LstmUnitFunctor<Place, T>::compute(device_ctx, lstm_value,
                                                frame_size, cur_batch_size,
                                                gate_act, cell_act, cand_act);
       lstm_value.prevStateValue = lstm_value.stateValue;
     }
 
     math::Batch2LoDTensorFunctor<Place, T> to_seq;
-    batch_out.set_lod(batch_gate->lod());
+    batch_hidden.set_lod(batch_gate->lod());
     // restore the output hidden in LoDTensor from the batch hidden
-    to_seq(ctx.device_context(), batch_out, *hidden_out);
+    to_seq(device_ctx, batch_hidden, *hidden_out);
 
     batch_cell.set_lod(batch_gate->lod());
     // restore the output cell state in LoDTensor from the batch cell
-    to_seq(ctx.device_context(), batch_cell, *cell_out);
+    to_seq(device_ctx, batch_cell, *cell_out);
   }
 };
 
 template <typename Place, typename T>
 class LSTMGradKernel : public framework::OpKernel<T> {
  public:
-  void Compute(const framework::ExecutionContext& ctx) const override {}
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* input = ctx.Input<LoDTensor>("Input");
+    auto* weight = ctx.Input<Tensor>("Weight");
+    auto* bias = ctx.Input<Tensor>("Bias");
+
+    auto* hidden_out = ctx.Input<LoDTensor>("Hidden");
+    auto* cell_out = ctx.Input<LoDTensor>("Cell");
+
+    auto* batch_gate = ctx.Input<LoDTensor>("BatchGate");
+    auto* batch_cell_pre_act = ctx.Input<LoDTensor>("BatchCellPreAct");
+
+    auto* hidden_g = ctx.Input<LoDTensor>(framework::GradVarName("Hidden"));
+
+    auto* in_g = ctx.Output<LoDTensor>(framework::GradVarName("Input"));
+    auto* weight_g = ctx.Output<Tensor>(framework::GradVarName("Weight"));
+    auto* bias_g = ctx.Output<Tensor>(framework::GradVarName("Bias"));
+
+    auto& device_ctx = ctx.device_context();
+    math::SetConstant<Place, T> zero;
+    if (weight_g) {
+      weight_g->mutable_data<T>(ctx.GetPlace());
+      zero(device_ctx, weight_g, static_cast<T>(0.0));
+    }
+
+    auto in_dims = input->dims();
+    auto out_dims = hidden_g->dims();
+    int frame_size = static_cast<int>(in_dims[1] / 4);
+    PADDLE_ENFORCE_EQ(frame_size, out_dims[1]);
+
+    math::LstmMetaValue<T> lstm_value;
+    if (bias) {
+      T* bias_data = const_cast<T*>(bias->data<T>());
+      lstm_value.checkIg = bias_data + 4 * frame_size;
+      lstm_value.checkFg = lstm_value.checkIg + frame_size;
+      lstm_value.checkOg = lstm_value.checkFg + frame_size;
+    } else {
+      lstm_value.checkIg = nullptr;
+      lstm_value.checkFg = nullptr;
+      lstm_value.checkOg = nullptr;
+    }
+
+    math::LstmMetaGrad<T> lstm_grad;
+    if (bias && bias_g) {
+      T* bias_g_data = const_cast<T*>(bias_g->mutable_data<T>(ctx.GetPlace()));
+      zero(device_ctx, bias_g, static_cast<T>(0.0));
+      lstm_grad.checkIgGrad = bias_g_data + 4 * frame_size;
+      lstm_grad.checkFgGrad = lstm_grad.checkIgGrad + frame_size;
+      lstm_grad.checkOgGrad = lstm_grad.checkFgGrad + frame_size;
+    } else {
+      lstm_grad.checkIgGrad = nullptr;
+      lstm_grad.checkFgGrad = nullptr;
+      lstm_grad.checkOgGrad = nullptr;
+    }
+
+    math::LoDTensor2BatchFunctor<Place, T> to_batch;
+
+    // use the local variable as here.
+    LoDTensor batch_hidden;
+    batch_hidden.mutable_data<T>(out_dims, ctx.GetPlace());
+    batch_hidden.set_lod(batch_gate->lod());
+    to_batch(device_ctx, *hidden_out, batch_hidden, false);
+
+    LoDTensor batch_hidden_g;
+    batch_hidden_g.mutable_data<T>(out_dims, ctx.GetPlace());
+    batch_hidden_g.set_lod(batch_gate->lod());
+    to_batch(device_ctx, *hidden_g, batch_hidden_g, false);
+
+    LoDTensor batch_cell;
+    batch_cell.mutable_data<T>(out_dims, ctx.GetPlace());
+    batch_cell.set_lod(batch_gate->lod());
+    to_batch(device_ctx, *cell_out, batch_cell, false);
+
+    LoDTensor batch_cell_g;
+    batch_cell_g.mutable_data<T>(out_dims, ctx.GetPlace());
+    batch_cell_g.set_lod(batch_gate->lod());
+    // TODO(qingqing) support the case output cell has gradient.
+    // to_batch(device_ctx, *cell_g, batch_cell_g, false);
+    zero(device_ctx, &batch_cell_g, static_cast<T>(0.0));
+
+    LoDTensor batch_gate_g;
+    batch_gate_g.mutable_data<T>(batch_gate->dims(), ctx.GetPlace());
+    batch_gate_g.set_lod(batch_gate->lod());
+
+    auto gate_act = ctx.Attr<std::string>("gateActivation");
+    auto cell_act = ctx.Attr<std::string>("cellActivation");
+    auto cand_act = ctx.Attr<std::string>("candidateActivation");
+
+    auto batch_starts = batch_gate->lod()[0];
+    size_t num_batch = batch_starts.size() - 1;
+    for (int n = static_cast<int>(num_batch) - 1; n >= 0; n--) {
+      int bstart = static_cast<int>(batch_starts[n]);
+      int bend = static_cast<int>(batch_starts[n + 1]);
+
+      Tensor gate = batch_gate->Slice(bstart, bend);
+      Tensor cell = batch_cell.Slice(bstart, bend);
+      Tensor cell_pre_act = batch_cell_pre_act->Slice(bstart, bend);
+      lstm_value.gateValue = gate.data<T>();
+      lstm_value.stateValue = cell.data<T>();
+      lstm_value.stateActiveValue = cell_pre_act.data<T>();
+
+      Tensor out_g = batch_hidden_g.Slice(bstart, bend);
+      Tensor gate_g = batch_gate_g.Slice(bstart, bend);
+      Tensor cell_g = batch_cell_g.Slice(bstart, bend);
+      lstm_grad.stateGrad = cell_g.data<T>();
+      lstm_grad.gateGrad = gate_g.data<T>();
+      lstm_grad.outputGrad = out_g.data<T>();
+
+      if (n) {
+        int bstart_pre = static_cast<int>(batch_starts[n - 1]);
+        Tensor cell_pre = batch_cell.Slice(bstart_pre, bstart);
+        Tensor cell_pre_g = batch_cell_g.Slice(bstart_pre, bstart);
+        lstm_value.prevStateValue = cell_pre.data<T>();
+        lstm_grad.prevStateGrad = cell_pre_g.data<T>();
+      } else {
+        lstm_value.prevStateValue = nullptr;
+        lstm_grad.prevStateGrad = nullptr;
+      }
+
+      int cur_batch_size = bend - bstart;
+      math::LstmUnitGradFunctor<Place, T>::compute(
+          device_ctx, lstm_value, lstm_grad, frame_size, cur_batch_size,
+          gate_act, cell_act, cand_act);
+
+      if (n != 0) {
+        int pre_h_start = static_cast<int>(batch_starts[n - 1]);
+        int pre_h_end = pre_h_start + cur_batch_size;
+        auto pre_hidden_g = batch_hidden_g.Slice(pre_h_start, pre_h_end);
+        math::matmul<Place, T>(device_ctx, gate_g, false, *weight, true,
+                               static_cast<T>(1.0), &pre_hidden_g,
+                               static_cast<T>(1.0));
+        if (weight_g) {
+          /* backward weight */
+          auto pre_hidden = batch_hidden.Slice(pre_h_start, pre_h_end);
+          math::matmul<Place, T>(device_ctx, pre_hidden, true, gate_g, false,
+                                 static_cast<T>(1.0), weight_g,
+                                 static_cast<T>(1.0));
+        }
+      }
+    }
+
+    math::Batch2LoDTensorFunctor<Place, T> to_seq;
+    if (in_g) {
+      /* backward data */
+      in_g->mutable_data<T>(ctx.GetPlace());
+      to_seq(device_ctx, batch_gate_g, *in_g);
+    }
+    if (bias && bias_g) {
+      /* backward bias */
+      int m = static_cast<int>(batch_gate_g.dims()[0]);
+      int n = static_cast<int>(batch_gate_g.dims()[1]);
+
+      Tensor ones;
+      ones.mutable_data<T>({m}, ctx.GetPlace());
+      math::SetConstant<Place, T> set;
+      set(device_ctx, &ones, static_cast<T>(1.0));
+
+      math::gemv<Place, T>(device_ctx, true, m, n, 1., batch_gate_g.data<T>(),
+                           ones.data<T>(), 0., bias_g->data<T>());
+    }
+  }
 };
 
 }  // namespace operators

paddle/operators/math/detail/lstm_cpu_kernel.h

@@ -26,10 +26,7 @@ namespace detail {
 
 template <class T, class Op>
 void naive_lstm_forward_one_sequence(Op op, LstmMetaValue<T> value,
-                                     int frameSize,
-                                     activation_mode_t active_node,
-                                     activation_mode_t active_gate,
-                                     activation_mode_t active_state) {
+                                     int frameSize) {
   T rValueIn;
   T rValueIg;
   T rValueFg;
@@ -60,10 +57,8 @@ void naive_lstm_forward_one_sequence(Op op, LstmMetaValue<T> value,
       rPrevState = value.prevStateValue[i];
     }
 
-    hppl::cpu::ForwardAct<T> act;
     op(rValueIn, rValueIg, rValueFg, rValueOg, rPrevState, rState, rStateAtv,
-       rOut, rCheckI, rCheckF, rCheckO, act(active_node), act(active_gate),
-       act(active_state));
+       rOut, rCheckI, rCheckF, rCheckO);
 
     valueIn[i] = rValueIn;
     valueIg[i] = rValueIg;
@@ -77,10 +72,7 @@ void naive_lstm_forward_one_sequence(Op op, LstmMetaValue<T> value,
 
 template <class T, class Op>
 void naive_lstm_backward_one_sequence(Op op, LstmMetaValue<T> value,
-                                      LstmMetaGrad<T> grad, int frameSize,
-                                      activation_mode_t active_node,
-                                      activation_mode_t active_gate,
-                                      activation_mode_t active_state) {
+                                      LstmMetaGrad<T> grad, int frameSize) {
   T rValueIn;
   T rValueIg;
   T rValueFg;
@@ -127,11 +119,10 @@ void naive_lstm_backward_one_sequence(Op op, LstmMetaValue<T> value,
       rPrevState = value.prevStateValue[i];
     }
 
-    hppl::cpu::BackwardAct<T> act;
     op(rValueIn, rValueIg, rValueFg, rValueOg, rGradIn, rGradIg, rGradFg,
        rGradOg, rPrevState, rPrevStateGrad, rState, rStateGrad, rStateAtv,
        rOutputGrad, rCheckI, rCheckF, rCheckO, rCheckIGrad, rCheckFGrad,
-       rCheckOGrad, act(active_node), act(active_gate), act(active_state));
+       rCheckOGrad);
 
     gradIn[i] = rGradIn;
     gradIg[i] = rGradIg;
@@ -283,8 +274,7 @@ void cpu_lstm_forward(Op op, LstmMetaValue<T> value, int frameSize,
     avx_lstm_forward_one_sequence<T>(op, value, frameSize, active_node,
                                      active_gate, active_state);
   } else {
-    naive_lstm_forward_one_sequence<T>(op, value, frameSize, active_node,
-                                       active_gate, active_state);
+    naive_lstm_forward_one_sequence<T>(op, value, frameSize);
   }
 }
 
@@ -297,8 +287,7 @@ void cpu_lstm_backward(Op op, LstmMetaValue<T> value, LstmMetaGrad<T> grad,
     avx_lstm_backward_one_sequence<T>(op, value, grad, frameSize, active_node,
                                       active_gate, active_state);
   } else {
-    naive_lstm_backward_one_sequence<T>(op, value, grad, frameSize, active_node,
-                                        active_gate, active_state);
+    naive_lstm_backward_one_sequence<T>(op, value, grad, frameSize);
   }
 }
 

paddle/operators/math/detail/lstm_gpu_kernel.h

@@ -32,9 +32,7 @@ namespace detail {
  */
 template <class T, class Op, bool isBatch>
 __global__ void KeLstmForward(Op op, LstmMetaValue<T> value, int frameSize,
-                              int batchSize, activation_mode_t active_node,
-                              activation_mode_t active_gate,
-                              activation_mode_t active_state) {
+                              int batchSize) {
   const int frameIdx = blockIdx.x * blockDim.x + threadIdx.x;
   if (frameIdx >= frameSize) return;
 
@@ -70,10 +68,8 @@ __global__ void KeLstmForward(Op op, LstmMetaValue<T> value, int frameSize,
     rPrevState = value.prevStateValue[frameIdx];
   }
 
-  hppl::gpu::ForwardAct<T> act;
   op(rValueIn, rValueIg, rValueFg, rValueOg, rPrevState, rState, rStateAtv,
-     rOut, rCheckI, rCheckF, rCheckO, act(active_node), act(active_gate),
-     act(active_state));
+     rOut, rCheckI, rCheckF, rCheckO);
 
   value.gateValue[frameIdx] = rValueIn;
   value.gateValue[frameIdx + frameSize] = rValueIg;
@@ -92,9 +88,7 @@ __global__ void KeLstmForward(Op op, LstmMetaValue<T> value, int frameSize,
 template <class T, class Op, bool isBatch>
 __global__ void KeLstmBackward(Op op, LstmMetaValue<T> value,
                                LstmMetaGrad<T> grad, int frameSize,
-                               int batchSize, activation_mode_t active_node,
-                               activation_mode_t active_gate,
-                               activation_mode_t active_state) {
+                               int batchSize) {
   const int frameIdx = blockIdx.x * blockDim.x + threadIdx.x;
   if (frameIdx >= frameSize) return;
 
@@ -145,11 +139,9 @@ __global__ void KeLstmBackward(Op op, LstmMetaValue<T> value,
     rPrevState = value.prevStateValue[frameIdx];
   }
 
-  hppl::gpu::BackwardAct<T> act;
   op(rValueIn, rValueIg, rValueFg, rValueOg, rGradIn, rGradIg, rGradFg, rGradOg,
      rPrevState, rPrevStateGrad, rState, rStateGrad, rStateAtv, rOutputGrad,
-     rCheckI, rCheckF, rCheckO, rCheckIGrad, rCheckFGrad, rCheckOGrad,
-     act(active_node), act(active_gate), act(active_state));
+     rCheckI, rCheckF, rCheckO, rCheckIGrad, rCheckFGrad, rCheckOGrad);
 
   grad.gateGrad[frameIdx] = rGradIn;
   grad.gateGrad[frameIdx + frameSize] = rGradIg;
@@ -205,13 +197,11 @@ void gpu_lstm_forward(const platform::DeviceContext& context, Op op,
   if (batchSize == 1) {
     KeLstmForward<T, Op,
                   /* isBatch= */ false><<<grid, threads, 0, stream>>>(
-        op, value, frameSize, batchSize, active_node, active_gate,
-        active_state);
+        op, value, frameSize, batchSize);
   } else {
     KeLstmForward<T, Op,
                   /* isBatch= */ true><<<grid, threads, 0, stream>>>(
-        op, value, frameSize, batchSize, active_node, active_gate,
-        active_state);
+        op, value, frameSize, batchSize);
   }
 }
 
@@ -240,13 +230,11 @@ void gpu_lstm_backward(const platform::DeviceContext& context, Op op,
   if (batchSize == 1) {
     KeLstmBackward<T, Op,
                    /* isBatch= */ false><<<grid, threads, 0, stream>>>(
-        op, value, grad, frameSize, batchSize, active_node, active_gate,
-        active_state);
+        op, value, grad, frameSize, batchSize);
   } else {
     KeLstmBackward<T, Op,
                    /* isBatch= */ true><<<grid, threads, 0, stream>>>(
-        op, value, grad, frameSize, batchSize, active_node, active_gate,
-        active_state);
+        op, value, grad, frameSize, batchSize);
   }
 }
 

paddle/operators/math/detail/lstm_kernel.h

@@ -24,15 +24,29 @@ namespace detail {
 
 namespace forward {
 
+template <typename T>
+DEVICE inline T sigmoid(const T a) {
+  const T min = SIGMOID_THRESHOLD_MIN;
+  const T max = SIGMOID_THRESHOLD_MAX;
+  T tmp = (a < min) ? min : ((a > max) ? max : a);
+  return static_cast<T>(1.0) / (static_cast<T>(1.0) + exp(-tmp));
+}
+
+template <typename T>
+DEVICE inline T tanh(const T a) {
+  T tmp = -2.0 * a;
+  tmp = (tmp > EXP_MAX_INPUT) ? EXP_MAX_INPUT : tmp;
+  return (2.0 / (1.0 + exp(tmp))) - 1.0;
+}
+
 template <class T>
 class lstm {
  public:
   HOSTDEVICE void operator()(T &valueIn, T &valueIg, T &valueFg, T &valueOg,
                              T &prevState, T &state, T &stateAtv, T &output,
-                             T &checkI, T &checkF, T &checkO,
-                             typename hppl::ForwardActType<T>::type actInput,
-                             typename hppl::ForwardActType<T>::type actGate,
-                             typename hppl::ForwardActType<T>::type actState) {
+                             T &checkI, T &checkF, T &checkO) {
+#if 0
+    // TODO(qingqing) support to activation speficed by users
     valueIn = actInput(valueIn);
     valueIg = actGate(valueIg + prevState * checkI);
     valueFg = actGate(valueFg + prevState * checkF);
@@ -40,6 +54,15 @@ class lstm {
     valueOg = actGate(valueOg + state * checkO);
     stateAtv = actState(state);
     output = valueOg * stateAtv;
+#else
+    valueIn = tanh<T>(valueIn);
+    valueIg = sigmoid<T>(valueIg + prevState * checkI);
+    valueFg = sigmoid<T>(valueFg + prevState * checkF);
+    state = valueIn * valueIg + prevState * valueFg;
+    valueOg = sigmoid<T>(valueOg + state * checkO);
+    stateAtv = tanh<T>(state);
+    output = valueOg * stateAtv;
+#endif
   }
 #ifndef __NVCC__
 #ifndef __AVX__  // If not compiled with AVX instructs. Disable AVX by default
@@ -72,6 +95,16 @@ class lstm {
 
 namespace backward {
 
+template <typename T>
+DEVICE inline T sigmoid(const T a, const T b) {
+  return a * b * (1.0 - b);
+}
+
+template <typename T>
+DEVICE inline T tanh(const T a, const T b) {
+  return a * (1.0 - b * b);
+}
+
 template <class T>
 class lstm {
  public:
@@ -80,10 +113,9 @@ class lstm {
                              T &prevState, T &prevStateGrad, T &state,
                              T &stateGrad, T &stateAtv, T &outputGrad,
                              T &checkI, T &checkF, T &checkO, T &checkIGrad,
-                             T &checkFGrad, T &checkOGrad,
-                             typename hppl::BackwardActType<T>::type actInput,
-                             typename hppl::BackwardActType<T>::type actGate,
-                             typename hppl::BackwardActType<T>::type actState) {
+                             T &checkFGrad, T &checkOGrad) {
+#if 0
+    // TODO(qingqing) support to activation speficed by users
     gradOg = actGate(outputGrad * stateAtv, valueOg);
     stateGrad += actState(outputGrad * valueOg, stateAtv) + gradOg * checkO;
     gradIn = actInput(stateGrad * valueIg, valueIn);
@@ -93,6 +125,17 @@ class lstm {
     checkIGrad = gradIg * prevState;
     checkFGrad = gradFg * prevState;
     checkOGrad = gradOg * state;
+#else
+    gradOg = sigmoid<T>(outputGrad * stateAtv, valueOg);
+    stateGrad += tanh<T>(outputGrad * valueOg, stateAtv) + gradOg * checkO;
+    gradIn = tanh<T>(stateGrad * valueIg, valueIn);
+    gradIg = sigmoid<T>(stateGrad * valueIn, valueIg);
+    gradFg = sigmoid<T>(stateGrad * prevState, valueFg);
+    prevStateGrad = gradIg * checkI + gradFg * checkF + stateGrad * valueFg;
+    checkIGrad = gradIg * prevState;
+    checkFGrad = gradFg * prevState;
+    checkOGrad = gradOg * state;
+#endif
   }
 #ifndef __NVCC__
 #ifndef __AVX__  // If not compiled with AVX instructs. Disable AVX by default

paddle/operators/math/math_function.cc

@@ -211,6 +211,26 @@ void batched_gemm<platform::CPUPlace, double>(
 }
 #endif
 
+template <>
+void gemv<platform::CPUPlace, float>(const platform::DeviceContext& context,
+                                     const bool trans_a, const int M,
+                                     const int N, const float alpha,
+                                     const float* A, const float* B,
+                                     const float beta, float* C) {
+  CBLAS_TRANSPOSE transA = (trans_a == false) ? CblasNoTrans : CblasTrans;
+  cblas_sgemv(CblasRowMajor, transA, M, N, alpha, A, N, B, 1, beta, C, 1);
+}
+
+template <>
+void gemv<platform::CPUPlace, double>(const platform::DeviceContext& context,
+                                      const bool trans_a, const int M,
+                                      const int N, const double alpha,
+                                      const double* A, const double* B,
+                                      const double beta, double* C) {
+  CBLAS_TRANSPOSE transA = (trans_a == false) ? CblasNoTrans : CblasTrans;
+  cblas_dgemv(CblasRowMajor, transA, M, N, alpha, A, N, B, 1, beta, C, 1);
+}
+
 template struct SetConstant<platform::CPUPlace, float>;
 
 }  // namespace math

paddle/operators/math/math_function.cu

@@ -203,6 +203,33 @@ void batched_gemm<platform::GPUPlace, double>(
       &beta, C, ldc, strideC, batchCount));
 }
 
+template <>
+void gemv<platform::GPUPlace, float>(const platform::DeviceContext& context,
+                                     const bool trans_a, const int M,
+                                     const int N, const float alpha,
+                                     const float* A, const float* B,
+                                     const float beta, float* C) {
+  cublasOperation_t cuTransA = (trans_a == false) ? CUBLAS_OP_T : CUBLAS_OP_N;
+
+  PADDLE_ENFORCE(platform::dynload::cublasSgemv(
+      reinterpret_cast<const platform::CUDADeviceContext&>(context)
+          .cublas_handle(),
+      cuTransA, N, M, &alpha, A, N, B, 1, &beta, C, 1));
+}
+
+template <>
+void gemv<platform::GPUPlace, double>(const platform::DeviceContext& context,
+                                      const bool trans_a, const int M,
+                                      const int N, const double alpha,
+                                      const double* A, const double* B,
+                                      const double beta, double* C) {
+  cublasOperation_t cuTransA = (trans_a == false) ? CUBLAS_OP_T : CUBLAS_OP_N;
+  PADDLE_ENFORCE(platform::dynload::cublasDgemv(
+      reinterpret_cast<const platform::CUDADeviceContext&>(context)
+          .cublas_handle(),
+      cuTransA, N, M, &alpha, A, N, B, 1, &beta, C, 1));
+}
+
 template struct SetConstant<platform::GPUPlace, float>;
 
 }  // namespace math

paddle/operators/math/math_function.h

@@ -93,6 +93,11 @@ void batched_gemm(const platform::DeviceContext& context,
                   const T* A, const T* B, const T beta, T* C,
                   const int batchCount, const int strideA, const int strideB);
 
+template <typename Place, typename T>
+void gemv(const platform::DeviceContext& context, const bool trans_a,
+          const int M, const int N, const T alpha, const T* A, const T* B,
+          const T beta, T* C);
+
 template <typename Place, typename T>
 struct SetConstant {
   void operator()(const platform::DeviceContext& context,

paddle/operators/math/math_function_test.cc

@@ -89,3 +89,53 @@ TEST(math_function, zero) {
   EXPECT_EQ(t[2], 1);
   EXPECT_EQ(t[3], 1);
 }
+
+template <typename T>
+void GemvTest(int m, int n, bool trans) {
+  paddle::framework::Tensor mat_a;
+  paddle::framework::Tensor vec_b;
+  paddle::framework::Tensor vec_c;
+  auto* cpu_place = new paddle::platform::CPUPlace();
+  int b_num = trans ? m : n;
+  int c_num = trans ? n : m;
+
+  T* data_a = mat_a.mutable_data<T>({m, n}, *cpu_place);
+  T* data_b = vec_b.mutable_data<T>({b_num}, *cpu_place);
+  T* data_c = vec_c.mutable_data<T>({c_num}, *cpu_place);
+  for (int i = 0; i < mat_a.numel(); ++i) {
+    data_a[i] = static_cast<T>(i);
+  }
+  for (int i = 0; i < vec_b.numel(); ++i) {
+    data_b[i] = static_cast<T>(i);
+  }
+
+  paddle::platform::CPUDeviceContext context(*cpu_place);
+  paddle::operators::math::gemv<paddle::platform::CPUPlace, T>(
+      context, trans, static_cast<int>(m), static_cast<int>(n), 1., data_a,
+      data_b, 0., data_c);
+
+  if (!trans) {
+    for (int i = 0; i < m; ++i) {
+      T sum = 0.0;
+      for (int j = 0; j < n; ++j) {
+        sum += data_a[i * n + j] * data_b[j];
+      }
+      ASSERT_FLOAT_EQ(data_c[i], sum);
+    }
+  } else {
+    for (int i = 0; i < n; ++i) {
+      T sum = 0.0;
+      for (int j = 0; j < m; ++j) {
+        sum += data_a[j * n + i] * data_b[j];
+      }
+      ASSERT_FLOAT_EQ(data_c[i], sum);
+    }
+  }
+}
+
+TEST(math_function, gemv) {
+  GemvTest<float>(3, 13, false);
+  GemvTest<double>(4, 5, false);
+  GemvTest<float>(12, 7, true);
+  GemvTest<double>(7, 9, true);
+}

paddle/operators/math/math_function_test.cu

@@ -177,3 +177,65 @@ TEST(math_function, gemm_trans_cublas) {
   EXPECT_EQ(input3_ptr[7], 99);
   delete gpu_place;
 }
+
+template <typename T>
+void GemvTest(int m, int n, bool trans) {
+  paddle::framework::Tensor mat_a;
+  paddle::framework::Tensor vec_b;
+  paddle::framework::Tensor vec_c;
+  auto* cpu_place = new paddle::platform::CPUPlace();
+
+  T* data_a = mat_a.mutable_data<T>({m, n}, *cpu_place);
+  T* data_b = vec_b.mutable_data<T>({trans ? m : n}, *cpu_place);
+  T* data_c = vec_c.mutable_data<T>({trans ? n : m}, *cpu_place);
+
+  auto* gpu_place = new paddle::platform::GPUPlace(0);
+  paddle::framework::Tensor g_mat_a;
+  paddle::framework::Tensor g_vec_b;
+  paddle::framework::Tensor g_vec_c;
+  T* g_data_a = g_mat_a.mutable_data<T>(mat_a.dims(), *gpu_place);
+  T* g_data_b = g_vec_b.mutable_data<T>(vec_b.dims(), *gpu_place);
+  T* g_data_c = g_vec_c.mutable_data<T>(vec_c.dims(), *gpu_place);
+
+  for (int i = 0; i < mat_a.numel(); ++i) {
+    data_a[i] = static_cast<T>(i);
+  }
+  for (int i = 0; i < vec_b.numel(); ++i) {
+    data_b[i] = static_cast<T>(i);
+  }
+
+  paddle::platform::CUDADeviceContext context(*gpu_place);
+  g_mat_a.CopyFrom(mat_a, *gpu_place, context);
+  g_vec_b.CopyFrom(vec_b, *gpu_place, context);
+
+  paddle::operators::math::gemv<paddle::platform::GPUPlace, T>(
+      context, trans, static_cast<int>(m), static_cast<int>(n), 1., g_data_a,
+      g_data_b, 0., g_data_c);
+
+  vec_c.CopyFrom(g_vec_c, paddle::platform::CPUPlace(), context);
+
+  if (!trans) {
+    for (int i = 0; i < m; ++i) {
+      T sum = 0.0;
+      for (int j = 0; j < n; ++j) {
+        sum += data_a[i * n + j] * data_b[j];
+      }
+      ASSERT_FLOAT_EQ(data_c[i], sum);
+    }
+  } else {
+    for (int i = 0; i < n; ++i) {
+      T sum = 0.0;
+      for (int j = 0; j < m; ++j) {
+        sum += data_a[j * n + i] * data_b[j];
+      }
+      ASSERT_FLOAT_EQ(data_c[i], sum);
+    }
+  }
+}
+
+TEST(math_function, gemv) {
+  GemvTest<float>(3, 13, false);
+  GemvTest<double>(3, 13, false);
+  GemvTest<float>(3, 13, true);
+  GemvTest<double>(3, 13, true);
+}

paddle/operators/math/sequence2batch.h

@@ -53,7 +53,18 @@ class LoDTensor2BatchFunctor {
  public:
   void operator()(const platform::DeviceContext& context,
                   const framework::LoDTensor& lod_tensor,
-                  framework::LoDTensor& batch, bool is_reverse) const {
+                  framework::LoDTensor& batch, bool is_cal_batch_lod,
+                  bool is_reverse = false) const {
+    if (!is_cal_batch_lod) {
+      auto lods = batch.lod();
+      PADDLE_ENFORCE_EQ(lods.size(), 2UL);
+      PADDLE_ENFORCE_EQ(lods[1].size(),
+                        static_cast<size_t>(lod_tensor.dims()[0]));
+      CopyMatrixRowsFunctor<Place, T> to_batch;
+      to_batch(context, lod_tensor, lods[1].data(), batch, true);
+      return;
+    }
+
     auto lods = lod_tensor.lod();
     PADDLE_ENFORCE_EQ(lods.size(), 1UL, "Only support one level sequence now.");
     auto lod = lods[0];
@@ -101,10 +112,10 @@ class LoDTensor2BatchFunctor {
     size_t* batch_starts = batch_lods[0].data();
     size_t* seq2batch_idx = batch_lods[1].data();
     batch_starts[0] = 0;
-    for (size_t n = 0; n < num_batch; n++) {
+    for (int n = 0; n < num_batch; n++) {
       auto batch_id = static_cast<int>(batch_starts[n]);
       for (size_t i = 0; i < seq_info.size(); ++i) {
-        size_t seq_len = seq_info[i].length;
+        int seq_len = seq_info[i].length;
         int start = seq_info[i].start;
         if (n < seq_len) {
           seq2batch_idx[batch_id] =
@@ -132,11 +143,8 @@ class Batch2LoDTensorFunctor {
     auto in_lod = batch.lod();
     PADDLE_ENFORCE_EQ(in_lod.size(), 2UL,
                       "The LoD size of input `batch` should be 2.");
-    auto out_lod = lod_tensor.lod()[0];
-    auto num = out_lod[out_lod.size() - 1];
-    PADDLE_ENFORCE_EQ(num, lod_tensor.dims()[0]);
-    PADDLE_ENFORCE_EQ(num, in_lod[1].size());
-    PADDLE_ENFORCE_EQ(num, batch.dims()[0]);
+    PADDLE_ENFORCE_EQ(in_lod[1].size(),
+                      static_cast<size_t>(lod_tensor.dims()[0]));
     CopyMatrixRowsFunctor<Place, T> to_seq;
     size_t* index = in_lod[1].data();
     to_seq(context, batch, index, lod_tensor, false);

paddle/operators/nccl_op_test.cu

@@ -185,7 +185,7 @@ TEST_F(NCCLTester, ncclAllReduceOp) {
         recv_tensor.numel() * sizeof(float),
         static_cast<p::CUDADeviceContext *>(dev_ctxs[i])->stream());
 
-    for (size_t j = 0; j < f::product(kDims); ++j) {
+    for (int64_t j = 0; j < f::product(kDims); ++j) {
       ASSERT_NEAR(ct[j], result, 1e-5);
     }
   }
@@ -234,7 +234,7 @@ TEST_F(NCCLTester, ncclReduceOp) {
       recv_tensor.numel() * sizeof(float),
       static_cast<p::CUDADeviceContext *>(dev_ctxs[kRoot])->stream());
 
-  for (int j = 0; j < f::product(kDims); ++j) {
+  for (int64_t j = 0; j < f::product(kDims); ++j) {
     ASSERT_NEAR(ct[j], result, 1e-5);
   }
 }
@@ -282,7 +282,7 @@ TEST_F(NCCLTester, ncclBcastOp) {
       recv_tensor.numel() * sizeof(float),
       static_cast<p::CUDADeviceContext *>(dev_ctxs[idx])->stream());
 
-  for (size_t j = 0; j < f::product(kDims); ++j) {
+  for (int64_t j = 0; j < f::product(kDims); ++j) {
     ASSERT_NEAR(ct[j], result, 1e-5);
   }
 }

paddle/operators/reshape_op.cc

@@ -36,7 +36,7 @@ class ReshapeOp : public framework::OperatorWithKernel {
     PADDLE_ENFORCE(shape.size() > 0, "Attr(shape) shouldn't be empty.");
     auto x_dims = ctx->GetInputDim("X");
     // TODO(qiao) change batch_size
-    for (int i = 1; i < shape.size(); ++i) {
+    for (size_t i = 1; i < shape.size(); ++i) {
       PADDLE_ENFORCE(shape[i] > 0,
                      "Each dimension of shape "
                      "must be positiv except the first.");

paddle/operators/save_load_op_test.cc

@@ -34,7 +34,7 @@ TEST(SaveLoadOp, CPU) {
 
   tensor->set_lod(expect_lod);
   int* expect = tensor->mutable_data<int>(place);
-  for (size_t i = 0; i < paddle::framework::product(tensor->dims()); ++i) {
+  for (int64_t i = 0; i < tensor->numel(); ++i) {
     expect[i] = static_cast<int>(i);
   }
   paddle::framework::AttributeMap attrs;
@@ -50,7 +50,7 @@ TEST(SaveLoadOp, CPU) {
       "load", {}, {{"Out", {"out_var"}}}, attrs);
   load_op->Run(scope, ctx);
   int* actual = target->data<int>();
-  for (size_t i = 0; i < paddle::framework::product(tensor->dims()); ++i) {
+  for (int64_t i = 0; i < tensor->numel(); ++i) {
     EXPECT_EQ(expect[i], actual[i]);
   }
   auto& actual_lod = target->lod();
@@ -60,4 +60,4 @@ TEST(SaveLoadOp, CPU) {
       EXPECT_EQ(expect_lod[i][j], actual_lod[i][j]);
     }
   }
-}
\ No newline at end of file
+}

paddle/operators/sequence_pool_op.cc

@@ -39,15 +39,14 @@ class SequencePoolOpMaker : public framework::OpProtoAndCheckerMaker {
     AddOutput("Out",
               "(Tensor), output of SequencePoolOp, which does not contain LoD "
               "infomation.");
-    AddAttr<int>(
-        "strategy",
-        "(int, default AVERAGE) the pooling strategy of SequencePoolOp.")
-        .SetDefault(AVERAGE)
-        .InEnum({AVERAGE, SUM, SQRT, MAX, LAST, FIRST});
+    AddAttr<std::string>(
+        "pooltype",
+        "(int, default AVERAGE) the pooling pooltype of SequencePoolOp.")
+        .SetDefault("AVERAGE");
     AddComment(R"DOC(
     SequencePoolOp pools features of all time-steps of each instance.
 
-    It supports six pooling strategy:
+    It supports six pooling pooltype:
     - AVERAGE: Out[i] = average_{for each instance in i-th sequence}{X[i]}
     - SUM:     Out[i] = sum_{for each instance in i-th sequence}{X[i]}
     - SQRT:    Out[i] = sum_{for each instance in i-th sequence}{X[i]} 
@@ -63,7 +62,7 @@ class SequencePoolOpMaker : public framework::OpProtoAndCheckerMaker {
     and the value of X = [[1, 3], [2, 4, 6], [5, 1]].
 
     Thus, Out is a [3,1,1] Tensor without LoD infomation.
-    And for different strategy, the value of Out is as follows:
+    And for different pooltype, the value of Out is as follows:
 
     - AVERAGE: [2, 4, 3], where 2=(1+3)/2, 4=(2+4+6)/3, 3=(5+1)/2
     - SUM: [4, 12, 6], where 4=1+3, 12=2+4+6, 6=5+1

paddle/operators/sequence_pool_op.h

@@ -29,22 +29,13 @@ template <typename T, int MajorType = Eigen::RowMajor,
           typename IndexType = Eigen::DenseIndex>
 using EigenMatrix = framework::EigenMatrix<T, MajorType, IndexType>;
 
-enum SeqPoolType {
-  AVERAGE = 0,
-  SUM = 1,
-  SQRT = 2,  // square_root_n
-  MAX = 3,
-  LAST = 4,
-  FIRST = 5
-};
-
 template <typename Place, typename T>
 class SequencePoolKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& context) const override {
     auto* in = context.Input<LoDTensor>("X");
     auto* out = context.Output<LoDTensor>("Out");
-    int strategy = context.Attr<int>("strategy");
+    std::string pooltype = context.Attr<std::string>("pooltype");
 
     auto dims = in->dims();
     auto lod = in->lod();
@@ -71,28 +62,21 @@ class SequencePoolKernel : public framework::OpKernel<T> {
       auto in_e = EigenMatrix<T>::From(in_t, framework::make_ddim({h, w}));
       auto out_e = EigenVector<T>::Flatten(out_t);
 
-      switch (strategy) {
-        case AVERAGE:
-          out_e.device(place) = in_e.mean(Eigen::array<int, 1>({{0}}));
-          break;
-        case SUM:
-          out_e.device(place) = in_e.sum(Eigen::array<int, 1>({{0}}));
-          break;
-        case SQRT:
-          out_e.device(place) = in_e.sum(Eigen::array<int, 1>({{0}})) /
-                                std::sqrt(static_cast<T>(h));
-          break;
-        case MAX:
-          out_e.device(place) = in_e.maximum(Eigen::array<int, 1>({{0}}));
-          break;
-        case LAST:
-          out_e.device(place) = in_e.chip(h - 1, 0);
-          break;
-        case FIRST:
-          out_e.device(place) = in_e.chip(0, 0);
-          break;
-        default:
-          PADDLE_THROW("unsupported pooling strategy");
+      if (pooltype == "AVERAGE") {
+        out_e.device(place) = in_e.mean(Eigen::array<int, 1>({{0}}));
+      } else if (pooltype == "SUM") {
+        out_e.device(place) = in_e.sum(Eigen::array<int, 1>({{0}}));
+      } else if (pooltype == "SQRT") {
+        out_e.device(place) = in_e.sum(Eigen::array<int, 1>({{0}})) /
+                              std::sqrt(static_cast<T>(h));
+      } else if (pooltype == "MAX") {
+        out_e.device(place) = in_e.maximum(Eigen::array<int, 1>({{0}}));
+      } else if (pooltype == "LAST") {
+        out_e.device(place) = in_e.chip(h - 1, 0);
+      } else if (pooltype == "FIRST") {
+        out_e.device(place) = in_e.chip(0, 0);
+      } else {
+        PADDLE_THROW("unsupported pooling pooltype");
       }
     }
   }
@@ -105,15 +89,15 @@ class SequencePoolGradKernel : public framework::OpKernel<T> {
     auto* in = context.Input<LoDTensor>("X");
     auto* in_g = context.Output<LoDTensor>(framework::GradVarName("X"));
     auto* out_g = context.Input<LoDTensor>(framework::GradVarName("Out"));
-    int strategy = context.Attr<int>("strategy");
+    std::string pooltype = context.Attr<std::string>("pooltype");
 
     auto dims = in->dims();
     auto lod = in->lod()[0];
     int64_t w = in->numel() / dims[0];
 
     in_g->mutable_data<T>(context.GetPlace());
-    if (strategy == LAST || strategy == FIRST) {
-      // set X@Grad be zero at first when strategy is LAST/FIRST
+    if (pooltype == "LAST" || pooltype == "FIRST") {
+      // set X@Grad be zero at first when pooltype is LAST/FIRST
       math::SetConstant<Place, T> functor;
       functor(context.device_context(), in_g, 0);
     }
@@ -127,41 +111,33 @@ class SequencePoolGradKernel : public framework::OpKernel<T> {
       auto out_g_e = EigenMatrix<T>::From(out_g_t, {1, w});
       Eigen::DSizes<int, 2> bcast(h, 1);
 
-      switch (strategy) {
-        case AVERAGE:
-          in_g_e.device(place) = (out_g_e / static_cast<T>(h)).broadcast(bcast);
-          break;
-        case SUM:
-          in_g_e.device(place) = (out_g_e).broadcast(bcast);
-          break;
-        case SQRT:
-          in_g_e.device(place) =
-              (out_g_e / std::sqrt(static_cast<T>(h))).broadcast(bcast);
-          break;
-        case MAX: {
-          auto in_t =
-              in->Slice(static_cast<int>(lod[i]), static_cast<int>(lod[i + 1]));
-          Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic>>
-              in_t_map(in_t.data<T>(), h, w);
-          int row_id;
-          Eigen::array<int, 2> extents{{1, 1}};
-          for (int col_id = 0; col_id < w; col_id++) {
-            in_t_map.col(col_id).maxCoeff(&row_id);
-            Eigen::array<int, 2> in_offsets{{row_id, col_id}};
-            Eigen::array<int, 2> out_offsets{{0, col_id}};
-            in_g_e.slice(in_offsets, extents).device(place) =
-                out_g_e.slice(out_offsets, extents);
-          }
-          break;
+      if (pooltype == "AVERAGE") {
+        in_g_e.device(place) = (out_g_e / static_cast<T>(h)).broadcast(bcast);
+      } else if (pooltype == "SUM") {
+        in_g_e.device(place) = (out_g_e).broadcast(bcast);
+      } else if (pooltype == "SQRT") {
+        in_g_e.device(place) =
+            (out_g_e / std::sqrt(static_cast<T>(h))).broadcast(bcast);
+      } else if (pooltype == "MAX") {
+        auto in_t =
+            in->Slice(static_cast<int>(lod[i]), static_cast<int>(lod[i + 1]));
+        Eigen::Map<const Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic>>
+            in_t_map(in_t.data<T>(), h, w);
+        int row_id;
+        Eigen::array<int, 2> extents{{1, 1}};
+        for (int col_id = 0; col_id < w; col_id++) {
+          in_t_map.col(col_id).maxCoeff(&row_id);
+          Eigen::array<int, 2> in_offsets{{row_id, col_id}};
+          Eigen::array<int, 2> out_offsets{{0, col_id}};
+          in_g_e.slice(in_offsets, extents).device(place) =
+              out_g_e.slice(out_offsets, extents);
         }
-        case LAST:
-          in_g_e.chip(h - 1, 0).device(place) = out_g_e;
-          break;
-        case FIRST:
-          in_g_e.chip(0, 0).device(place) = out_g_e;
-          break;
-        default:
-          PADDLE_THROW("unsupported pooling strategy");
+      } else if (pooltype == "LAST") {
+        in_g_e.chip(h - 1, 0).device(place) = out_g_e;
+      } else if (pooltype == "FIRST") {
+        in_g_e.chip(0, 0).device(place) = out_g_e;
+      } else {
+        PADDLE_THROW("unsupported pooling pooltype");
       }
     }
   }

paddle/operators/softmax_with_cross_entropy_op.cc

@@ -32,9 +32,9 @@ class SoftmaxWithCrossEntropyOpMaker
     AddInput("Label",
              "(Tensor, default: Tensor<int>), The ground truth which is a 2-D "
              "tensor. "
-             "If softLable is set to 0, Label is a Tensor<int> with shape [N x "
-             "1]. "
-             "If softLable is set to 1, Label is a Tensor<float/double> "
+             "If softLabel is set to false, Label is a Tensor<int> with shape "
+             "[N x 1]."
+             "If softLabel is set to true, Label is a Tensor<float/double> "
              "with shape [N x K].");
     AddOutput(
         "Softmax",
@@ -60,19 +60,23 @@ Because this operators performs a softmax on logits internally, it expects
 unscaled logits. Please do not call this op with the output of softmax operator,
 which will produce incorrect results.
 
-This operators expects mutually exclusive hard labels, each sample in a batch
-is in exactly one class with probabilities 1. Each sample in the batch with one
-and only one label.
+When the attribute softLabel is set false, this operators expects mutually
+exclusive hard labels, each sample in a batch is in exactly one class with
+probabilities 1. Each sample in the batch with one and only one label.
 
 Equation:
 
 1) hard label (one-hot label)
 
-Loss_j = -\text{Logit}_{Label_j} + \log\left(\sum_{i=0}^{K}\exp(\text{Logit}_i)\right), j = 1, ..., K
+Loss_j = \f$ -\text{Logit}_{Label_j} +
+\log\left(\sum_{i=0}^{K}\exp(\text{Logit}_i)\right),
+j = 1, ..., K $\f
 
 2) soft label (a distribution over all classes)
 
-Loss_j = -\sum_{i=0}^{K}\text{Label}_i\left(\text{Logit}_i-\log\left(\sum_{i=0}^{K}\exp(\text{Logit}_i)\right)\right), j = 1,...,K
+Loss_j = \f$ -\sum_{i=0}^{K}\text{Label}_i\left(\text{Logit}_i -
+\log\left(\sum_{i=0}^{K}\exp(\text{Logit}_i)\right)\right),
+j = 1,...,K $\f
 
 )DOC");
   }

paddle/scripts/docker/build_android.sh

@@ -4,6 +4,10 @@ set -xe
 
 if [ $ANDROID_ABI == "arm64-v8a" ]; then
   ANDROID_ARCH=arm64
+  if [ $ANDROID_API -lt 21 ]; then
+    echo "Warning: arm64-v8a requires ANDROID_API >= 21."
+    ANDROID_API=21
+  fi
 else # armeabi, armeabi-v7a
   ANDROID_ARCH=arm
 fi

paddle/trainer/MergeModel.cpp

@@ -27,6 +27,13 @@ using namespace paddle;  // NOLINT
 using namespace std;     // NOLINT
 
 int main(int argc, char** argv) {
+  if (FLAGS_model_dir.empty() || FLAGS_config_file.empty() ||
+      FLAGS_model_file.empty()) {
+    LOG(INFO) << "Usage: ./paddle_merge_model --model_dir=pass-00000 "
+                 "--config_file=config.py --model_file=out.paddle";
+    return 0;
+  }
+
   initMain(argc, argv);
   initPython(argc, argv);
 

python/paddle/v2/framework/layers.py

@@ -351,32 +351,21 @@ def conv2d(input,
     return helper.append_activation(pre_act)
 
 
-def sequence_pool(input, pool_type, program=None, init_program=None):
-    # FIXME(dzh) : want to unify the argument of python layer
-    # function. So we ignore some unecessary attributes
-
-    ENUM_POOL_TYPE = dict({
-        "AVERAGE": 0,
-        "SUM": 1,
-        "SQRT": 2,
-        "MAX": 3,
-        "LAST": 4,
-        "FIRST": 5
-    })
+def sequence_pool(input, pool_type, **kwargs):
+    ENUM_POOL_TYPE = set(["MAX", "AVG", "SQRT", "LAST", "FIRST"])
     if pool_type.upper() not in ENUM_POOL_TYPE:
         raise ValueError("Unknown pool_type: '%s'. It can only be %s.",
-                         str(pool_type), " ".join(ENUM_POOL_TYPE.keys()))
+                         str(pool_type), " ".join(ENUM_POOL_TYPE))
 
-    helper = LayerHelper('sequence_pool', **locals())
+    helper = LayerHelper('sequence_pool', **kwargs)
     dtype = helper.input_dtype()
     pool_out = helper.create_tmp_variable(dtype)
 
-    # FIXME(dzh): strategy
     helper.append_op(
         type="sequence_pool",
         inputs={"X": [input]},
         outputs={"Out": [pool_out]},
-        attrs={"strategy": ENUM_POOL_TYPE[pool_type.upper()]})
+        attrs={"pooltype": pool_type.upper()})
 
     return pool_out
 

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

+import unittest
+import random
+import numpy as np
+
+from op_test import OpTest
+
+
+class LinearChainCrfForward(object):
+    def __init__(self, seq_start_positions, emission_weights, emission_row_max,
+                 emission_exps, transition_weights, transition_exps, labels):
+        self.tag_num = emission_weights.shape[1]
+        self.seq_num = len(seq_start_positions) - 1
+
+        self.seq_start_positions = seq_start_positions
+        self.labels = labels
+        self.x = emission_weights
+
+        self.x_row_max = emission_row_max
+        self.x_exps = emission_exps
+
+        # unnormalized logits of the transition weights for the start mark.
+        self.a = transition_weights[0, :]
+        self.a_exps = transition_exps[0, :]
+        # unnormalized logits of the transition weights for the end mark.
+        self.b = transition_weights[1, :]
+        self.b_exps = transition_exps[1, :]
+        # unnormalized logits of the transition weights for all the other tags.
+        self.w = transition_weights[2:, :]
+        self.w_exps = transition_exps[2:, :]
+
+        # The output of linear chain crf operator.
+        # alpha is a memo table in dynamic programming to caculate
+        # nomalization factor.
+        self.alpha = np.zeros(
+            (seq_start_positions[-1], self.tag_num), dtype="float64")
+        self.log_likelihood = np.zeros((self.seq_num, 1))
+
+    def _l1_norm(self, x):
+        s = np.sum(x)
+        x /= s
+        return s
+
+    def _forward_a_sequence(self, x, x_row_max, x_exps, label, alpha):
+        seq_len = x_row_max.shape[0]
+        log_likelihood = 0.
+
+        for i in range(self.tag_num):
+            alpha[0, i] = self.a_exps[i] * x_exps[0, i]
+        log_likelihood = -x_row_max[0] - np.log(self._l1_norm(alpha[0, :]))
+
+        # calculate the unnormalized logits of the normalization factor.
+        for k in range(1, seq_len):
+            for i in range(self.tag_num):
+                s = 0.
+                for j in range(self.tag_num):
+                    s += alpha[k - 1, j] * self.w_exps[j, i]
+                alpha[k, i] = x_exps[k, i] * s
+            log_likelihood -= x_row_max[k] + np.log(self._l1_norm(alpha[k, :]))
+        s = 0.
+        for i in range(self.tag_num):
+            s += alpha[-1, i] * self.b_exps[i]
+        log_likelihood -= np.log(s)
+
+        # calculate the nominator part.
+        log_likelihood += (
+            self.a[label[0]] + x[0, label[0]] + self.b[label[-1]])
+
+        for k in range(1, seq_len):
+            log_likelihood += (x[k, label[k]] + self.w[label[k - 1], label[k]])
+        return -log_likelihood
+
+    def crf_forward_compute(self):
+        for i in range(self.seq_num):
+            start = self.seq_start_positions[i]
+            end = self.seq_start_positions[i + 1]
+
+            self.log_likelihood[i] = self._forward_a_sequence(
+                self.x[start:end, :], self.x_row_max[start:end, :],
+                self.x_exps[start:end, :], self.labels[start:end, :],
+                self.alpha[start:end, :])
+        return self.alpha, self.log_likelihood
+
+
+class TestLinearChainCrfOp(OpTest):
+    def set_test_data(self):
+        # TODO(caoying) Fix the unittest by: add the boundary cases when
+        # sequence lengths are 1, 2, and 3.
+
+        SEQ_NUM = 3
+        TAG_NUM = 17
+        MAX_SEQ_LEN = 5
+
+        # the linear_chain_crf operator only supports sequence (LoD level = 1)
+        lod = [[0]]
+        for i in range(SEQ_NUM):
+            lod[-1].append(lod[-1][-1] + random.randint(1, MAX_SEQ_LEN))
+        emission = np.random.uniform(-1, 1,
+                                     [lod[-1][-1], TAG_NUM]).astype("float64")
+        emission_row_max = np.amax(emission, axis=1, keepdims=True)
+        emission_exps = np.exp(emission - emission_row_max)
+
+        transition = np.random.uniform(-0.5, 0.5,
+                                       [TAG_NUM + 2, TAG_NUM]).astype("float64")
+        transition_exps = np.exp(transition)
+
+        labels = np.random.randint(
+            low=0, high=TAG_NUM, size=(lod[-1][-1], 1), dtype="int32")
+
+        self.inputs = {
+            "Emission": (emission, lod),
+            "Transition": transition,
+            "Label": (labels, lod)
+        }
+        crf = LinearChainCrfForward(lod[0], emission, emission_row_max,
+                                    emission_exps, transition, transition_exps,
+                                    labels)
+        alpha, log_likelihood = crf.crf_forward_compute()
+
+        self.outputs = {
+            "Alpha": alpha,
+            "EmissionExps": emission_exps,
+            "TransitionExps": transition_exps,
+            "LogLikelihood": log_likelihood
+        }
+
+    def setUp(self):
+        self.op_type = "linear_chain_crf"
+        self.set_test_data()
+
+    def test_check_output(self):
+        self.check_output()
+
+    def test_check_grad(self):
+        self.check_grad(["Emission", "Transition"], "LogLikelihood")
+
+    def test_check_grad_ignore_transition(self):
+        self.check_grad(
+            ["Emission"], "LogLikelihood", no_grad_set=set("Transition"))
+
+
+if __name__ == "__main__":
+    unittest.main()

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

@@ -52,7 +52,7 @@ def lstm(
         g = np.dot(h_pre, w_h)  # 1 x 4D
         g = g + x
         g = np.reshape(g, (1, g.size))
-        c_tmp, g_i, g_f, g_o = np.split(g, 4, axis=1)
+        c, g_i, g_f, g_o = np.split(g, 4, axis=1)
         if w_c is None:
             g_i = act_gate(g_i)  # 1 x D
             g_f = act_gate(g_f)  # 1 x D
@@ -60,7 +60,7 @@ def lstm(
             w_ic, w_fc, w_oc = np.split(w_c, 3, axis=1)
             g_i = act_gate(g_i + w_ic * c_pre)  # 1 x D
             g_f = act_gate(g_f + w_fc * c_pre)  # 1 x D
-        c = g_f * c_pre + g_i * act_cand(c_tmp)  # 1 x D
+        c = g_f * c_pre + g_i * act_cand(c)  # 1 x D
 
         if w_c is None:
             g_o = act_gate(g_o)  # 1 x D
@@ -68,8 +68,7 @@ def lstm(
             _, _, w_oc = np.split(w_c, 3, axis=1)
             g_o = act_gate(g_o + w_oc * c)  # 1 x D
         h = g_o * act_cell(c)
-        bg = np.concatenate((act_cand(c_tmp), g_i, g_f, g_o), axis=1)
-        return h, c, bg
+        return h, c
 
     def _reverse(x, lod):
         y = np.zeros_like(x)
@@ -82,7 +81,6 @@ def lstm(
     batch_size = len(offset) - 1
     hidden = []
     cell = []
-    gate = []
     input = _reverse(input, offset) if is_reverse else input
     if w_b is not None:
         input = input + np.tile(w_b, (offset[-1], 1))
@@ -94,96 +92,109 @@ def lstm(
         c_pre = c0[i]  # 1 x D
         for j in range(seq_len):
             # compute one step
-            h_pre, c_pre, g_pre = _step(x[j], w_h, w_c, h_pre, c_pre, act_gate,
-                                        act_cell, act_cand)
+            h_pre, c_pre = _step(x[j], w_h, w_c, h_pre, c_pre, act_gate,
+                                 act_cell, act_cand)
             hidden.append(h_pre.flatten())
             cell.append(c_pre.flatten())
-            gate.append(g_pre.flatten())
 
-    hidden = np.array(hidden).astype("float64")
-    cell = np.array(cell).astype("float64")
-    gate = np.array(gate).astype("float64")
+    hidden = np.array(hidden).astype('float64')
+    cell = np.array(cell).astype('float64')
 
     hidden = _reverse(hidden, offset) if is_reverse else hidden
     cell = _reverse(cell, offset) if is_reverse else cell
 
-    assert gate.shape == input.shape
     assert hidden.shape == (input.shape[0], input.shape[1] / 4)
     assert cell.shape == (input.shape[0], input.shape[1] / 4)
-    return hidden, cell, gate
+    return hidden, cell
 
 
 class TestLstmOp(OpTest):
-    def set_data(self):
-        self.lod = [[0, 2, 6, 9]]
-        self.D = 64
-        self.sort_idx = [2, 6, 0, 3, 7, 1, 4, 8, 5]
+    def set_argument(self):
+        self.lod = [[0, 2, 5, 7]]
+        self.D = 16
 
-        self.act_gate = "sigmoid"
-        self.act_cell = "tanh"
-        self.act_cand = "tanh"
+        self.act_gate = 'sigmoid'
+        self.act_cell = 'tanh'
+        self.act_cand = 'tanh'
 
+        self.has_initial_state = True
         self.is_reverse = False
 
     def setUp(self):
-        self.set_data()
-        self.op_type = "lstm"
+        self.set_argument()
+        self.op_type = 'lstm'
 
         T = self.lod[0][-1]
         N = len(self.lod[0]) - 1
 
-        x = np.random.normal(size=(T, 4 * self.D)).astype("float64")
-        h0 = np.zeros((N, self.D)).astype("float64")
-        c0 = np.zeros((N, self.D)).astype("float64")
-        w = np.random.normal(size=(self.D, 4 * self.D)).astype("float64")
-        b = np.random.normal(size=(1, 7 * self.D)).astype("float64")
+        x = np.random.normal(size=(T, 4 * self.D)).astype('float64')
+        h0 = np.zeros((N, self.D)).astype('float64')
+        c0 = np.zeros((N, self.D)).astype('float64')
+        w = np.random.normal(size=(self.D, 4 * self.D)).astype('float64')
+        b = np.random.normal(size=(1, 7 * self.D)).astype('float64')
 
         w_b = b[:, 0:4 * self.D]
         w_c = b[:, 4 * self.D:]
-        h, c, g = lstm(x, self.lod, h0, c0, w, w_b, w_c, self.is_reverse,
-                       ACTVATION[self.act_gate], ACTVATION[self.act_cell],
-                       ACTVATION[self.act_cand])
-
-        g_sort = np.zeros_like(x)
-        for i, j in enumerate(self.sort_idx):
-            g_sort[i, :] = g[j, :]
-
-        self.inputs = {
-            'Input': (x, self.lod),
-            'H0': h0,
-            'C0': c0,
-            'Weight': w,
-            'Bias': b
-        }
+        h, c = lstm(x, self.lod, h0, c0, w, w_b, w_c, self.is_reverse,
+                    ACTVATION[self.act_gate], ACTVATION[self.act_cell],
+                    ACTVATION[self.act_cand])
+
+        self.inputs = {'Input': (x, self.lod), 'Weight': w, 'Bias': b}
+        if self.has_initial_state:
+            self.inputs['H0'] = h0
+            self.inputs['C0'] = c0
+
         self.outputs = {
             'Hidden': (h, self.lod),
             'Cell': (c, self.lod),
-            'BatchGate': g_sort
         }
         self.attrs = {
             'usePeepholes': True,
             'isReverse': self.is_reverse,
-            'gateActivation': 'sigmoid',
-            'cellActivation': 'tanh',
-            'candidateActivation': 'tanh'
+            'gateActivation': self.act_gate,
+            'cellActivation': self.act_cell,
+            'candidateActivation': self.act_cand
         }
 
     def test_check_output(self):
-        self.check_output()
+        self.check_output(atol=1e-8)
+
+    #TODO(qingqing) add more unit testing case
+    def test_check_grad(self):
+        # TODO(qingqing) remove folowing lines after the check_grad is refined.
+        N = len(self.lod[0]) - 1
+        self.outputs['BatchGate'] = np.zeros((N, 4 * self.D)).astype('float64')
+        self.outputs['BatchCellPreAct'] = np.zeros(
+            (N, self.D)).astype('float64')
+        self.check_grad(
+            ['Input', 'Weight', 'Bias'], ['Hidden'], max_relative_error=5e-4)
+
+
+class TestLstmOpHasNoInitial(TestLstmOp):
+    def set_argument(self):
+        self.lod = [[0, 2, 5, 7]]
+        self.D = 16
+
+        self.act_gate = 'sigmoid'
+        self.act_cell = 'tanh'
+        self.act_cand = 'tanh'
+
+        self.has_initial_state = False
+        self.is_reverse = True
 
 
 class TestLstmOpRerverse(TestLstmOp):
-    def set_data(self):
-        self.lod = [[0, 2, 6, 9]]
-        self.D = 64
-        self.sort_idx = [2, 6, 0, 3, 7, 1, 4, 8, 5]
+    def set_argument(self):
+        self.lod = [[0, 2, 5, 7]]
+        self.D = 16
 
-        self.act_gate = "sigmoid"
-        self.act_cell = "tanh"
-        self.act_cand = "tanh"
+        self.act_gate = 'sigmoid'
+        self.act_cell = 'tanh'
+        self.act_cand = 'tanh'
 
+        self.has_initial_state = True
         self.is_reverse = True
 
 
-if __name__ == "__main__":
+if __name__ == '__main__':
     unittest.main()

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

@@ -3,15 +3,6 @@ import numpy as np
 from op_test import OpTest
 
 
-class SeqPoolType(OpTest):
-    AVERAGE = 0
-    SUM = 1
-    SQRT = 2
-    MAX = 3
-    LAST = 4
-    FIRST = 5
-
-
 class TestSeqAvgPool(OpTest):
     def set_data(self):
         self.op_type = 'sequence_pool'
@@ -25,7 +16,7 @@ class TestSeqAvgPool(OpTest):
         return x, lod, out
 
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.AVERAGE}
+        self.attrs = {'pooltype': "AVERAGE"}
         for i in range(4):
             sub_x = x[lod[0][i]:lod[0][i + 1], :]
             out[i] = sub_x.mean(axis=0)
@@ -54,7 +45,7 @@ class TestSeqAvgPool2D(TestSeqAvgPool):
         return x, lod, out
 
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.AVERAGE}
+        self.attrs = {'pooltype': "AVERAGE"}
         for i in range(4):
             sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17))
             out[i] = np.reshape(sub_x.mean(axis=0), (3, 17))
@@ -62,7 +53,7 @@ class TestSeqAvgPool2D(TestSeqAvgPool):
 
 class TestSeqSumPool(TestSeqAvgPool):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.SUM}
+        self.attrs = {'pooltype': "SUM"}
         for i in range(4):
             sub_x = x[lod[0][i]:lod[0][i + 1], :]
             out[i] = sub_x.sum(axis=0)
@@ -70,7 +61,7 @@ class TestSeqSumPool(TestSeqAvgPool):
 
 class TestSeqSumPool2D(TestSeqAvgPool2D):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.SUM}
+        self.attrs = {'pooltype': "SUM"}
         for i in range(4):
             sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17))
             out[i] = np.reshape(sub_x.sum(axis=0), (3, 17))
@@ -78,7 +69,7 @@ class TestSeqSumPool2D(TestSeqAvgPool2D):
 
 class TestSeqSqrtPool(TestSeqAvgPool):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.SQRT}
+        self.attrs = {'pooltype': "SQRT"}
         for i in range(4):
             sub_x = x[lod[0][i]:lod[0][i + 1], :]
             len = lod[0][i + 1] - lod[0][i]
@@ -87,7 +78,7 @@ class TestSeqSqrtPool(TestSeqAvgPool):
 
 class TestSeqSqrtPool2D(TestSeqAvgPool2D):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.SQRT}
+        self.attrs = {'pooltype': "SQRT"}
         for i in range(4):
             sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17))
             len = lod[0][i + 1] - lod[0][i]
@@ -99,7 +90,7 @@ class TestSeqSqrtPool2D(TestSeqAvgPool2D):
 
 class TestSeqMaxPool(TestSeqAvgPool):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.MAX}
+        self.attrs = {'pooltype': "MAX"}
         for i in range(4):
             sub_x = x[lod[0][i]:lod[0][i + 1], :]
             out[i] = np.amax(sub_x, axis=0)
@@ -111,7 +102,7 @@ class TestSeqMaxPool(TestSeqAvgPool):
 
 class TestSeqMaxPool2D(TestSeqAvgPool2D):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.MAX}
+        self.attrs = {'pooltype': "MAX"}
         for i in range(4):
             sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17))
             out[i] = np.reshape(np.amax(sub_x, axis=0), (3, 17))
@@ -123,7 +114,7 @@ class TestSeqMaxPool2D(TestSeqAvgPool2D):
 
 class TestSeqLastPool(TestSeqAvgPool):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.LAST}
+        self.attrs = {'pooltype': "LAST"}
         for i in range(4):
             sub_x = x[lod[0][i]:lod[0][i + 1], :]
             out[i] = sub_x[-1, :]
@@ -131,7 +122,7 @@ class TestSeqLastPool(TestSeqAvgPool):
 
 class TestSeqLastPool2D(TestSeqAvgPool2D):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.LAST}
+        self.attrs = {'pooltype': "LAST"}
         for i in range(4):
             sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17))
             out[i] = np.reshape(sub_x[-1, :], (3, 17))
@@ -139,7 +130,7 @@ class TestSeqLastPool2D(TestSeqAvgPool2D):
 
 class TestSeqFirstPool(TestSeqAvgPool):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.FIRST}
+        self.attrs = {'pooltype': "FIRST"}
         for i in range(4):
             sub_x = x[lod[0][i]:lod[0][i + 1], :]
             out[i] = sub_x[0, :]
@@ -147,7 +138,7 @@ class TestSeqFirstPool(TestSeqAvgPool):
 
 class TestSeqFirstPool2D(TestSeqAvgPool2D):
     def compute(self, x, lod, out):
-        self.attrs = {'strategy': SeqPoolType.FIRST}
+        self.attrs = {'pooltype': "FIRST"}
         for i in range(4):
             sub_x = np.reshape(x[lod[0][i]:lod[0][i + 1], :], (-1, 3 * 17))
             out[i] = np.reshape(sub_x[0, :], (3, 17))