Merge pull request #8482 from jacquesqiao/release/0.11.0

Release/0.11.0

Merge pull request #8482 from jacquesqiao/release/0.11.0
Release/0.11.0
1ff47b28 · Qiao Longfei · GitHub · 2c98becb · 67dce977 · 1ff47b28
1000 changed file
--- a/.clang-format
+++ b/.clang-format
@@ -25,4 +25,3 @@ AllowAllParametersOfDeclarationOnNextLine: true
 BinPackParameters: false
 BinPackArguments: false
 ...
--- a/.clang_format.hook
+++ b/.clang_format.hook
+#!/bin/bash
+set -e
+readonly VERSION="3.8"
+version=$(clang-format -version)
+if ! [[ $version == *"$VERSION"* ]]; then
+    echo "clang-format version check failed."
+    echo "a version contains '$VERSION' is needed, but get '$version'"
+    echo "you can install the right version, and make an soft-link to '\$PATH' env"
+    exit -1
+fi
+clang-format $@
--- a/.gitignore
+++ b/.gitignore
@@ -16,3 +16,16 @@ third_party/
 *~
 bazel-*
 third_party/
+# clion workspace.
+cmake-build-*
+# generated while compiling
+python/paddle/v2/fluid/core.so
+paddle/pybind/pybind.h
+CMakeFiles
+cmake_install.cmake
+paddle/.timestamp
+python/paddlepaddle.egg-info/
+paddle/pybind/pybind.h
+python/paddle/version.py
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -3,8 +3,8 @@
    hooks:
    -   id: remove-crlf
        files: (?!.*third_party)^.*$ | (?!.*book)^.*$
-   repo: https://github.com/reyoung/mirrors-yapf.git
+-   repo: https://github.com/PaddlePaddle/mirrors-yapf.git
-    sha: v0.13.2
+    sha: 0d79c0c469bab64f7229c9aca2b1186ef47f0e37
    hooks:
    -   id: yapf
        files: (.*\.(py|bzl)|BUILD|.*\.BUILD|WORKSPACE)$
@@ -17,7 +17,17 @@
    -   id: detect-private-key
        files: (?!.*third_party)^.*$ | (?!.*book)^.*$
    -   id: end-of-file-fixer
-   repo: https://github.com/PaddlePaddle/clang-format-pre-commit-hook.git
+-   repo: local
-    sha: 28c0ea8a67a3e2dbbf4822ef44e85b63a0080a29
    hooks:
-    -   id: clang-formater
+    -   id: clang-format-with-version-check
+        name: clang-format
+        description: Format files with ClangFormat.
+        entry: bash ./.clang_format.hook -i
+        language: system
+        files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|proto)$
+-   repo: https://github.com/PaddlePaddle/pre-commit-golang
+    sha: 8337620115c25ff8333f1b1a493bd031049bd7c0
+    hooks:
+    -   id: go-fmt
+        types:
+        - go
--- a/.travis.yml
+++ b/.travis.yml
 language: cpp
 cache:
  directories:
-    - $HOME/third_party
    - $HOME/.ccache
    - $HOME/.cache/pip
+    - $TRAVIS_BUILD_DIR/build/third_party
 sudo: required
 dist: trusty
 os:
  - linux
 env:
-  - JOB=DOCS
+  - JOB=build_doc
-  - JOB=BUILD_AND_TEST
+  - JOB=check_style
-  - JOB=PRE_COMMIT
 addons:
  apt:
    packages:
      - gcc-4.8
      - g++-4.8
-      - gfortran-4.8
      - git
      - build-essential
      - python
      - python-pip
      - python2.7-dev
-      - python-numpy
      - python-wheel
+      - libboost-dev
      - curl
      - swig
      - graphviz
@@ -32,28 +30,27 @@ addons:
      - automake
      - libtool
      - ccache
+  ssh_known_hosts: 52.76.173.135
 before_install:
-  - |
+  - if [[ "$JOB" == "check_style" ]]; then sudo ln -s /usr/bin/clang-format-3.8 /usr/bin/clang-format; fi
-    if [ ${JOB} == "BUILD_AND_TEST" ]; then
+  # Paddle is using protobuf 3.1 currently. Protobuf 3.2 breaks the compatibility. So we specify the python
-      local change_list=`git diff --name-only $TRAVIS_COMMIT_RANGE`
-      if [ $? -eq 0 ]; then  # if git diff return no zero, then rerun unit test.
-        if ! echo ${change_list} | grep -qvE '(\.md$)|(\.rst$)|(\.jpg$)|(\.png$)'
-        then
-          echo "Only markdown docs were updated, stopping build process."
-          exit
-        fi
-      fi
-    fi
-  - if [[ "$JOB" == "PRE_COMMIT" ]]; then sudo ln -s /usr/bin/clang-format-3.8 /usr/bin/clang-format; fi
-  # Paddle is using protobuf 3.1 currently. Protobuf 3.2 breaks the compatibility. So we specify the python 
  # protobuf version.
-  - pip install numpy wheel 'protobuf==3.1' sphinx recommonmark sphinx-rtd-theme==0.1.9 virtualenv pre-commit requests==2.9.2 LinkChecker
+  - sudo pip install -r $TRAVIS_BUILD_DIR/python/requirements.txt
+  - sudo pip install wheel sphinx==1.5.6 recommonmark sphinx-rtd-theme==0.1.9 virtualenv pre-commit LinkChecker
  - |
    function timeout() { perl -e 'alarm shift; exec @ARGV' "$@"; }
 script:
-  - | 
+  - |
-    timeout 2580 paddle/scripts/travis/main.sh  # 43min timeout
+    timeout 2580 paddle/scripts/travis/${JOB}.sh # 43min timeout
-    RESULT=$?; if [ $RESULT -eq 0 ] || [ $RESULT -eq 142 ]; then true; else false; fi;
+    RESULT=$?; if [ $RESULT -eq 0 ] || [ $RESULT -eq 142 ]; then true ;else exit 1; fi;
+  - |
+    if [[ "$JOB" != "build_doc" ]]; then exit 0; fi;
+    if [[ "$TRAVIS_PULL_REQUEST" != "false" ]]; then exit 0; fi;
+    if [[ "$TRAVIS_BRANCH" != "develop"  && ! "$TRAVIS_BRANCH" =~ ^v[[:digit:]]+\.[[:digit:]]+(\.[[:digit:]]+)?(-\S*)?$ ]]; then exit 0; fi;
+    export DEPLOY_DOCS_SH=https://raw.githubusercontent.com/PaddlePaddle/PaddlePaddle.org/master/scripts/deploy/deploy_docs.sh
+    export DOCS_DIR=`pwd`
+    cd ..
+    curl $DEPLOY_DOCS_SH | bash -s $CONTENT_DEC_PASSWD $TRAVIS_BRANCH $DOCS_DIR $DOCS_DIR/build/doc   
 notifications:
  email:
    on_success: change

--- a/AUTHORS.md
+++ b/AUTHORS.md
+| Github account | name |
+|---|---|
+| backyes | Yan-Fei Wang |
+| beckett1124 | Bin Qi |
+| Canpio | Jia-Yi Feng |
+| chengxiaohua1105 | Xiao-Hua Cheng |
+| cxwangyi, yiwangbaidu, wangkuiyi | Yi Wang |
+| cxysteven | Xing-Yi Cheng |
+| dzhwinter | Zhi-Hong Dong |
+| emailweixu | Wei Xu |
+| gangliao | Gang Liao |
+| gongweibao | Wei-Bao Gong |
+| Guo Sheng | Sheng Guo |
+| Haichao-Zhang | Hai-Chao Zhang |
+| hedaoyuan | Dao-Yuan He |
+| helinwang | He-Lin Wang |
+| jacquesqiao | Long-Fei Qiao |
+| kuke | Yi-Bing Liu |
+| lcy-seso | Ying Cao |
+| lipeng-unisound | Peng Li |
+| liuyuan | Yuan Liu |
+| livc | Zhao Li |
+| llxxxll | Yong-Feng Liu |
+| luotao01 | Tao Luo |
+| lzhao4ever | Liang Zhao |
+| NHZlX | Zhao-Long Xing |
+| pakchoi | Chuan-Jiang Song |
+| pengli09 | Peng Li |
+| pkuyym | Ya-Ming Yang |
+| QiJune | Jun Qi |
+| qingqing01 | Qing-Qing Dang |
+| reyoung | Yang Yu |
+| Superjom | Chun-Wei Yan |
+| tianbingsz | Tian-Bing Xu |
+| typhoonzero | Yi Wu |
+| wanghaoshuang | Hao-Shuang Wang |
+| wangyang59 | Yang Wang |
+| wangzhen-nlp | Zhen Wang |
+| wen-bo-yang | Wen-Bo Yang |
+| wwhu | Wei-Wei Hu |
+| xinghai-sun | Xing-Hai Sun |
+| Xreki | Yi-Qun Liu |
+| xujun05 | Jun Xu |
+| xushaoyong | Shao-Yong Xu |
+| Yancey1989 | Xu Yan |
+| zhaopu7 | Pu Zhao |
+| zhouxiao-coder | Xiao Zhou |
+| Zrachel | Rui-Qing Zhang |
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
 # 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
-set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/cmake")
+cmake_minimum_required(VERSION 3.0)
-set(PROJ_ROOT ${CMAKE_SOURCE_DIR})
+set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/cmake")
+set(PADDLE_SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR})
+set(PADDLE_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR})
+SET(CMAKE_CXX_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
+SET(CMAKE_C_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
 include(system)
-if(ANDROID)
+project(paddle CXX C Go)
-    cmake_minimum_required(VERSION 3.7)
+message(STATUS "CXX compiler: " ${CMAKE_CXX_COMPILER} ", version: " ${CMAKE_CXX_COMPILER_VERSION})
-else()
+message(STATUS "C compiler: " ${CMAKE_C_COMPILER} ", version: " ${CMAKE_C_COMPILER_VERSION})
-    cmake_minimum_required(VERSION 3.0)
-endif()
-project(paddle CXX C)
 find_package(Sphinx)
 if(NOT CMAKE_CROSSCOMPILING)
@@ -31,12 +31,16 @@ if(NOT CMAKE_CROSSCOMPILING)
 endif(NOT CMAKE_CROSSCOMPILING)
 find_package(Git REQUIRED)
 find_package(Threads REQUIRED)
+if(NOT ANDROID AND NOT IOS)
+    find_package(Boost QUIET)
+endif()
 include(simd)
 ################################ Configurations #######################################
 option(WITH_GPU         "Compile PaddlePaddle with NVIDIA GPU"          ${CUDA_FOUND})
 option(WITH_AVX         "Compile PaddlePaddle with AVX intrinsics"      ${AVX_FOUND})
+option(WITH_MKL         "Compile PaddlePaddle with MKL support."        ${AVX_FOUND})
 option(WITH_DSO         "Compile PaddlePaddle with dynamic linked CUDA" ON)
 option(WITH_TESTING     "Compile PaddlePaddle with unit testing"        ON)
 option(WITH_SWIG_PY     "Compile PaddlePaddle with inference api"       ON)
@@ -50,6 +54,13 @@ option(WITH_DOC         "Compile PaddlePaddle with documentation"       OFF)
 option(WITH_COVERAGE    "Compile PaddlePaddle with code coverage"       OFF)
 option(COVERALLS_UPLOAD "Package code coverage data to coveralls"       OFF)
 option(ON_TRAVIS        "Exclude special unit test on Travis CI"        OFF)
+option(WITH_C_API       "Compile PaddlePaddle with C-API(Prediction)"   OFF)
+option(WITH_GOLANG      "Compile PaddlePaddle with GOLANG"              OFF)
+option(GLIDE_INSTALL    "Download and install go dependencies "         ON)
+option(USE_NNPACK       "Compile PaddlePaddle with NNPACK library"      OFF)
+option(WITH_DISTRIBUTE  "Compile with grpc distributed support"         OFF)
+option(USE_EIGEN_FOR_BLAS   "Use matrix multiplication in Eigen"        OFF)
+option(WITH_ARM_FP16    "Use half precision support on armv8.2-a cpu"   OFF)
 # CMAKE_BUILD_TYPE
 if(NOT CMAKE_BUILD_TYPE)
@@ -58,25 +69,61 @@ if(NOT CMAKE_BUILD_TYPE)
      FORCE)
 endif()
-if(ANDROID)
+if(ANDROID OR IOS)
-    if(${CMAKE_SYSTEM_VERSION} VERSION_LESS "21")
+    if(ANDROID)
-        message(FATAL_ERROR "Unsupport standalone toolchains with Android API level lower than 21")
+        if(${CMAKE_SYSTEM_VERSION} VERSION_LESS "16")
+            message(FATAL_ERROR "Unsupport standalone toolchains with Android API level lower than 16")
+        endif()
    endif()
    set(WITH_GPU OFF CACHE STRING
-        "Disable GPU when cross-compiling for Android" FORCE)
+        "Disable GPU when cross-compiling for Android and iOS" FORCE)
    set(WITH_AVX OFF CACHE STRING
-        "Disable AVX when cross-compiling for Android" FORCE)
+        "Disable AVX when cross-compiling for Android and iOS" FORCE)
    set(WITH_PYTHON OFF CACHE STRING
-        "Disable PYTHON when cross-compiling for Android" FORCE)
+        "Disable PYTHON when cross-compiling for Android and iOS" FORCE)
    set(WITH_RDMA OFF CACHE STRING
-        "Disable RDMA when cross-compiling for Android" FORCE)
+        "Disable RDMA when cross-compiling for Android and iOS" FORCE)
-endif(ANDROID)
+    set(WITH_MKL OFF CACHE STRING
+        "Disable MKL when cross-compiling for Android and iOS" FORCE)
+    set(WITH_GOLANG OFF CACHE STRING
+        "Disable golang when cross-compiling for Android and iOS" FORCE)
+    # Compile PaddlePaddle mobile inference library
+    if (NOT WITH_C_API)
+        set(WITH_C_API ON CACHE STRING
+            "Always compile the C_API when cross-compiling for Android and iOS" FORCE)
+    endif()
+    set(MOBILE_INFERENCE ON)
+    add_definitions(-DPADDLE_MOBILE_INFERENCE)
+endif()
-set(THIRD_PARTY_PATH "${PROJ_ROOT}/third_party" CACHE STRING
+set(THIRD_PARTY_PATH "${CMAKE_BINARY_DIR}/third_party" CACHE STRING
  "A path setting third party libraries download & build directories.")
+if (WITH_C_API AND WITH_PYTHON)
+  message(WARNING "It is suggest not embedded a python interpreter in Paddle "
+    "when using C-API. It will give an unpredictable behavior when using a "
+    "different Python interpreter from compiling.")
+endif()
+if(MOBILE_INFERENCE)
+    set(THIRD_PARTY_BUILD_TYPE MinSizeRel)
+else()
+    set(THIRD_PARTY_BUILD_TYPE Release)
+endif()
+set(WITH_MKLML ${WITH_MKL})
+if (WITH_MKL AND AVX2_FOUND)
+    set(WITH_MKLDNN ON)
+else()
+    message(STATUS "Do not have AVX2 intrinsics and disabled MKL-DNN")
+    set(WITH_MKLDNN OFF)
+endif()
 ########################################################################################
+include(external/mklml)     # download mklml package
 include(external/zlib)      # download, build, install zlib
 include(external/gflags)    # download, build, install gflags
 include(external/glog)      # download, build, install glog
@@ -84,24 +131,34 @@ include(external/gtest)     # download, build, install gtest
 include(external/protobuf)  # download, build, install protobuf
 include(external/python)    # download, build, install python
 include(external/openblas)  # download, build, install openblas
+include(external/mkldnn)    # download, build, install mkldnn
 include(external/swig)      # download, build, install swig
 include(external/warpctc)   # download, build, install warpctc
 include(external/any)       # download libn::any
+include(external/eigen)     # download eigen3
+include(external/pybind11)  # download pybind11
+include(external/nccl)
+include(external/cares)
+include(external/grpc)
+include(cudnn)              # set cudnn libraries, must before configure
+include(configure)          # add paddle env configuration
+include(generic)            # simplify cmake module
 include(package)            # set paddle packages
 include(cpplint)            # set paddle c++ style
 include(ccache)             # set ccache for compilation
 include(util)               # set unittest and link libs
 include(rdma)               # set rdma libraries
 include(flags)              # set paddle compile flags
-include(cudnn)              # set cudnn libraries
 include(version)            # set PADDLE_VERSION
 include(coveralls)          # set code coverage
-include(configure)          # add paddle env configuration
-include_directories("${PROJ_ROOT}")
-include_directories("${PROJ_ROOT}/paddle/cuda/include")
+include_directories("${PADDLE_SOURCE_DIR}")
+include_directories("${PADDLE_SOURCE_DIR}/paddle/cuda/include")
 include_directories("${CMAKE_CURRENT_BINARY_DIR}/proto")
+include_directories("${CMAKE_CURRENT_BINARY_DIR}/go/pserver/client/c")
+include_directories(${Boost_INCLUDE_DIRS})
 set(EXTERNAL_LIBS
    ${GFLAGS_LIBRARIES}
@@ -113,15 +170,41 @@ set(EXTERNAL_LIBS
 )
 if(WITH_GPU)
-    list(APPEND EXTERNAL_LIB ${CUDA_LIBRARIES} ${CUDA_rt_LIBRARY})
+  include(cuda)
-    if(NOT WITH_DSO)
-        list(APPEND EXTERNAL_LIB ${CUDNN_LIBRARY} ${CUDA_CUBLAS_LIBRARIES} ${CUDA_curand_LIBRARY})
-    endif(NOT WITH_DSO)
 endif(WITH_GPU)
+if(WITH_MKLML)
+    list(APPEND EXTERNAL_LIBS ${MKLML_IOMP_LIB})
+endif()
+if(WITH_MKLDNN)
+    list(APPEND EXTERNAL_LIBS ${MKLDNN_LIB})
+endif()
+if(USE_NNPACK)
+    include(external/nnpack)
+    list(APPEND EXTERNAL_LIBS ${NNPACK_LIBS})
+endif(USE_NNPACK)
 add_subdirectory(proto)
+if(NOT MOBILE_INFERENCE)
+    # "add_subdirectory(go)" should be placed after the following loine,
+    # because it depends on paddle/optimizer.
+    add_subdirectory(paddle/optimizer)
+endif()
+# "add_subdirectory(paddle)" and "add_subdirectory(python)" should be
+# placed after this block, because they depends on it.
+if(WITH_GOLANG)
+    add_subdirectory(go)
+endif(WITH_GOLANG)
+set(PADDLE_PYTHON_BUILD_DIR "${CMAKE_CURRENT_BINARY_DIR}/python/build")
 add_subdirectory(paddle)
-add_subdirectory(python)
+if(WITH_PYTHON)
+  add_subdirectory(python)
+endif()
 if(WITH_DOC)
    add_subdirectory(doc)

--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
-./doc/howto/dev/contribute_to_paddle_en.md
+# Contribute Code
+We sincerely appreciate your contribution.  This document explains our workflow and work style.
+## Workflow
+PaddlePaddle uses this [Git branching model](http://nvie.com/posts/a-successful-git-branching-model/).  The following steps guide usual contributions.
+1. Fork
+   Our development community has been growing fastly; it doesn't make sense for everyone to write into the official repo.  So, please file Pull Requests from your fork.  To make a fork,  just head over to the GitHub page and click the ["Fork" button](https://help.github.com/articles/fork-a-repo/).
+1. Clone
+   To make a copy of your fork to your local computers, please run
+   ```bash
+   git clone https://github.com/your-github-account/paddle
+   cd paddle
+   ```
+1. Create the local feature branch
+   For daily works like adding a new feature or fixing a bug, please open your feature branch before coding:
+   ```bash
+   git checkout -b my-cool-stuff
+   ```
+1. Commit
+   Before issuing your first `git commit` command, please install [`pre-commit`](http://pre-commit.com/) by running the following commands:
+   ```bash
+   pip install pre-commit
+   pre-commit install
+   ```
+   Our pre-commit configuration requires clang-format 3.8 for auto-formating C/C++ code and yapf for Python.
+   Once installed, `pre-commit` checks the style of code and documentation in every commit.  We will see something like the following when you run `git commit`:
+   ```
+   ➜  git commit
+   CRLF end-lines remover...............................(no files to check)Skipped
+   yapf.................................................(no files to check)Skipped
+   Check for added large files..............................................Passed
+   Check for merge conflicts................................................Passed
+   Check for broken symlinks................................................Passed
+   Detect Private Key...................................(no files to check)Skipped
+   Fix End of Files.....................................(no files to check)Skipped
+   clang-formater.......................................(no files to check)Skipped
+   [my-cool-stuff c703c041] add test file
+    1 file changed, 0 insertions(+), 0 deletions(-)
+    create mode 100644 233
+   ```
+1. Build and test
+   Users can build PaddlePaddle natively on Linux and Mac OS X.  But to unify the building environment and to make it easy for debugging, the recommended way is [using Docker](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/howto/dev/build_en.md).
+1. Keep pulling
+   An experienced Git user pulls from the official repo often -- daily or even hourly, so they notice conflicts with others work early, and it's easier to resolve smaller conflicts.
+   ```bash
+   git remote add upstream https://github.com/PaddlePaddle/Paddle
+   git pull upstream develop
+   ```
+1. Push and file a pull request
+   You can "push" your local work into your forked repo:
+   ```bash
+   git push origin my-cool-stuff
+   ```
+   The push allows you to create a pull request, requesting owners of this [official repo](https://github.com/PaddlePaddle/Paddle) to pull your change into the official one.
+   To create a pull request, please follow [these steps](https://help.github.com/articles/creating-a-pull-request/).
+   If your change is for fixing an issue, please write ["Fixes <issue-URL>"](https://help.github.com/articles/closing-issues-using-keywords/) in the description section of your pull request.  Github would close the issue when the owners merge your pull request.
+   Please remember to specify some reviewers for your pull request.  If you don't know who are the right ones, please follow Github's recommendation.
+1. Delete local and remote branches
+   To keep your local workspace and your fork clean, you might want to remove merged branches:
+   ```bash
+   git push origin :my-cool-stuff
+   git checkout develop
+   git pull upstream develop
+   git branch -d my-cool-stuff
+   ```
+### Code Review
+-  Please feel free to ping your reviewers by sending them the URL of your pull request via IM or email.  Please do this after your pull request passes the CI.
+- Please answer reviewers' every comment.  If you are to follow the comment, please write "Done"; please give a reason otherwise.
+- If you don't want your reviewers to get overwhelmed by email notifications, you might reply their comments by [in a batch](https://help.github.com/articles/reviewing-proposed-changes-in-a-pull-request/).
+- Reduce the unnecessary commits.  Some developers commit often.  It is recommended to append a sequence of small changes into one commit by running `git commit --amend` instead of `git commit`.
+## Coding Standard
+### Code Style
+Our C/C++ code follows the [Google style guide](http://google.github.io/styleguide/cppguide.html).
+Our Python code follows the [PEP8 style guide](https://www.python.org/dev/peps/pep-0008/).
+Our build process helps to check the code style.  In [`build.sh`](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/paddle/scripts/docker/build.sh#L42), the entry point of our [builder Docker image](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/Dockerfile#L88), the CMake argument `WITH_STYLE_CHECK` is set to `ON` by default.  This flag is on
+Please install pre-commit, which automatically reformat the changes to C/C++ and Python code whenever we run `git commit`.  To check the whole codebase, we can run the command `pre-commit run -a`, as in the [`check_style.sh` file](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/paddle/scripts/travis/check_style.sh#L30), which is invoked by [our Travis CI configuration](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/.travis.yml#L43).
+### Unit Tests
+Please remember to add related unit tests.
+- For C/C++ code, please follow [`google-test` Primer](https://github.com/google/googletest/blob/master/googletest/docs/Primer.md).
+- For Python code, please use [Python's standard `unittest` package](http://pythontesting.net/framework/unittest/unittest-introduction/).
+### Writing Logs
+We use [glog](https://github.com/google/glog) for logging in our C/C++ code.
+For general information, please use `LOG`.  For debug information, please use [`VLOG`](http://htmlpreview.github.io/?https://github.com/google/glog/blob/master/doc/glog.html#verbose).  The reason is at [here](https://groups.google.com/a/chromium.org/d/msg/chromium-dev/3NDNd1KzXeY/AZKMMx37fdQJ).
+`VLOG` requires a *verbose level* parameter.  For example:
+```c++
+VLOG(3) << "Operator FC is taking " << num_inputs << "inputs."
+```
+When we run a PaddlePaddle application or test, we can specify a verbose threshold.  For example:
+```bash
+GLOG_vmodule=buddy_allocator=2 \
+GLOG_v=10 \
+python \
+../python/paddle/v2/framework/tests/test_recurrent_op.py
+```
+This will enable VLOG messages generated by `buddy_allocator.{h,cc}` and in the verbose range of 0 to 3, so you will see above example VLOG message, which is in level 3.  This suggests that we output overall messages in lower verbose levels, so they display with higher probability.  When coding C++, please follow the verbose level convention as follows:
+- verbose level 1: [framework](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/framework)
+- verbose level 3: [operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators)
+- verbose level 5: [memory](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/memory), [platform](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/platform)
+- verbose level 7: [math](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/math)
--- a/Dockerfile
+++ b/Dockerfile
 # A image for building paddle binaries
 # Use cuda devel base image for both cpu and gpu environment
-FROM nvidia/cuda:8.0-cudnn5-devel-ubuntu14.04
+FROM nvidia/cuda:8.0-cudnn5-devel-ubuntu16.04
 MAINTAINER PaddlePaddle Authors <paddle-dev@baidu.com>
 ARG UBUNTU_MIRROR
@@ -10,26 +10,40 @@ RUN /bin/bash -c 'if [[ -n ${UBUNTU_MIRROR} ]]; then sed -i 's#http://archive.ub
 ARG WITH_GPU
 ARG WITH_AVX
 ARG WITH_DOC
-ARG WITH_STYLE_CHECK
 ENV WOBOQ OFF
-ENV WITH_GPU=${WITH_GPU:-OFF}
+ENV WITH_GPU=${WITH_GPU:-ON}
 ENV WITH_AVX=${WITH_AVX:-ON}
 ENV WITH_DOC=${WITH_DOC:-OFF}
-ENV WITH_STYLE_CHECK=${WITH_STYLE_CHECK:-OFF}
 ENV HOME /root
 # Add bash enhancements
 COPY ./paddle/scripts/docker/root/ /root/
 RUN apt-get update && \
-    apt-get install -y git python-pip python-dev openssh-server bison && \
+    apt-get install -y \
-    apt-get install -y wget unzip tar xz-utils bzip2 gzip coreutils && \
+    git python-pip python-dev openssh-server bison libnccl-dev \
-    apt-get install -y curl sed grep graphviz libjpeg-dev zlib1g-dev && \
+    wget unzip unrar tar xz-utils bzip2 gzip coreutils ntp \
-    apt-get install -y python-numpy python-matplotlib gcc g++ gfortran && \
+    curl sed grep graphviz libjpeg-dev zlib1g-dev  \
-    apt-get install -y automake locales clang-format-3.8 swig doxygen && \
+    python-matplotlib gcc-4.8 g++-4.8 \
+    automake locales clang-format swig doxygen cmake  \
+    liblapack-dev liblapacke-dev libboost-dev \
+    clang-3.8 llvm-3.8 libclang-3.8-dev \
+    net-tools libtool && \
    apt-get clean -y
+# Install Go and glide
+RUN wget -qO- https://storage.googleapis.com/golang/go1.8.1.linux-amd64.tar.gz | \
+    tar -xz -C /usr/local && \
+    mkdir /root/gopath && \
+    mkdir /root/gopath/bin && \
+    mkdir /root/gopath/src
+ENV GOROOT=/usr/local/go GOPATH=/root/gopath
+# should not be in the same line with GOROOT definition, otherwise docker build could not find GOROOT.
+ENV PATH=${PATH}:${GOROOT}/bin:${GOPATH}/bin
+# install glide
+RUN curl -s -q https://glide.sh/get | sh
 # git credential to skip password typing
 RUN git config --global credential.helper store
@@ -39,18 +53,29 @@ RUN localedef -i en_US -f UTF-8 en_US.UTF-8
 # FIXME: due to temporary ipykernel dependency issue, specify ipykernel jupyter
 # version util jupyter fixes this issue.
 RUN pip install --upgrade pip && \
-    pip install -U 'protobuf==3.1.0' && \
+    pip install -U wheel && \
-    pip install -U wheel pillow BeautifulSoup && \
    pip install -U docopt PyYAML sphinx && \
-    pip install -U sphinx-rtd-theme==0.1.9 recommonmark && \
+    pip install -U sphinx-rtd-theme==0.1.9 recommonmark
-    pip install pre-commit 'requests==2.9.2' 'ipython==5.3.0' && \
-    pip install 'ipykernel==4.6.0' 'jupyter==1.0.0'
+RUN pip install pre-commit 'ipython==5.3.0' && \
+    pip install 'ipykernel==4.6.0' 'jupyter==1.0.0' && \
+    pip install opencv-python
+COPY ./python/requirements.txt /root/
+RUN pip install -r /root/requirements.txt
+# To fix https://github.com/PaddlePaddle/Paddle/issues/1954, we use
+# the solution in https://urllib3.readthedocs.io/en/latest/user-guide.html#ssl-py2
+RUN apt-get install -y libssl-dev libffi-dev
+RUN pip install certifi urllib3[secure]
-RUN curl -sSL https://cmake.org/files/v3.4/cmake-3.4.1.tar.gz | tar -xz && \
-    cd cmake-3.4.1 && ./bootstrap && make -j `nproc` && make install && \
-    cd .. && rm -rf cmake-3.4.1
-VOLUME ["/woboq_out"]
+# Install woboq_codebrowser to /woboq
+RUN git clone https://github.com/woboq/woboq_codebrowser /woboq && \
+    (cd /woboq \
+     cmake -DLLVM_CONFIG_EXECUTABLE=/usr/bin/llvm-config-3.8 \
+           -DCMAKE_BUILD_TYPE=Release . \
+     make)
 # Configure OpenSSH server. c.f. https://docs.docker.com/engine/examples/running_ssh_service
 RUN mkdir /var/run/sshd

--- a/Dockerfile.android
+++ b/Dockerfile.android
+FROM ubuntu:16.04
+MAINTAINER PaddlePaddle Authors <paddle-dev@baidu.com>
+ARG UBUNTU_MIRROR
+RUN /bin/bash -c 'if [[ -n ${UBUNTU_MIRROR} ]]; then sed -i 's#http://archive.ubuntu.com/ubuntu#${UBUNTU_MIRROR}#g' /etc/apt/sources.list; fi'
+# ENV variables
+ARG ANDROID_ABI
+ARG ANDROID_API
+ENV ANDROID_ABI=${ANDROID_ABI:-"armeabi-v7a"}
+ENV ANDROID_API=${ANDROID_API:-21}
+ENV HOME=/root \
+    ANDROID_NDK_HOME=/opt/android-ndk-linux \
+    ANDROID_TOOLCHAINS_DIR=/opt/toolchains
+RUN apt-get update && \
+    apt-get install -y \
+    git python-dev python-pip python-numpy \
+    wget curl tar unzip gcc g++ locales clang-format-3.8 swig cmake && \
+    apt-get clean -y
+# Install Go and glide
+RUN wget -qO- go.tgz https://storage.googleapis.com/golang/go1.8.1.linux-amd64.tar.gz | \
+    tar -xz -C /usr/local && \
+    mkdir /root/gopath && \
+    mkdir /root/gopath/bin && \
+    mkdir /root/gopath/src
+ENV GOROOT=/usr/local/go GOPATH=/root/gopath
+# should not be in the same line with GOROOT definition, otherwise docker build could not find GOROOT.
+ENV PATH=${PATH}:${GOROOT}/bin:${GOPATH}/bin
+# git credential to skip password typing
+RUN git config --global credential.helper store
+# Fix locales to en_US.UTF-8
+RUN localedef -i en_US -f UTF-8 en_US.UTF-8
+RUN pip install --upgrade pip && \
+    pip install -U 'protobuf==3.1.0' && \
+    pip install -U wheel sphinx && \
+    pip install pre-commit
+# Android NDK
+RUN mkdir -p ${ANDROID_TOOLCHAINS_DIR} && \
+    mkdir -p /opt/android-ndk-tmp && \
+    cd /opt/android-ndk-tmp && \
+    wget -q https://dl.google.com/android/repository/android-ndk-r14b-linux-x86_64.zip && \
+    unzip -q android-ndk-r14b-linux-x86_64.zip && \
+    mv android-ndk-r14b ${ANDROID_NDK_HOME} && \
+    rm -rf /opt/android-ndk-tmp
+CMD ["bash", "/paddle/paddle/scripts/docker/build_android.sh"]
--- a/README.md
+++ b/README.md
@@ -2,8 +2,8 @@
 [![Build Status](https://travis-ci.org/PaddlePaddle/Paddle.svg?branch=develop)](https://travis-ci.org/PaddlePaddle/Paddle)
-[![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://www.paddlepaddle.org/develop/doc/)
+[![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://doc.paddlepaddle.org/develop/doc/)
-[![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://www.paddlepaddle.org/doc_cn/)
+[![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://doc.paddlepaddle.org/develop/doc_cn/)
 [![Coverage Status](https://coveralls.io/repos/github/PaddlePaddle/Paddle/badge.svg?branch=develop)](https://coveralls.io/github/PaddlePaddle/Paddle?branch=develop)
 [![Release](https://img.shields.io/github/release/PaddlePaddle/Paddle.svg)](https://github.com/PaddlePaddle/Paddle/releases)
 [![License](https://img.shields.io/badge/license-Apache%202-blue.svg)](LICENSE)
@@ -51,45 +51,46 @@ Please refer to our [release announcement](https://github.com/PaddlePaddle/Paddl
 - **Connected to Products**
    In addition, PaddlePaddle is also designed to be easily deployable. At Baidu,
-    PaddlePaddle has been deployed into products or service with a vast number
+    PaddlePaddle has been deployed into products and services with a vast number
    of users, including ad click-through rate (CTR) prediction, large-scale image
    classification, optical character recognition(OCR), search ranking, computer
    virus detection, recommendation, etc. It is widely utilized in products at
-    Baidu and it has achieved a significant impact. We hope you can also exploit
+    Baidu and it has achieved a significant impact. We hope you can also explore
-    the capability of PaddlePaddle to make a huge impact for your product.
+    the capability of PaddlePaddle to make an impact on your product.
 ## Installation
 It is recommended to check out the
-[Docker installation guide](http://www.paddlepaddle.org/develop/doc/getstarted/build_and_install/docker_install_en.html)
+[Docker installation guide](http://doc.paddlepaddle.org/develop/doc/getstarted/build_and_install/docker_install_en.html)
 before looking into the
-[build from source guide](http://www.paddlepaddle.org/develop/doc/getstarted/build_and_install/build_from_source_en.html)
+[build from source guide](http://doc.paddlepaddle.org/develop/doc/getstarted/build_and_install/build_from_source_en.html).
 ## Documentation
-We provide [English](http://www.paddlepaddle.org/develop/doc/) and
+We provide [English](http://doc.paddlepaddle.org/develop/doc/) and
-[Chinese](http://www.paddlepaddle.org/doc_cn/) documentation.
+[Chinese](http://doc.paddlepaddle.org/doc_cn/) documentation.
- [Deep Learning 101](http://book.paddlepaddle.org/index.en.html)
+- [Deep Learning 101](http://book.paddlepaddle.org/index.html)
-  You might want to start from the this online interactive book that can run in Jupyter Notebook.
+  You might want to start from this online interactive book that can run in a Jupyter Notebook.
- [Distributed Training](http://www.paddlepaddle.org/develop/doc/howto/usage/cluster/cluster_train_en.html)
+- [Distributed Training](http://doc.paddlepaddle.org/develop/doc/howto/usage/cluster/cluster_train_en.html)
  You can run distributed training jobs on MPI clusters.
- [Distributed Training on Kubernetes](http://www.paddlepaddle.org/develop/doc/howto/usage/k8s/k8s_en.html)
+- [Distributed Training on Kubernetes](http://doc.paddlepaddle.org/develop/doc/howto/usage/k8s/k8s_en.html)
   You can also run distributed training jobs on Kubernetes clusters.
- [Python API](http://www.paddlepaddle.org/develop/doc/api/index_en.html)
+- [Python API](http://doc.paddlepaddle.org/develop/doc/api/index_en.html)
   Our new API enables much shorter programs.
- [How to Contribute](http://www.paddlepaddle.org/develop/doc/howto/dev/contribute_to_paddle_en.html)
+- [How to Contribute](http://doc.paddlepaddle.org/develop/doc/howto/dev/contribute_to_paddle_en.html)
   We appreciate your contributions!
 ## Ask Questions
 You are welcome to submit questions and bug reports as [Github Issues](https://github.com/PaddlePaddle/Paddle/issues).

--- a/RELEASE.cn.md
+++ b/RELEASE.cn.md
+# v0.11.0版本
+## PaddlePaddle Fluid
+- PaddlePaddle发布版本v0.11.0包含一个新的特性*PaddlePaddle Fluid*. Fluid 是设计用来让用户像Pytorch和Tensorflow Eager Execution一样执行程序。在这些系统中，不再有*模型*这个概念，应用也不再包含一个用于描述Operator图或者一系列层的符号描述，而是像通用程序那样描述训练或者预测的过程。而Fluid与PyTorch或Eager Execution的区别在于Fluid不依赖Python提供的控制流，例如 if-else-then或者for，而是提供了基于C++实现的控制流并暴露了对应的用with语法实现的Python接口。例如：
+  https://github.com/PaddlePaddle/Paddle/blob/3df78ed2a98d37f7ae6725894cc7514effd5664b/python/paddle/v2/fluid/tests/test_while_op.py#L36-L44
+- 在v0.11.0版本中，我们提供了一个C++类`Executor`用于运行一个Fluid程序。Executor类似一个解释器。在未来的版本中，我们将提升和优化Executor成为一个调试器，就像GDB。并可能提供一些编译器，这个编译器会读取一个上文所描述的应用然后编译成一个等价的
+源代码，这个源代码可以被nvcc编译成可以使用CUDA的二进制，或者被icc编译成可以充分利用Intel CPU的二进制。
+## 新特点
+* 发布 `PaddlePaddle Fluid`。
+* 增加了用于模型预测的C-API。
+* 用Fluid API实现了一个简单的GAN的例子。
+* 增加了关于性能调优的文档。
+* 为`paddle.v2.dataset`下载数据集提供了重试机制.
+* C++中使用protobuf-lite替换protobuf减少了二进制的大小。
+* 发布了新特性 [Elastic Deep Learning (EDL)](https://github.com/PaddlePaddle/cloud/tree/develop/doc/autoscale/experiment).
+* 基于Bazel API利用cmake实现了一个的新的构建系统函数库。
+* 当使用编译选项`WITH_MKL=ON`时自动下载和编译Intel® [MKLML](https://github.com/01org/mkl-dnn/releases/download/v0.11/mklml_lnx_2018.0.1.20171007.tgz) 函数库.
+* [Intel® MKL-DNN on PaddlePaddle](https://github.com/PaddlePaddle/Paddle/tree/develop/doc/design/mkldnn):
+  - 完成了 11个 MKL-DNN 层: Convolution, Fully connectivity, Pooling, ReLU, Tanh, ELU, Softmax, BatchNorm, AddTo, Concat, LRN。
+  - 完成了 3个 MKL-DNN 网络: VGG-19, ResNet-50, GoogleNet
+  - 基于Intel Skylake 6148 CPU的[性能测试](https://github.com/PaddlePaddle/Paddle/blob/develop/benchmark/IntelOptimizedPaddle.md) : 相对于MKLML有2~3倍的训练加速。
+* 增加 [softsign activation](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/activation.html#softsign)
+* 增加 [dot product layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#dot-prod)
+* 增加 [L2 distance layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#l2-distance)
+* 增加 [sub-nested sequence layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#sub-nested-seq)
+* 增加 [kmax sequence score layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#kmax-sequence-score)
+* 增加 [sequence slice layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#seq-slice)
+* 增加 [row convolution layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#row-conv)
+* 增加移动端友好的网页
+## 改进
+* 使用一个Python`whl`包即可安装.
+* [V2 API可以实现用户定制化评估](https://github.com/PaddlePaddle/models/tree/develop/ltr#训练过程中输出自定义评估指标)。
+* 将 `PADDLE_ONLY_CPU` 改为 `PADDLE_WITH_GPU`, 因为我们会支持多种设备。
+* 删除了有一些bug的BarrierStat。
+* 清理和删除了paddle::Parameter中未使用的函数。
+* 删除了ProtoDataProvider。
+* Huber loss同时支持回归和分类。
+* 为sequence pooling 层增加`stride`参数。
+* v2 API自动使用cudnn batch normalization。
+* 可以使用一个固定的参数名共享BN层的参数。
+* 2D convolution operation支持variable-dimension input特性。
+* 重构cmake中关于CUDA的部分并实现自动检测GPU架构的功能。
+* 优化网页导航。
+## 错误修复
+* 修复ROI pooling的Bug. cc9a761
+* 修复当label是dense vector是AUC变成0的问题. #5274
+* 修复WarpCTC 层的Bug.
 # v0.10.0版本
 我们非常高兴发布了PaddlePaddle V0.10.0版，并开发了新的[Python API](http://research.baidu.com/paddlepaddles-new-api-simplifies-deep-learning-programs/)。

--- a/RELEASE.md
+++ b/RELEASE.md
+# Release v0.11.0
+## PaddlePaddle Fluid
+- Release 0.11.0 includes a new feature *PaddlePaddle Fluid*.  Fluid is
+  designed to allow users to program like PyTorch and TensorFlow Eager Execution.
+  In these systems, there is no longer the concept *model* and applications
+  do not include a symbolic description of a graph of operators nor a sequence
+  of layers. Instead, applications look exactly like a usual program that
+  describes a process of training or inference.  The difference between
+  Fluid and PyTorch or Eager Execution is that Fluid doesn't rely on Python's
+  control-flow, `if-then-else` nor `for`.  Instead, Fluid provides its
+  C++ implementations and their Python binding using the `with` statement.  For an example
+  https://github.com/PaddlePaddle/Paddle/blob/3df78ed2a98d37f7ae6725894cc7514effd5664b/python/paddle/v2/fluid/tests/test_while_op.py#L36-L44
+- In 0.11.0, we provides a C++ class `Executor` to run a Fluid program.
+Executor works like an interpreter. In future version, we will improve
+`Executor` into a debugger like GDB, and we might provide some compilers,
+which, for example, takes an application like the above one, and outputs
+an equivalent C++ source program, which can be compiled using
+[`nvcc`](http://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html)
+to generate binaries that use CUDA, or using
+[`icc`](https://software.intel.com/en-us/c-compilers) to generate binaries
+that make full use of Intel CPUs.
+## New Features
+* Release `PaddlePaddle Fluid`.
+* Add C-API for model inference
+* Use fluid API to create a simple GAN demo.
+* Add develop guide about performance tunning.
+* Add retry when download `paddle.v2.dataset`.
+* Linking protobuf-lite not protobuf in C++. Reduce the binary size.
+* Feature [Elastic Deep Learning (EDL)](https://github.com/PaddlePaddle/cloud/tree/develop/doc/autoscale/experiment) released.
+* A new style cmake functions for Paddle. It is based on Bazel API.
+* Automatically download and compile with Intel® [MKLML](https://github.com/01org/mkl-dnn/releases/download/v0.11/mklml_lnx_2018.0.1.20171007.tgz) library as CBLAS when build `WITH_MKL=ON`.
+* [Intel® MKL-DNN on PaddlePaddle](https://github.com/PaddlePaddle/Paddle/tree/develop/doc/design/mkldnn):
+  - Complete 11 MKL-DNN layers: Convolution, Fully connectivity, Pooling, ReLU, Tanh, ELU, Softmax, BatchNorm, AddTo, Concat, LRN.
+  - Complete 3 MKL-DNN networks: VGG-19, ResNet-50, GoogleNet
+  - [Benchmark](https://github.com/PaddlePaddle/Paddle/blob/develop/benchmark/IntelOptimizedPaddle.md) on Intel Skylake 6148 CPU: 2~3x training speedup compared with MKLML.
+* Add the [`softsign` activation](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/activation.html#softsign).
+* Add the [dot product layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#dot-prod).
+* Add the [L2 distance layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#l2-distance).
+* Add the [sub-nested sequence layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#sub-nested-seq).
+* Add the [kmax sequence score layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#kmax-sequence-score).
+* Add the [sequence slice layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#seq-slice).
+* Add the [row convolution layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#row-conv)
+* Add mobile friendly webpages.
+## Improvements
+* Build and install using a single `whl` package.
+* [Custom evaluating in V2 API](https://github.com/PaddlePaddle/models/tree/develop/ltr#训练过程中输出自定义评估指标).
+* Change `PADDLE_ONLY_CPU` to `PADDLE_WITH_GPU`, since we will support many kinds of devices.
+* Remove buggy BarrierStat.
+* Clean and remove unused functions in paddle::Parameter.
+* Remove ProtoDataProvider.
+* Huber loss supports both regression and classification.
+* Add the `stride` parameter  for sequence pooling layers.
+* Enable v2 API use cudnn batch normalization automatically.
+* The BN layer's parameter can be shared by a fixed the parameter name.
+* Support variable-dimension input feature for 2D convolution operation.
+* Refine cmake about CUDA to automatically detect GPU architecture.
+* Improved website navigation.
+## Bug Fixes
+* Fix bug in ROI pooling. cc9a761
+* Fix AUC is zero when label is dense vector. #5274
+* Fix bug in WarpCTC layer.
 # Release v0.10.0
-We are glad to release version 0.10.0.  In this version, we are happy to
+We are glad to release version 0.10.0.  In this version, we are happy to release the new 
-release the
-new
 [Python API](http://research.baidu.com/paddlepaddles-new-api-simplifies-deep-learning-programs/).
 - Our old Python API is kind of out of date.  It's hard to learn and hard to

--- a/authors
+++ b/authors
-Cao, Ying
-Cheng, Yujuan
-Dang, Qingqing
-Dong, Tengfei
-Du, Dalong
-Feng, Shouqiang
-Gao, Haoyuan
-Han, Baochang
-Han, Jinchen
-Hao, Nanyu
-He, Daoyuan
-He, Zhengyan
-Hou, Jue
-Huang, Chang
-Huang, Zhiheng
-Hu, Na
-Kong, Qi
-Liao, Gang
-Li, Bo
-Li, Jiajie
-Li, Jing
-Li, Lei
-Li, Peng
-Liu, Sheng
-Liu, Yuan
-Li, Yuze
-Luo, Heng
-Luo, Tao
-Lyu, Qin
-Mao, Hongyue
-Qian, Xiaojun
-Qiao, Longfei
-Qi, Jun
-Qin, Duohao
-Shen, Guolong
-Shi, Guangchuan
-Song, Xiang
-Wang, Helin
-Wang, Jiang
-Wang, Yanfei
-Wang, Yi
-Wang, Yong
-Weng, Renliang
-Xu, Tianbing
-Xu, Wei
-Xu, Xingyu
-Yan, Chong
-Yan, Chunwei
-Yang, Yi
-Yu, Yang
-Yu, Yinan
-Zhang, Jian
-Zhang, Ruiqing
-Zhang, Weide
-Zhao, Liang
-Zhou, Jie
--- a/benchmark/IntelOptimizedPaddle.md
+++ b/benchmark/IntelOptimizedPaddle.md
+# Benchmark
+Machine:
+- Server: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz, 2 Sockets, 20 Cores per socket
+- Laptop: TBD
+System: CentOS release 6.3 (Final), Docker 1.12.1.
+PaddlePaddle: (TODO: will rerun after 0.11.0)
+- paddlepaddle/paddle:latest (for MKLML and MKL-DNN)
+  - MKL-DNN tag v0.11
+  - MKLML 2018.0.1.20171007
+- paddlepaddle/paddle:latest-openblas (for OpenBLAS)
+  - OpenBLAS v0.2.20
+On each machine, we will test and compare the performance of training on single node using MKL-DNN / MKLML / OpenBLAS respectively.
+## Benchmark Model
+### Server
+Test on batch size 64, 128, 256 on Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
+Input image size - 3 * 224 * 224, Time: images/second
+- VGG-19
+| BatchSize    | 64    | 128  | 256     |
+|--------------|-------| -----| --------|
+| OpenBLAS     | 7.80  | 9.00  | 10.80  | 
+| MKLML        | 12.12 | 13.70 | 16.18  |
+| MKL-DNN      | 28.46 | 29.83 | 30.44  |
+<img src="figs/vgg-cpu-train.png" width="500">
+ - ResNet-50
+| BatchSize    | 64    | 128   | 256    |
+|--------------|-------| ------| -------|
+| OpenBLAS     | 25.22 | 25.68 | 27.12  | 
+| MKLML        | 32.52 | 31.89 | 33.12  |
+| MKL-DNN      | 81.69 | 82.35 | 84.08  |
+<img src="figs/resnet-cpu-train.png" width="500">
+ - GoogLeNet
+| BatchSize    | 64    | 128   | 256    |
+|--------------|-------| ------| -------|
+| OpenBLAS     | 89.52 | 96.97 | 108.25 | 
+| MKLML        | 128.46| 137.89| 158.63 |
+| MKL-DNN      | 250.46| 264.83| 269.50 |
+<img src="figs/googlenet-cpu-train.png" width="500">
+### Laptop
+TBD
--- a/benchmark/figs/googlenet-cpu-train.png
+++ b/benchmark/figs/googlenet-cpu-train.png
--- a/benchmark/figs/resnet-cpu-train.png
+++ b/benchmark/figs/resnet-cpu-train.png
--- a/benchmark/figs/vgg-cpu-train.png
+++ b/benchmark/figs/vgg-cpu-train.png
--- a/benchmark/paddle/image/googlenet.py
+++ b/benchmark/paddle/image/googlenet.py
@@ -5,10 +5,22 @@ height = 224
 width = 224
 num_class = 1000
 batch_size = get_config_arg('batch_size', int, 128)
+use_gpu = get_config_arg('use_gpu', bool, True)
-args = {'height': height, 'width': width, 'color': True, 'num_class': num_class}
+is_infer = get_config_arg("is_infer", bool, False)
+args = {
+    'height': height,
+    'width': width,
+    'color': True,
+    'num_class': num_class,
+    'is_infer': is_infer
+}
 define_py_data_sources2(
-    "train.list", None, module="provider", obj="process", args=args)
+    "train.list" if not is_infer else None,
+    "test.list" if is_infer else None,
+    module="provider",
+    obj="process",
+    args=args)
 settings(
    batch_size=batch_size,
@@ -16,6 +28,8 @@ settings(
    learning_method=MomentumOptimizer(0.9),
    regularization=L2Regularization(0.0005 * batch_size))
+conv_projection = conv_projection if use_gpu else img_conv_layer
 def inception2(name, input, channels, \
    filter1,
    filter3R, filter3,
@@ -138,12 +152,11 @@ def inception(name, input, channels, \
    cat = concat_layer(
        name=name,
        input=[cov1, cov3, cov5, covprj],
-        bias_attr=True,
+        bias_attr=True if use_gpu else False,
        act=ReluActivation())
    return cat
-lab = data_layer(name="label", size=1000)
 data = data_layer(name="input", size=3 * height * width)
 # stage 1
@@ -221,6 +234,10 @@ pool5 = img_pool_layer(
 dropout = dropout_layer(name="dropout", input=pool5, dropout_rate=0.4)
 out3 = fc_layer(
    name="output3", input=dropout, size=1000, act=SoftmaxActivation())
-loss3 = cross_entropy(name='loss3', input=out3, label=lab)
-outputs(loss3)
+if is_infer:
+    outputs(out3)
+else:
+    lab = data_layer(name="label", size=num_class)
+    loss3 = cross_entropy(name='loss3', input=out3, label=lab)
+    outputs(loss3)
--- a/benchmark/paddle/image/provider.py
+++ b/benchmark/paddle/image/provider.py
@@ -13,14 +13,20 @@ def initHook(settings, height, width, color, num_class, **kwargs):
        settings.data_size = settings.height * settings.width * 3
    else:
        settings.data_size = settings.height * settings.width
+    settings.is_infer = kwargs.get('is_infer', False)
-    settings.slots = [dense_vector(settings.data_size), integer_value(1)]
+    if settings.is_infer:
+        settings.slots = [dense_vector(settings.data_size)]
+    else:
+        settings.slots = [dense_vector(settings.data_size), integer_value(1)]
 @provider(
    init_hook=initHook, min_pool_size=-1, cache=CacheType.CACHE_PASS_IN_MEM)
 def process(settings, file_list):
-    for i in xrange(1024):
+    for i in xrange(2560 if settings.is_infer else 1024):
        img = np.random.rand(1, settings.data_size).reshape(-1, 1).flatten()
-        lab = random.randint(0, settings.num_class)
+        if settings.is_infer:
-        yield img.astype('float32'), int(lab)
+            yield img.astype('float32')
+        else:
+            lab = random.randint(0, settings.num_class - 1)
+            yield img.astype('float32'), int(lab)
--- a/benchmark/paddle/image/resnet.py
+++ b/benchmark/paddle/image/resnet.py
+#!/usr/bin/env python
+from paddle.trainer_config_helpers import *
+height = 224
+width = 224
+num_class = 1000
+batch_size = get_config_arg('batch_size', int, 64)
+layer_num = get_config_arg("layer_num", int, 50)
+is_infer = get_config_arg("is_infer", bool, False)
+args = {
+    'height': height,
+    'width': width,
+    'color': True,
+    'num_class': num_class,
+    'is_infer': is_infer
+}
+define_py_data_sources2(
+    "train.list" if not is_infer else None,
+    "test.list" if is_infer else None,
+    module="provider",
+    obj="process",
+    args=args)
+settings(
+    batch_size=batch_size,
+    learning_rate=0.01 / batch_size,
+    learning_method=MomentumOptimizer(0.9),
+    regularization=L2Regularization(0.0005 * batch_size))
+#######################Network Configuration #############
+def conv_bn_layer(name,
+                  input,
+                  filter_size,
+                  num_filters,
+                  stride,
+                  padding,
+                  channels=None,
+                  active_type=ReluActivation()):
+    """
+    A wrapper for conv layer with batch normalization layers.
+    Note:
+    conv layer has no activation.
+    """
+    tmp = img_conv_layer(
+        name=name + "_conv",
+        input=input,
+        filter_size=filter_size,
+        num_channels=channels,
+        num_filters=num_filters,
+        stride=stride,
+        padding=padding,
+        act=LinearActivation(),
+        bias_attr=False)
+    return batch_norm_layer(
+        name=name + "_bn",
+        input=tmp,
+        act=active_type,
+        use_global_stats=is_infer)
+def bottleneck_block(name, input, num_filters1, num_filters2):
+    """
+    A wrapper for bottlenect building block in ResNet.
+    Last conv_bn_layer has no activation.
+    Addto layer has activation of relu.
+    """
+    last_name = conv_bn_layer(
+        name=name + '_branch2a',
+        input=input,
+        filter_size=1,
+        num_filters=num_filters1,
+        stride=1,
+        padding=0)
+    last_name = conv_bn_layer(
+        name=name + '_branch2b',
+        input=last_name,
+        filter_size=3,
+        num_filters=num_filters1,
+        stride=1,
+        padding=1)
+    last_name = conv_bn_layer(
+        name=name + '_branch2c',
+        input=last_name,
+        filter_size=1,
+        num_filters=num_filters2,
+        stride=1,
+        padding=0,
+        active_type=LinearActivation())
+    return addto_layer(
+        name=name + "_addto", input=[input, last_name], act=ReluActivation())
+def mid_projection(name, input, num_filters1, num_filters2, stride=2):
+    """
+    A wrapper for middile projection in ResNet.
+    projection shortcuts are used for increasing dimensions,
+    and other shortcuts are identity
+    branch1: projection shortcuts are used for increasing
+    dimensions, has no activation.
+    branch2x: bottleneck building block, shortcuts are identity.
+    """
+    # stride = 2
+    branch1 = conv_bn_layer(
+        name=name + '_branch1',
+        input=input,
+        filter_size=1,
+        num_filters=num_filters2,
+        stride=stride,
+        padding=0,
+        active_type=LinearActivation())
+    last_name = conv_bn_layer(
+        name=name + '_branch2a',
+        input=input,
+        filter_size=1,
+        num_filters=num_filters1,
+        stride=stride,
+        padding=0)
+    last_name = conv_bn_layer(
+        name=name + '_branch2b',
+        input=last_name,
+        filter_size=3,
+        num_filters=num_filters1,
+        stride=1,
+        padding=1)
+    last_name = conv_bn_layer(
+        name=name + '_branch2c',
+        input=last_name,
+        filter_size=1,
+        num_filters=num_filters2,
+        stride=1,
+        padding=0,
+        active_type=LinearActivation())
+    return addto_layer(
+        name=name + "_addto", input=[branch1, last_name], act=ReluActivation())
+img = data_layer(name='image', size=height * width * 3)
+def deep_res_net(res2_num=3, res3_num=4, res4_num=6, res5_num=3):
+    """
+    A wrapper for 50,101,152 layers of ResNet.
+    res2_num: number of blocks stacked in conv2_x
+    res3_num: number of blocks stacked in conv3_x
+    res4_num: number of blocks stacked in conv4_x
+    res5_num: number of blocks stacked in conv5_x
+    """
+    # For ImageNet
+    # conv1: 112x112
+    tmp = conv_bn_layer(
+        "conv1",
+        input=img,
+        filter_size=7,
+        channels=3,
+        num_filters=64,
+        stride=2,
+        padding=3)
+    tmp = img_pool_layer(name="pool1", input=tmp, pool_size=3, stride=2)
+    # conv2_x: 56x56
+    tmp = mid_projection(
+        name="res2_1", input=tmp, num_filters1=64, num_filters2=256, stride=1)
+    for i in xrange(2, res2_num + 1, 1):
+        tmp = bottleneck_block(
+            name="res2_" + str(i), input=tmp, num_filters1=64, num_filters2=256)
+    # conv3_x: 28x28
+    tmp = mid_projection(
+        name="res3_1", input=tmp, num_filters1=128, num_filters2=512)
+    for i in xrange(2, res3_num + 1, 1):
+        tmp = bottleneck_block(
+            name="res3_" + str(i),
+            input=tmp,
+            num_filters1=128,
+            num_filters2=512)
+    # conv4_x: 14x14
+    tmp = mid_projection(
+        name="res4_1", input=tmp, num_filters1=256, num_filters2=1024)
+    for i in xrange(2, res4_num + 1, 1):
+        tmp = bottleneck_block(
+            name="res4_" + str(i),
+            input=tmp,
+            num_filters1=256,
+            num_filters2=1024)
+    # conv5_x: 7x7
+    tmp = mid_projection(
+        name="res5_1", input=tmp, num_filters1=512, num_filters2=2048)
+    for i in xrange(2, res5_num + 1, 1):
+        tmp = bottleneck_block(
+            name="res5_" + str(i),
+            input=tmp,
+            num_filters1=512,
+            num_filters2=2048)
+    tmp = img_pool_layer(
+        name='avgpool',
+        input=tmp,
+        pool_size=7,
+        stride=1,
+        pool_type=AvgPooling())
+    return fc_layer(input=tmp, size=num_class, act=SoftmaxActivation())
+if layer_num == 50:
+    resnet = deep_res_net(3, 4, 6, 3)
+elif layer_num == 101:
+    resnet = deep_res_net(3, 4, 23, 3)
+elif layer_num == 152:
+    resnet = deep_res_net(3, 8, 36, 3)
+else:
+    print("Wrong layer number.")
+if is_infer:
+    outputs(resnet)
+else:
+    lbl = data_layer(name="label", size=num_class)
+    loss = cross_entropy(name='loss', input=resnet, label=lbl)
+    outputs(loss)
--- a/benchmark/paddle/image/run_mkldnn_infer.sh
+++ b/benchmark/paddle/image/run_mkldnn_infer.sh
+set -e
+function clock_to_seconds() {
+  hours=`echo $1 | awk -F ':' '{print $1}'`
+  mins=`echo $1 | awk -F ':' '{print $2}'`
+  secs=`echo $1 | awk -F ':' '{print $3}'`
+  echo `awk 'BEGIN{printf "%.2f",('$secs' + '$mins' * 60 + '$hours' * 3600)}'`
+}
+function infer() {
+  unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
+  topology=$1
+  layer_num=$2
+  bs=$3
+  use_mkldnn=$4
+  if [ $4 == "True" ]; then
+    thread=1
+    log="logs/infer-${topology}-${layer_num}-mkldnn-${bs}.log"
+  elif [ $4 == "False" ]; then
+    thread=`nproc`
+    if [ $thread -gt $bs ]; then
+      thread=$bs
+    fi
+    log="logs/infer-${topology}-${layer_num}-${thread}mklml-${bs}.log"
+  else
+    echo "Wrong input $4, use True or False."
+    exit 0
+  fi
+  models_in="models/${topology}-${layer_num}/pass-00000/"
+  if [ ! -d $models_in ]; then
+    echo "Training model ${topology}_${layer_num}"
+    paddle train --job=train \
+      --config="${topology}.py" \
+      --use_mkldnn=True \
+      --use_gpu=False \
+      --trainer_count=1 \
+      --num_passes=1 \
+      --save_dir="models/${topology}-${layer_num}" \
+      --config_args="batch_size=128,layer_num=${layer_num}" \
+      > /dev/null 2>&1
+    echo "Done"
+  fi
+  log_period=$((256 / bs))
+  paddle train --job=test \
+    --config="${topology}.py" \
+    --use_mkldnn=$use_mkldnn \
+    --use_gpu=False \
+    --trainer_count=$thread \
+    --log_period=$log_period \
+    --config_args="batch_size=${bs},layer_num=${layer_num},is_infer=True" \
+    --init_model_path=$models_in \
+    2>&1 | tee ${log}
+  # calculate the last 5 logs period time of 1280 samples,
+  # the time before are burning time.
+  start=`tail ${log} -n 7 | head -n 1 | awk -F ' ' '{print $2}' | xargs`
+  end=`tail ${log} -n 2 | head -n 1 | awk -F ' ' '{print $2}' | xargs`
+  start_sec=`clock_to_seconds $start`
+  end_sec=`clock_to_seconds $end`
+  fps=`awk 'BEGIN{printf "%.2f",(1280 / ('$end_sec' - '$start_sec'))}'`
+  echo "Last 1280 samples start: ${start}(${start_sec} sec), end: ${end}(${end_sec} sec;" >> ${log}
+  echo "FPS: $fps images/sec" 2>&1 | tee -a ${log}
+}
+if [ ! -f "train.list" ]; then
+  echo " " > train.list
+fi
+if [ ! -f "test.list" ]; then
+  echo " " > test.list
+fi
+if [ ! -d "logs" ]; then
+  mkdir logs
+fi
+if [ ! -d "models" ]; then
+  mkdir -p models
+fi
+# inference benchmark
+for use_mkldnn in True False; do
+  for batchsize in 1 2 4 8 16; do
+    infer googlenet v1 $batchsize $use_mkldnn
+    infer resnet 50 $batchsize $use_mkldnn
+    infer vgg 19 $batchsize $use_mkldnn
+  done
+done
--- a/benchmark/paddle/image/run_mkldnn_train.sh
+++ b/benchmark/paddle/image/run_mkldnn_train.sh
+set -e
+function train() {
+  unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
+  topology=$1
+  layer_num=$2
+  bs=$3
+  use_mkldnn=$4
+  if [ $4 == "True" ]; then
+    thread=1
+    log="logs/train-${topology}-${layer_num}-mkldnn-${bs}.log"
+  elif [ $4 == "False" ]; then
+    thread=`nproc`
+    # each trainer_count use only 1 core to avoid conflict
+    log="logs/train-${topology}-${layer_num}-${thread}mklml-${bs}.log"
+  else
+    echo "Wrong input $4, use True or False."
+    exit 0
+  fi
+  args="batch_size=${bs},layer_num=${layer_num}"
+  config="${topology}.py"
+  paddle train --job=time \
+    --config=$config \
+    --use_mkldnn=$use_mkldnn \
+    --use_gpu=False \
+    --trainer_count=$thread \
+    --log_period=10 \
+    --test_period=100 \
+    --config_args=$args \
+    2>&1 | tee ${log} 
+}
+if [ ! -f "train.list" ]; then
+  echo " " > train.list
+fi
+if [ ! -d "logs" ]; then
+  mkdir logs
+fi
+# training benchmark
+for use_mkldnn in True False; do
+  for batchsize in 64 128 256; do
+    train vgg 19 $batchsize $use_mkldnn
+    train resnet 50 $batchsize $use_mkldnn
+    train googlenet v1 $batchsize $use_mkldnn
+  done
+done
--- a/benchmark/paddle/image/vgg.py
+++ b/benchmark/paddle/image/vgg.py
+#!/usr/bin/env python
+from paddle.trainer_config_helpers import *
+height = 224
+width = 224
+num_class = 1000
+batch_size = get_config_arg('batch_size', int, 64)
+layer_num = get_config_arg('layer_num', int, 19)
+is_infer = get_config_arg("is_infer", bool, False)
+args = {
+    'height': height,
+    'width': width,
+    'color': True,
+    'num_class': num_class,
+    'is_infer': is_infer
+}
+define_py_data_sources2(
+    "train.list" if not is_infer else None,
+    "test.list" if is_infer else None,
+    module="provider",
+    obj="process",
+    args=args)
+settings(
+    batch_size=batch_size,
+    learning_rate=0.001 / batch_size,
+    learning_method=MomentumOptimizer(0.9),
+    regularization=L2Regularization(0.0005 * batch_size))
+img = data_layer(name='image', size=height * width * 3)
+def vgg_network(vgg_num=3):
+    tmp = img_conv_group(
+        input=img,
+        num_channels=3,
+        conv_padding=1,
+        conv_num_filter=[64, 64],
+        conv_filter_size=3,
+        conv_act=ReluActivation(),
+        pool_size=2,
+        pool_stride=2,
+        pool_type=MaxPooling())
+    tmp = img_conv_group(
+        input=tmp,
+        conv_num_filter=[128, 128],
+        conv_padding=1,
+        conv_filter_size=3,
+        conv_act=ReluActivation(),
+        pool_stride=2,
+        pool_type=MaxPooling(),
+        pool_size=2)
+    channels = []
+    for i in range(vgg_num):
+        channels.append(256)
+    tmp = img_conv_group(
+        input=tmp,
+        conv_num_filter=channels,
+        conv_padding=1,
+        conv_filter_size=3,
+        conv_act=ReluActivation(),
+        pool_stride=2,
+        pool_type=MaxPooling(),
+        pool_size=2)
+    channels = []
+    for i in range(vgg_num):
+        channels.append(512)
+    tmp = img_conv_group(
+        input=tmp,
+        conv_num_filter=channels,
+        conv_padding=1,
+        conv_filter_size=3,
+        conv_act=ReluActivation(),
+        pool_stride=2,
+        pool_type=MaxPooling(),
+        pool_size=2)
+    tmp = img_conv_group(
+        input=tmp,
+        conv_num_filter=channels,
+        conv_padding=1,
+        conv_filter_size=3,
+        conv_act=ReluActivation(),
+        pool_stride=2,
+        pool_type=MaxPooling(),
+        pool_size=2)
+    tmp = fc_layer(
+        input=tmp,
+        size=4096,
+        act=ReluActivation(),
+        layer_attr=ExtraAttr(drop_rate=0.5))
+    tmp = fc_layer(
+        input=tmp,
+        size=4096,
+        act=ReluActivation(),
+        layer_attr=ExtraAttr(drop_rate=0.5))
+    return fc_layer(input=tmp, size=num_class, act=SoftmaxActivation())
+if layer_num == 16:
+    vgg = vgg_network(3)
+elif layer_num == 19:
+    vgg = vgg_network(4)
+else:
+    print("Wrong layer number.")
+if is_infer:
+    outputs(vgg)
+else:
+    lab = data_layer('label', num_class)
+    loss = cross_entropy(input=vgg, label=lab)
+    outputs(loss)
--- a/cmake/CMakeDetermineGoCompiler.cmake
+++ b/cmake/CMakeDetermineGoCompiler.cmake
+if(NOT CMAKE_Go_COMPILER)
+  if(NOT $ENV{GO_COMPILER} STREQUAL "")
+    get_filename_component(CMAKE_Go_COMPILER_INIT $ENV{GO_COMPILER} PROGRAM PROGRAM_ARGS CMAKE_Go_FLAGS_ENV_INIT)
+    if(CMAKE_Go_FLAGS_ENV_INIT)
+      set(CMAKE_Go_COMPILER_ARG1 "${CMAKE_Go_FLAGS_ENV_INIT}" CACHE STRING "First argument to Go compiler")
+    endif()
+    if(NOT EXISTS ${CMAKE_Go_COMPILER_INIT})
+      message(SEND_ERROR "Could not find compiler set in environment variable GO_COMPILER:\n$ENV{GO_COMPILER}.")
+    endif()
+  endif()
+  set(Go_BIN_PATH
+    $ENV{GOPATH}
+    $ENV{GOROOT}
+    $ENV{GOROOT}/bin
+    $ENV{GO_COMPILER}
+    /usr/bin
+    /usr/local/bin
+    )
+  if(CMAKE_Go_COMPILER_INIT)
+    set(CMAKE_Go_COMPILER ${CMAKE_Go_COMPILER_INIT} CACHE PATH "Go Compiler")
+  else()
+    find_program(CMAKE_Go_COMPILER
+      NAMES go
+      PATHS ${Go_BIN_PATH}
+    )
+    if(CMAKE_Go_COMPILER)
+      EXEC_PROGRAM(${CMAKE_Go_COMPILER} ARGS version OUTPUT_VARIABLE GOLANG_VERSION)
+      STRING(REGEX MATCH "go[0-9]+[.0-9]*[ /A-Za-z0-9]*" VERSION "${GOLANG_VERSION}")
+      message("-- The Golang compiler identification is ${VERSION}")
+      message("-- Check for working Golang compiler: ${CMAKE_Go_COMPILER}")
+    endif()
+  endif()
+endif()
+mark_as_advanced(CMAKE_Go_COMPILER)
+configure_file(${CMAKE_CURRENT_SOURCE_DIR}/cmake/CMakeGoCompiler.cmake.in
+  ${CMAKE_PLATFORM_INFO_DIR}/CMakeGoCompiler.cmake @ONLY)
+set(CMAKE_Go_COMPILER_ENV_VAR "GO_COMPILER")
--- a/cmake/CMakeGoCompiler.cmake.in
+++ b/cmake/CMakeGoCompiler.cmake.in
+set(CMAKE_Go_COMPILER "@CMAKE_Go_COMPILER@")
+set(CMAKE_Go_COMPILER_LOADED 1)
+set(CMAKE_Go_SOURCE_FILE_EXTENSIONS go)
+set(CMAKE_Go_LINKER_PREFERENCE 40)
+set(CMAKE_Go_OUTPUT_EXTENSION .o)
+set(CMAKE_Go_OUTPUT_EXTENSION_REPLACE 1)
+set(CMAKE_Go_COMPILER_ENV_VAR "GO_COMPILER")
--- a/cmake/CMakeGoInformation.cmake
+++ b/cmake/CMakeGoInformation.cmake
+if(NOT CMAKE_Go_COMPILE_OBJECT)
+  set(CMAKE_Go_COMPILE_OBJECT "go tool compile -l -N -o <OBJECT> <SOURCE> ")
+endif()
+if(NOT CMAKE_Go_LINK_EXECUTABLE)
+  set(CMAKE_Go_LINK_EXECUTABLE "go tool link -o <TARGET> <OBJECTS>  ")
+endif()
--- a/cmake/CMakeTestGoCompiler.cmake
+++ b/cmake/CMakeTestGoCompiler.cmake
+set(CMAKE_Go_COMPILER_WORKS 1 CACHE INTERNAL "")
--- a/cmake/cblas.cmake
+++ b/cmake/cblas.cmake
 # Find the CBlas and lapack libraries
 #
-# It will search MKL, atlas, OpenBlas, reference-cblas in order.
+# It will search MKLML, atlas, OpenBlas, reference-cblas in order.
 #
 # If any cblas implementation found, the following variable will be set.
-#    CBLAS_PROVIDER  # one of MKL, ATLAS, OPENBLAS, REFERENCE
+#    CBLAS_PROVIDER  # one of MKLML, ATLAS, OPENBLAS, REFERENCE
 #    CBLAS_INC_DIR   # the include directory for cblas.
-#    CBLAS_LIBS      # a list of libraries should be linked by paddle. 
+#    CBLAS_LIBS      # a list of libraries should be linked by paddle.
 #                    # Each library should be full path to object file.
-#
-# User should set one of MKL_ROOT, ATLAS_ROOT, OPENBLAS_ROOT, REFERENCE_CBLAS_ROOT
-# during cmake. If none of them set, it will try to find cblas implementation in
-# system paths.
-#
 set(CBLAS_FOUND OFF)
-## Find MKL First.
+## Find MKLML First.
-set(INTEL_ROOT "/opt/intel" CACHE PATH "Folder contains intel libs")
+if(WITH_MKLML AND MKLML_INC_DIR AND MKLML_LIB)
-set(MKL_ROOT ${INTEL_ROOT}/mkl CACHE PATH "Folder contains MKL")
-find_path(MKL_INC_DIR mkl.h PATHS
-  ${MKL_ROOT}/include)
-find_path(MKL_LAPACK_INC_DIR mkl_lapacke.h PATHS
-  ${MKL_ROOT}/include)
-find_library(MKL_CORE_LIB NAMES mkl_core PATHS
-  ${MKL_ROOT}/lib
-  ${MKL_ROOT}/lib/intel64)
-find_library(MKL_SEQUENTIAL_LIB NAMES mkl_sequential PATHS
-  ${MKL_ROOT}/lib
-  ${MKL_ROOT}/lib/intel64)
-find_library(MKL_INTEL_LP64 NAMES mkl_intel_lp64 PATHS
-  ${MKL_ROOT}/lib
-  ${MKL_ROOT}/lib/intel64)
-if(MKL_INC_DIR AND MKL_CORE_LIB AND MKL_SEQUENTIAL_LIB AND MKL_INTEL_LP64)
-  set(CBLAS_PROVIDER MKL)
-  set(CBLAS_INC_DIR ${MKL_INC_DIR})
-  set(CBLAS_LIBRARIES ${MKL_INTEL_LP64}
-          ${MKL_SEQUENTIAL_LIB}
-          ${MKL_CORE_LIB})
-  add_definitions(-DPADDLE_USE_MKL)
-  message(STATUS "Found MKL (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
  set(CBLAS_FOUND ON)
-  if(${MKL_LAPACK_INC_DIR})
+  set(CBLAS_PROVIDER MKLML)
-    add_definitions(-DPADDLE_USE_LAPACK)
+  set(CBLAS_INC_DIR ${MKLML_INC_DIR})
-    message(STATUS "Found lapack in MKL (include: ${MKL_LAPACK_INC_DIR})")
+  set(CBLAS_LIBRARIES ${MKLML_LIB})
-  endif()
-  return() # return file.
+  add_definitions(-DPADDLE_USE_MKLML)
+  add_definitions(-DLAPACK_FOUND)
+  message(STATUS "Found cblas and lapack in MKLML "
+    "(include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
+  return()
 endif()
 ## Then find atlas.
@@ -63,26 +38,26 @@ set(ATLAS_LIB_SEARCH_PATHS
        /usr/lib/atlas
        /usr/lib/atlas-base   # special for ubuntu 14.04.
    )
-find_path(ATLAS_INC_DIR NAMES cblas.h 
+find_path(ATLAS_INC_DIR NAMES cblas.h
  PATHS ${ATLAS_INCLUDE_SEARCH_PATHS})
 find_path(ATLAS_CLAPACK_INC_DIR NAMES clapack.h
  PATHS ${ATLAS_INCLUDE_SEARCH_PATHS})
-find_library(ATLAS_CBLAS_LIB NAMES cblas libcblas.so.3 
+find_library(ATLAS_CBLAS_LIB NAMES cblas libcblas.so.3
  PATHS ${ATLAS_LIB_SEARCH_PATHS})
-find_library(ATLAS_LIB NAMES lapack_atlas liblapack_atlas.so.3
+find_library(ATLAS_CLAPACK_LIB NAMES lapack_atlas liblapack_atlas.so.3
  PATHS ${ATLAS_LIB_SEARCH_PATHS})
-if(ATLAS_INC_DIR AND ATLAS_CBLAS_LIB AND ATLAS_LIB AND NOT CBLAS_FOUND)
+if(ATLAS_CLAPACK_INC_DIR AND ATLAS_INC_DIR AND ATLAS_CBLAS_LIB AND ATLAS_CLAPACK_LIB)
-  set(CBLAS_PROVIDER ATLAS)
-  set(CBLAS_INC_DIR ${ATLAS_INC_DIR})
-  set(CBLAS_LIBRARIES ${ATLAS_LIB} ${ATLAS_CBLAS_LIB})
-  add_definitions(-DPADDLE_USE_ATLAS)  
-  message(STATUS "Found ATLAS (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
  set(CBLAS_FOUND ON)
-  if(ATLAS_CLAPACK_INC_DIR)
+  set(CBLAS_PROVIDER ATLAS)
-    add_definitions(-DPADDLE_USE_LAPACK)
+  set(CBLAS_INC_DIR ${ATLAS_INC_DIR} ${ATLAS_CLAPACK_INC_DIR})
-    message(STATUS "Found lapack in ATLAS (include: ${ATLAS_CLAPACK_INC_DIR})")
+  set(CBLAS_LIBRARIES ${ATLAS_CLAPACK_LIB} ${ATLAS_CBLAS_LIB})
-  endif()
+  add_definitions(-DPADDLE_USE_ATLAS)
+  add_definitions(-DLAPACK_FOUND)
+  message(STATUS "Found ATLAS (include: ${ATLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
+  message(STATUS "Found lapack in ATLAS (include: ${ATLAS_CLAPACK_INC_DIR})")
  return()
 endif()
@@ -107,16 +82,17 @@ find_path(OPENBLAS_LAPACKE_INC_DIR NAMES lapacke.h
 find_library(OPENBLAS_LIB NAMES openblas
  PATHS ${OPENBLAS_LIB_SEARCH_PATHS})
-if(OPENBLAS_INC_DIR AND OPENBLAS_LIB)
+if(OPENBLAS_LAPACKE_INC_DIR AND OPENBLAS_INC_DIR AND OPENBLAS_LIB)
+  set(CBLAS_FOUND ON)
  set(CBLAS_PROVIDER OPENBLAS)
-  set(CBLAS_INC_DIR ${OPENBLAS_INC_DIR})
+  set(CBLAS_INC_DIR ${OPENBLAS_INC_DIR} ${OPENBLAS_LAPACKE_INC_DIR})
  set(CBLAS_LIBRARIES ${OPENBLAS_LIB})
-  message(STATUS "Found OpenBLAS (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
-  set(CBLAS_FOUND ON)
+  add_definitions(-DPADDLE_USE_OPENBLAS)
-  if(OPENBLAS_LAPACKE_INC_DIR)
+  add_definitions(-DLAPACK_FOUND)
-    add_definitions(-DPADDLE_USE_LAPACK)
-    message(STATUS "Found lapack in OpenBLAS (include: ${OPENBLAS_LAPACKE_INC_DIR})")
+  message(STATUS "Found OpenBLAS (include: ${OPENBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
-  endif()
+  message(STATUS "Found lapack in OpenBLAS (include: ${OPENBLAS_LAPACKE_INC_DIR})")
  return()
 endif()
@@ -124,7 +100,7 @@ endif()
 ## Then find the reference-cblas.  www.netlib.org/blas/
-set(REFERENCE_CBLAS_ROOT $ENV{REFERENCE_CBLAS_ROOT} CACHE PATH 
+set(REFERENCE_CBLAS_ROOT $ENV{REFERENCE_CBLAS_ROOT} CACHE PATH
  "Folder contains reference-cblas")
 set(REFERENCE_CBLAS_INCLUDE_SEARCH_PATHS
  ${REFERENCE_CBLAS_ROOT}/include
@@ -145,9 +121,17 @@ find_library(REFERENCE_CBLAS_LIBRARY NAMES cblas PATHS
        ${REFERENCE_CBLAS_LIB_SEARCH_PATHS})
 if (REFERENCE_CBLAS_INCLUDE_DIR AND REFERENCE_CBLAS_LIBRARY)
+  set(CBLAS_FOUND ON)
  set(CBLAS_PROVIDER REFERENCE)
  set(CBLAS_INC_DIR ${REFERENCE_CBLAS_INCLUDE_DIR})
  set(CBLAS_LIBRARIES ${REFERENCE_CBLAS_LIBRARY})
-  message(STATUS "Found reference-cblas (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBS})")
+  add_definitions(-DPADDLE_USE_REFERENCE_CBLAS)
+  message(STATUS "Found reference-cblas (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
+endif()
+if(IOS_USE_VECLIB_FOR_BLAS AND VECLIB_FOUND)
  set(CBLAS_FOUND ON)
+  set(CBLAS_PROVIDER vecLib)
+  set(CBLAS_INC_DIR ${VECLIB_INC_DIR})
+  add_definitions(-DPADDLE_USE_VECLIB)
 endif()
--- a/cmake/configure.cmake
+++ b/cmake/configure.cmake
 # Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
-# 
+#
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at
-# 
+#
 # http://www.apache.org/licenses/LICENSE-2.0
-# 
+#
 # Unless required by applicable law or agreed to in writing, software
 # distributed under the License is distributed on an "AS IS" BASIS,
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
@@ -24,10 +24,23 @@ if(WITH_DOUBLE)
    add_definitions(-DPADDLE_TYPE_DOUBLE)
 endif(WITH_DOUBLE)
+if(WITH_ARM_FP16)
+    add_definitions(-DPADDLE_ARM_FP16)
+    add_definitions("-march=armv8.2-a+fp16+simd")
+endif(WITH_ARM_FP16)
+if(WITH_TESTING)
+    add_definitions(-DPADDLE_WITH_TESTING)
+endif(WITH_TESTING)
 if(NOT WITH_TIMER)
    add_definitions(-DPADDLE_DISABLE_TIMER)
 endif(NOT WITH_TIMER)
+if(USE_EIGEN_FOR_BLAS)
+    add_definitions(-DPADDLE_USE_EIGEN_FOR_BLAS)
+endif(USE_EIGEN_FOR_BLAS)
 if(NOT WITH_PROFILER)
    add_definitions(-DPADDLE_DISABLE_PROFILER)
 endif(NOT WITH_PROFILER)
@@ -40,20 +53,25 @@ if(NOT CMAKE_CROSSCOMPILING)
    endif()
 endif()
+if(NOT WITH_GOLANG)
+    add_definitions(-DPADDLE_WITHOUT_GOLANG)
+endif(NOT WITH_GOLANG)
 if(NOT WITH_GPU)
-    add_definitions(-DPADDLE_ONLY_CPU)
    add_definitions(-DHPPL_STUB_FUNC)
    list(APPEND CMAKE_CXX_SOURCE_FILE_EXTENSIONS cu)
 else()
+    add_definitions(-DPADDLE_WITH_CUDA)
    FIND_PACKAGE(CUDA REQUIRED)
    if(${CUDA_VERSION_MAJOR} VERSION_LESS 7)
-        message(FATAL_ERROR "Paddle need CUDA >= 7.0 to compile")
+        message(FATAL_ERROR "Paddle needs CUDA >= 7.0 to compile")
    endif()
    if(NOT CUDNN_FOUND)
-        message(FATAL_ERROR "Paddle need cudnn to compile")
+        message(FATAL_ERROR "Paddle needs cudnn to compile")
    endif()
    set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler ${SIMD_FLAG}")
@@ -63,5 +81,63 @@ else()
    include_directories(${CUDA_TOOLKIT_INCLUDE})
 endif(NOT WITH_GPU)
+if (WITH_MKLML AND MKLML_IOMP_LIB)
+    message(STATUS "Enable Intel OpenMP with ${MKLML_IOMP_LIB}")
+    set(OPENMP_FLAGS "-fopenmp")
+    set(CMAKE_C_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
+    set(CMAKE_CXX_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OPENMP_FLAGS}")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OPENMP_FLAGS}")
+endif()
 set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SIMD_FLAG}")
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SIMD_FLAG}")
+if(WITH_GOLANG)
+  # we need to symlink Paddle directory into GOPATH. If we
+  # don't do it and we have code that depends on Paddle, go
+  # get ./... will download a new Paddle repo from Github,
+  # without the changes in our current Paddle repo that we
+  # want to build.
+  set(GOPATH "${CMAKE_CURRENT_BINARY_DIR}/go")
+  file(MAKE_DIRECTORY ${GOPATH})
+  set(PADDLE_IN_GOPATH "${GOPATH}/src/github.com/PaddlePaddle/Paddle")
+  file(MAKE_DIRECTORY "${PADDLE_IN_GOPATH}")
+  set(PADDLE_GO_PATH "${CMAKE_SOURCE_DIR}/go")
+  add_custom_target(go_path)
+  add_custom_command(TARGET go_path
+    # Symlink Paddle directory into GOPATH
+    COMMAND mkdir -p ${PADDLE_IN_GOPATH}
+    COMMAND rm -rf ${PADDLE_IN_GOPATH}
+    COMMAND ln -sf ${CMAKE_SOURCE_DIR} ${PADDLE_IN_GOPATH}
+    # Automatically get all dependencies specified in the source code
+    # We can't run `go get -d ./...` for every target, because
+    # multiple `go get` can not run concurrently, but make need to be
+    # able to run with multiple jobs.
+    WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
+  )
+  if (GLIDE_INSTALL)
+    if(EXISTS $ENV{GOPATH}/bin/glide)
+      set(GLIDE "$ENV{GOPATH}/bin/glide")
+    else()
+      message(FATAL_ERROR "no glide executeble found: $ENV{GOPATH}/bin/glide")
+    endif()
+    # this command will only run when the file it depends is missing
+    # or has changed, or the output is missing.
+    add_custom_command(OUTPUT ${CMAKE_BINARY_DIR}/glide
+      COMMAND env GOPATH=${GOPATH} ${GLIDE} install
+      COMMAND touch ${CMAKE_BINARY_DIR}/glide
+      DEPENDS ${PADDLE_SOURCE_DIR}/go/glide.lock
+      WORKING_DIRECTORY "${PADDLE_IN_GOPATH}/go"
+      )
+    # depends on the custom command which outputs
+    # ${CMAKE_BINARY_DIR}/glide, the custom command does not need to
+    # run every time this target is built.
+    add_custom_target(go_vendor DEPENDS ${CMAKE_BINARY_DIR}/glide go_path)
+  endif()
+endif(WITH_GOLANG)
--- a/cmake/cpplint.cmake
+++ b/cmake/cpplint.cmake
@@ -25,8 +25,10 @@ set(STYLE_FILTER "${STYLE_FILTER}-readability/casting")
 set(IGNORE_PATTERN
    .*ImportanceSampler.*
    .*cblas\\.h.*
-    .*LtrDataProvider.*
+    .*\\.pb\\.txt
-    .*MultiDataProvider.*)
+    .*MultiDataProvider.*
+    .*pb.*
+    .*pybind.h)
 # add_style_check_target
 #
@@ -34,29 +36,27 @@ set(IGNORE_PATTERN
 #
 # first argument: target name to attach
 # rest arguments: source list to check code style.
-# 
+#
 # NOTE: If WITH_STYLE_CHECK is OFF, then this macro just do nothing.
 macro(add_style_check_target TARGET_NAME)
    if(WITH_STYLE_CHECK)
        set(SOURCES_LIST ${ARGN})
        list(REMOVE_DUPLICATES SOURCES_LIST)
-        list(SORT SOURCES_LIST)
        foreach(filename ${SOURCES_LIST})
-            set(LINT ON)
            foreach(pattern ${IGNORE_PATTERN})
                if(filename MATCHES ${pattern})
-                    message(STATUS "DROP LINT ${filename}")
+                    list(REMOVE_ITEM SOURCES_LIST ${filename})
-                    set(LINT OFF)
+                endif()
-                endif() 
            endforeach()
-            if(LINT MATCHES ON)
-                add_custom_command(TARGET ${TARGET_NAME}
-                    PRE_BUILD
-                    COMMAND env ${py_env} "${PYTHON_EXECUTABLE}" "${PROJ_ROOT}/paddle/scripts/cpplint.py"
-                                "--filter=${STYLE_FILTER}" ${filename}
-                    WORKING_DIRECTORY ${CMAKE_CURRENT_LIST_DIR})
-            endif()
        endforeach()
+        if(SOURCES_LIST)
+            add_custom_command(TARGET ${TARGET_NAME} POST_BUILD
+                COMMAND "${PYTHON_EXECUTABLE}" "${PADDLE_SOURCE_DIR}/paddle/scripts/cpplint.py"
+                        "--filter=${STYLE_FILTER}"
+                        ${SOURCES_LIST}
+                COMMENT "cpplint: Checking source code style"
+                WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})        
+        endif()
    endif()
 endmacro()
--- a/cmake/cross_compiling/android.cmake
+++ b/cmake/cross_compiling/android.cmake
+# 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.
+# This is a toolchain file for cross-compiling for Android, and the
+# configuration refers to the open-source resposity:
+#     https://github.com/taka-no-me/android-cmake
+# Most of the variables are compatible with that used in
+#     https://developer.android.com/ndk/guides/cmake.html
+# The supported variables are listed belows:
+# 
+# ANDROID_STANDALONE_TOOLCHAIN
+# ANDROID_TOOLCHAIN
+# ANDROID_ABI
+# ANDROID_NATIVE_API_LEVEL
+# ANDROID_ARM_MODE
+# ANDROID_ARM_NEON
+#
+# For CMake >= 3.7.0, all the settings will be delivered to CMake system
+# variables to let CMake do the cross-compiling configurations itself.
+# More detail of cross-compiling settings
+#     https://cmake.org/cmake/help/v3.7/manual/cmake-toolchains.7.html
+IF(NOT ANDROID)
+    return()
+ENDIF()
+# check the exist of android standalone toolchain
+IF(NOT DEFINED ANDROID_STANDALONE_TOOLCHAIN)
+    SET(ANDROID_STANDALONE_TOOLCHAIN $ENV{ANDROID_STANDALONE_TOOLCHAIN}
+        CACHE PATH "Folder holds the standalone toolchain of Android NDK")
+ENDIF()
+IF(NOT ANDROID_STANDALONE_TOOLCHAIN)
+    MESSAGE(WARNING "It is recommended to set ANDROID_STANDALONE_TOOLCHAIN to "
+            "use a standalone toolchain.\n"
+            "To cross-compile for Android, you need to:\n"
+            "1. Download an Android NDK from"
+            " https://developer.android.com/ndk/downloads/index.html\n"
+            "2. Setup a standalone toolchain"
+            "https://developer.android.google.cn/ndk/guides/standalone_toolchain.html?hl=zh-cn\n")
+ENDIF()
+IF(NOT DEFINED CMAKE_SYSTEM_VERSION AND ANDROID_NATIVE_API_LEVEL)
+    IF(ANDROID_NATIVE_API_LEVEL MATCHES "^android-[0-9]+$")
+        STRING(REPLACE "android-" "" CMAKE_SYSTEM_VERSION "${CMAKE_MATCH_0}")
+    ELSEIF(ANDROID_NATIVE_API_LEVEL MATCHES "^[0-9]+$")
+        SET(CMAKE_SYSTEM_VERSION ${ANDROID_NATIVE_API_LEVEL})
+    ENDIF()
+ENDIF()
+IF(NOT DEFINED ANDROID_TOOLCHAIN)
+    SET(ANDROID_TOOLCHAIN clang)
+ENDIF()
+IF(NOT DEFINED ANDROID_ABI)
+    SET(ANDROID_ABI "armeabi-v7a")
+ENDIF()
+IF(NOT DEFINED ANDROID_ARM_MODE)
+    SET(ANDROID_ARM_MODE ON)
+ENDIF()
+IF(ANDROID_ARM_MODE)
+    SET(ANDROID_ARM_MODE_NAME "arm")
+ELSE(ANDROID_ARM_MODE)
+    SET(ANDROID_ARM_MODE_NAME "thumb")
+ENDIF(ANDROID_ARM_MODE)
+IF(NOT DEFINED ANDROID_ARM_NEON)
+    SET(ANDROID_ARM_NEON ON)
+ENDIF()
+IF("${CMAKE_VERSION}" VERSION_LESS "3.7.0")
+    IF("${CMAKE_VERSION}" VERSION_LESS "3.1.0")
+        SET(CMAKE_SYSTEM_NAME "Linux")
+    ENDIF()
+    MESSAGE(WARNING "It is recommended to use CMake >= 3.7.0 (current version: "
+            "${CMAKE_VERSION}), when cross-compiling for Android.")
+    IF(ANDROID_STANDALONE_TOOLCHAIN)
+        # Use standalone toolchain
+        SET(CMAKE_SYSROOT "${ANDROID_STANDALONE_TOOLCHAIN}/sysroot")
+        IF(NOT CMAKE_SYSTEM_VERSION)
+            SET(ANDROID_STANDALONE_TOOLCHAIN_API "")
+            SET(ANDROID_API_LEVEL_H_REGEX "^[\t ]*#[\t ]*define[\t ]+__ANDROID_API__[\t ]+([0-9]+)")
+            FILE(STRINGS "${ANDROID_STANDALONE_TOOLCHAIN}/sysroot/usr/include/android/api-level.h"
+                ANDROID_API_LEVEL_H_CONTENT REGEX "${ANDROID_API_LEVEL_H_REGEX}")
+            IF(ANDROID_API_LEVEL_H_CONTENT MATCHES "${ANDROID_API_LEVEL_H_REGEX}")
+                SET(ANDROID_STANDALONE_TOOLCHAIN_API "${CMAKE_MATCH_1}")
+            ENDIF()
+            SET(CMAKE_SYSTEM_VERSION ${ANDROID_STANDALONE_TOOLCHAIN_API})
+        ENDIF()
+        # Toolchain
+        SET(ANDROID_TOOLCHAIN_ROOT ${ANDROID_STANDALONE_TOOLCHAIN})
+    ELSE(ANDROID_NDK)
+        # TODO: use android ndk
+    ENDIF()
+    IF(ANDROID_ABI MATCHES "^armeabi(-v7a)?$")
+        SET(ANDROID_TOOLCHAIN_NAME arm-linux-androideabi)
+        IF(ANDROID_ABI STREQUAL "armeabi")
+            SET(CMAKE_SYSTEM_PROCESSOR armv5te)
+            SET(ANDROID_CLANG_TRIPLE armv5te-none-linux-androideabi)
+        ELSEIF(ANDROID_ABI STREQUAL "armeabi-v7a")
+            SET(CMAKE_SYSTEM_PROCESSOR armv7-a)
+            SET(ANDROID_CLANG_TRIPLE armv7-none-linux-androideabi)
+        ENDIF()
+    ELSEIF(ANDROID_ABI STREQUAL "arm64-v8a")
+        SET(ANDROID_TOOLCHAIN_NAME aarch64-linux-android)
+        SET(CMAKE_SYSTEM_PROCESSOR aarch64)
+        SET(ANDROID_CLANG_TRIPLE aarch64-none-linux-android)
+    ELSE()
+        MESSAGE(FATAL_ERROR "Invalid Android ABI: ${ANDROID_ABI}.")
+    ENDIF()
+    SET(ANDROID_TOOLCHAIN_PREFIX "${ANDROID_TOOLCHAIN_ROOT}/bin/${ANDROID_TOOLCHAIN_NAME}-")
+    IF(ANDROID_TOOLCHAIN STREQUAL clang)
+        SET(ANDROID_C_COMPILER_NAME clang)
+        SET(ANDROID_CXX_COMPILER_NAME clang++)
+        SET(CMAKE_C_COMPILER_TARGET   ${ANDROID_CLANG_TRIPLE})
+        SET(CMAKE_CXX_COMPILER_TARGET ${ANDROID_CLANG_TRIPLE})
+    ELSEIF(ANDROID_TOOLCHAIN STREQUAL gcc)
+        SET(ANDROID_C_COMPILER_NAME gcc)
+        SET(ANDROID_CXX_COMPILER_NAME g++)
+    ELSE()
+        MESSAGE(FATAL_ERROR "Invalid Android toolchain: ${ANDROID_TOOLCHAIN}")
+    ENDIF()
+    # C compiler
+    IF(NOT CMAKE_C_COMPILER)
+        SET(ANDROID_C_COMPILER "${ANDROID_TOOLCHAIN_PREFIX}${ANDROID_C_COMPILER_NAME}")
+    ELSE()
+        GET_FILENAME_COMPONENT(ANDROID_C_COMPILER ${CMAKE_C_COMPILER} PROGRAM)
+    ENDIF()
+    IF(NOT EXISTS ${ANDROID_C_COMPILER})
+        MESSAGE(FATAL_ERROR "Cannot find C compiler: ${ANDROID_C_COMPILER}")
+    ENDIF()
+    # CXX compiler
+    IF(NOT CMAKE_CXX_COMPILER)
+        SET(ANDROID_CXX_COMPILER "${ANDROID_TOOLCHAIN_PREFIX}${ANDROID_CXX_COMPILER_NAME}")
+    ELSE()
+        GET_FILENAME_COMPONENT(ANDROID_CXX_COMPILER ${CMAKE_CXX_COMPILER} PROGRAM)
+    ENDIF()
+    IF(NOT EXISTS ${ANDROID_CXX_COMPILER})
+        MESSAGE(FATAL_ERROR "Cannot find CXX compiler: ${ANDROID_CXX_COMPILER}")
+    ENDIF()
+    SET(CMAKE_C_COMPILER ${ANDROID_C_COMPILER} CACHE PATH "C compiler" FORCE)
+    SET(CMAKE_CXX_COMPILER ${ANDROID_CXX_COMPILER} CACHE PATH "CXX compiler" FORCE)
+    # Toolchain and ABI specific flags.
+    SET(ANDROID_COMPILER_FLAGS "-ffunction-sections -fdata-sections")
+    SET(ANDROID_LINKER_FLAGS "-Wl,--gc-sections")
+    IF(ANDROID_ABI STREQUAL "armeabi")
+        LIST(APPEND ANDROID_COMPILER_FLAGS
+             -march=armv5te
+             -mtune=xscale
+             -msoft-float)
+    ELSEIF(ANDROID_ABI STREQUAL "armeabi-v7a")
+        LIST(APPEND ANDROID_COMPILER_FLAGS
+             -march=armv7-a
+             -mfloat-abi=softfp)
+        IF(ANDROID_ARM_NEON)
+            LIST(APPEND ANDROID_COMPILER_FLAGS -mfpu=neon)
+        ELSE()
+            LIST(APPEND ANDROID_COMPILER_FLAGS -mfpu=vfpv3-d16)
+        ENDIF()
+        LIST(APPEND ANDROID_LINKER_FLAGS -Wl,--fix-cortex-a8)
+    ELSEIF(ANDROID_ABI STREQUAL "arm64-v8a")
+        LIST(APPEND ANDROID_COMPILER_FLAGS -march=armv8-a)
+    ENDIF()
+    IF(ANDROID_ABI MATCHES "^armeabi(-v7a)?$")
+        IF(ANDROID_ARM_MODE)
+            LIST(APPEND ANDROID_COMPILER_FLAGS -marm)
+        ELSE()
+            LIST(APPEND ANDROID_COMPILER_FLAGS -mthumb)
+        ENDIF()
+        IF(ANDROID_TOOLCHAIN STREQUAL clang)
+            # Disable integrated-as for better compatibility.
+            LIST(APPEND ANDROID_COMPILER_FLAGS -fno-integrated-as)
+        ENDIF()
+    ENDIF()
+    IF(ANDROID_TOOLCHAIN STREQUAL clang)
+        # CMake automatically forwards all compiler flags to the linker,
+        # and clang doesn't like having -Wa flags being used for linking.
+        # To prevent CMake from doing this would require meddling with
+        # the CMAKE_<LANG>_COMPILE_OBJECT rules, which would get quite messy.
+        LIST(APPEND ANDROID_LINKER_FLAGS -Qunused-arguments)
+    ENDIF()
+    STRING(REPLACE ";" " " ANDROID_COMPILER_FLAGS "${ANDROID_COMPILER_FLAGS}")
+    STRING(REPLACE ";" " " ANDROID_LINKER_FLAGS "${ANDROID_LINKER_FLAGS}")
+    SET(CMAKE_C_FLAGS "${ANDROID_COMPILER_FLAGS} ${CMAKE_C_FLAGS}"
+        CACHE STRING "C flags")
+    SET(CMAKE_CXX_FLAGS "${ANDROID_COMPILER_FLAGS} ${CMAKE_CXX_FLAGS}"
+        CACHE STRING "CXX flags")
+    SET(CMAKE_SHARED_LINKER_FLAGS "${ANDROID_LINKER_FLAGS} ${CMAKE_SHARED_LINKER_FLAGS}"
+        CACHE STRING "shared linker flags")
+    SET(CMAKE_POSITION_INDEPENDENT_CODE TRUE)
+    SET(CMAKE_EXE_LINKER_FLAGS "-pie -fPIE ${ANDROID_LINKER_FLAGS} ${CMAKE_EXE_LINKER_FLAGS}"
+        CACHE STRING "executable linker flags")
+    MESSAGE(STATUS "Android: Targeting API '${CMAKE_SYSTEM_VERSION}' "
+            "with architecture '${ANDROID_ARM_MODE_NAME}', "
+            "ABI '${ANDROID_ABI}', and processor '${CMAKE_SYSTEM_PROCESSOR}'")
+    MESSAGE(STATUS "System CMAKE_C_FLAGS: " ${CMAKE_C_FLAGS})
+    MESSAGE(STATUS "System CMAKE_CXX_FLAGS: " ${CMAKE_CXX_FLAGS})
+ELSE()
+    IF(ANDROID_STANDALONE_TOOLCHAIN)
+        SET(CMAKE_ANDROID_STANDALONE_TOOLCHAIN ${ANDROID_STANDALONE_TOOLCHAIN})
+    ENDIF()
+    SET(CMAKE_ANDROID_ARCH_ABI ${ANDROID_ABI})
+    IF(ANDROID_ABI MATCHES "^armeabi(-v7a)?$")
+        SET(CMAKE_ANDROID_ARM_MODE ${ANDROID_ARM_MODE})
+        IF(ANDROID_ABI STREQUAL "armeabi-v7a")
+            SET(CMAKE_ANDROID_ARM_NEON ${ANDROID_ARM_NEON})
+        ENDIF()
+    ENDIF()
+ENDIF()
--- a/cmake/cross_compiling/host.cmake
+++ b/cmake/cross_compiling/host.cmake
+# 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.
+# find host C compiler
+IF(HOST_C_COMPILER)
+    SET(HOST_C_COMPILER_NAME ${HOST_C_COMPILER})
+ELSEIF(NOT $ENV{CC} STREQUAL "")
+    SET(HOST_C_COMPILER_NAME $ENV{CC})
+ELSE()
+    SET(HOST_C_COMPILER_NAME cc)
+ENDIF()
+GET_FILENAME_COMPONENT(HOST_C_COMPILER_PATH ${HOST_C_COMPILER_NAME} PROGRAM)
+IF(NOT HOST_C_COMPILER_PATH OR NOT EXISTS ${HOST_C_COMPILER_PATH})
+    MESSAGE(FATAL_ERROR "Cannot find host C compiler, set host C compiler:\n"
+            "\tcmake .. -DHOST_C_COMPILER=...")
+ENDIF()
+# find host CXX compiler
+IF(HOST_CXX_COMPILER)
+    SET(HOST_CXX_COMPILER_NAME ${HOST_CXX_COMPILER})
+ELSEIF(NOT $ENV{CXX} STREQUAL "")
+    SET(HOST_CXX_COMPILER_NAME $ENV{CXX})
+ELSE()
+    SET(HOST_CXX_COMPILER_NAME c++)
+ENDIF()
+GET_FILENAME_COMPONENT(HOST_CXX_COMPILER_PATH ${HOST_CXX_COMPILER_NAME} PROGRAM)
+IF(NOT HOST_CXX_COMPILER_PATH OR NOT EXISTS ${HOST_CXX_COMPILER_PATH})
+    MESSAGE(FATAL_ERROR "Cannot find host CXX compiler, set host CXX compiler:\n"
+            "\tcmake .. -DHOST_CXX_COMPILER=...")
+ENDIF()
+SET(HOST_C_COMPILER ${HOST_C_COMPILER_PATH} CACHE PATH "Host C compiler")
+SET(HOST_CXX_COMPILER ${HOST_CXX_COMPILER_PATH} CACHE PATH "Host CXX compiler")
+MESSAGE(STATUS "Found host C compiler: " ${HOST_C_COMPILER})
+MESSAGE(STATUS "Found host CXX compiler: " ${HOST_CXX_COMPILER})
--- a/cmake/cross_compiling/ios.cmake
+++ b/cmake/cross_compiling/ios.cmake
+# 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.
+# This is a toolchain file for cross-compiling for iOS, and the
+# configuration largely refers to public toolchain file:
+#    https://raw.githubusercontent.com/leetal/ios-cmake/master/ios.toolchain.cmake
+# and
+#    https://github.com/cristeab/ios-cmake
+#
+# Supports options:
+# IOS_PLATFORM = OS (default) or SIMULATOR
+#   This decides if SDKS will be selected from the iPhoneOS.platform or iPhoneSimulator.platform folders
+#   OS - the default, used to build for iPhone and iPad physical devices, which have an arm arch.
+#   SIMULATOR - used to build for the Simulator platforms, which have an x86 arch.
+# IOS_ARCH
+#   The archectures wanted to support, such "arm64", "armv7;arm64"
+# IOS_DEPLOYMENT_TARGET
+#   The minimum iOS deployment version, such as "7.0"
+# IOS_ENABLE_BITCODE = ON (default) or OFF
+# IOS_USE_VECLIB_FOR_BLAS = OFF (default) or ON
+# IOS_DEVELOPER_ROOT = automatic(default) or /path/to/platform/Developer folder
+#   By default this location is automatcially chosen based on the IOS_PLATFORM value above.
+#   If set manually, it will override the default location and force the user of a particular Developer Platform
+# IOS_SDK_ROOT = automatic(default) or /path/to/platform/Developer/SDKs/SDK folder
+#   By default this location is automatcially chosen based on the IOS_DEVELOPER_ROOT value.
+#   In this case it will always be the most up-to-date SDK found in the IOS_DEVELOPER_ROOT path.
+#   If set manually, this will force the use of a specific SDK version
+# Macros:
+# set_xcode_property (TARGET XCODE_PROPERTY XCODE_VALUE)
+#  A convenience macro for setting xcode specific properties on targets
+#  example: set_xcode_property (myioslib IPHONEOS_DEPLOYMENT_TARGET "3.1")
+# find_host_package (PROGRAM ARGS)
+#  A macro used to find executable programs on the host system, not within the iOS environment.
+#  Thanks to the android-cmake project for providing the command
+if(NOT IOS)
+  return()
+endif()
+set(CMAKE_SYSTEM_NAME Darwin)
+# Get the Xcode version being used.
+execute_process(COMMAND xcodebuild -version
+                OUTPUT_VARIABLE XCODE_VERSION
+                RESULT_VARIABLE XCODE_VERSION_RESULT
+                ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
+if(NOT ${XCODE_VERSION_RESULT})
+  string(REGEX MATCH "Xcode [0-9\\.]+" XCODE_VERSION "${XCODE_VERSION}")
+  string(REGEX REPLACE "Xcode ([0-9\\.]+)" "\\1" XCODE_VERSION "${XCODE_VERSION}")
+  message(STATUS "Building with Xcode version: ${XCODE_VERSION}")
+else()
+  message(FATAL_ERROR "Cannot execute xcodebuild, please check whether xcode is installed.")
+endif()
+# Required as of cmake 2.8.10
+set(CMAKE_OSX_DEPLOYMENT_TARGET "" CACHE STRING "Force unset of the deployment target for iOS" FORCE)
+# Setup iOS platform unless specified manually with IOS_PLATFORM
+if(NOT DEFINED IOS_PLATFORM)
+  set(IOS_PLATFORM "OS")
+endif()
+set(IOS_PLATFORM ${IOS_PLATFORM} CACHE STRING "Type of iOS Platform")
+# Set the architecture for iOS
+if(NOT DEFINED IOS_ARCH)
+  if(IOS_PLATFORM STREQUAL "OS")
+    set(IOS_ARCH "armv7;armv7s;arm64")
+  elseif(IOS_PLATFORM STREQUAL "SIMULATOR")
+    set(IOS_ARCH "i386;x86_64")
+  endif()
+endif()
+set(CMAKE_OSX_ARCHITECTURES ${IOS_ARCH} CACHE string  "Build architecture for iOS")
+# Specify minimum iOS deployment version
+if(NOT DEFINED IOS_DEPLOYMENT_TARGET)
+  set(IOS_DEPLOYMENT_TARGET "7.0")
+endif()
+set(IOS_DEPLOYMENT_TARGET ${IOS_DEPLOYMENT_TARGET} CACHE STRING "Minimum iOS version")
+# Whether to enable bitcode
+if(NOT DEFINED IOS_ENABLE_BITCODE)
+  set(IOS_ENABLE_BITCODE ON)
+endif()
+set(IOS_ENABLE_BITCODE ${IOS_ENABLE_BITCODE} CACHE BOOL "Whether to enable bitcode")
+if(NOT DEFINED IOS_USE_VECLIB_FOR_BLAS)
+  set(IOS_USE_VECLIB_FOR_BLAS OFF)
+endif()
+set(IOS_USE_VECLIB_FOR_BLAS ${IOS_UES_VECLIB_FOR_BLAS} CACHE BOOL "Whether to use veclib")
+# Check the platform selection and setup for developer root
+if(${IOS_PLATFORM} STREQUAL "OS")
+  set(IOS_PLATFORM_LOCATION "iPhoneOS.platform")
+  set(XCODE_IOS_PLATFORM iphoneos)
+  # This causes the installers to properly locate the output libraries
+  set(CMAKE_XCODE_EFFECTIVE_PLATFORMS "-iphoneos")
+elseif(${IOS_PLATFORM} STREQUAL "SIMULATOR")
+  set(IOS_PLATFORM_LOCATION "iPhoneSimulator.platform")
+  set(XCODE_IOS_PLATFORM iphonesimulator)
+  # This causes the installers to properly locate the output libraries
+  set(CMAKE_XCODE_EFFECTIVE_PLATFORMS "-iphonesimulator")
+elseif(${IOS_PLATFORM} STREQUAL "WATCHOS")
+  set(IOS_PLATFORM_LOCATION "WatchOS.platform")
+  set(XCODE_IOS_PLATFORM watchos)
+  # This causes the installers to properly locate the output libraries
+  set(CMAKE_XCODE_EFFECTIVE_PLATFORMS "-watchos")
+else(${IOS_PLATFORM} STREQUAL "OS")
+  message(FATAL_ERROR "Unsupported IOS_PLATFORM value selected. Please set to\n"
+          "\t OS, SIMULATOR, or WATCHOS.")
+endif()
+# Check iOS developer toolchain
+if(NOT DEFINED IOS_DEVELOPER_ROOT)
+  # Setup iOS developer location
+  execute_process(COMMAND xcode-select -print-path
+                  OUTPUT_VARIABLE XCODE_DEVELOPER_DIR
+                  RESULT_VARIABLE XCODE_DEVELOPER_DIR_RESULT
+                  ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
+  # Xcode 4.3 changed the installation location, choose the most recent one available
+  if(${XCODE_VERSION} VERSION_LESS "4.3.0")
+    set(IOS_DEVELOPER_ROOT "/Developer/Platforms/${IOS_PLATFORM_LOCATION}/Developer")
+  else()
+    set(IOS_DEVELOPER_ROOT "${XCODE_DEVELOPER_DIR}/Platforms/${IOS_PLATFORM_LOCATION}/Developer")
+  endif()
+endif()
+if(EXISTS ${IOS_DEVELOPER_ROOT})
+  set(IOS_DEVELOPER_ROOT ${IOS_DEVELOPER_ROOT} CACHE PATH "Location of iOS Platform")
+else()
+  message(FATAL_ERROR "Invalid IOS_DEVELOPER_ROOT: ${IOS_DEVELOPER_ROOT} does not exist.")
+endif()
+# Check iOS SDK
+if(NOT DEFINED IOS_SDK_ROOT)
+  # Find and use the most recent iOS sdk
+  file(GLOB IOS_SDK_LISTS "${IOS_DEVELOPER_ROOT}/SDKs/*")
+  if(IOS_SDK_LISTS)
+    list(SORT IOS_SDK_LISTS)
+    list(REVERSE IOS_SDK_LISTS)
+    list(GET IOS_SDK_LISTS 0 IOS_SDK_ROOT)
+  else(IOS_SDK_LISTS)
+    message(FATAL_ERROR "No iOS SDK's found in default search path ${IOS_DEVELOPER_ROOT}."
+            " Please manually set IOS_SDK_ROOT or install the iOS SDK.")
+  endif(IOS_SDK_LISTS)
+endif()
+if(EXISTS ${IOS_SDK_ROOT})
+  set(IOS_SDK_ROOT ${IOS_SDK_ROOT} CACHE PATH "Location of the selected iOS SDK")
+  message(STATUS "iOS toolchain: ${IOS_SDK_ROOT}")
+else()
+  message(FATAL_ERROR "Invalid IOS_SDK_ROOT: ${IOS_SDK_ROOT} does not exist.")
+endif()
+# Set the sysroot default to the most recent SDK
+set(CMAKE_OSX_SYSROOT ${IOS_SDK_ROOT} CACHE PATH "Sysroot used for iOS support")
+# Get version of iOS SDK
+execute_process(COMMAND xcodebuild -sdk ${CMAKE_OSX_SYSROOT} -version SDKVersion
+                OUTPUT_VARIABLE IOS_SDK_VERSION
+                RESULT_VARIABLE IOS_SDK_VERSION_RESULT
+                ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
+if(${IOS_SDK_VERSION_RESULT})
+  string(REGEX MATCH "(([0-9]+)\\.)+([0-9]+)" IOS_SDK_VERSION "${IOS_SDK_ROOT}")
+endif()
+if(NOT IOS_SDK_VERSION)
+  message(WARNING "Cannot get SDK's version.")
+  set(IOS_SDK_VERSION 1)
+endif()
+set(CMAKE_SYSTEM_VERSION ${IOS_SDK_VERSION})
+# Find the C & C++ compilers for the specified SDK.
+if(NOT CMAKE_C_COMPILER)
+  # Default to use clang
+  execute_process(COMMAND xcrun -sdk ${CMAKE_OSX_SYSROOT} -find clang
+                  OUTPUT_VARIABLE IOS_C_COMPILER
+                  RESULT_VARIABLE IOS_C_COMPILER_RESULT
+                  ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
+  if(${IOS_C_COMPILER_RESULT})
+    get_filename_component(IOS_C_COMPILER clang PROGRAM)
+  endif()
+else(NOT CMAKE_C_COMPILER)
+  # User can set it in cmake command
+  get_filename_component(IOS_C_COMPILER ${CMAKE_C_COMPILER} PROGRAM)
+endif(NOT CMAKE_C_COMPILER)
+if(NOT EXISTS ${IOS_C_COMPILER})
+  message(FATAL_ERROR "Cannot find C compiler: ${IOS_C_COMPILER}")
+endif()
+if(NOT CMAKE_CXX_COMPILER)
+  # Default to use clang++
+  execute_process(COMMAND xcrun -sdk ${CMAKE_OSX_SYSROOT} -find clang++
+                  OUTPUT_VARIABLE IOS_CXX_COMPILER
+                  RESULT_VARIABLE IOS_CXX_COMPILER_RESULT
+                  ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
+  if(${IOS_CXX_COMPILER_RESULT})
+    get_filename_component(IOS_CXX_COMPILER clang++ PROGRAM)
+  endif()
+else(NOT CMAKE_CXX_COMPILER)
+  # User can set it in cmake command
+  get_filename_component(IOS_CXX_COMPILER ${CMAKE_CXX_COMPILER} PROGRAM)
+endif(NOT CMAKE_CXX_COMPILER)
+if(NOT EXISTS ${IOS_CXX_COMPILER})
+  message(FATAL_ERROR "Cannot find CXX compiler: ${IOS_CXX_COMPILER}")
+endif()
+set(CMAKE_C_COMPILER ${IOS_C_COMPILER} CACHE PATH "C compiler" FORCE)
+set(CMAKE_CXX_COMPILER ${IOS_CXX_COMPILER} CACHE PATH "CXX compiler" FORCE)
+set(CMAKE_C_OSX_COMPATIBILITY_VERSION_FLAG "-compatibility_version ")
+set(CMAKE_C_OSX_CURRENT_VERSION_FLAG "-current_version ")
+set(CMAKE_CXX_OSX_COMPATIBILITY_VERSION_FLAG "${CMAKE_C_OSX_COMPATIBILITY_VERSION_FLAG}")
+set(CMAKE_CXX_OSX_CURRENT_VERSION_FLAG "${CMAKE_C_OSX_CURRENT_VERSION_FLAG}")
+# Set iOS specific C/C++ flags
+if(IOS_PLATFORM STREQUAL "OS")
+  if(XCODE_VERSION VERSION_LESS "7.0")
+    set(XCODE_IOS_PLATFORM_VERSION_FLAGS "-mios-version-min=${IOS_DEPLOYMENT_TARGET}")
+  else()
+    # Xcode 7.0+ uses flags we can build directly from XCODE_IOS_PLATFORM.
+    set(XCODE_IOS_PLATFORM_VERSION_FLAGS "-m${XCODE_IOS_PLATFORM}-version-min=${IOS_DEPLOYMENT_TARGET}")
+  endif()
+else()
+  set(XCODE_IOS_FLATFORM_VERSION_FLAGS "-mios-simulator-version-min=${IOS_DEPLOYMENT_TARGET}")
+endif()
+if(IOS_ENABLE_BITCODE)
+  set(XCODE_IOS_BITCODE_FLAGS "${IOS_COMPILER_FLAGS} -fembed-bitcode")
+else()
+  set(XCODE_IOS_BITCODE_FLAGS "")
+endif()
+set(IOS_COMPILER_FLAGS "${XCODE_IOS_PLATFORM_VERSION_FLAGS} ${XCODE_IOS_BITCODE_FLAGS}")
+# Hidden visibilty is required for cxx on iOS 
+set(CMAKE_C_FLAGS "${IOS_COMPILER_FLAGS} ${CMAKE_C_FLAGS}" CACHE STRING "C flags")
+set(CMAKE_CXX_FLAGS "${IOS_COMPILER_FLAGS} -fvisibility=hidden -fvisibility-inlines-hidden ${CMAKE_CXX_FLAGS}" CACHE STRING "CXX flags")
+set(IOS_LINK_FLAGS "${XCODE_IOS_PLATFORM_VERSION_FLAGS} -Wl,-search_paths_first")
+if(IOS_USE_VECLIB_FOR_BLAS)
+  # Find vecLib for iOS
+  set(VECLIB_SEARCH_DIRS
+      ${IOS_SDK_ROOT}/System/Library/Frameworks/Accelerate.framework/Versions/Current/Frameworks
+      ${IOS_SDK_ROOT}/System/Library/Frameworks/Accelerate.framework/Frameworks
+      )
+  find_path(VECLIB_INC_DIR vecLib.h PATHS ${VECLIB_SEARCH_DIRS}/vecLib.framework/Headers)
+  include(FindPackageHandleStandardArgs)
+  find_package_handle_standard_args(vecLib DEFAULT_MSG VECLIB_INC_DIR)
+  if(VECLIB_FOUND)
+    if(VECLIB_INC_DIR MATCHES "^/System/Library/Frameworks/vecLib.framework.*")
+      set(IOS_LINK_FLAGS ${IOS_LINK_FLAGS} -lcblas "-framework vecLib")
+      message(STATUS "Found standalone vecLib.framework")
+    else()
+      set(IOS_LINK_FLAGS ${IOS_LINK_FLAGS} -lcblas "-framework Accelerate")
+      message(STATUS "Found vecLib as part of Accelerate.framework")
+    endif()
+  endif()
+endif()
+set(CMAKE_C_LINK_FLAGS "${IOS_LINK_FLAGS} ${CMAKE_C_LINK_FLAGS}")
+set(CMAKE_CXX_LINK_FLAGS "${IOS_LINK_FLAGS} ${CMAKE_CXX_LINK_FLAGS}")
+set(CMAKE_PLATFORM_HAS_INSTALLNAME 1)
+if(NOT IOS_ENABLE_BITCODE)
+  set(CMAKE_SHARED_LIBRARY_CREATE_C_FLAGS "-dynamiclib -headerpad_max_install_names")
+  set(CMAKE_SHARED_MODULE_CREATE_C_FLAGS "-bundle -headerpad_max_install_names")
+else()
+  set(CMAKE_SHARED_LIBRARY_CREATE_C_FLAGS "-dynamiclib")
+  set(CMAKE_SHARED_MODULE_CREATE_C_FLAGS "-bundle")
+endif()
+set(CMAKE_SHARED_MODULE_LOADER_C_FLAG "-Wl,-bundle_loader,")
+set(CMAKE_SHARED_MODULE_LOADER_CXX_FLAG "-Wl,-bundle_loader,")
+set(CMAKE_FIND_LIBRARY_SUFFIXES ".dylib" ".so" ".a")
+# hack: if a new cmake (which uses CMAKE_INSTALL_NAME_TOOL) runs on an old build tree
+# (where install_name_tool was hardcoded) and where CMAKE_INSTALL_NAME_TOOL isn't in the cache
+# and still cmake didn't fail in CMakeFindBinUtils.cmake (because it isn't rerun)
+# hardcode CMAKE_INSTALL_NAME_TOOL here to install_name_tool, so it behaves as it did before, Alex
+if(NOT DEFINED CMAKE_INSTALL_NAME_TOOL)
+  find_program(CMAKE_INSTALL_NAME_TOOL install_name_tool)
+endif()
+# Set the find root to the iOS developer roots and to user defined paths
+set(CMAKE_FIND_ROOT_PATH ${IOS_DEVELOPER_ROOT} ${IOS_SDK_ROOT} ${CMAKE_PREFIX_PATH}
+    CACHE string  "iOS find search path root")
+# default to searching for frameworks first
+set(CMAKE_FIND_FRAMEWORK FIRST)
+# set up the default search directories for frameworks
+set(CMAKE_SYSTEM_FRAMEWORK_PATH
+    ${IOS_SDK_ROOT}/System/Library/Frameworks
+    ${IOS_SDK_ROOT}/System/Library/PrivateFrameworks
+    ${IOS_SDK_ROOT}/Developer/Library/Frameworks
+    )
+# only search the iOS sdks, not the remainder of the host filesystem
+set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
+set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
+set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
+message(STATUS "iOS: Targeting iOS '${CMAKE_SYSTEM_VERSION}', "
+        "building for '${IOS_PLATFORM}' platform, with architecture '${CMAKE_OSX_ARCHITECTURES}'")
+message(STATUS "System CMAKE_C_FLAGS: ${CMAKE_C_FLAGS}")
+message(STATUS "System CMAKE_CXX_FLAGS: ${CMAKE_CXX_FLAGS}")
+# Used in ExternalProject command
+string(REPLACE ";" "\\$<SEMICOLON>" EXTERNAL_IOS_ARCHITECTURES "${CMAKE_OSX_ARCHITECTURES}")
+set(EXTERNAL_OPTIONAL_ARGS
+    -DCMAKE_OSX_SYSROOT=${CMAKE_OSX_SYSROOT}
+    -DCMAKE_OSX_ARCHITECTURES=${EXTERNAL_IOS_ARCHITECTURES})
+# This little macro lets you set any XCode specific property
+macro(set_xcode_property TARGET XCODE_PROPERTY XCODE_VALUE)
+  set_property (TARGET ${TARGET} PROPERTY XCODE_ATTRIBUTE_${XCODE_PROPERTY} ${XCODE_VALUE})
+endmacro(set_xcode_property)
+# This macro lets you find executable programs on the host system
+macro(find_host_package)
+  set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
+  set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY NEVER)
+  set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE NEVER)
+  set(IOS FALSE)
+  find_package(${ARGN})
+  set(IOS TRUE)
+  set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM ONLY)
+  set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
+  set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
+endmacro(find_host_package)
--- a/cmake/cross_compiling/raspberry_pi.cmake
+++ b/cmake/cross_compiling/raspberry_pi.cmake
+# 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.
+# This is a toolchain file for cross-compiling for Raspberry Pi.
+# The supported variables are listed belows:
+#
+# RPI_TOOLCHAIN
+# RPI_ARM_NEON
+#
+# Also you can set CMAKE_C/CXX_COMPILER yourself, through cmake arguments.
+IF(NOT RPI)
+    return()
+ENDIF()
+SET(CMAKE_SYSTEM_NAME Linux)
+SET(CMAKE_SYSTEM_VERSION 1)
+SET(CMAKE_SYSTEM_PROCESSOR arm)
+# check the exist of raspberry pi toolchain
+IF(NOT DEFINED RPI_TOOLCHAIN)
+    SET(RPI_TOOLCHAIN $ENV{RPI_TOOLCHAIN}
+        CACHE PATH "Folder holds the toolchain of Raspberr Pi")
+ENDIF()
+IF(NOT RPI_TOOLCHAIN)
+    MESSAGE(WARNING "It is recommended to set RPI_TOOLCHAIN to use toolchain.\n"
+            "To cross-compile for Raspberry Pi, you need to download the tools using:\n"
+            " git clone https://github.com/raspberrypi/tools\n")
+ENDIF()
+IF(NOT DEFINED RPI_ARM_NEON)
+    SET(RPI_ARM_NEON ON)
+ENDIF()
+IF(RPI_TOOLCHAIN)
+    SET(RPI_TOOLCHAIN_ROOT ${RPI_TOOLCHAIN})
+    IF(RPI_TOOLCHAIN_ROOT MATCHES "gcc-linaro-arm-linux-gnueabihf-raspbian(-x64)?$")
+        # gcc-linaro-arm-linux-gnueabihf-raspbian
+        # gcc-linaro-arm-linux-gnueabihf-raspbian-x64
+        SET(RPI_TOOLCHAIN_NAME arm-linux-gnueabihf)
+    ENDIF()
+    SET(RPI_TOOLCHAIN_PREFIX "${RPI_TOOLCHAIN_ROOT}/bin/${RPI_TOOLCHAIN_NAME}-")
+ENDIF()
+# C compiler
+IF(NOT CMAKE_C_COMPILER)
+    SET(RPI_C_COMPILER "${RPI_TOOLCHAIN_PREFIX}gcc")
+ELSE()
+    GET_FILENAME_COMPONENT(RPI_C_COMPILER ${CMAKE_C_COMPILER} PROGRAM)
+ENDIF()
+IF(NOT EXISTS ${RPI_C_COMPILER})
+    MESSAGE(FATAL_ERROR "Cannot find C compiler: ${RPI_C_COMPILER}")
+ENDIF()
+# CXX compiler
+IF(NOT CMAKE_CXX_COMPILER)
+    SET(RPI_CXX_COMPILER "${RPI_TOOLCHAIN_PREFIX}g++")
+ELSE()
+    GET_FILENAME_COMPONENT(RPI_CXX_COMPILER ${CMAKE_CXX_COMPILER} PROGRAM)
+ENDIF()
+IF(NOT EXISTS ${RPI_CXX_COMPILER})
+    MESSAGE(FATAL_ERROR "Cannot find CXX compiler: ${RPI_CXX_COMPILER}")
+ENDIF()
+SET(CMAKE_C_COMPILER ${RPI_C_COMPILER} CACHE PATH "C compiler" FORCE)
+SET(CMAKE_CXX_COMPILER ${RPI_CXX_COMPILER} CACHE PATH "CXX compiler" FORCE)
+IF(RPI_ARM_NEON)
+    SET(RPI_C_FLAGS "${RPI_C_FLAGS} -mfpu=neon")
+ENDIF()
+SET(CMAKE_C_FLAGS "${RPI_C_FLAGS} ${CMAKE_C_FLAGS}" CACHE STRING "C flags")
+SET(CMAKE_CXX_FLAGS "${RPI_C_FLAGS} ${CMAKE_CXX_FLAGS}" CACHE STRING "CXX flags")
--- a/cmake/cuda.cmake
+++ b/cmake/cuda.cmake
+if(NOT WITH_GPU)
+    return()
+endif()
+set(paddle_known_gpu_archs "30 35 50 52 60 61 70")
+set(paddle_known_gpu_archs7 "30 35 50 52")
+set(paddle_known_gpu_archs8 "30 35 50 52 60 61")
+######################################################################################
+# A function for automatic detection of GPUs installed  (if autodetection is enabled)
+# Usage:
+#   detect_installed_gpus(out_variable)
+function(detect_installed_gpus out_variable)
+  if(NOT CUDA_gpu_detect_output)
+    set(cufile ${PROJECT_BINARY_DIR}/detect_cuda_archs.cu)
+    file(WRITE ${cufile} ""
+      "#include <cstdio>\n"
+      "int main() {\n"
+      "  int count = 0;\n"
+      "  if (cudaSuccess != cudaGetDeviceCount(&count)) return -1;\n"
+      "  if (count == 0) return -1;\n"
+      "  for (int device = 0; device < count; ++device) {\n"
+      "    cudaDeviceProp prop;\n"
+      "    if (cudaSuccess == cudaGetDeviceProperties(&prop, device))\n"
+      "      std::printf(\"%d.%d \", prop.major, prop.minor);\n"
+      "  }\n"
+      "  return 0;\n"
+      "}\n")
+    execute_process(COMMAND "${CUDA_NVCC_EXECUTABLE}" "-ccbin=${CUDA_HOST_COMPILER}"
+                    "--run" "${cufile}"
+                    WORKING_DIRECTORY "${PROJECT_BINARY_DIR}/CMakeFiles/"
+                    RESULT_VARIABLE nvcc_res OUTPUT_VARIABLE nvcc_out
+                    ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
+    if(nvcc_res EQUAL 0)
+      # only keep the last line of nvcc_out
+      STRING(REGEX REPLACE ";" "\\\\;" nvcc_out "${nvcc_out}")
+      STRING(REGEX REPLACE "\n" ";" nvcc_out "${nvcc_out}")
+      list(GET nvcc_out -1 nvcc_out)
+      string(REPLACE "2.1" "2.1(2.0)" nvcc_out "${nvcc_out}")
+      set(CUDA_gpu_detect_output ${nvcc_out} CACHE INTERNAL "Returned GPU architetures from detect_installed_gpus tool" FORCE)
+    endif()
+  endif()
+  if(NOT CUDA_gpu_detect_output)
+    message(STATUS "Automatic GPU detection failed. Building for all known architectures.")
+    set(${out_variable} ${paddle_known_gpu_archs} PARENT_SCOPE)
+  else()
+    set(${out_variable} ${CUDA_gpu_detect_output} PARENT_SCOPE)
+  endif()
+endfunction()
+########################################################################
+# Function for selecting GPU arch flags for nvcc based on CUDA_ARCH_NAME
+# Usage:
+#   select_nvcc_arch_flags(out_variable)
+function(select_nvcc_arch_flags out_variable)
+  # List of arch names
+  set(archs_names "Kepler" "Maxwell" "Pascal" "All" "Manual")
+  set(archs_name_default "All")
+  if(NOT CMAKE_CROSSCOMPILING)
+    list(APPEND archs_names "Auto")
+  endif()
+  # set CUDA_ARCH_NAME strings (so it will be seen as dropbox in CMake-Gui)
+  set(CUDA_ARCH_NAME ${archs_name_default} CACHE STRING "Select target NVIDIA GPU achitecture.")
+  set_property( CACHE CUDA_ARCH_NAME PROPERTY STRINGS "" ${archs_names} )
+  mark_as_advanced(CUDA_ARCH_NAME)
+  # verify CUDA_ARCH_NAME value
+  if(NOT ";${archs_names};" MATCHES ";${CUDA_ARCH_NAME};")
+    string(REPLACE ";" ", " archs_names "${archs_names}")
+    message(FATAL_ERROR "Only ${archs_names} architeture names are supported.")
+  endif()
+  if(${CUDA_ARCH_NAME} STREQUAL "Manual")
+    set(CUDA_ARCH_BIN ${paddle_known_gpu_archs} CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
+    set(CUDA_ARCH_PTX "50"                     CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
+    mark_as_advanced(CUDA_ARCH_BIN CUDA_ARCH_PTX)
+  else()
+    unset(CUDA_ARCH_BIN CACHE)
+    unset(CUDA_ARCH_PTX CACHE)
+  endif()
+  if(${CUDA_ARCH_NAME} STREQUAL "Kepler")
+    set(cuda_arch_bin "30 35")
+  elseif(${CUDA_ARCH_NAME} STREQUAL "Maxwell")
+    set(cuda_arch_bin "50")
+  elseif(${CUDA_ARCH_NAME} STREQUAL "Pascal")
+    set(cuda_arch_bin "60 61")
+  elseif(${CUDA_ARCH_NAME} STREQUAL "Volta")
+    set(cuda_arch_bin "70")
+  elseif(${CUDA_ARCH_NAME} STREQUAL "All")
+    set(cuda_arch_bin ${paddle_known_gpu_archs})
+  elseif(${CUDA_ARCH_NAME} STREQUAL "Auto")
+    detect_installed_gpus(cuda_arch_bin)
+  else()  # (${CUDA_ARCH_NAME} STREQUAL "Manual")
+    set(cuda_arch_bin ${CUDA_ARCH_BIN})
+  endif()
+  # remove dots and convert to lists
+  string(REGEX REPLACE "\\." "" cuda_arch_bin "${cuda_arch_bin}")
+  string(REGEX REPLACE "\\." "" cuda_arch_ptx "${CUDA_ARCH_PTX}")
+  string(REGEX MATCHALL "[0-9()]+" cuda_arch_bin "${cuda_arch_bin}")
+  string(REGEX MATCHALL "[0-9]+"   cuda_arch_ptx "${cuda_arch_ptx}")
+  list(REMOVE_DUPLICATES cuda_arch_bin)
+  list(REMOVE_DUPLICATES cuda_arch_ptx)
+  set(nvcc_flags "")
+  set(nvcc_archs_readable "")
+  # Tell NVCC to add binaries for the specified GPUs
+  foreach(arch ${cuda_arch_bin})
+    if(arch MATCHES "([0-9]+)\\(([0-9]+)\\)")
+      # User explicitly specified PTX for the concrete BIN
+      list(APPEND nvcc_flags -gencode arch=compute_${CMAKE_MATCH_2},code=sm_${CMAKE_MATCH_1})
+      list(APPEND nvcc_archs_readable sm_${CMAKE_MATCH_1})
+    else()
+      # User didn't explicitly specify PTX for the concrete BIN, we assume PTX=BIN
+      list(APPEND nvcc_flags -gencode arch=compute_${arch},code=sm_${arch})
+      list(APPEND nvcc_archs_readable sm_${arch})
+    endif()
+  endforeach()
+  # Tell NVCC to add PTX intermediate code for the specified architectures
+  foreach(arch ${cuda_arch_ptx})
+    list(APPEND nvcc_flags -gencode arch=compute_${arch},code=compute_${arch})
+    list(APPEND nvcc_archs_readable compute_${arch})
+  endforeach()
+  string(REPLACE ";" " " nvcc_archs_readable "${nvcc_archs_readable}")
+  set(${out_variable}          ${nvcc_flags}          PARENT_SCOPE)
+  set(${out_variable}_readable ${nvcc_archs_readable} PARENT_SCOPE)
+endfunction()
+message(STATUS "CUDA detected: " ${CUDA_VERSION})
+if (${CUDA_VERSION} LESS 7.0)
+  set(paddle_known_gpu_archs ${paddle_known_gpu_archs})
+elseif (${CUDA_VERSION} LESS 8.0) # CUDA 7.x
+  set(paddle_known_gpu_archs ${paddle_known_gpu_archs7})
+  list(APPEND CUDA_NVCC_FLAGS "-D_MWAITXINTRIN_H_INCLUDED")
+  list(APPEND CUDA_NVCC_FLAGS "-D__STRICT_ANSI__")
+elseif (${CUDA_VERSION} LESS 9.0) # CUDA 8.x
+  set(paddle_known_gpu_archs ${paddle_known_gpu_archs8})
+  list(APPEND CUDA_NVCC_FLAGS "-D_MWAITXINTRIN_H_INCLUDED")
+  list(APPEND CUDA_NVCC_FLAGS "-D__STRICT_ANSI__")
+  # CUDA 8 may complain that sm_20 is no longer supported. Suppress the
+  # warning for now.
+  list(APPEND CUDA_NVCC_FLAGS "-Wno-deprecated-gpu-targets")
+endif()
+include_directories(${CUDA_INCLUDE_DIRS})
+list(APPEND EXTERNAL_LIBS ${CUDA_LIBRARIES} ${CUDA_rt_LIBRARY})
+if(NOT WITH_DSO)
+    list(APPEND EXTERNAL_LIBS ${CUDNN_LIBRARY} ${CUDA_CUBLAS_LIBRARIES} ${CUDA_curand_LIBRARY} ${NCCL_LIBRARY})
+endif(NOT WITH_DSO)
+# setting nvcc arch flags
+select_nvcc_arch_flags(NVCC_FLAGS_EXTRA)
+list(APPEND CUDA_NVCC_FLAGS ${NVCC_FLAGS_EXTRA})
+message(STATUS "Added CUDA NVCC flags for: ${NVCC_FLAGS_EXTRA_readable}")
+# Set C++11 support
+set(CUDA_PROPAGATE_HOST_FLAGS OFF)
+# Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
+# So, don't set these flags here.
+list(APPEND CUDA_NVCC_FLAGS "-std=c++11")
+list(APPEND CUDA_NVCC_FLAGS "--use_fast_math")
+list(APPEND CUDA_NVCC_FLAGS "-Xcompiler -fPIC")
+# Set :expt-relaxed-constexpr to suppress Eigen warnings
+list(APPEND CUDA_NVCC_FLAGS "--expt-relaxed-constexpr")
+if(CMAKE_BUILD_TYPE  STREQUAL "Debug")
+    list(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_DEBUG})
+elseif(CMAKE_BUILD_TYPE  STREQUAL "Release")
+    list(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_RELEASE})
+elseif(CMAKE_BUILD_TYPE  STREQUAL "RelWithDebInfo")
+    list(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_RELWITHDEBINFO})
+elseif(CMAKE_BUILD_TYPE  STREQUAL "MinSizeRel")
+    list(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_MINSIZEREL})
+endif()
+mark_as_advanced(CUDA_BUILD_CUBIN CUDA_BUILD_EMULATION CUDA_VERBOSE_BUILD)
+mark_as_advanced(CUDA_SDK_ROOT_DIR CUDA_SEPARABLE_COMPILATION)
--- a/cmake/cudnn.cmake
+++ b/cmake/cudnn.cmake
@@ -2,7 +2,7 @@ if(NOT WITH_GPU)
    return()
 endif()
-set(CUDNN_ROOT "" CACHE PATH "CUDNN ROOT")
+set(CUDNN_ROOT "/usr" CACHE PATH "CUDNN ROOT")
 find_path(CUDNN_INCLUDE_DIR cudnn.h
    PATHS ${CUDNN_ROOT} ${CUDNN_ROOT}/include
    $ENV{CUDNN_ROOT} $ENV{CUDNN_ROOT}/include ${CUDA_TOOLKIT_INCLUDE}
@@ -11,11 +11,16 @@ find_path(CUDNN_INCLUDE_DIR cudnn.h
 get_filename_component(__libpath_hist ${CUDA_CUDART_LIBRARY} PATH)
+set(TARGET_ARCH "x86_64")
+if(NOT ${CMAKE_SYSTEM_PROCESSOR})
+    set(TARGET_ARCH ${CMAKE_SYSTEM_PROCESSOR})
+endif()
 list(APPEND CUDNN_CHECK_LIBRARY_DIRS
    ${CUDNN_ROOT}
    ${CUDNN_ROOT}/lib64
    ${CUDNN_ROOT}/lib
-    ${CUDNN_ROOT}/lib/x86_64-linux-gnu
+    ${CUDNN_ROOT}/lib/${TARGET_ARCH}-linux-gnu
    $ENV{CUDNN_ROOT}
    $ENV{CUDNN_ROOT}/lib64
    $ENV{CUDNN_ROOT}/lib

--- a/cmake/external/any.cmake
+++ b/cmake/external/any.cmake
@@ -2,13 +2,13 @@ INCLUDE(ExternalProject)
 SET(ANY_SOURCE_DIR ${THIRD_PARTY_PATH}/any)
-INCLUDE_DIRECTORIES(${ANY_SOURCE_DIR}/src/linb_any)
+INCLUDE_DIRECTORIES(${ANY_SOURCE_DIR}/src/extern_lib_any)
 ExternalProject_Add(
-    linb_any
+    extern_lib_any
    ${EXTERNAL_PROJECT_LOG_ARGS}
-    GIT_REPOSITORY  "https://github.com/thelink2012/any.git"
+    GIT_REPOSITORY  "https://github.com/PaddlePaddle/any.git"
-    GIT_TAG         "8fef1e93710a0edf8d7658999e284a1142c4c020"
+    GIT_TAG         "15595d8324be9e8a9a80d9ae442fdd12bd66df5d"
    PREFIX          ${ANY_SOURCE_DIR}
    UPDATE_COMMAND  ""
    CONFIGURE_COMMAND ""
@@ -17,4 +17,15 @@ ExternalProject_Add(
    TEST_COMMAND      ""
 )
+if (${CMAKE_VERSION} VERSION_LESS "3.3.0")
+    set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/lib_any_dummy.c)
+    file(WRITE ${dummyfile} "const char * dummy_any = \"${dummyfile}\";")
+    add_library(lib_any STATIC ${dummyfile})
+else()
+    add_library(lib_any INTERFACE)
+endif()
+add_dependencies(lib_any extern_lib_any)
 add_definitions(-DANY_IMPL_ANY_CAST_MOVEABLE)
+LIST(APPEND external_project_dependencies lib_any)
--- a/cmake/external/cares.cmake
+++ b/cmake/external/cares.cmake
+# 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.
+#
+IF(MOBILE_INFERENCE OR NOT WITH_DISTRIBUTE)
+    return()
+ENDIF()
+include (ExternalProject)
+# NOTE: c-ares is needed when linking with grpc.
+SET(CARES_SOURCES_DIR ${THIRD_PARTY_PATH}/cares)
+SET(CARES_INSTALL_DIR ${THIRD_PARTY_PATH}/install/cares)
+SET(CARES_INCLUDE_DIR "${CARES_INSTALL_DIR}/include/" CACHE PATH "cares include directory." FORCE)
+ExternalProject_Add(
+    extern_cares
+    GIT_REPOSITORY "https://github.com/c-ares/c-ares.git"
+    GIT_TAG "cares-1_13_0"
+    PREFIX          ${CARES_SOURCES_DIR}
+    UPDATE_COMMAND  ""
+    CONFIGURE_COMMAND ./buildconf && ./configure --disable-shared --prefix=${CARES_INSTALL_DIR}
+    BUILD_IN_SOURCE 1
+    BUILD_COMMAND   make -j8
+    INSTALL_COMMAND make install
+)
+ADD_LIBRARY(cares STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET cares PROPERTY IMPORTED_LOCATION
+             "${CARES_INSTALL_DIR}/lib/libcares.a")
+include_directories(${CARES_INCLUDE_DIR})
+ADD_DEPENDENCIES(cares extern_cares)
--- a/cmake/external/eigen.cmake
+++ b/cmake/external/eigen.cmake
+INCLUDE(ExternalProject)
+SET(EIGEN_SOURCE_DIR ${THIRD_PARTY_PATH}/eigen3)
+INCLUDE_DIRECTORIES(${EIGEN_SOURCE_DIR}/src/extern_eigen3)
+ExternalProject_Add(
+    extern_eigen3
+    ${EXTERNAL_PROJECT_LOG_ARGS}
+    GIT_REPOSITORY  "https://github.com/RLovelett/eigen.git"
+    GIT_TAG         70661066beef694cadf6c304d0d07e0758825c10
+    PREFIX          ${EIGEN_SOURCE_DIR}
+    UPDATE_COMMAND  ""
+    CONFIGURE_COMMAND ""
+    BUILD_COMMAND     ""
+    INSTALL_COMMAND   ""
+    TEST_COMMAND      ""
+)
+if (${CMAKE_VERSION} VERSION_LESS "3.3.0")
+    set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/eigen3_dummy.c)
+    file(WRITE ${dummyfile} "const char * dummy_eigen3 = \"${dummyfile}\";")
+    add_library(eigen3 STATIC ${dummyfile})
+else()
+    add_library(eigen3 INTERFACE)
+endif()
+add_dependencies(eigen3 extern_eigen3)
+LIST(APPEND external_project_dependencies eigen3)
--- a/cmake/external/gflags.cmake
+++ b/cmake/external/gflags.cmake
 # 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.
@@ -18,30 +18,45 @@ SET(GFLAGS_SOURCES_DIR ${THIRD_PARTY_PATH}/gflags)
 SET(GFLAGS_INSTALL_DIR ${THIRD_PARTY_PATH}/install/gflags)
 SET(GFLAGS_INCLUDE_DIR "${GFLAGS_INSTALL_DIR}/include" CACHE PATH "gflags include directory." FORCE)
 IF(WIN32)
-    set(GFLAGS_LIBRARIES "${GFLAGS_INSTALL_DIR}/lib/gflags.lib" CACHE FILEPATH "GFLAGS_LIBRARIES" FORCE)
+  set(GFLAGS_LIBRARIES "${GFLAGS_INSTALL_DIR}/lib/gflags.lib" CACHE FILEPATH "GFLAGS_LIBRARIES" FORCE)
 ELSE(WIN32)
-    set(GFLAGS_LIBRARIES "${GFLAGS_INSTALL_DIR}/lib/libgflags.a" CACHE FILEPATH "GFLAGS_LIBRARIES" FORCE)
+  set(GFLAGS_LIBRARIES "${GFLAGS_INSTALL_DIR}/lib/libgflags.a" CACHE FILEPATH "GFLAGS_LIBRARIES" FORCE)
 ENDIF(WIN32)
 INCLUDE_DIRECTORIES(${GFLAGS_INCLUDE_DIR})
 ExternalProject_Add(
-    gflags
+    extern_gflags
    ${EXTERNAL_PROJECT_LOG_ARGS}
    GIT_REPOSITORY  "https://github.com/gflags/gflags.git"
+    GIT_TAG         77592648e3f3be87d6c7123eb81cbad75f9aef5a
    PREFIX          ${GFLAGS_SOURCES_DIR}
    UPDATE_COMMAND  ""
    CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-    CMAKE_ARGS      -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
+                    -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
-    CMAKE_ARGS      -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
+                    -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
-    CMAKE_ARGS      -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
+                    -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
-    CMAKE_ARGS      -DCMAKE_INSTALL_PREFIX=${GFLAGS_INSTALL_DIR}
+                    -DCMAKE_INSTALL_PREFIX=${GFLAGS_INSTALL_DIR}
-    CMAKE_ARGS      -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+                    -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-    CMAKE_ARGS      -DBUILD_TESTING=OFF
+                    -DBUILD_TESTING=OFF
-    CMAKE_ARGS      -DCMAKE_BUILD_TYPE=Release
+                    -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
+                    ${EXTERNAL_OPTIONAL_ARGS}
    CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GFLAGS_INSTALL_DIR}
                     -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-                     -DCMAKE_BUILD_TYPE:STRING=Release
+                     -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
 )
+ADD_LIBRARY(gflags STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET gflags PROPERTY IMPORTED_LOCATION ${GFLAGS_LIBRARIES})
+ADD_DEPENDENCIES(gflags extern_gflags)
 LIST(APPEND external_project_dependencies gflags)
+IF(WITH_C_API)
+  INSTALL(DIRECTORY ${GFLAGS_INCLUDE_DIR} DESTINATION third_party/gflags)
+  IF(ANDROID)
+    INSTALL(FILES ${GFLAGS_LIBRARIES} DESTINATION third_party/gflags/lib/${ANDROID_ABI})
+  ELSE()
+    INSTALL(FILES ${GFLAGS_LIBRARIES} DESTINATION third_party/gflags/lib)
+  ENDIF()
+ENDIF()
--- a/cmake/external/glog.cmake
+++ b/cmake/external/glog.cmake
@@ -19,33 +19,60 @@ SET(GLOG_INSTALL_DIR ${THIRD_PARTY_PATH}/install/glog)
 SET(GLOG_INCLUDE_DIR "${GLOG_INSTALL_DIR}/include" CACHE PATH "glog include directory." FORCE)
 IF(WIN32)
-    SET(GLOG_LIBRARIES "${GLOG_INSTALL_DIR}/lib/libglog.lib" CACHE FILEPATH "glog library." FORCE)
+  SET(GLOG_LIBRARIES "${GLOG_INSTALL_DIR}/lib/libglog.lib" CACHE FILEPATH "glog library." FORCE)
 ELSE(WIN32)
-    SET(GLOG_LIBRARIES "${GLOG_INSTALL_DIR}/lib/libglog.a" CACHE FILEPATH "glog library." FORCE)
+  SET(GLOG_LIBRARIES "${GLOG_INSTALL_DIR}/lib/libglog.a" CACHE FILEPATH "glog library." FORCE)
 ENDIF(WIN32)
 INCLUDE_DIRECTORIES(${GLOG_INCLUDE_DIR})
+IF(ANDROID AND ${CMAKE_SYSTEM_VERSION} VERSION_LESS "21")
+  # Using the unofficial glog for Android API < 21
+  SET(GLOG_REPOSITORY "https://github.com/Xreki/glog.git")
+  SET(GLOG_TAG "8a547150548b284382ccb6582408e9140ff2bea8")
+ELSE()
+  SET(GLOG_REPOSITORY "https://github.com/google/glog.git")
+  SET(GLOG_TAG "v0.3.5")
+ENDIF()
 ExternalProject_Add(
-    glog
+    extern_glog
    ${EXTERNAL_PROJECT_LOG_ARGS}
    DEPENDS gflags
-    GIT_REPOSITORY  "https://github.com/google/glog.git"
+    GIT_REPOSITORY  ${GLOG_REPOSITORY}
+    GIT_TAG         ${GLOG_TAG}
    PREFIX          ${GLOG_SOURCES_DIR}
    UPDATE_COMMAND  ""
    CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-    CMAKE_ARGS      -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
+                    -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
-    CMAKE_ARGS      -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
+                    -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
-    CMAKE_ARGS      -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
+                    -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
-    CMAKE_ARGS      -DCMAKE_INSTALL_PREFIX=${GLOG_INSTALL_DIR}
+                    -DCMAKE_INSTALL_PREFIX=${GLOG_INSTALL_DIR}
-    CMAKE_ARGS      -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+                    -DCMAKE_INSTALL_LIBDIR=${GLOG_INSTALL_DIR}/lib
-    CMAKE_ARGS      -DWITH_GFLAGS=ON
+                    -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-    CMAKE_ARGS      -Dgflags_DIR=${GFLAGS_INSTALL_DIR}/lib/cmake/gflags
+                    -DWITH_GFLAGS=ON
-    CMAKE_ARGS      -DBUILD_TESTING=OFF
+                    -Dgflags_DIR=${GFLAGS_INSTALL_DIR}/lib/cmake/gflags
-    CMAKE_ARGS      -DCMAKE_BUILD_TYPE=Release
+                    -DBUILD_TESTING=OFF
+                    -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
+                    ${EXTERNAL_OPTIONAL_ARGS}
    CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GLOG_INSTALL_DIR}
+                     -DCMAKE_INSTALL_LIBDIR:PATH=${GLOG_INSTALL_DIR}/lib
                     -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-                     -DCMAKE_BUILD_TYPE:STRING=Release
+                     -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
 )
+ADD_LIBRARY(glog STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET glog PROPERTY IMPORTED_LOCATION ${GLOG_LIBRARIES})
+ADD_DEPENDENCIES(glog extern_glog gflags)
+LINK_LIBRARIES(glog gflags)
 LIST(APPEND external_project_dependencies glog)
+IF(WITH_C_API)
+  INSTALL(DIRECTORY ${GLOG_INCLUDE_DIR} DESTINATION third_party/glog)
+  IF(ANDROID)
+    INSTALL(FILES ${GLOG_LIBRARIES} DESTINATION third_party/glog/lib/${ANDROID_ABI})
+  ELSE()
+    INSTALL(FILES ${GLOG_LIBRARIES} DESTINATION third_party/glog/lib)
+  ENDIF()
+ENDIF()
--- a/cmake/external/grpc.cmake
+++ b/cmake/external/grpc.cmake
+# 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.
+#
+IF(MOBILE_INFERENCE OR NOT WITH_DISTRIBUTE)
+    return()
+ENDIF()
+include (ExternalProject)
+SET(GRPC_SOURCES_DIR ${THIRD_PARTY_PATH}/grpc)
+SET(GRPC_INSTALL_DIR ${THIRD_PARTY_PATH}/install/grpc)
+SET(GRPC_INCLUDE_DIR "${GRPC_INSTALL_DIR}/include/" CACHE PATH "grpc include directory." FORCE)
+SET(GRPC_CPP_PLUGIN "${GRPC_INSTALL_DIR}/bin/grpc_cpp_plugin" CACHE FILEPATH "GRPC_CPP_PLUGIN" FORCE)
+IF(APPLE)
+  SET(BUILD_CMD make -n HAS_SYSTEM_PROTOBUF=false -s -j8 static grpc_cpp_plugin | sed "s/-Werror//g" | sh)
+ELSE()
+  SET(BUILD_CMD make HAS_SYSTEM_PROTOBUF=false -s -j8 static grpc_cpp_plugin)
+ENDIF()
+ExternalProject_Add(
+    extern_grpc
+    DEPENDS protobuf zlib
+    GIT_REPOSITORY "https://github.com/grpc/grpc.git"
+    GIT_TAG "v1.7.x"
+    PREFIX          ${GRPC_SOURCES_DIR}
+    UPDATE_COMMAND  ""
+    CONFIGURE_COMMAND ""
+    BUILD_IN_SOURCE 1
+    # NOTE(yuyang18):
+    # Disable -Werror, otherwise the compile will fail in MacOS.
+    # It seems that we cannot configure that by make command.
+    # Just dry run make command and remove `-Werror`, then use a shell to run make commands
+    BUILD_COMMAND  ${BUILD_CMD}
+    INSTALL_COMMAND make prefix=${GRPC_INSTALL_DIR} install
+)
+# FIXME(typhoonzero): hack to get static lib path, try a better way like merge them.
+ADD_LIBRARY(grpc++_unsecure STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET grpc++_unsecure PROPERTY IMPORTED_LOCATION
+             "${GRPC_INSTALL_DIR}/lib/libgrpc++_unsecure.a")
+ADD_LIBRARY(grpc++ STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET grpc++ PROPERTY IMPORTED_LOCATION
+            "${GRPC_INSTALL_DIR}/lib/libgrpc++.a")
+ADD_LIBRARY(gpr STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET gpr PROPERTY IMPORTED_LOCATION
+            "${GRPC_INSTALL_DIR}/lib/libgpr.a")
+ADD_LIBRARY(grpc_unsecure STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET grpc_unsecure PROPERTY IMPORTED_LOCATION
+            "${GRPC_INSTALL_DIR}/lib/libgrpc_unsecure.a")
+include_directories(${GRPC_INCLUDE_DIR})
+ADD_DEPENDENCIES(grpc++_unsecure extern_grpc)
--- a/cmake/external/gtest.cmake
+++ b/cmake/external/gtest.cmake
 # Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
-# 
+#
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at
-# 
+#
 # http://www.apache.org/licenses/LICENSE-2.0
-# 
+#
 # Unless required by applicable law or agreed to in writing, software
 # distributed under the License is distributed on an "AS IS" BASIS,
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
@@ -34,26 +34,42 @@ IF(WITH_TESTING)
            "${GTEST_INSTALL_DIR}/lib/libgtest_main.a" CACHE FILEPATH "gtest main libraries." FORCE)
    ENDIF(WIN32)
+    IF(WITH_MKLML)
+        # wait for mklml downloading completed
+        SET(GTEST_DEPENDS   ${MKLML_PROJECT})
+    ENDIF()
    ExternalProject_Add(
-        gtest
+        extern_gtest
        ${EXTERNAL_PROJECT_LOG_ARGS}
+        DEPENDS         ${GTEST_DEPENDS}
        GIT_REPOSITORY  "https://github.com/google/googletest.git"
        GIT_TAG         "release-1.8.0"
        PREFIX          ${GTEST_SOURCES_DIR}
        UPDATE_COMMAND  ""
        CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-        CMAKE_ARGS      -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
+                        -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
-        CMAKE_ARGS      -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
+                        -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
-        CMAKE_ARGS      -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
+                        -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
-        CMAKE_ARGS      -DCMAKE_INSTALL_PREFIX=${GTEST_INSTALL_DIR}
+                        -DCMAKE_INSTALL_PREFIX=${GTEST_INSTALL_DIR}
-        CMAKE_ARGS      -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+                        -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-        CMAKE_ARGS      -DBUILD_GMOCK=ON
+                        -DBUILD_GMOCK=ON
-        CMAKE_ARGS      -Dgtest_disable_pthreads=ON
+                        -Dgtest_disable_pthreads=ON
-        CMAKE_ARGS      -Dgtest_force_shared_crt=ON
+                        -Dgtest_force_shared_crt=ON
-        CMAKE_ARGS      -DCMAKE_BUILD_TYPE=Release
+                        -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
+                        ${EXTERNAL_OPTIONAL_ARGS}
        CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GTEST_INSTALL_DIR}
                         -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-                         -DCMAKE_BUILD_TYPE:STRING=Release
+                         -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
    )
-    LIST(APPEND external_project_dependencies gtest)
+    ADD_LIBRARY(gtest STATIC IMPORTED GLOBAL)
+    SET_PROPERTY(TARGET gtest PROPERTY IMPORTED_LOCATION ${GTEST_LIBRARIES})
+    ADD_DEPENDENCIES(gtest extern_gtest)
+    ADD_LIBRARY(gtest_main STATIC IMPORTED GLOBAL)
+    SET_PROPERTY(TARGET gtest_main PROPERTY IMPORTED_LOCATION ${GTEST_MAIN_LIBRARIES})
+    ADD_DEPENDENCIES(gtest_main extern_gtest)
+    LIST(APPEND external_project_dependencies gtest gtest_main)
 ENDIF(WITH_TESTING)
--- a/cmake/external/mkldnn.cmake
+++ b/cmake/external/mkldnn.cmake
+# Copyright (c) 2017 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.
+IF(NOT ${WITH_MKLDNN})
+  return()
+ENDIF(NOT ${WITH_MKLDNN})
+INCLUDE(ExternalProject)
+SET(MKLDNN_PROJECT        "extern_mkldnn")
+SET(MKLDNN_SOURCES_DIR    ${THIRD_PARTY_PATH}/mkldnn)
+SET(MKLDNN_INSTALL_DIR    ${THIRD_PARTY_PATH}/install/mkldnn)
+SET(MKLDNN_INC_DIR        "${MKLDNN_INSTALL_DIR}/include" CACHE PATH "mkldnn include directory." FORCE)
+IF(WIN32 OR APPLE)
+    MESSAGE(WARNING 
+        "Windows or Mac is not supported with MKLDNN in Paddle yet."
+        "Force WITH_MKLDNN=OFF")
+    SET(WITH_MKLDNN OFF CACHE STRING "Disable MKLDNN in Windows and MacOS" FORCE)
+    return()
+ENDIF()
+SET(MKLDNN_LIB "${MKLDNN_INSTALL_DIR}/lib/libmkldnn.so" CACHE FILEPATH "mkldnn library." FORCE)
+MESSAGE(STATUS "Set ${MKLDNN_INSTALL_DIR}/lib to runtime path")
+SET(CMAKE_INSTALL_RPATH_USE_LINK_PATH TRUE)
+SET(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_RPATH}" "${MKLDNN_INSTALL_DIR}/lib")
+INCLUDE_DIRECTORIES(${MKLDNN_INC_DIR})
+IF(${CBLAS_PROVIDER} STREQUAL "MKLML")
+    SET(MKLDNN_DEPENDS   ${MKLML_PROJECT})
+    MESSAGE(STATUS "Build MKLDNN with MKLML ${MKLML_ROOT}")
+ELSE()
+    MESSAGE(FATAL_ERROR "Should enable MKLML when build MKLDNN")
+ENDIF()
+SET(MKLDNN_CFLAG "${CMAKE_C_FLAGS} -Wno-error=strict-overflow")
+SET(MKLDNN_CXXFLAG "${CMAKE_CXX_FLAGS} -Wno-error=strict-overflow")
+ExternalProject_Add(
+    ${MKLDNN_PROJECT}
+    ${EXTERNAL_PROJECT_LOG_ARGS}
+    DEPENDS             ${MKLDNN_DEPENDS}
+    GIT_REPOSITORY      "https://github.com/01org/mkl-dnn.git"
+    GIT_TAG             "v0.11"
+    PREFIX              ${MKLDNN_SOURCES_DIR}
+    UPDATE_COMMAND      ""
+    CMAKE_ARGS          -DCMAKE_INSTALL_PREFIX=${MKLDNN_INSTALL_DIR}
+    CMAKE_ARGS          -DMKLROOT=${MKLML_ROOT}
+    CMAKE_ARGS          -DCMAKE_C_FLAGS=${MKLDNN_CFLAG}
+    CMAKE_ARGS          -DCMAKE_CXX_FLAGS=${MKLDNN_CXXFLAG}
+    CMAKE_CACHE_ARGS    -DCMAKE_INSTALL_PREFIX:PATH=${MKLDNN_INSTALL_DIR}
+                        -DMKLROOT:PATH=${MKLML_ROOT}
+)
+ADD_LIBRARY(mkldnn SHARED IMPORTED GLOBAL)
+SET_PROPERTY(TARGET mkldnn PROPERTY IMPORTED_LOCATION ${MKLDNN_LIB})
+ADD_DEPENDENCIES(mkldnn ${MKLDNN_PROJECT})
+MESSAGE(STATUS "MKLDNN library: ${MKLDNN_LIB}")
+add_definitions(-DPADDLE_USE_MKLDNN)
+LIST(APPEND external_project_dependencies mkldnn)
--- a/cmake/external/mklml.cmake
+++ b/cmake/external/mklml.cmake
+# Copyright (c) 2017 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.
+IF(NOT ${WITH_MKLML})
+  return()
+ENDIF(NOT ${WITH_MKLML})
+IF(WIN32 OR APPLE)
+    MESSAGE(WARNING
+        "Windows or Mac is not supported with MKLML in Paddle yet."
+        "Force WITH_MKLML=OFF")
+    SET(WITH_MKLML OFF CACHE STRING "Disable MKLML package in Windows and MacOS" FORCE)
+    return()
+ENDIF()
+INCLUDE(ExternalProject)
+SET(MKLML_PROJECT       "extern_mklml")
+SET(MKLML_VER           "mklml_lnx_2018.0.1.20171007")
+SET(MKLML_URL           "https://github.com/01org/mkl-dnn/releases/download/v0.11/${MKLML_VER}.tgz")
+SET(MKLML_SOURCE_DIR    "${THIRD_PARTY_PATH}/mklml")
+SET(MKLML_DOWNLOAD_DIR  "${MKLML_SOURCE_DIR}/src/${MKLML_PROJECT}")
+SET(MKLML_DST_DIR       "mklml")
+SET(MKLML_INSTALL_ROOT  "${THIRD_PARTY_PATH}/install")
+SET(MKLML_INSTALL_DIR   ${MKLML_INSTALL_ROOT}/${MKLML_DST_DIR})
+SET(MKLML_ROOT          ${MKLML_INSTALL_DIR}/${MKLML_VER})
+SET(MKLML_INC_DIR       ${MKLML_ROOT}/include)
+SET(MKLML_LIB_DIR       ${MKLML_ROOT}/lib)
+SET(MKLML_LIB           ${MKLML_LIB_DIR}/libmklml_intel.so)
+SET(MKLML_IOMP_LIB      ${MKLML_LIB_DIR}/libiomp5.so)
+SET(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_RPATH}" "${MKLML_ROOT}/lib")
+INCLUDE_DIRECTORIES(${MKLML_INC_DIR})
+FILE(WRITE ${MKLML_DOWNLOAD_DIR}/CMakeLists.txt
+  "PROJECT(MKLML)\n"
+  "cmake_minimum_required(VERSION 3.0)\n"
+  "install(DIRECTORY ${MKLML_VER}\n"
+  "        DESTINATION ${MKLML_DST_DIR})\n")
+ExternalProject_Add(
+    ${MKLML_PROJECT}
+    ${EXTERNAL_PROJECT_LOG_ARGS}
+    PREFIX                ${MKLML_SOURCE_DIR}
+    DOWNLOAD_DIR          ${MKLML_DOWNLOAD_DIR}
+    DOWNLOAD_COMMAND      wget --no-check-certificate ${MKLML_URL} -c -q -O ${MKLML_VER}.tgz 
+                          && tar zxf ${MKLML_VER}.tgz
+    DOWNLOAD_NO_PROGRESS  1
+    UPDATE_COMMAND        ""
+    CMAKE_ARGS            -DCMAKE_INSTALL_PREFIX=${MKLML_INSTALL_ROOT}
+    CMAKE_CACHE_ARGS      -DCMAKE_INSTALL_PREFIX:PATH=${MKLML_INSTALL_ROOT}
+)
+ADD_LIBRARY(mklml SHARED IMPORTED GLOBAL)
+SET_PROPERTY(TARGET mklml PROPERTY IMPORTED_LOCATION ${MKLML_LIB})
+ADD_DEPENDENCIES(mklml ${MKLML_PROJECT})
+LIST(APPEND external_project_dependencies mklml)
--- a/cmake/external/nccl.cmake
+++ b/cmake/external/nccl.cmake
+# 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.
+if(NOT WITH_GPU)
+  return()
+endif()
+include(ExternalProject)
+set(NCCL_SOURCE_DIR ${THIRD_PARTY_PATH}/nccl)
+include_directories(${NCCL_SOURCE_DIR}/src/extern_nccl/src)
+if(WITH_DSO)
+  # If we use DSO, we do not build nccl, just download the dependencies
+  set(NCCL_BUILD_COMMAND "")
+  set(NCCL_INSTALL_COMMAND "")
+  set(NCCL_INSTALL_DIR "")
+else()
+  # otherwise, we build nccl and link it.
+  set(NCCL_INSTALL_DIR ${THIRD_PARTY_PATH}/install/nccl)
+  # Note: cuda 8.0 is needed to make nccl
+  # When cuda is not installed on the system directory, need to set CUDA_HOME to your cuda root
+  set(NCCL_BUILD_COMMAND "make -j 8")
+  set(NCCL_INSTALL_COMMAND  "make install PREFIX=${NCCL_INSTALL_DIR}")
+endif()
+ExternalProject_Add(
+    extern_nccl
+    ${EXTERNAL_PROJECT_LOG_ARGS}
+    GIT_REPOSITORY  "https://github.com/NVIDIA/nccl.git"
+    GIT_TAG         "v1.3.4-1"
+    PREFIX          "${NCCL_SOURCE_DIR}"
+    UPDATE_COMMAND  ""
+    CONFIGURE_COMMAND ""
+    BUILD_COMMAND     "${NCCL_BUILD_COMMAND}"
+    INSTALL_COMMAND   "${NCCL_INSTALL_COMMAND}"
+    INSTALL_DIR       "${NCCL_INSTALL_DIR}"
+    TEST_COMMAND      ""
+)
+if(WITH_DSO)
+  if(${CMAKE_VERSION} VERSION_LESS "3.3.0")
+    set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/lib_nccl_dummy.c)
+    file(WRITE ${dummyfile} "const char * dummy_nccl = \"${dummyfile}\";")
+    add_library(nccl STATIC ${dummyfile})
+  else()
+    add_library(nccl INTERFACE)
+  endif()
+else()
+  add_library(nccl STATIC IMPORTED GLOBAL)
+  set_property(TARGET nccl PROPERTY IMPORTED_LOCATION
+               ${NCCL_INSTALL_DIR}/lib/libnccl_static.a)
+endif()
+add_dependencies(nccl extern_nccl)
--- a/cmake/external/nnpack.cmake
+++ b/cmake/external/nnpack.cmake
+# Find the NNPACK library
+#  NNPACK_ROOT - where to find NNPACK include and library.
+#
+set(NNPACK_FOUND OFF)
+set(NNPACK_ROOT $ENV{NNPACK_ROOT} CACHE PATH "Folder contains NNPACK")
+find_path(NNPACK_INC_DIR nnpack.h PATHS ${NNPACK_ROOT}/include)
+find_library(NNPACK_LIB NAMES nnpack PATHS ${NNPACK_ROOT}/lib)
+find_library(PTHREADPOOL_LIB NAMES pthreadpool PATHS ${NNPACK_ROOT}/lib)
+find_library(NNPACK_UKERNELS_LIB NAMES nnpack_ukernels PATHS ${NNPACK_ROOT}/lib)
+find_library(NNPACK_CPUFEATURES_LIB NAMES cpufeatures PATHS ${NNPACK_ROOT}/lib)
+if(NNPACK_INC_DIR AND NNPACK_LIB AND PTHREADPOOL_LIB)
+  set(NNPACK_FOUND ON)
+  INCLUDE_DIRECTORIES(${NNPACK_INC_DIR})
+  set(NNPACK_LIBS)
+  list(APPEND NNPACK_LIBS ${NNPACK_LIB} ${PTHREADPOOL_LIB})
+  if (NNPACK_UKERNELS_LIB)
+    list(APPEND NNPACK_LIBS ${NNPACK_UKERNELS_LIB})
+  endif()
+  if (NNPACK_CPUFEATURES_LIB)
+    list(APPEND NNPACK_LIBS ${NNPACK_CPUFEATURES_LIB})
+  endif()
+  if(NOT ANDROID)
+    list(APPEND NNPACK_LIBS "rt")
+  endif()
+else()
+  message(FATAL_ERROR "Cannot find NNPACK in (${NNPACK_ROOT})")
+endif()
--- a/cmake/external/openblas.cmake
+++ b/cmake/external/openblas.cmake
 # 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.
+IF(USE_EIGEN_FOR_BLAS)
+    return()
+ENDIF(USE_EIGEN_FOR_BLAS)
 INCLUDE(cblas)
 IF(NOT ${CBLAS_FOUND})
@@ -21,65 +25,103 @@ IF(NOT ${CBLAS_FOUND})
    SET(CBLAS_INSTALL_DIR ${THIRD_PARTY_PATH}/install/openblas)
    SET(CBLAS_INC_DIR "${CBLAS_INSTALL_DIR}/include" CACHE PATH "openblas include directory." FORCE)
-    IF(WIN32)
+    SET(CBLAS_LIBRARIES
-        SET(CBLAS_LIBRARIES "${CBLAS_INSTALL_DIR}/lib/openblas.lib" CACHE FILEPATH "openblas library." FORCE)
+        "${CBLAS_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}openblas${CMAKE_STATIC_LIBRARY_SUFFIX}"
-    ELSE(WIN32)
+        CACHE FILEPATH "openblas library." FORCE)
-        SET(CBLAS_LIBRARIES "${CBLAS_INSTALL_DIR}/lib/libopenblas.a" CACHE FILEPATH "openblas library" FORCE)
-    ENDIF(WIN32)
-    IF(CMAKE_COMPILER_IS_GNUCC)
+    SET(OPENBLAS_CC "${CMAKE_C_COMPILER} -Wno-unused-but-set-variable -Wno-unused-variable")
-        ENABLE_LANGUAGE(Fortran)
-        if (NOT CMAKE_Fortran_COMPILER_VERSION)
-          # cmake < 3.4 cannot get CMAKE_Fortran_COMPILER_VERSION directly.
-          execute_process(COMMAND ${CMAKE_Fortran_COMPILER} -dumpversion
-                    OUTPUT_VARIABLE CMAKE_Fortran_COMPILER_VERSION)
-        endif()
-        string(REGEX MATCHALL "[0-9]+" Fortran_VERSION ${CMAKE_Fortran_COMPILER_VERSION})
-        list(GET Fortran_VERSION 0 Fortran_MAJOR)
-        list(GET Fortran_VERSION 1 Fortran_MINOR)
-        find_library(GFORTRAN_LIBRARY NAMES gfortran PATHS 
-                     /lib
-                     /usr/lib
-                     /usr/lib/gcc/x86_64-linux-gnu/${Fortran_MAJOR}.${Fortran_MINOR}/
-                     /usr/lib/gcc/x86_64-linux-gnu/${Fortran_MAJOR}/)
-        if (NOT GFORTRAN_LIBRARY)
-            message(FATAL_ERROR "Cannot found gfortran library which it is used by openblas")
-        endif()
-        find_package(Threads REQUIRED)
-        LIST(APPEND CBLAS_LIBRARIES ${GFORTRAN_LIBRARY} ${CMAKE_THREAD_LIBS_INIT})
-    ENDIF(CMAKE_COMPILER_IS_GNUCC)
-    IF(NOT CMAKE_Fortran_COMPILER)
+    IF(CMAKE_CROSSCOMPILING)
-        MESSAGE(FATAL_ERROR "To build lapack in libopenblas, "
+        SET(OPTIONAL_ARGS HOSTCC=${HOST_C_COMPILER})
-                "you need to set gfortran compiler: cmake .. -DCMAKE_Fortran_COMPILER=...")
+        GET_FILENAME_COMPONENT(CROSS_SUFFIX ${CMAKE_C_COMPILER} DIRECTORY)
-    ENDIF(NOT CMAKE_Fortran_COMPILER)
+        SET(CROSS_SUFFIX ${CROSS_SUFFIX}/)
+        IF(ANDROID)
+            # arm_soft_fp_abi branch of OpenBLAS to support softfp
+            #   https://github.com/xianyi/OpenBLAS/tree/arm_soft_fp_abi
+            SET(OPENBLAS_COMMIT "b5c96fcfcdc82945502a2303116a64d89985daf5")
+            IF(ANDROID_ABI MATCHES "^armeabi(-v7a)?$")
+                SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} TARGET=ARMV7 ARM_SOFTFP_ABI=1 USE_THREAD=0)
+            ELSEIF(ANDROID_ABI STREQUAL "arm64-v8a")
+                SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} TARGET=ARMV8 BINARY=64 USE_THREAD=0)
+            ENDIF()
+        ELSEIF(IOS)
+            IF(CMAKE_OSX_ARCHITECTURES MATCHES "arm64")
+                SET(OPENBLAS_COMMIT "b5c96fcfcdc82945502a2303116a64d89985daf5")
+                SET(OPENBLAS_CC "${OPENBLAS_CC} ${CMAKE_C_FLAGS} -isysroot ${CMAKE_OSX_SYSROOT}")
+                SET(OPENBLAS_CC "${OPENBLAS_CC} -arch arm64")
+                SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} TARGET=ARMV8 BINARY=64 USE_THREAD=0 CROSS_SUFFIX=${CROSS_SUFFIX})
+            ELSE()
+                MESSAGE(FATAL_ERROR "OpenBLAS only support arm64 architectures on iOS. "
+                       "You can set IOS_USE_VECLIB_FOR_BLAS=ON or USE_EIGEN_FOR_BLAS=ON to use other blas library instead.")
+            ENDIF()
+        ELSEIF(RPI)
+            # use hardfp
+            SET(OPENBLAS_COMMIT "v0.2.20")
+            SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} TARGET=ARMV7 USE_THREAD=0)
+        ENDIF()
+    ELSE()
+        IF(APPLE)
+            SET(OPENBLAS_CC "${CMAKE_C_COMPILER} -isysroot ${CMAKE_OSX_SYSROOT}")
+        ENDIF()
+        SET(OPENBLAS_COMMIT "v0.2.20")
+        SET(OPTIONAL_ARGS "")
+        IF(CMAKE_SYSTEM_PROCESSOR MATCHES "^x86(_64)?$")
+            SET(OPTIONAL_ARGS DYNAMIC_ARCH=1 NUM_THREADS=64)
+        ENDIF()
+    ENDIF()
-    ADD_DEFINITIONS(-DPADDLE_USE_LAPACK)
+    SET(COMMON_ARGS CC=${OPENBLAS_CC} NO_SHARED=1 NO_LAPACK=1 libs)
    ExternalProject_Add(
-        openblas
+        extern_openblas
        ${EXTERNAL_PROJECT_LOG_ARGS}
        GIT_REPOSITORY      https://github.com/xianyi/OpenBLAS.git
-        GIT_TAG             v0.2.19
+        GIT_TAG             ${OPENBLAS_COMMIT}
        PREFIX              ${CBLAS_SOURCES_DIR}
        INSTALL_DIR         ${CBLAS_INSTALL_DIR}
        BUILD_IN_SOURCE     1
-        BUILD_COMMAND       ${CMAKE_MAKE_PROGRAM} FC=${CMAKE_Fortran_COMPILER} CC=${CMAKE_C_COMPILER} HOSTCC=${CMAKE_C_COMPILER} DYNAMIC_ARCH=1 NO_SHARED=1 libs netlib
+        BUILD_COMMAND       ${CMAKE_MAKE_PROGRAM} ${COMMON_ARGS} ${OPTIONAL_ARGS}
-        INSTALL_COMMAND     ${CMAKE_MAKE_PROGRAM} install NO_SHARED=1 PREFIX=<INSTALL_DIR>
+        INSTALL_COMMAND     ${CMAKE_MAKE_PROGRAM} install NO_SHARED=1 NO_LAPACK=1 PREFIX=<INSTALL_DIR>
        UPDATE_COMMAND      ""
        CONFIGURE_COMMAND   ""
    )
+    SET(CBLAS_PROVIDER openblas)
-    ExternalProject_Add_Step(
+    IF(WITH_C_API)
-        openblas lapacke_install
+        INSTALL(DIRECTORY ${CBLAS_INC_DIR} DESTINATION third_party/openblas)
-        COMMAND ${CMAKE_COMMAND} -E copy "${CBLAS_SOURCES_DIR}/src/openblas/lapack-netlib/LAPACKE/include/lapacke_mangling_with_flags.h" "${CBLAS_INSTALL_DIR}/include/lapacke_mangling.h"
+        # Because libopenblas.a is a symbolic link of another library, thus need to
-        COMMAND ${CMAKE_COMMAND} -E copy "${CBLAS_SOURCES_DIR}/src/openblas/lapack-netlib/LAPACKE/include/lapacke.h" "${CBLAS_INSTALL_DIR}/include/lapacke.h"
+        # install the whole directory.
-        COMMAND ${CMAKE_COMMAND} -E copy "${CBLAS_SOURCES_DIR}/src/openblas/lapack-netlib/LAPACKE/include/lapacke_config.h" "${CBLAS_INSTALL_DIR}/include/lapacke_config.h"
+        IF(ANDROID)
-        COMMAND ${CMAKE_COMMAND} -E copy "${CBLAS_SOURCES_DIR}/src/openblas/lapack-netlib/LAPACKE/include/lapacke_utils.h" "${CBLAS_INSTALL_DIR}/include/lapacke_utils.h"
+            SET(TMP_INSTALL_DIR third_party/openblas/lib/${ANDROID_ABI})
-        DEPENDEES install
+        ELSE()
-    )
+            SET(TMP_INSTALL_DIR third_party/openblas/lib)
+        ENDIF()
-    LIST(APPEND external_project_dependencies openblas)
+        INSTALL(CODE "execute_process(
+            COMMAND ${CMAKE_COMMAND} -E copy_directory ${CBLAS_INSTALL_DIR}/lib
+                    ${CMAKE_INSTALL_PREFIX}/${TMP_INSTALL_DIR}
+            )"
+        )
+        INSTALL(CODE "MESSAGE(STATUS \"Installing: \"
+                \"${CBLAS_INSTALL_DIR}/lib -> ${CMAKE_INSTALL_PREFIX}/${TMP_INSTALL_DIR}\"
+            )"
+        )
+    ENDIF()
 ENDIF(NOT ${CBLAS_FOUND})
+MESSAGE(STATUS "BLAS library: ${CBLAS_LIBRARIES}")
 INCLUDE_DIRECTORIES(${CBLAS_INC_DIR})
+# FIXME(gangliao): generate cblas target to track all high performance
+# linear algebra libraries for cc_library(xxx SRCS xxx.c DEPS cblas)
+SET(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/cblas_dummy.c)
+FILE(WRITE ${dummyfile} "const char * dummy = \"${dummyfile}\";")
+ADD_LIBRARY(cblas STATIC ${dummyfile})
+TARGET_LINK_LIBRARIES(cblas ${CBLAS_LIBRARIES})
+IF(NOT ${CBLAS_FOUND})
+    ADD_DEPENDENCIES(cblas extern_openblas)
+    LIST(APPEND external_project_dependencies cblas)
+ELSE()
+    IF("${CBLAS_PROVIDER}" STREQUAL "MKLML")
+        ADD_DEPENDENCIES(cblas mklml)
+    ENDIF()
+ENDIF(NOT ${CBLAS_FOUND})
--- a/cmake/external/protobuf.cmake
+++ b/cmake/external/protobuf.cmake
 # Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
-# 
+#
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
 # You may obtain a copy of the License at
-# 
+#
 # http://www.apache.org/licenses/LICENSE-2.0
-# 
+#
 # Unless required by applicable law or agreed to in writing, software
 # distributed under the License is distributed on an "AS IS" BASIS,
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
@@ -13,68 +13,257 @@
 # limitations under the License.
 INCLUDE(ExternalProject)
+# Always invoke `FIND_PACKAGE(Protobuf)` for importing function protobuf_generate_cpp
+FIND_PACKAGE(Protobuf QUIET)
+macro(UNSET_VAR VAR_NAME)
+    UNSET(${VAR_NAME} CACHE)
+    UNSET(${VAR_NAME})
+endmacro()
+UNSET_VAR(PROTOBUF_INCLUDE_DIR)
+UNSET_VAR(PROTOBUF_FOUND)
+UNSET_VAR(PROTOBUF_PROTOC_EXECUTABLE)
+UNSET_VAR(PROTOBUF_PROTOC_LIBRARY)
+UNSET_VAR(PROTOBUF_LITE_LIBRARY)
+UNSET_VAR(PROTOBUF_LIBRARY)
+UNSET_VAR(PROTOBUF_INCLUDE_DIR)
+UNSET_VAR(Protobuf_PROTOC_EXECUTABLE)
-set(PROTOBUF_VERSION 3.1)
+if(NOT COMMAND protobuf_generate_python)  # before cmake 3.4, protobuf_genrerate_python is not defined.
-FIND_PACKAGE(Protobuf ${PROTOBUF_VERSION})
+    function(protobuf_generate_python SRCS)
+        # shameless copy from https://github.com/Kitware/CMake/blob/master/Modules/FindProtobuf.cmake
+        if(NOT ARGN)
+            message(SEND_ERROR "Error: PROTOBUF_GENERATE_PYTHON() called without any proto files")
+            return()
+        endif()
-IF(PROTOBUF_FOUND)
+        if(PROTOBUF_GENERATE_CPP_APPEND_PATH)
+            # Create an include path for each file specified
+            foreach(FIL ${ARGN})
+                get_filename_component(ABS_FIL ${FIL} ABSOLUTE)
+                get_filename_component(ABS_PATH ${ABS_FIL} PATH)
+                list(FIND _protobuf_include_path ${ABS_PATH} _contains_already)
+                if(${_contains_already} EQUAL -1)
+                    list(APPEND _protobuf_include_path -I ${ABS_PATH})
+                endif()
+            endforeach()
+        else()
+            set(_protobuf_include_path -I ${CMAKE_CURRENT_SOURCE_DIR})
+        endif()
+        if(DEFINED PROTOBUF_IMPORT_DIRS AND NOT DEFINED Protobuf_IMPORT_DIRS)
+            set(Protobuf_IMPORT_DIRS "${PROTOBUF_IMPORT_DIRS}")
+        endif()
+        if(DEFINED Protobuf_IMPORT_DIRS)
+            foreach(DIR ${Protobuf_IMPORT_DIRS})
+                get_filename_component(ABS_PATH ${DIR} ABSOLUTE)
+                list(FIND _protobuf_include_path ${ABS_PATH} _contains_already)
+                if(${_contains_already} EQUAL -1)
+                    list(APPEND _protobuf_include_path -I ${ABS_PATH})
+                endif()
+            endforeach()
+        endif()
+        set(${SRCS})
+        foreach(FIL ${ARGN})
+            get_filename_component(ABS_FIL ${FIL} ABSOLUTE)
+            get_filename_component(FIL_WE ${FIL} NAME_WE)
+            if(NOT PROTOBUF_GENERATE_CPP_APPEND_PATH)
+                get_filename_component(FIL_DIR ${FIL} DIRECTORY)
+                if(FIL_DIR)
+                    set(FIL_WE "${FIL_DIR}/${FIL_WE}")
+                endif()
+            endif()
+            list(APPEND ${SRCS} "${CMAKE_CURRENT_BINARY_DIR}/${FIL_WE}_pb2.py")
+            add_custom_command(
+                    OUTPUT "${CMAKE_CURRENT_BINARY_DIR}/${FIL_WE}_pb2.py"
+                    COMMAND  ${Protobuf_PROTOC_EXECUTABLE} --python_out ${CMAKE_CURRENT_BINARY_DIR} ${_protobuf_include_path} ${ABS_FIL}
+                    DEPENDS ${ABS_FIL} ${Protobuf_PROTOC_EXECUTABLE}
+                    COMMENT "Running Python protocol buffer compiler on ${FIL}"
+                    VERBATIM )
+        endforeach()
+        set(${SRCS} ${${SRCS}} PARENT_SCOPE)
+    endfunction()
+endif()
+# Print and set the protobuf library information,
+# finish this cmake process and exit from this file.
+macro(PROMPT_PROTOBUF_LIB)
+    SET(protobuf_DEPS ${ARGN})
+    MESSAGE(STATUS "Protobuf protoc executable: ${PROTOBUF_PROTOC_EXECUTABLE}")
+    MESSAGE(STATUS "Protobuf library: ${PROTOBUF_LIBRARY}")
+    MESSAGE(STATUS "Protobuf version: ${PROTOBUF_VERSION}")
+    INCLUDE_DIRECTORIES(${PROTOBUF_INCLUDE_DIR})
+    # Assuming that all the protobuf libraries are of the same type.
+    IF(${PROTOBUF_LIBRARY} MATCHES "${CMAKE_STATIC_LIBRARY_SUFFIX}$")
+        SET(protobuf_LIBTYPE STATIC)
+    ELSEIF(${PROTOBUF_LIBRARY} MATCHES "${CMAKE_SHARED_LIBRARY_SUFFIX}$")
+        SET(protobuf_LIBTYPE SHARED)
+    ELSE()
+        MESSAGE(FATAL_ERROR "Unknown library type: ${PROTOBUF_LIBRARY}")
+    ENDIF()
+    ADD_LIBRARY(protobuf ${protobuf_LIBTYPE} IMPORTED GLOBAL)
+    SET_PROPERTY(TARGET protobuf PROPERTY IMPORTED_LOCATION ${PROTOBUF_LIBRARY})
+    ADD_LIBRARY(protobuf_lite ${protobuf_LIBTYPE} IMPORTED GLOBAL)
+    SET_PROPERTY(TARGET protobuf_lite PROPERTY IMPORTED_LOCATION ${PROTOBUF_LITE_LIBRARY})
+    ADD_LIBRARY(libprotoc ${protobuf_LIBTYPE} IMPORTED GLOBAL)
+    SET_PROPERTY(TARGET libprotoc PROPERTY IMPORTED_LOCATION ${PROTOC_LIBRARY})
+    ADD_EXECUTABLE(protoc IMPORTED GLOBAL)
+    SET_PROPERTY(TARGET protoc PROPERTY IMPORTED_LOCATION ${PROTOBUF_PROTOC_EXECUTABLE})
+    # FIND_Protobuf.cmake uses `Protobuf_PROTOC_EXECUTABLE`.
+    # make `protobuf_generate_cpp` happy.
+    SET(Protobuf_PROTOC_EXECUTABLE ${PROTOBUF_PROTOC_EXECUTABLE})
+    FOREACH(dep ${protobuf_DEPS})
+        ADD_DEPENDENCIES(protobuf ${dep})
+        ADD_DEPENDENCIES(protobuf_lite ${dep})
+        ADD_DEPENDENCIES(libprotoc ${dep})
+        ADD_DEPENDENCIES(protoc ${dep})
+    ENDFOREACH()
+    LIST(APPEND external_project_dependencies protobuf)
+    RETURN()
+endmacro()
+macro(SET_PROTOBUF_VERSION)
    EXEC_PROGRAM(${PROTOBUF_PROTOC_EXECUTABLE} ARGS --version OUTPUT_VARIABLE PROTOBUF_VERSION)
    STRING(REGEX MATCH "[0-9]+.[0-9]+" PROTOBUF_VERSION "${PROTOBUF_VERSION}")
-    IF (${PROTOBUF_VERSION} VERSION_LESS "3.1.0")
+endmacro()
-        SET(PROTOBUF_FOUND OFF)
+set(PROTOBUF_ROOT "" CACHE PATH "Folder contains protobuf")
+if (NOT "${PROTOBUF_ROOT}" STREQUAL "")
+    find_path(PROTOBUF_INCLUDE_DIR google/protobuf/message.h PATHS ${PROTOBUF_ROOT}/include NO_DEFAULT_PATH)
+    find_library(PROTOBUF_LIBRARY protobuf PATHS ${PROTOBUF_ROOT}/lib NO_DEFAULT_PATH)
+    find_library(PROTOBUF_LITE_LIBRARY protobuf-lite PATHS ${PROTOBUF_ROOT}/lib NO_DEFAULT_PATH)
+    find_library(PROTOBUF_PROTOC_LIBRARY protoc PATHS ${PROTOBUF_ROOT}/lib NO_DEFAULT_PATH)
+    find_program(PROTOBUF_PROTOC_EXECUTABLE protoc PATHS ${PROTOBUF_ROOT}/bin NO_DEFAULT_PATH)
+    if (PROTOBUF_INCLUDE_DIR AND PROTOBUF_LIBRARY AND PROTOBUF_LITE_LIBRARY AND PROTOBUF_PROTOC_LIBRARY AND PROTOBUF_PROTOC_EXECUTABLE)
+        message(STATUS "Using custom protobuf library in ${PROTOBUF_ROOT}.")
+        SET_PROTOBUF_VERSION()
+        PROMPT_PROTOBUF_LIB()
+    else()
+        message(WARNING "Cannot find protobuf library in ${PROTOBUF_ROOT}.")
+    endif()
+endif()
+FUNCTION(build_protobuf TARGET_NAME BUILD_FOR_HOST)
+    STRING(REPLACE "extern_" "" TARGET_DIR_NAME "${TARGET_NAME}")
+    SET(PROTOBUF_SOURCES_DIR ${THIRD_PARTY_PATH}/${TARGET_DIR_NAME})
+    SET(PROTOBUF_INSTALL_DIR ${THIRD_PARTY_PATH}/install/${TARGET_DIR_NAME})
+    SET(${TARGET_NAME}_INCLUDE_DIR "${PROTOBUF_INSTALL_DIR}/include" PARENT_SCOPE)
+    SET(PROTOBUF_INCLUDE_DIR "${PROTOBUF_INSTALL_DIR}/include" PARENT_SCOPE)
+    SET(${TARGET_NAME}_LITE_LIBRARY
+        "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf-lite${CMAKE_STATIC_LIBRARY_SUFFIX}"
+         PARENT_SCOPE)
+    SET(${TARGET_NAME}_LIBRARY
+        "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf${CMAKE_STATIC_LIBRARY_SUFFIX}"
+         PARENT_SCOPE)
+    SET(${TARGET_NAME}_PROTOC_LIBRARY
+        "${PROTOBUF_INSTALL_DIR}/lib/libprotoc${CMAKE_STATIC_LIBRARY_SUFFIX}"
+         PARENT_SCOPE)
+    SET(${TARGET_NAME}_PROTOC_EXECUTABLE
+        "${PROTOBUF_INSTALL_DIR}/bin/protoc${CMAKE_EXECUTABLE_SUFFIX}"
+         PARENT_SCOPE)
+    SET(OPTIONAL_CACHE_ARGS "")
+    SET(OPTIONAL_ARGS "")
+    IF(BUILD_FOR_HOST)
+        SET(OPTIONAL_ARGS "-Dprotobuf_WITH_ZLIB=OFF")
+    ELSE()
+        SET(OPTIONAL_ARGS
+            "-DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}"
+            "-DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}"
+            "-DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}"
+            "-DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}"
+            "-Dprotobuf_WITH_ZLIB=ON"
+            "-DZLIB_ROOT:FILEPATH=${ZLIB_ROOT}"
+            ${EXTERNAL_OPTIONAL_ARGS})
+        SET(OPTIONAL_CACHE_ARGS "-DZLIB_ROOT:STRING=${ZLIB_ROOT}")
    ENDIF()
-ENDIF(PROTOBUF_FOUND)
-IF(NOT PROTOBUF_FOUND)
+    SET(PROTOBUF_REPO "https://github.com/google/protobuf.git")
-    SET(PROTOBUF_SOURCES_DIR ${THIRD_PARTY_PATH}/protobuf)
+    SET(PROTOBUF_TAG "9f75c5aa851cd877fb0d93ccc31b8567a6706546")
-    SET(PROTOBUF_INSTALL_DIR ${THIRD_PARTY_PATH}/install/protobuf)
+    IF(MOBILE_INFERENCE)
-    SET(PROTOBUF_INCLUDE_DIR "${PROTOBUF_INSTALL_DIR}/include" CACHE PATH "protobuf include directory." FORCE)
+        # The reason why the official version is not used is described in
+        # https://github.com/PaddlePaddle/Paddle/issues/6114
-    IF(WIN32)
+        SET(PROTOBUF_REPO "https://github.com/qingqing01/protobuf.git")
-        SET(PROTOBUF_LITE_LIBRARY
+        SET(PROTOBUF_TAG "v3.2.0")
-            "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf-lite.lib" CACHE FILEPATH "protobuf lite library." FORCE)
+        IF(NOT BUILD_FOR_HOST)
-        SET(PROTOBUF_LIBRARY
+            SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} "-Dprotobuf_BUILD_PROTOC_BINARIES=OFF")
-            "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf.lib" CACHE FILEPATH "protobuf library." FORCE)
+        ENDIF()
-        SET(PROTOBUF_PROTOC_LIBRARY
+    ENDIF()
-            "${PROTOBUF_INSTALL_DIR}/lib/libprotoc.lib" CACHE FILEPATH "protoc library." FORCE)
-        SET(PROTOBUF_PROTOC_EXECUTABLE "${PROTOBUF_INSTALL_DIR}/bin/protoc.exe" CACHE FILEPATH "protobuf executable." FORCE)
-    ELSE(WIN32)
-        SET(PROTOBUF_LITE_LIBRARY
-            "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf-lite.a" CACHE FILEPATH "protobuf lite library." FORCE)
-        SET(PROTOBUF_LIBRARY
-            "${PROTOBUF_INSTALL_DIR}/lib/libprotobuf.a" CACHE FILEPATH "protobuf library." FORCE)
-        SET(PROTOBUF_PROTOC_LIBRARY
-            "${PROTOBUF_INSTALL_DIR}/lib/libprotoc.a" CACHE FILEPATH "protoc library." FORCE)
-        SET(PROTOBUF_PROTOC_EXECUTABLE "${PROTOBUF_INSTALL_DIR}/bin/protoc" CACHE FILEPATH "protobuf executable." FORCE)
-    ENDIF(WIN32)
    ExternalProject_Add(
-        protobuf
+        ${TARGET_NAME}
        ${EXTERNAL_PROJECT_LOG_ARGS}
        PREFIX          ${PROTOBUF_SOURCES_DIR}
        UPDATE_COMMAND  ""
        DEPENDS         zlib
-        GIT_REPOSITORY  "https://github.com/google/protobuf.git"
+        GIT_REPOSITORY  ${PROTOBUF_REPO}
-        GIT_TAG         "9f75c5aa851cd877fb0d93ccc31b8567a6706546"
+        GIT_TAG         ${PROTOBUF_TAG}
        CONFIGURE_COMMAND
-        ${CMAKE_COMMAND} ${PROTOBUF_SOURCES_DIR}/src/protobuf/cmake
+        ${CMAKE_COMMAND} ${PROTOBUF_SOURCES_DIR}/src/${TARGET_NAME}/cmake
+            ${OPTIONAL_ARGS}
            -Dprotobuf_BUILD_TESTS=OFF
-            -DZLIB_ROOT:FILEPATH=${ZLIB_ROOT}
-            -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-            -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
            -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-            -DCMAKE_BUILD_TYPE=Release
+            -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
            -DCMAKE_INSTALL_PREFIX=${PROTOBUF_INSTALL_DIR}
            -DCMAKE_INSTALL_LIBDIR=lib
        CMAKE_CACHE_ARGS
            -DCMAKE_INSTALL_PREFIX:PATH=${PROTOBUF_INSTALL_DIR}
-            -DCMAKE_BUILD_TYPE:STRING=Release
+            -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
            -DCMAKE_VERBOSE_MAKEFILE:BOOL=OFF
            -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-            -DZLIB_ROOT:STRING=${ZLIB_ROOT}
+            ${OPTIONAL_CACHE_ARGS}
    )
+ENDFUNCTION()
-    LIST(APPEND external_project_dependencies protobuf)
+IF(NOT MOBILE_INFERENCE)
-ENDIF(NOT PROTOBUF_FOUND)
+    SET(PROTOBUF_VERSION 3.1)
+ELSE()
+    SET(PROTOBUF_VERSION 3.2)
+ENDIF()
+IF(CMAKE_CROSSCOMPILING)
+    build_protobuf(protobuf_host TRUE)
+    LIST(APPEND external_project_dependencies protobuf_host)
+    SET(PROTOBUF_PROTOC_EXECUTABLE ${protobuf_host_PROTOC_EXECUTABLE}
+        CACHE FILEPATH "protobuf executable." FORCE)
+ENDIF()
-INCLUDE_DIRECTORIES(${PROTOBUF_INCLUDE_DIR})
+IF(NOT PROTOBUF_FOUND)
+    build_protobuf(extern_protobuf FALSE)
+    SET(PROTOBUF_INCLUDE_DIR ${extern_protobuf_INCLUDE_DIR}
+        CACHE PATH "protobuf include directory." FORCE)
+    SET(PROTOBUF_LITE_LIBRARY ${extern_protobuf_LITE_LIBRARY}
+        CACHE FILEPATH "protobuf lite library." FORCE)
+    SET(PROTOBUF_LIBRARY ${extern_protobuf_LIBRARY}
+        CACHE FILEPATH "protobuf library." FORCE)
+    SET(PROTOBUF_PROTOC_LIBRARY ${extern_protobuf_PROTOC_LIBRARY}
+        CACHE FILEPATH "protoc library." FORCE)
+    IF(WITH_C_API)
+        INSTALL(DIRECTORY ${PROTOBUF_INCLUDE_DIR} DESTINATION third_party/protobuf)
+        IF(ANDROID)
+            INSTALL(FILES ${PROTOBUF_LIBRARY} DESTINATION third_party/protobuf/lib/${ANDROID_ABI})
+        ELSE()
+            INSTALL(FILES ${PROTOBUF_LIBRARY} DESTINATION third_party/protobuf/lib)
+        ENDIF()
+    ENDIF()
+    IF(CMAKE_CROSSCOMPILING)
+        PROMPT_PROTOBUF_LIB(protobuf_host extern_protobuf)
+    ELSE()
+        SET(PROTOBUF_PROTOC_EXECUTABLE ${extern_protobuf_PROTOC_EXECUTABLE}
+            CACHE FILEPATH "protobuf executable." FORCE)
+        PROMPT_PROTOBUF_LIB(extern_protobuf)
+    ENDIF()
+ENDIF(NOT PROTOBUF_FOUND)
--- a/cmake/external/pybind11.cmake
+++ b/cmake/external/pybind11.cmake
+# 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.
+if(NOT WITH_PYTHON)
+    return()
+endif()
+include(ExternalProject)
+set(PYBIND_SOURCE_DIR ${THIRD_PARTY_PATH}/pybind)
+include_directories(${PYBIND_SOURCE_DIR}/src/extern_pybind/include)
+ExternalProject_Add(
+        extern_pybind
+        ${EXTERNAL_PROJECT_LOG_ARGS}
+        GIT_REPOSITORY  "https://github.com/pybind/pybind11.git"
+        GIT_TAG         "v2.1.1"
+        PREFIX          ${PYBIND_SOURCE_DIR}
+        UPDATE_COMMAND  ""
+        CONFIGURE_COMMAND ""
+        BUILD_COMMAND     ""
+        INSTALL_COMMAND   ""
+        TEST_COMMAND      ""
+)
+if(${CMAKE_VERSION} VERSION_LESS "3.3.0")
+    set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/pybind_dummy.c)
+    file(WRITE ${dummyfile} "const char * dummy_pybind = \"${dummyfile}\";")
+    add_library(pybind STATIC ${dummyfile})
+else()
+    add_library(pybind INTERFACE)
+endif()
+add_dependencies(pybind extern_pybind)
--- a/cmake/external/python.cmake
+++ b/cmake/external/python.cmake
 # 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(ExternalProject)
+IF(NOT WITH_PYTHON)
+    return()
+ENDIF()
 INCLUDE(python_module)
 FIND_PACKAGE(PythonInterp 2.7)
 FIND_PACKAGE(PythonLibs 2.7)
+# Fixme: Maybe find a static library. Get SHARED/STATIC by FIND_PACKAGE.
+ADD_LIBRARY(python SHARED IMPORTED GLOBAL)
+SET_PROPERTY(TARGET python PROPERTY IMPORTED_LOCATION ${PYTHON_LIBRARIES})
 SET(py_env "")
+IF(PYTHONINTERP_FOUND)
-IF(PYTHONLIBS_FOUND AND PYTHONINTERP_FOUND)
    find_python_module(pip REQUIRED)
    find_python_module(numpy REQUIRED)
    find_python_module(wheel REQUIRED)
@@ -30,198 +35,7 @@ IF(PYTHONLIBS_FOUND AND PYTHONINTERP_FOUND)
        MESSAGE(FATAL_ERROR "Found Python Protobuf ${PY_GOOGLE.PROTOBUF_VERSION} < 3.0.0, "
        "please use pip to upgrade protobuf. pip install -U protobuf")
    ENDIF()
-ELSE(PYTHONLIBS_FOUND AND PYTHONINTERP_FOUND)
+ENDIF(PYTHONINTERP_FOUND)
-    MESSAGE(FATAL_ERROR "Please install python 2.7 before building PaddlePaddle.")
-    ##################################### PYTHON ########################################
-    SET(PYTHON_SOURCES_DIR ${THIRD_PARTY_PATH}/python)
-    SET(PYTHON_INSTALL_DIR ${THIRD_PARTY_PATH}/install/python)
-    SET(_python_DIR ${PYTHON_INSTALL_DIR})
-    IF(UNIX)
-        SET(PYTHON_FOUND ON)
-        SET(PYTHON_INCLUDE_DIR "${PYTHON_INSTALL_DIR}/include/python2.7" CACHE PATH "Python include dir" FORCE)
-        SET(PYTHON_LIBRARIES "${PYTHON_INSTALL_DIR}/lib/libpython2.7.a" CACHE FILEPATH "Python library" FORCE)
-        SET(PYTHON_EXECUTABLE ${PYTHON_INSTALL_DIR}/bin/python CACHE FILEPATH "Python executable" FORCE)
-        SET(PY_SITE_PACKAGES_PATH "${PYTHON_INSTALL_DIR}/lib/python2.7/site-packages" CACHE PATH "Python site-packages path" FORCE)
-    ELSEIF(WIN32)
-        SET(PYTHON_FOUND ON)
-        SET(PYTHON_INCLUDE_DIR "${PYTHON_INSTALL_DIR}/include" CACHE PATH "Python include dir" FORCE)
-        SET(PYTHON_LIBRARIES "${PYTHON_INSTALL_DIR}/libs/python27.lib" CACHE FILEPATH "Python library" FORCE)
-        SET(PYTHON_EXECUTABLE "${PYTHON_INSTALL_DIR}/bin/python.exe" CACHE FILEPATH "Python executable" FORCE)
-        SET(PY_SITE_PACKAGES_PATH "${PYTHON_INSTALL_DIR}/Lib/site-packages" CACHE PATH "Python site-packages path" FORCE)
-    ELSE()
-        MESSAGE(FATAL_ERROR "Unknown system !")
-    ENDIF()
-    IF(APPLE)
-        LIST(APPEND EXTERNAL_PROJECT_OPTIONAL_CMAKE_ARGS
-            -DCMAKE_BUILD_WITH_INSTALL_RPATH:BOOL=ON
-            )
-    ENDIF()
-    SET(EXTERNAL_PROJECT_OPTIONAL_CMAKE_CACHE_ARGS)
-    # Force Python build to "Release".
-    IF(CMAKE_CONFIGURATION_TYPES)
-        SET(SAVED_CMAKE_CFG_INTDIR ${CMAKE_CFG_INTDIR})
-        SET(CMAKE_CFG_INTDIR "Release")
-    ELSE()
-        LIST(APPEND EXTERNAL_PROJECT_OPTIONAL_CMAKE_CACHE_ARGS
-            -DCMAKE_BUILD_TYPE:STRING=Release
-            )
-    ENDIF()
-    ExternalProject_Add(python
-        ${EXTERNAL_PROJECT_LOG_ARGS}
-        GIT_REPOSITORY    "https://github.com/python-cmake-buildsystem/python-cmake-buildsystem.git"
-        PREFIX            ${PYTHON_SOURCES_DIR}
-        UPDATE_COMMAND    ""
-        CMAKE_ARGS        -DPYTHON_VERSION=2.7.12
-        CMAKE_ARGS        -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-        CMAKE_ARGS        -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
-        CMAKE_CACHE_ARGS
-            -DCMAKE_INSTALL_PREFIX:PATH=${PYTHON_INSTALL_DIR}
-            -DBUILD_LIBPYTHON_SHARED:BOOL=OFF
-            -DUSE_SYSTEM_LIBRARIES:BOOL=OFF
-            -DZLIB_ROOT:FILEPATH=${ZLIB_ROOT}
-            -DZLIB_INCLUDE_DIR:PATH=${ZLIB_INCLUDE_DIR}
-            -DZLIB_LIBRARY:FILEPATH=${ZLIB_LIBRARIES}
-            -DDOWNLOAD_SOURCES:BOOL=ON
-            -DINSTALL_WINDOWS_TRADITIONAL:BOOL=OFF
-            ${EXTERNAL_PROJECT_OPTIONAL_CMAKE_CACHE_ARGS}
-            ${EXTERNAL_PROJECT_OPTIONAL_CMAKE_ARGS}
-        DEPENDS zlib
-    )
-    SET(py_env
-        PATH=${PYTHON_INSTALL_DIR}/bin
-        PYTHONHOME=${PYTHON_INSTALL_DIR}
-        PYTHONPATH=${PYTHON_INSTALL_DIR}/lib:${PYTHON_INSTALL_DIR}/lib/python2.7:${PY_SITE_PACKAGES_PATH})
-    ####################################################################################
-    ##################################### SETUPTOOLS ###################################
+INCLUDE_DIRECTORIES(${PYTHON_INCLUDE_DIR})
-    SET(SETUPTOOLS_SOURCES_DIR ${PYTHON_SOURCES_DIR}/setuptools)
+INCLUDE_DIRECTORIES(${PYTHON_NUMPY_INCLUDE_DIR})
-    ExternalProject_Add(setuptools
-        ${EXTERNAL_PROJECT_LOG_ARGS}
-        PREFIX              ${SETUPTOOLS_SOURCES_DIR}
-        URL                 "https://pypi.python.org/packages/source/s/setuptools/setuptools-18.3.2.tar.gz"
-        BUILD_IN_SOURCE     1
-        PATCH_COMMAND       ""
-        UPDATE_COMMAND      ""
-        CONFIGURE_COMMAND   ""
-        INSTALL_COMMAND     ""
-        BUILD_COMMAND       env ${py_env} ${PYTHON_EXECUTABLE} setup.py install
-        DEPENDS             python zlib
-    )
-    #####################################################################################
-    ##################################### SIX ###########################################
-    SET(SIX_SOURCES_DIR ${PYTHON_SOURCES_DIR}/six)
-    ExternalProject_Add(six
-        ${EXTERNAL_PROJECT_LOG_ARGS}
-        PREFIX              ${SIX_SOURCES_DIR}
-        URL                 https://pypi.python.org/packages/source/s/six/six-1.10.0.tar.gz
-        BUILD_IN_SOURCE     1
-        PATCH_COMMAND       ""
-        UPDATE_COMMAND      ""
-        CONFIGURE_COMMAND   ""
-        INSTALL_COMMAND     ""
-        BUILD_COMMAND       env ${py_env} ${PYTHON_EXECUTABLE} setup.py install
-        DEPENDS             python setuptools
-    )
-    #####################################################################################
-    ##################################### CYTHON ########################################
-    SET(CYTHON_SOURCES_DIR ${PYTHON_SOURCES_DIR}/cython)
-    ExternalProject_Add(cython
-        ${EXTERNAL_PROJECT_LOG_ARGS}
-        PREFIX                ${CYTHON_SOURCES_DIR}
-        URL                   https://github.com/cython/cython/archive/0.25.2.tar.gz
-        GIT_TAG               0.25.2
-        BUILD_IN_SOURCE       1
-        CONFIGURE_COMMAND     ""
-        PATCH_COMMAND         ""
-        UPDATE_COMMAND        ""
-        INSTALL_COMMAND       ""
-        BUILD_COMMAND         env ${py_env} ${PYTHON_EXECUTABLE} setup.py install
-        DEPENDS               python
-    )
-    ####################################################################################
-    ##################################### NUMPY ########################################
-    SET(NUMPY_SOURCES_DIR ${PYTHON_SOURCES_DIR}/numpy)
-    SET(NUMPY_TAG_VERSION "v1.11.3")
-    SET(NUMPY_VERSION "1.11.3")
-    SET(EGG_NAME "")
-    SET(PYTHON_NUMPY_INCLUDE_DIR "")
-    IF(WIN32)
-        SET(EGG_NAME "numpy-${NUMPY_VERSION}-py2.7-${HOST_SYSTEM}.egg")
-    ELSE(WIN32)
-        IF(APPLE)
-            SET(EGG_NAME "numpy-${NUMPY_VERSION}-py2.7-${HOST_SYSTEM}-${MACOS_VERSION}")
-        ELSE(APPLE)
-            SET(EGG_NAME "numpy-${NUMPY_VERSION}-py2.7-linux")
-            SET(EGG_NAME "numpy-${NUMPY_VERSION}-py2.7-linux")
-        ENDIF(APPLE)
-        FOREACH(suffix x86_64 intel fat64 fat32 universal)
-            LIST(APPEND PYTHON_NUMPY_INCLUDE_DIR ${PY_SITE_PACKAGES_PATH}/${EGG_NAME}-${suffix}.egg/numpy/core/include)
-        ENDFOREACH()
-    ENDIF(WIN32)
-    ExternalProject_Add(numpy
-        ${EXTERNAL_PROJECT_LOG_ARGS}
-        GIT_REPOSITORY      https://github.com/numpy/numpy.git
-        GIT_TAG             ${NUMPY_TAG_VERSION}
-        CONFIGURE_COMMAND   ""
-        UPDATE_COMMAND      ""
-        PREFIX              ${NUMPY_SOURCES_DIR}
-        BUILD_COMMAND       env ${py_env} ${PYTHON_EXECUTABLE} setup.py build
-        INSTALL_COMMAND     env ${py_env} ${PYTHON_EXECUTABLE} setup.py install
-        BUILD_IN_SOURCE     1
-        DEPENDS             python setuptools cython
-    )
-    ####################################################################################
-    ##################################### WHEEL ########################################
-    SET(WHEEL_SOURCES_DIR ${PYTHON_SOURCES_DIR}/wheel)
-    ExternalProject_Add(wheel
-        ${EXTERNAL_PROJECT_LOG_ARGS}
-        URL                 https://pypi.python.org/packages/source/w/wheel/wheel-0.29.0.tar.gz
-        PREFIX              ${WHEEL_SOURCES_DIR}
-        CONFIGURE_COMMAND   ""
-        UPDATE_COMMAND      ""
-        BUILD_COMMAND       ""
-        INSTALL_COMMAND     env ${py_env} ${PYTHON_EXECUTABLE} setup.py install
-        BUILD_IN_SOURCE     1
-        DEPENDS             python setuptools
-    )
-    ####################################################################################
-    ################################### PROTOBUF #######################################
-    SET(PY_PROTOBUF_SOURCES_DIR ${PYTHON_SOURCES_DIR}/protobuf)
-    ExternalProject_Add(python-protobuf
-        ${EXTERNAL_PROJECT_LOG_ARGS}
-        URL                   https://pypi.python.org/packages/e0/b0/0a1b364fe8a7d177b4b7d4dca5b798500dc57a7273b93cca73931b305a6a/protobuf-3.1.0.post1.tar.gz
-        URL_MD5               38b5fb160c768d2f8444d0c6d637ff91
-        PREFIX                ${PY_PROTOBUF_SOURCES_DIR}
-        BUILD_IN_SOURCE       1
-        PATCH_COMMAND         ""
-        CONFIGURE_COMMAND     ""
-        BUILD_COMMAND         env ${py_env} ${PYTHON_EXECUTABLE} setup.py build
-        INSTALL_COMMAND       env ${py_env} ${PYTHON_EXECUTABLE} setup.py install
-        DEPENDS               python setuptools six
-    )
-    ####################################################################################
-    LIST(APPEND external_project_dependencies python setuptools six cython wheel python-protobuf numpy)
-ENDIF(PYTHONLIBS_FOUND AND PYTHONINTERP_FOUND)
-IF(WITH_PYTHON)
-    INCLUDE_DIRECTORIES(${PYTHON_INCLUDE_DIR})
-    INCLUDE_DIRECTORIES(${PYTHON_NUMPY_INCLUDE_DIR})
-ELSE()
-    SET(PYTHON_LIBRARIES "")
-ENDIF()
--- a/cmake/external/swig.cmake
+++ b/cmake/external/swig.cmake
 # 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.
+IF(NOT WITH_SWIG_PY)
+    return()
+ENDIF()
 FIND_PACKAGE(SWIG)
 IF(NOT SWIG_FOUND)

--- a/cmake/external/warpctc.cmake
+++ b/cmake/external/warpctc.cmake
 # 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.
+IF(MOBILE_INFERENCE)
+    return()
+ENDIF()
 INCLUDE(ExternalProject)
 SET(WARPCTC_SOURCES_DIR ${THIRD_PARTY_PATH}/warpctc)
 SET(WARPCTC_INSTALL_DIR ${THIRD_PARTY_PATH}/install/warpctc)
-SET(WARPCTC_INCLUDE_DIR "${WARPCTC_INSTALL_DIR}/include" CACHE PATH "Warp-ctc Directory" FORCE)
-INCLUDE_DIRECTORIES(${WARPCTC_INCLUDE_DIR})
-SET(WARPCTC_LIB_DIR "${WARPCTC_INSTALL_DIR}/lib" CACHE PATH "Warp-ctc Library Directory" FORCE)
-IF(WIN32)
+SET(WARPCTC_INCLUDE_DIR "${WARPCTC_INSTALL_DIR}/include"
-    SET(WARPCTC_LIBRARIES
+    CACHE PATH "Warp-ctc Directory" FORCE)
-        "${WARPCTC_INSTALL_DIR}/lib/warpctc.dll" CACHE FILEPATH "Warp-ctc Library" FORCE)
+# Used in unit test test_WarpCTCLayer
-ELSE(WIN32)
+SET(WARPCTC_LIB_DIR "${WARPCTC_INSTALL_DIR}/lib"
-    IF(APPLE)
+    CACHE PATH "Warp-ctc Library Directory" FORCE)
-        SET(_warpctc_SHARED_SUFFIX dylib)
+SET(WARPCTC_LIBRARIES "${WARPCTC_INSTALL_DIR}/lib/libwarpctc${CMAKE_SHARED_LIBRARY_SUFFIX}"
-    ELSE(APPLE)
+    CACHE FILEPATH "Warp-ctc Library" FORCE)
-        SET(_warpctc_SHARED_SUFFIX so)
-    ENDIF(APPLE)
-    SET(WARPCTC_LIBRARIES
-        "${WARPCTC_INSTALL_DIR}/lib/libwarpctc.${_warpctc_SHARED_SUFFIX}" CACHE FILEPATH "Warp-ctc Library" FORCE)
-ENDIF(WIN32)
 IF(CMAKE_CXX_COMPILER_ID STREQUAL "Clang" OR CMAKE_CXX_COMPILER_ID STREQUAL "AppleClang" )
    SET(USE_OMP OFF)
@@ -43,26 +36,35 @@ ELSE()
 ENDIF()
 ExternalProject_Add(
-    warpctc
+    extern_warpctc
    ${EXTERNAL_PROJECT_LOG_ARGS}
    GIT_REPOSITORY  "https://github.com/gangliao/warp-ctc.git"
+    GIT_TAG         b63a0644654a3e0ed624c85a1767bc8193aead09
    PREFIX          ${WARPCTC_SOURCES_DIR}
    UPDATE_COMMAND  ""
    CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-    CMAKE_ARGS      -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
+                    -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
-    CMAKE_ARGS      -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
+                    -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
-    CMAKE_ARGS      -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
+                    -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
-    CMAKE_ARGS      -DCMAKE_INSTALL_PREFIX=${WARPCTC_INSTALL_DIR}
+                    -DCMAKE_INSTALL_PREFIX=${WARPCTC_INSTALL_DIR}
-    CMAKE_ARGS      -DWITH_GPU=${WITH_GPU}
+                    -DWITH_GPU=${WITH_GPU}
-    CMAKE_ARGS      -DWITH_OMP=${USE_OMP}
+                    -DWITH_OMP=${USE_OMP}
-    CMAKE_ARGS      -DWITH_TORCH=OFF
+                    -DWITH_TORCH=OFF
-    CMAKE_ARGS      -DCMAKE_DISABLE_FIND_PACKAGE_Torch=ON
+                    -DCMAKE_DISABLE_FIND_PACKAGE_Torch=ON
-    CMAKE_ARGS      -DBUILD_SHARED=ON
+                    -DBUILD_SHARED=ON
-    CMAKE_ARGS      -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+                    -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-    CMAKE_ARGS      -DCMAKE_BUILD_TYPE=Release
+                    -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
-    CMAKE_CACHE_ARGS -DCMAKE_BUILD_TYPE:STRING=Release
+                    ${EXTERNAL_OPTIONAL_ARGS}
+    CMAKE_CACHE_ARGS -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
                     -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
                     -DCMAKE_INSTALL_PREFIX:PATH=${WARPCTC_INSTALL_DIR}
 )
+MESSAGE(STATUS "warp-ctc library: ${WARPCTC_LIBRARIES}")
+INCLUDE_DIRECTORIES(${WARPCTC_INCLUDE_DIR})
+ADD_LIBRARY(warpctc STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET warpctc PROPERTY IMPORTED_LOCATION ${WARPCTC_LIBRARIES})
+ADD_DEPENDENCIES(warpctc extern_warpctc)
 LIST(APPEND external_project_dependencies warpctc)
--- a/cmake/external/zlib.cmake
+++ b/cmake/external/zlib.cmake
 # 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.
@@ -34,18 +34,30 @@ ExternalProject_Add(
    GIT_TAG         "v1.2.8"
    PREFIX          ${ZLIB_SOURCES_DIR}
    UPDATE_COMMAND  ""
-    CMAKE_ARGS      -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
    CMAKE_ARGS      -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
-    CMAKE_ARGS      -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
+                    -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
-    CMAKE_ARGS      -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
+                    -DCMAKE_C_FLAGS=${CMAKE_C_FLAGS}
-    CMAKE_ARGS      -DCMAKE_INSTALL_PREFIX=${ZLIB_INSTALL_DIR}
+                    -DCMAKE_CXX_FLAGS=${CMAKE_CXX_FLAGS}
-    CMAKE_ARGS      -DBUILD_SHARED_LIBS=OFF
+                    -DCMAKE_INSTALL_PREFIX=${ZLIB_INSTALL_DIR}
-    CMAKE_ARGS      -DCMAKE_POSITION_INDEPENDENT_CODE=ON
+                    -DBUILD_SHARED_LIBS=OFF
-    CMAKE_ARGS      -DCMAKE_MACOSX_RPATH=ON
+                    -DCMAKE_POSITION_INDEPENDENT_CODE=ON
-    CMAKE_ARGS      -DCMAKE_BUILD_TYPE=Release
+                    -DCMAKE_MACOSX_RPATH=ON
+                    -DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
+                    ${EXTERNAL_OPTIONAL_ARGS}
    CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${ZLIB_INSTALL_DIR}
                     -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-                     -DCMAKE_BUILD_TYPE:STRING=Release
+                     -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
 )
 LIST(APPEND external_project_dependencies zlib)
+ADD_LIBRARY(zlib_target STATIC IMPORTED GLOBAL)
+SET_PROPERTY(TARGET zlib_target PROPERTY IMPORTED_LOCATION ${ZLIB_LIBRARIES})
+IF(WITH_C_API)
+  INSTALL(DIRECTORY ${ZLIB_INCLUDE_DIR} DESTINATION third_party/zlib)
+  IF(ANDROID)
+    INSTALL(FILES ${ZLIB_LIBRARIES} DESTINATION third_party/zlib/lib/${ANDROID_ABI})
+  ELSE()
+    INSTALL(FILES ${ZLIB_LIBRARIES} DESTINATION third_party/zlib/lib)
+  ENDIF()
+ENDIF()
--- a/cmake/flags.cmake
+++ b/cmake/flags.cmake
@@ -109,7 +109,11 @@ set(COMMON_FLAGS
    -Wno-unused-function
    -Wno-error=literal-suffix
    -Wno-error=sign-compare
-    -Wno-error=unused-local-typedefs)
+    -Wno-error=unused-local-typedefs
+    -Wno-error=parentheses-equality # Warnings in pybind11
+    -Wno-error=ignored-attributes  # Warnings in Eigen, gcc 6.3
+    -Wno-error=terminate  # Warning in PADDLE_ENFORCE
+)
 set(GPU_COMMON_FLAGS
    -fPIC
@@ -122,11 +126,14 @@ set(GPU_COMMON_FLAGS
    -Wno-error=literal-suffix
    -Wno-error=unused-local-typedefs
    -Wno-error=unused-function  # Warnings in Numpy Header.
+    -Wno-error=array-bounds # Warnings in Eigen::array
 )
 if (APPLE)
-    # On Mac OS X build fat binaries with x86_64 architectures by default.
+    if(NOT CMAKE_CROSSCOMPILING)
-    set (CMAKE_OSX_ARCHITECTURES "x86_64" CACHE STRING "Build architectures for OSX" FORCE)
+        # On Mac OS X build fat binaries with x86_64 architectures by default.
+        set (CMAKE_OSX_ARCHITECTURES "x86_64" CACHE STRING "Build architectures for OSX" FORCE)
+    endif()
 else()
    set(GPU_COMMON_FLAGS
        -Wall
@@ -144,56 +151,3 @@ endforeach()
 foreach(flag ${GPU_COMMON_FLAGS})
    safe_set_nvflag(${flag})
 endforeach()
-set(CUDA_PROPAGATE_HOST_FLAGS OFF)
-# Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
-# So, don't set these flags here.
-LIST(APPEND CUDA_NVCC_FLAGS -std=c++11)
-LIST(APPEND CUDA_NVCC_FLAGS --use_fast_math)
-if(CMAKE_BUILD_TYPE  STREQUAL "Debug")
-    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_DEBUG})
-elseif(CMAKE_BUILD_TYPE  STREQUAL "Release")
-    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_RELEASE})
-elseif(CMAKE_BUILD_TYPE  STREQUAL "RelWithDebInfo")
-    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_RELWITHDEBINFO})
-elseif(CMAKE_BUILD_TYPE  STREQUAL "MinSizeRel")
-    LIST(APPEND CUDA_NVCC_FLAGS  ${CMAKE_CXX_FLAGS_MINSIZEREL})
-endif()
-function(specify_cuda_arch cuda_version cuda_arch)
-    if(${cuda_version} VERSION_GREATER "8.0")
-        foreach(capability 61 62)
-          if(${cuda_arch} STREQUAL ${capability})
-            list(APPEND __arch_flags " -gencode arch=compute_${cuda_arch},code=sm_${cuda_arch}")
-          endif()
-        endforeach()
-    elseif(${cuda_version} VERSION_GREATER "7.0" and ${cuda_arch} STREQUAL "53")
-        list(APPEND __arch_flags " -gencode arch=compute_${cuda_arch},code=sm_${cuda_arch}")
-    endif()
-endfunction()
-# Common gpu architectures: Kepler, Maxwell
-foreach(capability 30 35 50)
-      list(APPEND __arch_flags " -gencode arch=compute_${capability},code=sm_${capability}")
-endforeach()
-if (CUDA_VERSION VERSION_GREATER "7.0" OR CUDA_VERSION VERSION_EQUAL "7.0")
-      list(APPEND __arch_flags " -gencode arch=compute_52,code=sm_52")
-endif()
-# Modern gpu architectures: Pascal
-if (CUDA_VERSION VERSION_GREATER "8.0" OR CUDA_VERSION VERSION_EQUAL "8.0")
-      list(APPEND __arch_flags " -gencode arch=compute_60,code=sm_60")
-endif()
-# Custom gpu architecture
-set(CUDA_ARCH)
-if(CUDA_ARCH)
-  specify_cuda_arch(${CUDA_VERSION} ${CUDA_ARCH})
-endif()
-set(CUDA_NVCC_FLAGS ${__arch_flags} ${CUDA_NVCC_FLAGS})
--- a/cmake/generic.cmake
+++ b/cmake/generic.cmake
--- a/cmake/make_resource.py
+++ b/cmake/make_resource.py
+import os
+import re
+import sys
+res = sys.argv[1]
+out = sys.argv[2]
+var = re.sub(r'[ .-]', '_', os.path.basename(res))
+open(out, "w").write("const unsigned char " + var + "[] = {" + ",".join([
+    "0x%02x" % ord(c) for c in open(res).read()
+]) + ",0};\n" + "const unsigned " + var + "_size = sizeof(" + var + ");\n")
--- a/cmake/package.cmake
+++ b/cmake/package.cmake
@@ -12,7 +12,7 @@ set(CPACK_PACKAGE_DESCRIPTION "")
 set(CPACK_DEBIAN_PACKAGE_DEPENDS "libpython2.7-dev, libstdc++6, python-pip, curl, libgfortran3, python-pip-whl")
 set(CPACK_DEBIAN_PACKAGE_SECTION Devel)
 set(CPACK_DEBIAN_PACKAGE_VERSION ${PADDLE_VERSION})
-set(CPACK_DEBIAN_PACKAGE_CONTROL_EXTRA "${PROJ_ROOT}/paddle/scripts/deb/postinst")
+set(CPACK_DEBIAN_PACKAGE_CONTROL_EXTRA "${PADDLE_SOURCE_DIR}/paddle/scripts/deb/postinst")
 #set(CPACK_GENERATOR "DEB")
 # Start cpack
 include (CMakePackageConfigHelpers)

--- a/cmake/rdma.cmake
+++ b/cmake/rdma.cmake
@@ -10,7 +10,7 @@ if(WITH_RDMA)
  function(generate_rdma_links)
    #redirect to current DIR to isolate the pollution from system runtime environment
-    #it can benifits unified control for different gcc environment. 
+    #it can benifits unified control for different gcc environment.
    #e.g, by default gcc48 did not refer /usr/lib64 which could contain low version
    #runtime libraries that will crash process while loading it. That redirect trick
    #can fix it.
@@ -19,7 +19,9 @@ if(WITH_RDMA)
      COMMAND ln -s -f /usr/lib64/libibverbs.so.1.0.0 librdma/libibverbs.so.1
      COMMAND ln -s -f /usr/lib64/libibverbs.so.1.0.0 librdma/libibverbs.so
      COMMAND ln -s -f /usr/lib64/librdmacm.so.1.0.0 librdma/librdmacm.so.1
-      COMMAND ln -s -f /usr/lib64/librdmacm.so.1.0.0 librdma/librdmacm.so 
+      COMMAND ln -s -f /usr/lib64/librdmacm.so.1.0.0 librdma/librdmacm.so
+      COMMAND ln -s -f /lib64/libnl.so.1.1.4 librdma/libnl.so.1
+      COMMAND ln -s -f /lib64/libnl.so.1.1.4 librdma/libnl.so
      WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
    )
  endfunction(generate_rdma_links)
@@ -44,7 +46,7 @@ if(WITH_RDMA)
      RDMA_INC_XIO AND
      RDMA_INC_EVENT AND
      RDMA_INC_NUMA AND
-      RDMA_LIB_SXISOCK AND 
+      RDMA_LIB_SXISOCK AND
      RDMA_LIB_XIO AND
      RDMA_LIB_EVENT AND
      RDMA_LIB_EVENT_CORE AND
@@ -53,19 +55,19 @@ if(WITH_RDMA)
      RDMA_LIB_NUMA
      )
-    set(RDMA_INC_DIR 
+    set(RDMA_INC_DIR
-      ${RDMA_INC_SXISOCK} 
+      ${RDMA_INC_SXISOCK}
      ${RDMA_INC_XIO}
      ${RDMA_INC_EVENT}
      ${RDMA_INC_NUMA})
-    set(RDMA_LIBS  
+    set(RDMA_LIBS
-      ${RDMA_LIB_SXISOCK} 
+      ${RDMA_LIB_SXISOCK}
-      ${RDMA_LIB_XIO} 
+      ${RDMA_LIB_XIO}
-      ${RDMA_LIB_EVENT} 
+      ${RDMA_LIB_EVENT}
-      ${RDMA_LIB_EVENT_CORE} 
+      ${RDMA_LIB_EVENT_CORE}
-      ${RDMA_LIB_EVENT_EXTRA} 
+      ${RDMA_LIB_EVENT_EXTRA}
-      ${RDMA_LIB_EVENT_PTHREADS} 
+      ${RDMA_LIB_EVENT_PTHREADS}
-      ${RDMA_LIB_NUMA} 
+      ${RDMA_LIB_NUMA}
      )
    set(RDMA_LD_FLAGS "-L./librdma -libverbs -lrdmacm -Xlinker -rpath ./librdma")
    include_directories("${RDMA_INC_DIR}")

--- a/cmake/simd.cmake
+++ b/cmake/simd.cmake
 # This file is use to check all support level of AVX on your machine
 # so that PaddlePaddle can unleash the vectorization power of muticore.
-INCLUDE(CheckCXXSourceRuns)
+include(CheckCXXSourceRuns)
-INCLUDE(CheckCXXSourceCompiles)
+include(CheckCXXSourceCompiles)
-IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+if(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
    set(MMX_FLAG "-mmmx")
    set(SSE2_FLAG "-msse2")
    set(SSE3_FLAG "-msse3")
-    SET(AVX_FLAG "-mavx")
+    set(AVX_FLAG "-mavx")
-    SET(AVX2_FLAG "-mavx2")
+    set(AVX2_FLAG "-mavx2")
-ELSEIF(MSVC)
+elseif(MSVC)
    set(MMX_FLAG "/arch:MMX")
    set(SSE2_FLAG "/arch:SSE2")
    set(SSE3_FLAG "/arch:SSE3")
    SET(AVX_FLAG "/arch:AVX")
    SET(AVX2_FLAG "/arch:AVX2")
-ENDIF()
+endif()
 set(CMAKE_REQUIRED_FLAGS_RETAINED ${CMAKE_REQUIRED_FLAGS})
 # Check  MMX
 set(CMAKE_REQUIRED_FLAGS ${MMX_FLAG})
+set(MMX_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
 CHECK_CXX_SOURCE_RUNS("
 #include <mmintrin.h>
 int main()
@@ -32,6 +33,7 @@ int main()
 # Check SSE2
 set(CMAKE_REQUIRED_FLAGS ${SSE2_FLAG})
+set(SSE2_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
 CHECK_CXX_SOURCE_RUNS("
 #include <emmintrin.h>
 int main()
@@ -42,6 +44,7 @@ int main()
 # Check SSE3
 set(CMAKE_REQUIRED_FLAGS ${SSE3_FLAG})
+set(SSE3_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
 CHECK_CXX_SOURCE_RUNS("
 #include <pmmintrin.h>
 int main()
@@ -55,6 +58,7 @@ int main()
 # Check AVX
 set(CMAKE_REQUIRED_FLAGS ${AVX_FLAG})
+set(AVX_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
 CHECK_CXX_SOURCE_RUNS("
 #include <immintrin.h>
 int main()
@@ -67,6 +71,7 @@ int main()
 # Check AVX 2
 set(CMAKE_REQUIRED_FLAGS ${AVX2_FLAG})
+set(AVX2_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
 CHECK_CXX_SOURCE_RUNS("
 #include <immintrin.h>
 int main()

--- a/cmake/system.cmake
+++ b/cmake/system.cmake
@@ -13,9 +13,9 @@
 # limitations under the License.
 # Detects the OS and sets appropriate variables.
-# CMAKE_SYSTEM_NAME only give us a coarse-grained name,
+# CMAKE_SYSTEM_NAME only give us a coarse-grained name of the OS CMake is
-# but the name like centos is necessary in some scenes
+# building for, but the host processor name like centos is necessary
-# to distinguish system for customization.
+# in some scenes to distinguish system for customization.
 #
 # for instance, protobuf libs path is <install_dir>/lib64
 # on CentOS, but <install_dir>/lib on other systems.
@@ -24,10 +24,15 @@ IF(WIN32)
    SET(HOST_SYSTEM "win32")
 ELSE(WIN32)
    IF(APPLE)
-        EXEC_PROGRAM (sw_vers ARGS -productVersion OUTPUT_VARIABLE MACOSX_VERSION)
-        STRING(REGEX MATCH "[0-9]+.[0-9]+" VERSION "${MACOSX_VERSION}")
-        SET(MACOS_VERSION ${VERSION})
        SET(HOST_SYSTEM "macosx")
+        EXEC_PROGRAM(sw_vers ARGS -productVersion OUTPUT_VARIABLE HOST_SYSTEM_VERSION)
+        STRING(REGEX MATCH "[0-9]+.[0-9]+" MACOS_VERSION "${HOST_SYSTEM_VERSION}")
+        IF(NOT DEFINED $ENV{MACOSX_DEPLOYMENT_TARGET})
+            # Set cache variable - end user may change this during ccmake or cmake-gui configure.
+            SET(CMAKE_OSX_DEPLOYMENT_TARGET ${MACOS_VERSION} CACHE STRING
+                "Minimum OS X version to target for deployment (at runtime); newer APIs weak linked. Set to empty string for default value.")
+        ENDIF()
+        set(CMAKE_EXE_LINKER_FLAGS "-framework CoreFoundation -framework Security")
    ELSE(APPLE)
        IF(EXISTS "/etc/issue")
@@ -43,6 +48,8 @@ ELSE(WIN32)
            ELSEIF(LINUX_ISSUE MATCHES "Fedora")
                SET(HOST_SYSTEM "fedora")
            ENDIF()
+            STRING(REGEX MATCH "(([0-9]+)\\.)+([0-9]+)" HOST_SYSTEM_VERSION "${LINUX_ISSUE}")
        ENDIF(EXISTS "/etc/issue")
        IF(EXISTS "/etc/redhat-release")
@@ -64,12 +71,21 @@ CMAKE_HOST_SYSTEM_INFORMATION(RESULT CPU_CORES QUERY NUMBER_OF_LOGICAL_CORES)
 MARK_AS_ADVANCED(HOST_SYSTEM CPU_CORES)
-MESSAGE(STATUS "Found Paddle host system: ${HOST_SYSTEM}")
+MESSAGE(STATUS "Found Paddle host system: ${HOST_SYSTEM}, version: ${HOST_SYSTEM_VERSION}")
 MESSAGE(STATUS "Found Paddle host system's CPU: ${CPU_CORES} cores")
+# configuration for cross-compiling
 IF(DEFINED CMAKE_SYSTEM_NAME)
+    INCLUDE(cross_compiling/host)
    IF(${CMAKE_SYSTEM_NAME} STREQUAL "Android")
        SET(ANDROID TRUE)
+        INCLUDE(cross_compiling/android)
+    ELSEIF(${CMAKE_SYSTEM_NAME} STREQUAL "RPi")
+        SET(RPI TRUE)
+        INCLUDE(cross_compiling/raspberry_pi)
+    ELSEIF(${CMAKE_SYSTEM_NAME} STREQUAL "iOS")
+        SET(IOS TRUE)
+        INCLUDE(cross_compiling/ios)
    ENDIF()
 ENDIF()

--- a/cmake/util.cmake
+++ b/cmake/util.cmake
@@ -25,7 +25,9 @@ function(target_circle_link_libraries TARGET_NAME)
            endif()
        endforeach()
        if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang" OR "${CMAKE_CXX_COMPILER_ID}" STREQUAL "AppleClang")
-            list(APPEND LIBS "-undefined dynamic_lookup")
+            if(NOT IOS_ENABLE_BITCODE)
+                list(APPEND LIBS "-undefined dynamic_lookup")
+            endif()
        endif()
        list(REVERSE libsInArgn)
        target_link_libraries(${TARGET_NAME}
@@ -71,29 +73,52 @@ function(link_paddle_exe TARGET_NAME)
        generate_rdma_links()
    endif()
-    target_circle_link_libraries(${TARGET_NAME}
+    if(MOBILE_INFERENCE)
-        ARCHIVE_START
+        target_circle_link_libraries(${TARGET_NAME}
-        paddle_gserver
+            ARCHIVE_START
-        paddle_function
+            paddle_gserver
-        ARCHIVE_END
+            paddle_function
-        paddle_pserver
+            ARCHIVE_END
-        paddle_trainer_lib
+            paddle_math
-        paddle_network
+            paddle_utils
-        paddle_math
+            paddle_parameter
-        paddle_utils
+            paddle_proto
-        paddle_parameter
+            paddle_cuda
-        paddle_proto
+            ${EXTERNAL_LIBS}
-        paddle_cuda
+            ${CMAKE_THREAD_LIBS_INIT}
-        ${EXTERNAL_LIBS}
+            ${CMAKE_DL_LIBS}
-        ${CMAKE_THREAD_LIBS_INIT}
+            ${RDMA_LD_FLAGS}
-        ${CMAKE_DL_LIBS}
+            ${RDMA_LIBS})
-        ${RDMA_LD_FLAGS}
+    else()
-        ${RDMA_LIBS})
+        target_circle_link_libraries(${TARGET_NAME}
+            ARCHIVE_START
+            paddle_gserver
+            paddle_function
+            ARCHIVE_END
+            paddle_pserver
+            paddle_trainer_lib
+            paddle_network
+            paddle_math
+            paddle_utils
+            paddle_parameter
+            paddle_proto
+            paddle_cuda
+            paddle_optimizer
+            ${EXTERNAL_LIBS}
+            ${CMAKE_THREAD_LIBS_INIT}
+            ${CMAKE_DL_LIBS}
+            ${RDMA_LD_FLAGS}
+            ${RDMA_LIBS})
+    endif()
    if(ANDROID)
        target_link_libraries(${TARGET_NAME} log)
    endif(ANDROID)
+    if(WITH_MKLML AND MKLML_LIB_DIR AND MKLML_IOMP_LIB)
+      target_link_libraries(${TARGET_NAME} "-L${MKLML_LIB_DIR} -liomp5 -Wl,--as-needed")
+    endif()
    add_dependencies(${TARGET_NAME} ${external_project_dependencies})
 endfunction()
@@ -117,7 +142,6 @@ endfunction()
 macro(add_unittest_without_exec TARGET_NAME)
    add_executable(${TARGET_NAME} ${ARGN})
    link_paddle_test(${TARGET_NAME})
-    add_style_check_target(${TARGET_NAME} ${ARGN})
 endmacro()
 # add_unittest
@@ -138,17 +162,9 @@ macro(add_simple_unittest TARGET_NAME)
 endmacro()
 # Creates C resources file from files in given resource file
-function(create_resources res_file output)
+function(create_resources res_file output_file)
-    # Create empty output file
+  add_custom_command(
-    file(WRITE ${output} "")
+    OUTPUT ${output_file}
-    # Get short filename
+    COMMAND python ARGS ${PADDLE_SOURCE_DIR}/cmake/make_resource.py ${res_file} ${output_file}
-    string(REGEX MATCH "([^/]+)$" filename ${res_file})
+    DEPENDS ${res_file} ${PADDLE_SOURCE_DIR}/cmake/make_resource.py)
-    # Replace filename spaces & extension separator for C compatibility
-    string(REGEX REPLACE "\\.| |-" "_" filename ${filename})
-    # Read hex data from file
-    file(READ ${res_file} filedata HEX)
-    # Convert hex data for C compatibility
-    string(REGEX REPLACE "([0-9a-f][0-9a-f])" "0x\\1," filedata ${filedata})
-    # Append data to output file
-    file(APPEND ${output} "const unsigned char ${filename}[] = {${filedata}0};\nconst unsigned ${filename}_size = sizeof(${filename});\n")
 endfunction()
--- a/cmake/version.cmake
+++ b/cmake/version.cmake
@@ -4,7 +4,7 @@ set(tmp_version "HEAD")
 while ("${PADDLE_VERSION}" STREQUAL "")
  execute_process(
    COMMAND ${GIT_EXECUTABLE} describe --tags --abbrev=0 ${tmp_version}
-    WORKING_DIRECTORY ${PROJ_ROOT}
+    WORKING_DIRECTORY ${PADDLE_SOURCE_DIR}
    OUTPUT_VARIABLE GIT_TAG_NAME
    RESULT_VARIABLE GIT_RESULT
    ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)

--- a/demo/image_classification/.gitignore
+++ b/demo/image_classification/.gitignore
-data/cifar-10-batches-py
-data/cifar-out
-cifar_vgg_model/*
-plot.png
-train.log
-image_provider_copy_1.py
-*pyc
-train.list
-test.list
--- a/demo/image_classification/api_v2_resnet.py
+++ b/demo/image_classification/api_v2_resnet.py
-# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-import paddle.v2 as paddle
-__all__ = ['resnet_cifar10']
-def conv_bn_layer(input,
-                  ch_out,
-                  filter_size,
-                  stride,
-                  padding,
-                  active_type=paddle.activation.Relu(),
-                  ch_in=None):
-    tmp = paddle.layer.img_conv(
-        input=input,
-        filter_size=filter_size,
-        num_channels=ch_in,
-        num_filters=ch_out,
-        stride=stride,
-        padding=padding,
-        act=paddle.activation.Linear(),
-        bias_attr=False)
-    return paddle.layer.batch_norm(input=tmp, act=active_type)
-def shortcut(ipt, n_in, n_out, stride):
-    if n_in != n_out:
-        return conv_bn_layer(ipt, n_out, 1, stride, 0,
-                             paddle.activation.Linear())
-    else:
-        return ipt
-def basicblock(ipt, ch_out, stride):
-    ch_in = ch_out * 2
-    tmp = conv_bn_layer(ipt, ch_out, 3, stride, 1)
-    tmp = conv_bn_layer(tmp, ch_out, 3, 1, 1, paddle.activation.Linear())
-    short = shortcut(ipt, ch_in, ch_out, stride)
-    return paddle.layer.addto(input=[tmp, short], act=paddle.activation.Relu())
-def layer_warp(block_func, ipt, features, count, stride):
-    tmp = block_func(ipt, features, stride)
-    for i in range(1, count):
-        tmp = block_func(tmp, features, 1)
-    return tmp
-def resnet_cifar10(ipt, depth=32):
-    # depth should be one of 20, 32, 44, 56, 110, 1202
-    assert (depth - 2) % 6 == 0
-    n = (depth - 2) / 6
-    nStages = {16, 64, 128}
-    conv1 = conv_bn_layer(
-        ipt, ch_in=3, ch_out=16, filter_size=3, stride=1, padding=1)
-    res1 = layer_warp(basicblock, conv1, 16, n, 1)
-    res2 = layer_warp(basicblock, res1, 32, n, 2)
-    res3 = layer_warp(basicblock, res2, 64, n, 2)
-    pool = paddle.layer.img_pool(
-        input=res3, pool_size=8, stride=1, pool_type=paddle.pooling.Avg())
-    return pool
--- a/demo/image_classification/api_v2_train.py
+++ b/demo/image_classification/api_v2_train.py
--- a/demo/image_classification/api_v2_vgg.py
+++ b/demo/image_classification/api_v2_vgg.py
--- a/demo/image_classification/data/download_cifar.sh
+++ b/demo/image_classification/data/download_cifar.sh
--- a/demo/image_classification/data/process_cifar.py
+++ b/demo/image_classification/data/process_cifar.py
--- a/demo/image_classification/image_provider.py
+++ b/demo/image_classification/image_provider.py
--- a/demo/image_classification/image_util.py
+++ b/demo/image_classification/image_util.py
--- a/demo/image_classification/predict.sh
+++ b/demo/image_classification/predict.sh
--- a/demo/image_classification/prediction.py
+++ b/demo/image_classification/prediction.py
--- a/demo/image_classification/preprocess.py
+++ b/demo/image_classification/preprocess.py
--- a/demo/image_classification/preprocess.sh
+++ b/demo/image_classification/preprocess.sh
--- a/demo/image_classification/train.sh
+++ b/demo/image_classification/train.sh
--- a/demo/image_classification/vgg_16_cifar.py
+++ b/demo/image_classification/vgg_16_cifar.py
--- a/demo/introduction/.gitignore
+++ b/demo/introduction/.gitignore
-dataprovider.pyc
-empty.list
-train.log
-output
-train.list
--- a/demo/introduction/README.md
+++ b/demo/introduction/README.md
-This folder contains scripts used in PaddlePaddle introduction.
- use `bash train.sh` to train a simple linear regression model
- use `python evaluate_model.py` to read model parameters. You can see that `w` and `b` are very close to [2, 0.3].
--- a/demo/introduction/api_train_v2.py
+++ b/demo/introduction/api_train_v2.py
--- a/demo/introduction/dataprovider.py
+++ b/demo/introduction/dataprovider.py
--- a/demo/introduction/evaluate_model.py
+++ b/demo/introduction/evaluate_model.py
--- a/demo/introduction/train.sh
+++ b/demo/introduction/train.sh
--- a/demo/introduction/trainer_config.py
+++ b/demo/introduction/trainer_config.py
--- a/demo/mnist/api_train_v2.py
+++ b/demo/mnist/api_train_v2.py
--- a/demo/recommendation/.gitignore
+++ b/demo/recommendation/.gitignore
--- a/demo/recommendation/api_train_v2.py
+++ b/demo/recommendation/api_train_v2.py
--- a/demo/recommendation/common_utils.py
+++ b/demo/recommendation/common_utils.py
--- a/demo/recommendation/data/config.json
+++ b/demo/recommendation/data/config.json
--- a/demo/recommendation/data/config_generator.py
+++ b/demo/recommendation/data/config_generator.py
--- a/demo/recommendation/data/meta_config.json
+++ b/demo/recommendation/data/meta_config.json
--- a/demo/recommendation/data/meta_generator.py
+++ b/demo/recommendation/data/meta_generator.py
--- a/demo/recommendation/data/ml_data.sh
+++ b/demo/recommendation/data/ml_data.sh
--- a/demo/recommendation/data/split.py
+++ b/demo/recommendation/data/split.py
--- a/demo/recommendation/dataprovider.py
+++ b/demo/recommendation/dataprovider.py
--- a/demo/recommendation/evaluate.py
+++ b/demo/recommendation/evaluate.py
--- a/demo/recommendation/evaluate.sh
+++ b/demo/recommendation/evaluate.sh
--- a/demo/recommendation/prediction.py
+++ b/demo/recommendation/prediction.py
--- a/demo/recommendation/preprocess.sh
+++ b/demo/recommendation/preprocess.sh
--- a/demo/recommendation/requirements.txt
+++ b/demo/recommendation/requirements.txt
--- a/demo/recommendation/run.sh
+++ b/demo/recommendation/run.sh
--- a/demo/recommendation/trainer_config.py
+++ b/demo/recommendation/trainer_config.py
--- a/demo/semantic_role_labeling/.gitignore
+++ b/demo/semantic_role_labeling/.gitignore
--- a/demo/semantic_role_labeling/api_train_v2.py
+++ b/demo/semantic_role_labeling/api_train_v2.py
--- a/demo/semantic_role_labeling/data/extract_dict_feature.py
+++ b/demo/semantic_role_labeling/data/extract_dict_feature.py
--- a/demo/semantic_role_labeling/data/extract_pairs.py
+++ b/demo/semantic_role_labeling/data/extract_pairs.py
--- a/demo/semantic_role_labeling/data/get_data.sh
+++ b/demo/semantic_role_labeling/data/get_data.sh
--- a/demo/semantic_role_labeling/data/test.list
+++ b/demo/semantic_role_labeling/data/test.list
--- a/demo/semantic_role_labeling/data/train.list
+++ b/demo/semantic_role_labeling/data/train.list
--- a/demo/semantic_role_labeling/dataprovider.py
+++ b/demo/semantic_role_labeling/dataprovider.py
--- a/demo/semantic_role_labeling/db_lstm.py
+++ b/demo/semantic_role_labeling/db_lstm.py
--- a/demo/semantic_role_labeling/predict.py
+++ b/demo/semantic_role_labeling/predict.py
--- a/demo/semantic_role_labeling/predict.sh
+++ b/demo/semantic_role_labeling/predict.sh
--- a/demo/semantic_role_labeling/test.sh
+++ b/demo/semantic_role_labeling/test.sh
--- a/demo/semantic_role_labeling/train.sh
+++ b/demo/semantic_role_labeling/train.sh
--- a/demo/sentiment/.gitignore
+++ b/demo/sentiment/.gitignore
--- a/demo/sentiment/data/get_imdb.sh
+++ b/demo/sentiment/data/get_imdb.sh
--- a/demo/sentiment/dataprovider.py
+++ b/demo/sentiment/dataprovider.py
--- a/demo/sentiment/predict.py
+++ b/demo/sentiment/predict.py
--- a/demo/sentiment/predict.sh
+++ b/demo/sentiment/predict.sh
--- a/demo/sentiment/preprocess.py
+++ b/demo/sentiment/preprocess.py
--- a/demo/sentiment/preprocess.sh
+++ b/demo/sentiment/preprocess.sh
--- a/demo/sentiment/sentiment_net.py
+++ b/demo/sentiment/sentiment_net.py
--- a/demo/sentiment/test.sh
+++ b/demo/sentiment/test.sh
--- a/demo/sentiment/train.sh
+++ b/demo/sentiment/train.sh
--- a/demo/sentiment/train_v2.py
+++ b/demo/sentiment/train_v2.py
--- a/demo/sentiment/trainer_config.py
+++ b/demo/sentiment/trainer_config.py
--- a/demo/seqToseq/.gitignore
+++ b/demo/seqToseq/.gitignore
--- a/demo/seqToseq/api_train_v2.py
+++ b/demo/seqToseq/api_train_v2.py
--- a/demo/seqToseq/data/paraphrase_data.sh
+++ b/demo/seqToseq/data/paraphrase_data.sh
--- a/demo/seqToseq/data/paraphrase_model.sh
+++ b/demo/seqToseq/data/paraphrase_model.sh
--- a/demo/seqToseq/data/wmt14_data.sh
+++ b/demo/seqToseq/data/wmt14_data.sh
--- a/demo/seqToseq/data/wmt14_model.sh
+++ b/demo/seqToseq/data/wmt14_model.sh
--- a/demo/seqToseq/dataprovider.py
+++ b/demo/seqToseq/dataprovider.py
--- a/demo/seqToseq/paraphrase/train.conf
+++ b/demo/seqToseq/paraphrase/train.conf
--- a/demo/seqToseq/paraphrase/train.sh
+++ b/demo/seqToseq/paraphrase/train.sh
--- a/demo/seqToseq/preprocess.py
+++ b/demo/seqToseq/preprocess.py
--- a/demo/seqToseq/seqToseq_net.py
+++ b/demo/seqToseq/seqToseq_net.py
--- a/demo/seqToseq/translation/eval_bleu.sh
+++ b/demo/seqToseq/translation/eval_bleu.sh
--- a/demo/seqToseq/translation/gen.conf
+++ b/demo/seqToseq/translation/gen.conf
--- a/demo/seqToseq/translation/gen.sh
+++ b/demo/seqToseq/translation/gen.sh
--- a/demo/seqToseq/translation/moses_bleu.sh
+++ b/demo/seqToseq/translation/moses_bleu.sh
--- a/demo/seqToseq/translation/train.conf
+++ b/demo/seqToseq/translation/train.conf
--- a/demo/seqToseq/translation/train.sh
+++ b/demo/seqToseq/translation/train.sh
--- a/demo/word2vec/train_v2.py
+++ b/demo/word2vec/train_v2.py
--- a/doc/CMakeLists.txt
+++ b/doc/CMakeLists.txt
--- a/doc/about/index_cn.md
+++ b/doc/about/index_cn.md
--- a/doc/about/index_en.rst
+++ b/doc/about/index_en.rst
--- a/doc/api/index_en.rst
+++ b/doc/api/index_en.rst
--- a/doc/api/v1/index_cn.rst
+++ b/doc/api/v1/index_cn.rst
--- a/doc/api/v1/trainer_config_helpers/activations.rst
+++ b/doc/api/v1/trainer_config_helpers/activations.rst
--- a/doc/api/v1/trainer_config_helpers/attrs.rst
+++ b/doc/api/v1/trainer_config_helpers/attrs.rst
--- a/doc/api/v1/trainer_config_helpers/data_sources.rst
+++ b/doc/api/v1/trainer_config_helpers/data_sources.rst
--- a/doc/api/v1/trainer_config_helpers/evaluators.rst
+++ b/doc/api/v1/trainer_config_helpers/evaluators.rst
--- a/doc/api/v1/trainer_config_helpers/layers.rst
+++ b/doc/api/v1/trainer_config_helpers/layers.rst
--- a/doc/api/v1/trainer_config_helpers/networks.rst
+++ b/doc/api/v1/trainer_config_helpers/networks.rst
--- a/doc/api/v1/trainer_config_helpers/optimizers.rst
+++ b/doc/api/v1/trainer_config_helpers/optimizers.rst
--- a/doc/api/v1/trainer_config_helpers/poolings.rst
+++ b/doc/api/v1/trainer_config_helpers/poolings.rst
--- a/doc/api/v2/config/activation.rst
+++ b/doc/api/v2/config/activation.rst
--- a/doc/api/v2/config/evaluators.rst
+++ b/doc/api/v2/config/evaluators.rst
--- a/doc/api/v2/config/layer.rst
+++ b/doc/api/v2/config/layer.rst
--- a/doc/api/v2/config/networks.rst
+++ b/doc/api/v2/config/networks.rst
--- a/doc/api/v2/data.rst
+++ b/doc/api/v2/data.rst
--- a/doc/api/v2/data/data_reader.rst
+++ b/doc/api/v2/data/data_reader.rst
--- a/doc/api/v2/data/dataset.rst
+++ b/doc/api/v2/data/dataset.rst
--- a/doc/api/v2/data/image.rst
+++ b/doc/api/v2/data/image.rst
--- a/doc/api/v2/fluid.rst
+++ b/doc/api/v2/fluid.rst
--- a/doc/api/v2/fluid/data_feeder.rst
+++ b/doc/api/v2/fluid/data_feeder.rst
--- a/doc/api/v2/fluid/evaluator.rst
+++ b/doc/api/v2/fluid/evaluator.rst
--- a/doc/api/v2/fluid/executor.rst
+++ b/doc/api/v2/fluid/executor.rst
--- a/doc/api/v2/fluid/initializer.rst
+++ b/doc/api/v2/fluid/initializer.rst
--- a/doc/api/v2/fluid/layers.rst
+++ b/doc/api/v2/fluid/layers.rst
--- a/doc/api/v2/fluid/nets.rst
+++ b/doc/api/v2/fluid/nets.rst
--- a/doc/api/v2/fluid/optimizer.rst
+++ b/doc/api/v2/fluid/optimizer.rst
--- a/doc/api/v2/fluid/param_attr.rst
+++ b/doc/api/v2/fluid/param_attr.rst
--- a/doc/api/v2/fluid/profiler.rst
+++ b/doc/api/v2/fluid/profiler.rst
--- a/doc/api/v2/fluid/regularizer.rst
+++ b/doc/api/v2/fluid/regularizer.rst
--- a/doc/api/v2/model_configs.rst
+++ b/doc/api/v2/model_configs.rst
--- a/doc/design/api.md
+++ b/doc/design/api.md
--- a/doc/design/auto_gradient_check.md
+++ b/doc/design/auto_gradient_check.md
--- a/doc/design/block.md
+++ b/doc/design/block.md
--- a/doc/design/build_system/README.md
+++ b/doc/design/build_system/README.md
--- a/doc/design/cluster_train/README.md
+++ b/doc/design/cluster_train/README.md
--- a/doc/design/cluster_train/checkpointing.md
+++ b/doc/design/cluster_train/checkpointing.md
--- a/doc/design/cluster_train/data_dispatch.md
+++ b/doc/design/cluster_train/data_dispatch.md
--- a/doc/design/cluster_train/large_model_dist_train.md
+++ b/doc/design/cluster_train/large_model_dist_train.md
--- a/doc/design/cluster_train/master_server.md
+++ b/doc/design/cluster_train/master_server.md
--- a/doc/design/cluster_train/pserver_client.md
+++ b/doc/design/cluster_train/pserver_client.md
--- a/doc/design/cluster_train/remote_parameter_updater.md
+++ b/doc/design/cluster_train/remote_parameter_updater.md
--- a/doc/design/cluster_train/save_model.md
+++ b/doc/design/cluster_train/save_model.md
--- a/doc/design/cluster_train/src/checkpointing.png
+++ b/doc/design/cluster_train/src/checkpointing.png
--- a/doc/design/cluster_train/src/data_dispatch.png
+++ b/doc/design/cluster_train/src/data_dispatch.png
--- a/doc/design/cluster_train/src/dataset.graffle
+++ b/doc/design/cluster_train/src/dataset.graffle
--- a/doc/design/cluster_train/src/dataset.png
+++ b/doc/design/cluster_train/src/dataset.png
--- a/doc/design/cluster_train/src/file_storage.graffle
+++ b/doc/design/cluster_train/src/file_storage.graffle
--- a/doc/design/cluster_train/src/file_storage.png
+++ b/doc/design/cluster_train/src/file_storage.png
--- a/doc/design/cluster_train/src/init_lock.graffle
+++ b/doc/design/cluster_train/src/init_lock.graffle
--- a/doc/design/cluster_train/src/init_lock.png
+++ b/doc/design/cluster_train/src/init_lock.png
--- a/doc/design/cluster_train/src/paddle-cloud-in-data-center.png
+++ b/doc/design/cluster_train/src/paddle-cloud-in-data-center.png
--- a/doc/design/cluster_train/src/paddle-etcd.graffle
+++ b/doc/design/cluster_train/src/paddle-etcd.graffle
--- a/doc/design/cluster_train/src/paddle-etcd.png
+++ b/doc/design/cluster_train/src/paddle-etcd.png
--- a/doc/design/dist/src/paddle-model-sharding.graffle
+++ b/doc/design/dist/src/paddle-model-sharding.graffle
--- a/doc/design/dist/src/paddle-model-sharding.png
+++ b/doc/design/dist/src/paddle-model-sharding.png
--- a/doc/design/dist/src/paddle-ps-0.png
+++ b/doc/design/dist/src/paddle-ps-0.png
--- a/doc/design/dist/src/paddle-ps-1.png
+++ b/doc/design/dist/src/paddle-ps-1.png
--- a/doc/design/dist/src/paddle-ps.graffle
+++ b/doc/design/dist/src/paddle-ps.graffle
--- a/doc/design/dist/src/paddle-task-queues.graffle
+++ b/doc/design/dist/src/paddle-task-queues.graffle
--- a/doc/design/dist/src/paddle-task-queues.png
+++ b/doc/design/dist/src/paddle-task-queues.png
--- a/doc/design/dist/src/paddle-task-states.graffle
+++ b/doc/design/dist/src/paddle-task-states.graffle
--- a/doc/design/dist/src/paddle-task-states.png
+++ b/doc/design/dist/src/paddle-task-states.png
--- a/doc/design/cluster_train/src/pserver_init.graffle
+++ b/doc/design/cluster_train/src/pserver_init.graffle
--- a/doc/design/cluster_train/src/pserver_init.png
+++ b/doc/design/cluster_train/src/pserver_init.png
--- a/doc/design/cluster_train/src/submit-job.graffle
+++ b/doc/design/cluster_train/src/submit-job.graffle
--- a/doc/design/cluster_train/src/submit-job.png
+++ b/doc/design/cluster_train/src/submit-job.png
--- a/doc/design/cluster_train/src/trainer.graffle
+++ b/doc/design/cluster_train/src/trainer.graffle
--- a/doc/design/cluster_train/src/trainer.png
+++ b/doc/design/cluster_train/src/trainer.png
--- a/doc/design/cluster_train/submit-job.md
+++ b/doc/design/cluster_train/submit-job.md
--- a/doc/design/dcgan.png
+++ b/doc/design/dcgan.png
--- a/doc/design/dist/README.md
+++ b/doc/design/dist/README.md
--- a/doc/design/dist/src/paddle-etcd.graffle
+++ b/doc/design/dist/src/paddle-etcd.graffle
--- a/doc/design/dist/src/paddle-etcd.png
+++ b/doc/design/dist/src/paddle-etcd.png
--- a/doc/design/evaluator.md
+++ b/doc/design/evaluator.md
--- a/doc/design/executor.md
+++ b/doc/design/executor.md
--- a/doc/design/file_manager/README.md
+++ b/doc/design/file_manager/README.md
--- a/doc/design/file_manager/pfs/pfsclient.md
+++ b/doc/design/file_manager/pfs/pfsclient.md
--- a/doc/design/file_manager/src/filemanager.graffle
+++ b/doc/design/file_manager/src/filemanager.graffle
--- a/doc/design/file_manager/src/filemanager.png
+++ b/doc/design/file_manager/src/filemanager.png
--- a/doc/design/float16.md
+++ b/doc/design/float16.md
--- a/doc/design/functions_operators_layers.md
+++ b/doc/design/functions_operators_layers.md
--- a/doc/design/gan_api.md
+++ b/doc/design/gan_api.md
--- a/doc/design/graph.md
+++ b/doc/design/graph.md
--- a/doc/design/graph_survey.md
+++ b/doc/design/graph_survey.md
--- a/doc/design/if_else_op.md
+++ b/doc/design/if_else_op.md
--- a/doc/design/images/asgd.gif
+++ b/doc/design/images/asgd.gif
--- a/doc/design/images/feed_forward.png
+++ b/doc/design/images/feed_forward.png
--- a/doc/design/images/feed_forward_regularized.png
+++ b/doc/design/images/feed_forward_regularized.png
--- a/doc/design/images/graph_construction_example.bash
+++ b/doc/design/images/graph_construction_example.bash
--- a/doc/design/images/graph_construction_example.dot
+++ b/doc/design/images/graph_construction_example.dot
--- a/doc/design/images/graph_construction_example_all.png
+++ b/doc/design/images/graph_construction_example_all.png
--- a/doc/design/images/graph_construction_example_forward_backward.png
+++ b/doc/design/images/graph_construction_example_forward_backward.png
--- a/doc/design/images/graph_construction_example_forward_only.png
+++ b/doc/design/images/graph_construction_example_forward_only.png
--- a/doc/design/images/l1_regularization.png
+++ b/doc/design/images/l1_regularization.png
--- a/doc/design/images/l2_regularization.png
+++ b/doc/design/images/l2_regularization.png
--- a/doc/design/images/loss_equation.png
+++ b/doc/design/images/loss_equation.png
--- a/doc/design/images/replica.png
+++ b/doc/design/images/replica.png
--- a/doc/design/images/theta_star.gif
+++ b/doc/design/images/theta_star.gif
--- a/doc/design/images/two_phase_commit.png
+++ b/doc/design/images/two_phase_commit.png
--- a/doc/design/infer_var_type.md
+++ b/doc/design/infer_var_type.md
--- a/doc/design/mkldnn/README.MD
+++ b/doc/design/mkldnn/README.MD
--- a/doc/design/mkldnn/image/engine.png
+++ b/doc/design/mkldnn/image/engine.png
--- a/doc/design/mkldnn/image/gradients.png
+++ b/doc/design/mkldnn/image/gradients.png
--- a/doc/design/mkldnn/image/layers.png
+++ b/doc/design/mkldnn/image/layers.png
--- a/doc/design/mkldnn/image/matrix.png
+++ b/doc/design/mkldnn/image/matrix.png
--- a/doc/design/mkldnn/image/overview.png
+++ b/doc/design/mkldnn/image/overview.png
--- a/doc/design/model_format.md
+++ b/doc/design/model_format.md
--- a/doc/design/multi_language_interface/00.why_plain_c.md
+++ b/doc/design/multi_language_interface/00.why_plain_c.md
--- a/doc/design/multi_language_interface/01.inference_implementation.md
+++ b/doc/design/multi_language_interface/01.inference_implementation.md
--- a/doc/design/multi_language_interface/why_plain_c.md
+++ b/doc/design/multi_language_interface/why_plain_c.md
--- a/doc/design/ops/images/2_level_rnn.dot
+++ b/doc/design/ops/images/2_level_rnn.dot
--- a/doc/design/ops/images/2_level_rnn.png
+++ b/doc/design/ops/images/2_level_rnn.png
--- a/doc/design/ops/images/LOD-and-shape-changes-during-decoding.jpg
+++ b/doc/design/ops/images/LOD-and-shape-changes-during-decoding.jpg
--- a/doc/design/ops/images/rnn.dot
+++ b/doc/design/ops/images/rnn.dot
--- a/doc/design/ops/images/rnn.jpg
+++ b/doc/design/ops/images/rnn.jpg
--- a/doc/design/ops/images/rnn.png
+++ b/doc/design/ops/images/rnn.png
--- a/doc/design/ops/images/rnn_2level_data.dot
+++ b/doc/design/ops/images/rnn_2level_data.dot
--- a/doc/design/ops/images/rnn_2level_data.png
+++ b/doc/design/ops/images/rnn_2level_data.png
--- a/doc/design/ops/rnn.md
+++ b/doc/design/ops/rnn.md
--- a/doc/design/ops/sequence_decoder.md
+++ b/doc/design/ops/sequence_decoder.md
--- a/doc/design/optimizer.md
+++ b/doc/design/optimizer.md
--- a/doc/design/parameter_average.md
+++ b/doc/design/parameter_average.md
--- a/doc/design/parameters_in_cpp.md
+++ b/doc/design/parameters_in_cpp.md
--- a/doc/design/program.md
+++ b/doc/design/program.md
--- a/doc/design/prune.md
+++ b/doc/design/prune.md
--- a/doc/design/python_api.md
+++ b/doc/design/python_api.md
--- a/doc/design/reader/README.md
+++ b/doc/design/reader/README.md
--- a/doc/design/refactor/distributed_architecture.md
+++ b/doc/design/refactor/distributed_architecture.md
--- a/doc/design/refactor/parameter_server.md
+++ b/doc/design/refactor/parameter_server.md
--- a/doc/design/refactor/session.md
+++ b/doc/design/refactor/session.md
--- a/doc/design/refactor/src/compiler.graffle
+++ b/doc/design/refactor/src/compiler.graffle
--- a/doc/design/refactor/src/compiler.png
+++ b/doc/design/refactor/src/compiler.png
--- a/doc/design/refactor/src/dist-graph.graffle
+++ b/doc/design/refactor/src/dist-graph.graffle
--- a/doc/design/refactor/src/dist-graph.png
+++ b/doc/design/refactor/src/dist-graph.png
--- a/doc/design/refactor/src/distributed_architecture.graffle
+++ b/doc/design/refactor/src/distributed_architecture.graffle
--- a/doc/design/refactor/src/distributed_architecture.png
+++ b/doc/design/refactor/src/distributed_architecture.png
--- a/doc/design/refactor/src/local-graph.graffle
+++ b/doc/design/refactor/src/local-graph.graffle
--- a/doc/design/refactor/src/local-graph.png
+++ b/doc/design/refactor/src/local-graph.png
--- a/doc/design/refactor/src/local_architecture.graffle
+++ b/doc/design/refactor/src/local_architecture.graffle
--- a/doc/design/refactor/src/local_architecture.png
+++ b/doc/design/refactor/src/local_architecture.png
--- a/doc/design/refactor/src/paddle-compile.graffle
+++ b/doc/design/refactor/src/paddle-compile.graffle
--- a/doc/design/refactor/src/paddle-compile.png
+++ b/doc/design/refactor/src/paddle-compile.png
--- a/doc/design/refactorization.md
+++ b/doc/design/refactorization.md
--- a/doc/design/register_grad_op.md
+++ b/doc/design/register_grad_op.md
--- a/doc/design/regularization.md
+++ b/doc/design/regularization.md
--- a/doc/design/releasing_process.md
+++ b/doc/design/releasing_process.md
--- a/doc/design/scope.md
+++ b/doc/design/scope.md
--- a/doc/design/selected_rows.md
+++ b/doc/design/selected_rows.md
--- a/doc/design/simple_op_design.md
+++ b/doc/design/simple_op_design.md
--- a/doc/design/speech/README.MD
+++ b/doc/design/speech/README.MD
--- a/doc/design/speech/image/ds2_network.png
+++ b/doc/design/speech/image/ds2_network.png
--- a/doc/design/tensor_array.md
+++ b/doc/design/tensor_array.md
--- a/doc/design/test.dot
+++ b/doc/design/test.dot
--- a/doc/design/test.dot.png
+++ b/doc/design/test.dot.png
--- a/doc/design/var_desc.md
+++ b/doc/design/var_desc.md
--- a/doc/faq/build_and_install/index_cn.rst
+++ b/doc/faq/build_and_install/index_cn.rst
--- a/doc/faq/cluster/index_cn.rst
+++ b/doc/faq/cluster/index_cn.rst
--- a/doc/faq/index_cn.rst
+++ b/doc/faq/index_cn.rst
--- a/doc/faq/local/index_cn.rst
+++ b/doc/faq/local/index_cn.rst
--- a/doc/faq/src/reduce_min_pool_size.py
+++ b/doc/faq/src/reduce_min_pool_size.py
--- a/doc/faq/src/word2vec_config.py
+++ b/doc/faq/src/word2vec_config.py
--- a/doc/faq/src/word2vec_dataprovider.py
+++ b/doc/faq/src/word2vec_dataprovider.py
--- a/doc/faq/model/index_cn.rst
+++ b/doc/faq/model/index_cn.rst
--- a/doc/faq/parameter/index_cn.rst
+++ b/doc/faq/parameter/index_cn.rst
--- a/doc/getstarted/basic_usage/index_cn.rst
+++ b/doc/getstarted/basic_usage/index_cn.rst
--- a/doc/getstarted/basic_usage/index_en.rst
+++ b/doc/getstarted/basic_usage/index_en.rst
--- a/doc/getstarted/basic_usage/parameters.png
+++ b/doc/getstarted/basic_usage/parameters.png
--- a/doc/getstarted/build_and_install/build_from_source_cn.rst
+++ b/doc/getstarted/build_and_install/build_from_source_cn.rst
--- a/doc/getstarted/build_and_install/build_from_source_en.md
+++ b/doc/getstarted/build_and_install/build_from_source_en.md
--- a/doc/getstarted/build_and_install/build_from_source_en.rst
+++ b/doc/getstarted/build_and_install/build_from_source_en.rst
--- a/doc/getstarted/build_and_install/cmake.png
+++ b/doc/getstarted/build_and_install/cmake.png
--- a/doc/getstarted/build_and_install/cmake/build_from_source_cn.rst
+++ b/doc/getstarted/build_and_install/cmake/build_from_source_cn.rst
--- a/doc/getstarted/build_and_install/cmake/cblas_settings.csv
+++ b/doc/getstarted/build_and_install/cmake/cblas_settings.csv
--- a/doc/getstarted/build_and_install/cmake/compile_options.csv
+++ b/doc/getstarted/build_and_install/cmake/compile_options.csv
--- a/doc/getstarted/build_and_install/docker_install_cn.rst
+++ b/doc/getstarted/build_and_install/docker_install_cn.rst
--- a/doc/getstarted/build_and_install/docker_install_en.rst
+++ b/doc/getstarted/build_and_install/docker_install_en.rst
--- a/doc/getstarted/build_and_install/index_cn.rst
+++ b/doc/getstarted/build_and_install/index_cn.rst
--- a/doc/getstarted/build_and_install/index_en.rst
+++ b/doc/getstarted/build_and_install/index_en.rst
--- a/doc/getstarted/build_and_install/paddleci.png
+++ b/doc/getstarted/build_and_install/paddleci.png
--- a/doc/getstarted/build_and_install/pip_install_cn.rst
+++ b/doc/getstarted/build_and_install/pip_install_cn.rst
--- a/doc/getstarted/build_and_install/pip_install_en.rst
+++ b/doc/getstarted/build_and_install/pip_install_en.rst
--- a/doc/getstarted/build_and_install/ubuntu_install_cn.rst
+++ b/doc/getstarted/build_and_install/ubuntu_install_cn.rst
--- a/doc/getstarted/build_and_install/ubuntu_install_en.rst
+++ b/doc/getstarted/build_and_install/ubuntu_install_en.rst
--- a/doc/getstarted/concepts/src/train.py
+++ b/doc/getstarted/concepts/src/train.py
--- a/doc/getstarted/concepts/use_concepts_cn.rst
+++ b/doc/getstarted/concepts/use_concepts_cn.rst
--- a/doc/getstarted/index_cn.rst
+++ b/doc/getstarted/index_cn.rst
--- a/doc/getstarted/index_en.rst
+++ b/doc/getstarted/index_en.rst
--- a/doc/howto/deep_model/rnn/hierarchical_layer_cn.rst
+++ b/doc/howto/deep_model/rnn/hierarchical_layer_cn.rst
--- a/doc/howto/deep_model/rnn/hrnn_rnn_api_compare_cn.rst
+++ b/doc/howto/deep_model/rnn/hrnn_rnn_api_compare_cn.rst
--- a/doc/howto/deep_model/rnn/index_cn.rst
+++ b/doc/howto/deep_model/rnn/index_cn.rst
--- a/doc/howto/deep_model/rnn/index_en.rst
+++ b/doc/howto/deep_model/rnn/index_en.rst
--- a/doc/howto/deep_model/rnn/rnn_config_cn.rst
+++ b/doc/howto/deep_model/rnn/rnn_config_cn.rst
--- a/doc/howto/deep_model/rnn/rnn_config_en.rst
+++ b/doc/howto/deep_model/rnn/rnn_config_en.rst
--- a/doc/tutorials/sentiment_analysis/bi_lstm.jpg
+++ b/doc/tutorials/sentiment_analysis/bi_lstm.jpg
--- a/doc/tutorials/text_generation/encoder-decoder-attention-model.png
+++ b/doc/tutorials/text_generation/encoder-decoder-attention-model.png
--- a/doc/howto/dev/build_cn.md
+++ b/doc/howto/dev/build_cn.md
--- a/doc/howto/dev/build_en.md
+++ b/doc/howto/dev/build_en.md
--- a/doc/howto/dev/contribute_to_paddle_cn.md
+++ b/doc/howto/dev/contribute_to_paddle_cn.md
--- a/doc/howto/dev/contribute_to_paddle_en.md
+++ b/doc/howto/dev/contribute_to_paddle_en.md
--- a/doc/howto/dev/contribute_to_paddle_en.md
+++ b/doc/howto/dev/contribute_to_paddle_en.md
--- a/doc/howto/dev/new_layer_cn.rst
+++ b/doc/howto/dev/new_layer_cn.rst
--- a/doc/howto/dev/new_layer_en.rst
+++ b/doc/howto/dev/new_layer_en.rst
--- a/doc/howto/dev/new_op_cn.md
+++ b/doc/howto/dev/new_op_cn.md
--- a/doc/howto/dev/new_op_en.md
+++ b/doc/howto/dev/new_op_en.md
--- a/doc/howto/dev/use_eigen_cn.md
+++ b/doc/howto/dev/use_eigen_cn.md
--- a/doc/howto/dev/use_eigen_en.md
+++ b/doc/howto/dev/use_eigen_en.md
--- a/doc/howto/dev/write_docs_cn.rst
+++ b/doc/howto/dev/write_docs_cn.rst
--- a/doc/howto/dev/write_docs_en.rst
+++ b/doc/howto/dev/write_docs_en.rst
--- a/doc/howto/index_cn.rst
+++ b/doc/howto/index_cn.rst
--- a/doc/howto/index_en.rst
+++ b/doc/howto/index_en.rst
--- a/doc/howto/optimization/cpu_profiling.md
+++ b/doc/howto/optimization/cpu_profiling.md
--- a/doc/howto/optimization/cpu_profiling_cn.md
+++ b/doc/howto/optimization/cpu_profiling_cn.md
--- a/doc/howto/optimization/pprof_1.png
+++ b/doc/howto/optimization/pprof_1.png
--- a/doc/howto/optimization/pprof_2.png
+++ b/doc/howto/optimization/pprof_2.png
--- a/doc/howto/usage/cluster/cluster_train_cn.md
+++ b/doc/howto/usage/cluster/cluster_train_cn.md
--- a/doc/howto/usage/cluster/cluster_train_en.md
+++ b/doc/howto/usage/cluster/cluster_train_en.md
--- a/doc/howto/usage/cluster/src/trainer.png
+++ b/doc/howto/usage/cluster/src/trainer.png
--- a/doc/howto/usage/cluster/src/trainer_cn.png
+++ b/doc/howto/usage/cluster/src/trainer_cn.png
--- a/doc/howto/usage/cluster/src/word2vec/api_train_v2.py
+++ b/doc/howto/usage/cluster/src/word2vec/api_train_v2.py
--- a/doc/howto/usage/cluster/src/word2vec/api_train_v2_cluster.py
+++ b/doc/howto/usage/cluster/src/word2vec/api_train_v2_cluster.py
--- a/doc/howto/usage/cluster/src/word2vec/prepare.py
+++ b/doc/howto/usage/cluster/src/word2vec/prepare.py
--- a/doc/howto/usage/cmd_parameter/arguments_cn.md
+++ b/doc/howto/usage/cmd_parameter/arguments_cn.md
--- a/doc/howto/usage/k8s/k8s_distributed_cn.md
+++ b/doc/howto/usage/k8s/k8s_distributed_cn.md
--- a/doc/index_cn.rst
+++ b/doc/index_cn.rst
--- a/doc/index_en.rst
+++ b/doc/index_en.rst
--- a/doc/mobile/cross_compiling_for_android_cn.md
+++ b/doc/mobile/cross_compiling_for_android_cn.md
--- a/doc/mobile/cross_compiling_for_android_en.md
+++ b/doc/mobile/cross_compiling_for_android_en.md
--- a/doc/mobile/cross_compiling_for_ios_cn.md
+++ b/doc/mobile/cross_compiling_for_ios_cn.md
--- a/doc/mobile/cross_compiling_for_raspberry_cn.md
+++ b/doc/mobile/cross_compiling_for_raspberry_cn.md
--- a/doc/mobile/cross_compiling_for_raspberry_en.md
+++ b/doc/mobile/cross_compiling_for_raspberry_en.md
--- a/doc/mobile/index_cn.rst
+++ b/doc/mobile/index_cn.rst
--- a/doc/mobile/index_en.rst
+++ b/doc/mobile/index_en.rst
--- a/doc/survey/cluster_bootstrapping_tools.md
+++ b/doc/survey/cluster_bootstrapping_tools.md
--- a/doc/templates/conf.py.cn.in
+++ b/doc/templates/conf.py.cn.in
--- a/doc/templates/conf.py.en.in
+++ b/doc/templates/conf.py.en.in
--- a/doc/tutorials/image_classification/cifar.png
+++ b/doc/tutorials/image_classification/cifar.png
--- a/doc/tutorials/image_classification/image_classification.png
+++ b/doc/tutorials/image_classification/image_classification.png
--- a/doc/tutorials/image_classification/index_cn.md
+++ b/doc/tutorials/image_classification/index_cn.md
--- a/doc/tutorials/image_classification/index_en.md
+++ b/doc/tutorials/image_classification/index_en.md
--- a/doc/tutorials/image_classification/lenet.png
+++ b/doc/tutorials/image_classification/lenet.png
--- a/doc/tutorials/image_classification/plot.png
+++ b/doc/tutorials/image_classification/plot.png
--- a/doc/tutorials/image_classification/src/cifar.png
+++ b/doc/tutorials/image_classification/src/cifar.png
--- a/doc/tutorials/image_classification/src/image_classification.png
+++ b/doc/tutorials/image_classification/src/image_classification.png
--- a/doc/tutorials/image_classification/src/lenet.png
+++ b/doc/tutorials/image_classification/src/lenet.png
--- a/doc/tutorials/image_classification/src/plot.png
+++ b/doc/tutorials/image_classification/src/plot.png
--- a/doc/tutorials/index_cn.md
+++ b/doc/tutorials/index_cn.md
--- a/doc/tutorials/index_en.md
+++ b/doc/tutorials/index_en.md
--- a/doc/tutorials/rec/ml_dataset_cn.md
+++ b/doc/tutorials/rec/ml_dataset_cn.md
--- a/doc/tutorials/rec/ml_dataset_en.md
+++ b/doc/tutorials/rec/ml_dataset_en.md
--- a/doc/tutorials/rec/ml_regression_cn.rst
+++ b/doc/tutorials/rec/ml_regression_cn.rst
--- a/doc/tutorials/rec/ml_regression_en.rst
+++ b/doc/tutorials/rec/ml_regression_en.rst
--- a/doc/tutorials/rec/rec_regression_network.png
+++ b/doc/tutorials/rec/rec_regression_network.png
--- a/doc/tutorials/semantic_role_labeling/feature.jpg
+++ b/doc/tutorials/semantic_role_labeling/feature.jpg
--- a/doc/tutorials/semantic_role_labeling/index_cn.md
+++ b/doc/tutorials/semantic_role_labeling/index_cn.md
--- a/doc/tutorials/semantic_role_labeling/index_en.md
+++ b/doc/tutorials/semantic_role_labeling/index_en.md
--- a/doc/tutorials/semantic_role_labeling/network_arch.png
+++ b/doc/tutorials/semantic_role_labeling/network_arch.png
--- a/doc/tutorials/semantic_role_labeling/src/curve.jpg
+++ b/doc/tutorials/semantic_role_labeling/src/curve.jpg
--- a/doc/tutorials/semantic_role_labeling/src/feature.jpg
+++ b/doc/tutorials/semantic_role_labeling/src/feature.jpg
--- a/doc/tutorials/semantic_role_labeling/src/network_arch.png
+++ b/doc/tutorials/semantic_role_labeling/src/network_arch.png
--- a/doc/tutorials/sentiment_analysis/index_cn.md
+++ b/doc/tutorials/sentiment_analysis/index_cn.md
--- a/doc/tutorials/sentiment_analysis/index_en.md
+++ b/doc/tutorials/sentiment_analysis/index_en.md
--- a/doc/tutorials/sentiment_analysis/lstm.png
+++ b/doc/tutorials/sentiment_analysis/lstm.png
--- a/doc/tutorials/sentiment_analysis/src/bi_lstm.jpg
+++ b/doc/tutorials/sentiment_analysis/src/bi_lstm.jpg
--- a/doc/tutorials/sentiment_analysis/src/lstm.png
+++ b/doc/tutorials/sentiment_analysis/src/lstm.png
--- a/doc/tutorials/sentiment_analysis/src/stacked_lstm.jpg
+++ b/doc/tutorials/sentiment_analysis/src/stacked_lstm.jpg
--- a/doc/tutorials/sentiment_analysis/stacked_lstm.jpg
+++ b/doc/tutorials/sentiment_analysis/stacked_lstm.jpg
--- a/doc/tutorials/text_generation/index_cn.md
+++ b/doc/tutorials/text_generation/index_cn.md
--- a/doc/tutorials/text_generation/index_en.md
+++ b/doc/tutorials/text_generation/index_en.md
--- a/doc/v1_api_tutorials/README.md
+++ b/doc/v1_api_tutorials/README.md
--- a/doc/tutorials/embedding_model/index_cn.md
+++ b/doc/tutorials/embedding_model/index_cn.md
--- a/doc/tutorials/embedding_model/index_en.md
+++ b/doc/tutorials/embedding_model/index_en.md
--- a/doc/tutorials/embedding_model/neural-n-gram-model.png
+++ b/doc/tutorials/embedding_model/neural-n-gram-model.png
--- a/doc/tutorials/gan/gan.png
+++ b/doc/tutorials/gan/gan.png
--- a/doc/tutorials/gan/index_en.md
+++ b/doc/tutorials/gan/index_en.md
--- a/doc/tutorials/gan/mnist_sample.png
+++ b/doc/tutorials/gan/mnist_sample.png
--- a/doc/tutorials/gan/uniform_sample.png
+++ b/doc/tutorials/gan/uniform_sample.png
--- a/doc/tutorials/imagenet_model/resnet_block.jpg
+++ b/doc/tutorials/imagenet_model/resnet_block.jpg
--- a/doc/tutorials/imagenet_model/resnet_model_cn.md
+++ b/doc/tutorials/imagenet_model/resnet_model_cn.md
--- a/doc/tutorials/imagenet_model/resnet_model_en.md
+++ b/doc/tutorials/imagenet_model/resnet_model_en.md
--- a/doc/tutorials/quick_start/index_cn.rst
+++ b/doc/tutorials/quick_start/index_cn.rst
--- a/doc/tutorials/quick_start/index_en.md
+++ b/doc/tutorials/quick_start/index_en.md
--- a/doc/tutorials/quick_start/src/NetContinuous_cn.jpg
+++ b/doc/tutorials/quick_start/src/NetContinuous_cn.jpg
--- a/doc/tutorials/quick_start/src/NetContinuous_en.png
+++ b/doc/tutorials/quick_start/src/NetContinuous_en.png
--- a/doc/tutorials/quick_start/src/NetConv_cn.jpg
+++ b/doc/tutorials/quick_start/src/NetConv_cn.jpg
--- a/doc/tutorials/quick_start/src/NetConv_en.png
+++ b/doc/tutorials/quick_start/src/NetConv_en.png
--- a/doc/tutorials/quick_start/src/NetLR_cn.jpg
+++ b/doc/tutorials/quick_start/src/NetLR_cn.jpg
--- a/doc/tutorials/quick_start/src/NetLR_en.png
+++ b/doc/tutorials/quick_start/src/NetLR_en.png
--- a/doc/tutorials/quick_start/src/NetRNN_cn.jpg
+++ b/doc/tutorials/quick_start/src/NetRNN_cn.jpg
--- a/doc/tutorials/quick_start/src/NetRNN_en.png
+++ b/doc/tutorials/quick_start/src/NetRNN_en.png
--- a/doc/tutorials/quick_start/src/PipelineNetwork_cn.jpg
+++ b/doc/tutorials/quick_start/src/PipelineNetwork_cn.jpg
--- a/doc/tutorials/quick_start/src/PipelineNetwork_en.jpg
+++ b/doc/tutorials/quick_start/src/PipelineNetwork_en.jpg
--- a/doc/tutorials/quick_start/src/PipelineTest_cn.jpg
+++ b/doc/tutorials/quick_start/src/PipelineTest_cn.jpg
--- a/doc/tutorials/quick_start/src/PipelineTest_en.png
+++ b/doc/tutorials/quick_start/src/PipelineTest_en.png
--- a/doc/tutorials/quick_start/src/PipelineTrain_cn.jpg
+++ b/doc/tutorials/quick_start/src/PipelineTrain_cn.jpg
--- a/doc/tutorials/quick_start/src/PipelineTrain_en.png
+++ b/doc/tutorials/quick_start/src/PipelineTrain_en.png
--- a/doc/tutorials/quick_start/src/Pipeline_cn.jpg
+++ b/doc/tutorials/quick_start/src/Pipeline_cn.jpg
--- a/doc/tutorials/quick_start/src/Pipeline_en.jpg
+++ b/doc/tutorials/quick_start/src/Pipeline_en.jpg
--- a/doc_theme/templates/layout.html
+++ b/doc_theme/templates/layout.html
--- a/go/.gitignore
+++ b/go/.gitignore
--- a/go/CMakeLists.txt
+++ b/go/CMakeLists.txt
--- a/go/cmd/master/CMakeLists.txt
+++ b/go/cmd/master/CMakeLists.txt
--- a/go/cmd/master/master.go
+++ b/go/cmd/master/master.go
--- a/go/cmd/pserver/.gitignore
+++ b/go/cmd/pserver/.gitignore
--- a/go/cmd/pserver/CMakeLists.txt
+++ b/go/cmd/pserver/CMakeLists.txt
--- a/go/cmd/pserver/pserver.go
+++ b/go/cmd/pserver/pserver.go
--- a/go/connection/conn.go
+++ b/go/connection/conn.go
--- a/go/glide.lock
+++ b/go/glide.lock
--- a/go/glide.yaml
+++ b/go/glide.yaml
--- a/go/master/CMakeLists.txt
+++ b/go/master/CMakeLists.txt
--- a/go/master/c/CMakeLists.txt
+++ b/go/master/c/CMakeLists.txt
--- a/go/master/c/client.go
+++ b/go/master/c/client.go
--- a/go/master/client.go
+++ b/go/master/client.go
--- a/go/master/client_internal_test.go
+++ b/go/master/client_internal_test.go
--- a/go/master/client_test.go
+++ b/go/master/client_test.go
--- a/go/master/etcd_client.go
+++ b/go/master/etcd_client.go
--- a/go/master/inmem_store.go
+++ b/go/master/inmem_store.go
--- a/go/master/service.go
+++ b/go/master/service.go
--- a/go/master/service_internal_test.go
+++ b/go/master/service_internal_test.go
--- a/go/master/service_test.go
+++ b/go/master/service_test.go
--- a/go/proto/.gitignore
+++ b/go/proto/.gitignore
--- a/go/pserver/CMakeLists.txt
+++ b/go/pserver/CMakeLists.txt
--- a/go/pserver/client/CMakeLists.txt
+++ b/go/pserver/client/CMakeLists.txt
--- a/go/pserver/client/c/.gitignore
+++ b/go/pserver/client/c/.gitignore
--- a/go/pserver/client/c/CMakeLists.txt
+++ b/go/pserver/client/c/CMakeLists.txt
--- a/go/pserver/client/c/cclient.go
+++ b/go/pserver/client/c/cclient.go
--- a/go/pserver/client/c/test/CMakeLists.txt
+++ b/go/pserver/client/c/test/CMakeLists.txt
--- a/go/pserver/client/c/test/test_cclient.c
+++ b/go/pserver/client/c/test/test_cclient.c
--- a/go/pserver/client/c/test/test_mnist.py
+++ b/go/pserver/client/c/test/test_mnist.py
--- a/go/pserver/client/c/test/test_train.py
+++ b/go/pserver/client/c/test/test_train.py
--- a/go/pserver/client/c/test/testdata/optimizer.pb
+++ b/go/pserver/client/c/test/testdata/optimizer.pb
--- a/go/pserver/client/client.go
+++ b/go/pserver/client/client.go
--- a/go/pserver/client/client_test.go
+++ b/go/pserver/client/client_test.go
--- a/go/pserver/client/etcd_client.go
+++ b/go/pserver/client/etcd_client.go
--- a/go/pserver/client/etcd_client_test.go
+++ b/go/pserver/client/etcd_client_test.go
--- a/go/pserver/etcd_client.go
+++ b/go/pserver/etcd_client.go
--- a/go/pserver/optimizer.go
+++ b/go/pserver/optimizer.go
--- a/go/pserver/optimizer_test.go
+++ b/go/pserver/optimizer_test.go
--- a/go/pserver/service.go
+++ b/go/pserver/service.go
--- a/go/pserver/service_internal_test.go
+++ b/go/pserver/service_internal_test.go
--- a/go/pserver/service_test.go
+++ b/go/pserver/service_test.go
--- a/go/utils/networkhelper/CMakeLists.txt
+++ b/go/utils/networkhelper/CMakeLists.txt
--- a/go/utils/networkhelper/helper.go
+++ b/go/utils/networkhelper/helper.go
--- a/go/utils/networkhelper/helper_test.go
+++ b/go/utils/networkhelper/helper_test.go
--- a/paddle/.set_python_path.sh
+++ b/paddle/.set_python_path.sh
--- a/paddle/CMakeLists.txt
+++ b/paddle/CMakeLists.txt
--- a/paddle/api/Arguments.cpp
+++ b/paddle/api/Arguments.cpp
--- a/paddle/api/CMakeLists.txt
+++ b/paddle/api/CMakeLists.txt
--- a/paddle/api/ConfigParser.cpp
+++ b/paddle/api/ConfigParser.cpp
--- a/paddle/api/Evaluator.cpp
+++ b/paddle/api/Evaluator.cpp
--- a/paddle/api/Paddle.i
+++ b/paddle/api/Paddle.i
--- a/paddle/api/PaddleAPI.h
+++ b/paddle/api/PaddleAPI.h
--- a/paddle/api/ParameterOptimizer.cpp
+++ b/paddle/api/ParameterOptimizer.cpp
--- a/paddle/api/ParameterUpdater.cpp
+++ b/paddle/api/ParameterUpdater.cpp
--- a/paddle/api/Util.cpp
+++ b/paddle/api/Util.cpp
--- a/paddle/api/Vector.cpp
+++ b/paddle/api/Vector.cpp
--- a/paddle/api/test/CMakeLists.txt
+++ b/paddle/api/test/CMakeLists.txt
--- a/paddle/api/test/run_tests.sh
+++ b/paddle/api/test/run_tests.sh
--- a/paddle/api/test/testArguments.py
+++ b/paddle/api/test/testArguments.py
--- a/paddle/capi/Arguments.cpp
+++ b/paddle/capi/Arguments.cpp
--- a/paddle/capi/CMakeLists.txt
+++ b/paddle/capi/CMakeLists.txt
--- a/paddle/capi/Main.cpp
+++ b/paddle/capi/Main.cpp
--- a/paddle/capi/Matrix.cpp
+++ b/paddle/capi/Matrix.cpp
--- a/paddle/capi/Vector.cpp
+++ b/paddle/capi/Vector.cpp
--- a/paddle/capi/arguments.h
+++ b/paddle/capi/arguments.h
--- a/paddle/capi/capi.h
+++ b/paddle/capi/capi.h
--- a/paddle/capi/capi_private.h
+++ b/paddle/capi/capi_private.h
--- a/paddle/capi/config.h.in
+++ b/paddle/capi/config.h.in
--- a/paddle/capi/error.cpp
+++ b/paddle/capi/error.cpp
--- a/paddle/capi/error.h
+++ b/paddle/capi/error.h
--- a/paddle/capi/examples/.gitignore
+++ b/paddle/capi/examples/.gitignore
--- a/paddle/capi/examples/README.md
+++ b/paddle/capi/examples/README.md
--- a/paddle/capi/examples/model_inference/README.md
+++ b/paddle/capi/examples/model_inference/README.md
--- a/paddle/capi/examples/model_inference/common/common.h
+++ b/paddle/capi/examples/model_inference/common/common.h
--- a/paddle/capi/examples/model_inference/dense/CMakeLists.txt
+++ b/paddle/capi/examples/model_inference/dense/CMakeLists.txt
--- a/paddle/capi/examples/model_inference/dense/convert_protobin.sh
+++ b/paddle/capi/examples/model_inference/dense/convert_protobin.sh
--- a/paddle/capi/examples/model_inference/dense/main.c
+++ b/paddle/capi/examples/model_inference/dense/main.c
--- a/paddle/capi/examples/model_inference/dense/trainer_config.py
+++ b/paddle/capi/examples/model_inference/dense/trainer_config.py
--- a/paddle/capi/examples/model_inference/multi_thread/.gitignore
+++ b/paddle/capi/examples/model_inference/multi_thread/.gitignore
--- a/paddle/capi/examples/model_inference/multi_thread/CMakeLists.txt
+++ b/paddle/capi/examples/model_inference/multi_thread/CMakeLists.txt
--- a/paddle/capi/examples/model_inference/multi_thread/convert_protobin.sh
+++ b/paddle/capi/examples/model_inference/multi_thread/convert_protobin.sh
--- a/paddle/capi/examples/model_inference/multi_thread/main.c
+++ b/paddle/capi/examples/model_inference/multi_thread/main.c
--- a/paddle/capi/examples/model_inference/multi_thread/main_gpu.c
+++ b/paddle/capi/examples/model_inference/multi_thread/main_gpu.c
--- a/paddle/capi/examples/model_inference/multi_thread/trainer_config.py
+++ b/paddle/capi/examples/model_inference/multi_thread/trainer_config.py
--- a/paddle/capi/examples/model_inference/sequence/.gitignore
+++ b/paddle/capi/examples/model_inference/sequence/.gitignore
--- a/paddle/capi/examples/model_inference/sequence/CMakeLists.txt
+++ b/paddle/capi/examples/model_inference/sequence/CMakeLists.txt
--- a/paddle/capi/examples/model_inference/sequence/convert_protobin.sh
+++ b/paddle/capi/examples/model_inference/sequence/convert_protobin.sh
--- a/paddle/capi/examples/model_inference/sequence/main.c
+++ b/paddle/capi/examples/model_inference/sequence/main.c
--- a/paddle/capi/examples/model_inference/sequence/trainer_config.py
+++ b/paddle/capi/examples/model_inference/sequence/trainer_config.py
--- a/paddle/capi/examples/model_inference/sparse_binary/.gitignore
+++ b/paddle/capi/examples/model_inference/sparse_binary/.gitignore
--- a/paddle/capi/examples/model_inference/sparse_binary/CMakeLists.txt
+++ b/paddle/capi/examples/model_inference/sparse_binary/CMakeLists.txt
--- a/paddle/capi/examples/model_inference/sparse_binary/convert_protobin.sh
+++ b/paddle/capi/examples/model_inference/sparse_binary/convert_protobin.sh
--- a/paddle/capi/examples/model_inference/sparse_binary/main.c
+++ b/paddle/capi/examples/model_inference/sparse_binary/main.c
--- a/paddle/capi/examples/model_inference/sparse_binary/trainer_config.py
+++ b/paddle/capi/examples/model_inference/sparse_binary/trainer_config.py
--- a/paddle/capi/gradient_machine.cpp
+++ b/paddle/capi/gradient_machine.cpp
--- a/paddle/capi/gradient_machine.h
+++ b/paddle/capi/gradient_machine.h
--- a/paddle/capi/main.h
+++ b/paddle/capi/main.h
--- a/paddle/capi/matrix.h
+++ b/paddle/capi/matrix.h
--- a/paddle/capi/paddle_capi.map
+++ b/paddle/capi/paddle_capi.map
--- a/paddle/capi/tests/.gitignore
+++ b/paddle/capi/tests/.gitignore
--- a/paddle/capi/tests/CMakeLists.txt
+++ b/paddle/capi/tests/CMakeLists.txt
--- a/paddle/capi/tests/test_Arguments.cpp
+++ b/paddle/capi/tests/test_Arguments.cpp
--- a/paddle/capi/tests/test_GradientMachine.cpp
+++ b/paddle/capi/tests/test_GradientMachine.cpp
--- a/paddle/capi/tests/test_Matrix.cpp
+++ b/paddle/capi/tests/test_Matrix.cpp
--- a/paddle/capi/tests/test_Vector.cpp
+++ b/paddle/capi/tests/test_Vector.cpp
--- a/paddle/capi/tests/test_predict_network.py
+++ b/paddle/capi/tests/test_predict_network.py
--- a/paddle/capi/vector.h
+++ b/paddle/capi/vector.h
--- a/paddle/cuda/CMakeLists.txt
+++ b/paddle/cuda/CMakeLists.txt
--- a/paddle/cuda/include/hl_activation_functions.h
+++ b/paddle/cuda/include/hl_activation_functions.h
--- a/paddle/cuda/include/hl_batch_norm.h
+++ b/paddle/cuda/include/hl_batch_norm.h
--- a/paddle/cuda/include/hl_cnn.h
+++ b/paddle/cuda/include/hl_cnn.h
--- a/paddle/cuda/include/hl_cpu_gru.cuh
+++ b/paddle/cuda/include/hl_cpu_gru.cuh
--- a/paddle/cuda/include/hl_cpu_matrix_kernel.cuh
+++ b/paddle/cuda/include/hl_cpu_matrix_kernel.cuh
--- a/paddle/cuda/include/hl_cpu_matrix_kernel_detail.cuh
+++ b/paddle/cuda/include/hl_cpu_matrix_kernel_detail.cuh
--- a/paddle/cuda/include/hl_cpu_scalar.cuh
+++ b/paddle/cuda/include/hl_cpu_scalar.cuh
--- a/paddle/cuda/include/hl_cpu_simd_neon.cuh
+++ b/paddle/cuda/include/hl_cpu_simd_neon.cuh
--- a/paddle/cuda/include/hl_cpu_simd_sse.cuh
+++ b/paddle/cuda/include/hl_cpu_simd_sse.cuh
--- a/paddle/cuda/include/hl_cuda_cudnn.h
+++ b/paddle/cuda/include/hl_cuda_cudnn.h
--- a/paddle/cuda/include/hl_dso_loader.h
+++ b/paddle/cuda/include/hl_dso_loader.h
--- a/paddle/cuda/include/hl_gpu.h
+++ b/paddle/cuda/include/hl_gpu.h
--- a/paddle/cuda/include/hl_matrix.h
+++ b/paddle/cuda/include/hl_matrix.h
--- a/paddle/cuda/include/hl_matrix_base.cuh
+++ b/paddle/cuda/include/hl_matrix_base.cuh
--- a/paddle/cuda/include/hl_matrix_base_detail.cuh
+++ b/paddle/cuda/include/hl_matrix_base_detail.cuh
--- a/paddle/cuda/include/hl_matrix_base_neon.cuh
+++ b/paddle/cuda/include/hl_matrix_base_neon.cuh
--- a/paddle/cuda/include/hl_matrix_base_sse.cuh
+++ b/paddle/cuda/include/hl_matrix_base_sse.cuh
--- a/paddle/cuda/include/hl_matrix_type.cuh
+++ b/paddle/cuda/include/hl_matrix_type.cuh
--- a/paddle/cuda/include/hl_neon_matrix_kernel.cuh
+++ b/paddle/cuda/include/hl_neon_matrix_kernel.cuh
--- a/paddle/cuda/include/hl_sse_matrix_kernel.cuh
+++ b/paddle/cuda/include/hl_sse_matrix_kernel.cuh
--- a/paddle/cuda/include/hl_tensor_ops.h
+++ b/paddle/cuda/include/hl_tensor_ops.h
--- a/paddle/cuda/include/stub/hl_cnn_stub.h
+++ b/paddle/cuda/include/stub/hl_cnn_stub.h
--- a/paddle/cuda/include/stub/hl_cuda_cudnn_stub.h
+++ b/paddle/cuda/include/stub/hl_cuda_cudnn_stub.h
--- a/paddle/cuda/include/stub/hl_matrix_stub.h
+++ b/paddle/cuda/include/stub/hl_matrix_stub.h
--- a/paddle/cuda/src/hl_batch_norm.cu
+++ b/paddle/cuda/src/hl_batch_norm.cu
--- a/paddle/cuda/src/hl_batch_transpose.cu
+++ b/paddle/cuda/src/hl_batch_transpose.cu
--- a/paddle/cuda/src/hl_cuda_aggregate.cu
+++ b/paddle/cuda/src/hl_cuda_aggregate.cu
--- a/paddle/cuda/src/hl_cuda_cnn.cu
+++ b/paddle/cuda/src/hl_cuda_cnn.cu
--- a/paddle/cuda/src/hl_cuda_cublas.cc
+++ b/paddle/cuda/src/hl_cuda_cublas.cc
--- a/paddle/cuda/src/hl_cuda_cudnn.cc
+++ b/paddle/cuda/src/hl_cuda_cudnn.cc
--- a/paddle/cuda/src/hl_cuda_device.cc
+++ b/paddle/cuda/src/hl_cuda_device.cc
--- a/paddle/cuda/src/hl_cuda_lstm.cu
+++ b/paddle/cuda/src/hl_cuda_lstm.cu
--- a/paddle/cuda/src/hl_cuda_matrix.cu
+++ b/paddle/cuda/src/hl_cuda_matrix.cu
--- a/paddle/cuda/src/hl_cuda_sequence.cu
+++ b/paddle/cuda/src/hl_cuda_sequence.cu
--- a/paddle/cuda/src/hl_cuda_sparse.cu
+++ b/paddle/cuda/src/hl_cuda_sparse.cu
--- a/paddle/cuda/src/hl_dso_loader.cc
+++ b/paddle/cuda/src/hl_dso_loader.cc
--- a/paddle/cuda/src/hl_perturbation_util.cu
+++ b/paddle/cuda/src/hl_perturbation_util.cu
--- a/paddle/cuda/src/hl_table_apply.cu
+++ b/paddle/cuda/src/hl_table_apply.cu
--- a/paddle/cuda/src/hl_top_k.cu
+++ b/paddle/cuda/src/hl_top_k.cu
--- a/paddle/cuda/src/hl_warpctc_wrap.cc
+++ b/paddle/cuda/src/hl_warpctc_wrap.cc
--- a/paddle/framework/.clang-format
+++ b/paddle/framework/.clang-format
--- a/paddle/framework/CMakeLists.txt
+++ b/paddle/framework/CMakeLists.txt
--- a/paddle/framework/attribute.cc
+++ b/paddle/framework/attribute.cc
--- a/paddle/framework/attribute.h
+++ b/paddle/framework/attribute.h
--- a/paddle/framework/backward.cc
+++ b/paddle/framework/backward.cc
--- a/paddle/framework/backward.h
+++ b/paddle/framework/backward.h
--- a/paddle/framework/backward.md
+++ b/paddle/framework/backward.md
--- a/paddle/framework/backward_test.cc
+++ b/paddle/framework/backward_test.cc
--- a/paddle/framework/block_desc.cc
+++ b/paddle/framework/block_desc.cc
--- a/paddle/framework/block_desc.h
+++ b/paddle/framework/block_desc.h
--- a/paddle/framework/data_type.h
+++ b/paddle/framework/data_type.h
--- a/paddle/framework/ddim.cc
+++ b/paddle/framework/ddim.cc
--- a/paddle/framework/ddim.h
+++ b/paddle/framework/ddim.h
--- a/paddle/framework/ddim_test.cc
+++ b/paddle/framework/ddim_test.cc
--- a/paddle/framework/details/op_registry.h
+++ b/paddle/framework/details/op_registry.h
--- a/paddle/framework/dim.h
+++ b/paddle/framework/dim.h
--- a/paddle/framework/dim_test.cu
+++ b/paddle/framework/dim_test.cu
--- a/paddle/framework/eigen.h
+++ b/paddle/framework/eigen.h
--- a/paddle/framework/eigen_test.cc
+++ b/paddle/framework/eigen_test.cc
--- a/paddle/framework/executor.cc
+++ b/paddle/framework/executor.cc
--- a/paddle/framework/executor.h
+++ b/paddle/framework/executor.h
--- a/paddle/framework/feed_fetch_method.h
+++ b/paddle/framework/feed_fetch_method.h
--- a/paddle/framework/feed_fetch_type.h
+++ b/paddle/framework/feed_fetch_type.h
--- a/paddle/framework/framework.proto
+++ b/paddle/framework/framework.proto
--- a/paddle/framework/grad_op_desc_maker.h
+++ b/paddle/framework/grad_op_desc_maker.h
--- a/paddle/framework/images/duplicate_op.graffle
+++ b/paddle/framework/images/duplicate_op.graffle
--- a/paddle/framework/images/duplicate_op.png
+++ b/paddle/framework/images/duplicate_op.png
--- a/paddle/framework/images/duplicate_op2.graffle
+++ b/paddle/framework/images/duplicate_op2.graffle
--- a/paddle/framework/images/duplicate_op2.png
+++ b/paddle/framework/images/duplicate_op2.png
--- a/paddle/framework/lod_rank_table.cc
+++ b/paddle/framework/lod_rank_table.cc
--- a/paddle/framework/lod_rank_table.h
+++ b/paddle/framework/lod_rank_table.h
--- a/paddle/framework/lod_tensor.cc
+++ b/paddle/framework/lod_tensor.cc
--- a/paddle/framework/lod_tensor.h
+++ b/paddle/framework/lod_tensor.h
--- a/paddle/framework/lod_tensor.md
+++ b/paddle/framework/lod_tensor.md
--- a/paddle/framework/lod_tensor_array.h
+++ b/paddle/framework/lod_tensor_array.h
--- a/paddle/framework/lod_tensor_test.cc
+++ b/paddle/framework/lod_tensor_test.cc
--- a/paddle/framework/lod_tensor_test.cu
+++ b/paddle/framework/lod_tensor_test.cu
--- a/paddle/framework/op_desc.cc
+++ b/paddle/framework/op_desc.cc
--- a/paddle/framework/op_desc.h
+++ b/paddle/framework/op_desc.h
--- a/paddle/framework/op_info.cc
+++ b/paddle/framework/op_info.cc
--- a/paddle/framework/op_info.h
+++ b/paddle/framework/op_info.h
--- a/paddle/framework/op_proto_maker.cc
+++ b/paddle/framework/op_proto_maker.cc
--- a/paddle/framework/op_proto_maker.h
+++ b/paddle/framework/op_proto_maker.h
--- a/paddle/framework/op_proto_maker_test.cc
+++ b/paddle/framework/op_proto_maker_test.cc
--- a/paddle/framework/op_registry.cc
+++ b/paddle/framework/op_registry.cc
--- a/paddle/framework/op_registry.h
+++ b/paddle/framework/op_registry.h
--- a/paddle/framework/op_registry_test.cc
+++ b/paddle/framework/op_registry_test.cc
--- a/paddle/framework/operator.cc
+++ b/paddle/framework/operator.cc
--- a/paddle/framework/operator.h
+++ b/paddle/framework/operator.h
--- a/paddle/framework/operator_test.cc
+++ b/paddle/framework/operator_test.cc
--- a/paddle/framework/program_desc.cc
+++ b/paddle/framework/program_desc.cc
--- a/paddle/framework/program_desc.h
+++ b/paddle/framework/program_desc.h
--- a/paddle/framework/program_desc_test.cc
+++ b/paddle/framework/program_desc_test.cc
--- a/paddle/framework/proto_desc.h
+++ b/paddle/framework/proto_desc.h
--- a/paddle/framework/prune.cc
+++ b/paddle/framework/prune.cc
--- a/paddle/framework/prune.h
+++ b/paddle/framework/prune.h
--- a/paddle/framework/prune_test.cc
+++ b/paddle/framework/prune_test.cc
--- a/paddle/framework/scope.cc
+++ b/paddle/framework/scope.cc
--- a/paddle/framework/scope.h
+++ b/paddle/framework/scope.h
--- a/paddle/framework/scope_test.cc
+++ b/paddle/framework/scope_test.cc
--- a/paddle/framework/selected_rows.cc
+++ b/paddle/framework/selected_rows.cc
--- a/paddle/framework/selected_rows.h
+++ b/paddle/framework/selected_rows.h
--- a/paddle/framework/selected_rows_test.cc
+++ b/paddle/framework/selected_rows_test.cc
--- a/paddle/framework/shape_inference.cc
+++ b/paddle/framework/shape_inference.cc
--- a/paddle/framework/shape_inference.h
+++ b/paddle/framework/shape_inference.h
--- a/paddle/framework/tensor.cc
+++ b/paddle/framework/tensor.cc
--- a/paddle/framework/tensor.h
+++ b/paddle/framework/tensor.h
--- a/paddle/framework/tensor.md
+++ b/paddle/framework/tensor.md
--- a/paddle/framework/tensor_impl.h
+++ b/paddle/framework/tensor_impl.h
--- a/paddle/framework/tensor_test.cc
+++ b/paddle/framework/tensor_test.cc
--- a/paddle/framework/tensor_util.h
+++ b/paddle/framework/tensor_util.h
--- a/paddle/framework/tensor_util_test.cc
+++ b/paddle/framework/tensor_util_test.cc
--- a/paddle/framework/type_defs.h
+++ b/paddle/framework/type_defs.h
--- a/paddle/framework/var_desc.cc
+++ b/paddle/framework/var_desc.cc
--- a/paddle/framework/var_desc.h
+++ b/paddle/framework/var_desc.h
--- a/paddle/framework/var_type.h
+++ b/paddle/framework/var_type.h
--- a/paddle/framework/var_type_inference.h
+++ b/paddle/framework/var_type_inference.h
--- a/paddle/framework/var_type_inference_test.cc
+++ b/paddle/framework/var_type_inference_test.cc
--- a/paddle/framework/variable.h
+++ b/paddle/framework/variable.h
--- a/paddle/framework/variable.md
+++ b/paddle/framework/variable.md
--- a/paddle/framework/variable_test.cc
+++ b/paddle/framework/variable_test.cc
--- a/paddle/function/BlockExpandOp.cpp
+++ b/paddle/function/BlockExpandOp.cpp
--- a/paddle/function/BlockExpandOpTest.cpp
+++ b/paddle/function/BlockExpandOpTest.cpp
--- a/paddle/function/BufferArgTest.cpp
+++ b/paddle/function/BufferArgTest.cpp
--- a/paddle/function/CMakeLists.txt
+++ b/paddle/function/CMakeLists.txt
--- a/paddle/function/ContextProjectionOp.cpp
+++ b/paddle/function/ContextProjectionOp.cpp
--- a/paddle/function/ContextProjectionOpGpu.cu
+++ b/paddle/function/ContextProjectionOpGpu.cu
--- a/paddle/function/ContextProjectionOpTest.cpp
+++ b/paddle/function/ContextProjectionOpTest.cpp
--- a/paddle/function/ConvOp.h
+++ b/paddle/function/ConvOp.h
--- a/paddle/function/ConvOpTest.h
+++ b/paddle/function/ConvOpTest.h
--- a/paddle/function/CosSimOp.cpp
+++ b/paddle/function/CosSimOp.cpp
--- a/paddle/function/CosSimOpGpu.cu
+++ b/paddle/function/CosSimOpGpu.cu
--- a/paddle/function/CosSimOpTest.cpp
+++ b/paddle/function/CosSimOpTest.cpp
--- a/paddle/function/CropOp.cpp
+++ b/paddle/function/CropOp.cpp
--- a/paddle/function/CropOp.h
+++ b/paddle/function/CropOp.h
--- a/paddle/function/CropOpGpu.cu
+++ b/paddle/function/CropOpGpu.cu
--- a/paddle/function/CropOpTest.cpp
+++ b/paddle/function/CropOpTest.cpp
--- a/paddle/function/CrossMapNormalOp.cpp
+++ b/paddle/function/CrossMapNormalOp.cpp
--- a/paddle/function/CrossMapNormalOpGpu.cu
+++ b/paddle/function/CrossMapNormalOpGpu.cu
--- a/paddle/function/CrossMapNormalOpTest.cpp
+++ b/paddle/function/CrossMapNormalOpTest.cpp
--- a/paddle/function/DepthwiseConvOp.cpp
+++ b/paddle/function/DepthwiseConvOp.cpp
--- a/paddle/function/DepthwiseConvOp.h
+++ b/paddle/function/DepthwiseConvOp.h
--- a/paddle/function/DepthwiseConvOpGpu.cu
+++ b/paddle/function/DepthwiseConvOpGpu.cu
--- a/paddle/function/DepthwiseConvOpTest.cpp
+++ b/paddle/function/DepthwiseConvOpTest.cpp
--- a/paddle/function/EigenGemm.cpp
+++ b/paddle/function/EigenGemm.cpp
--- a/paddle/function/FunctionTest.cpp
+++ b/paddle/function/FunctionTest.cpp
--- a/paddle/function/FunctionTest.h
+++ b/paddle/function/FunctionTest.h
--- a/paddle/function/GemmConvOp.cpp
+++ b/paddle/function/GemmConvOp.cpp
--- a/paddle/function/GemmConvOpTest.cpp
+++ b/paddle/function/GemmConvOpTest.cpp
--- a/paddle/function/GemmFunctor.cpp
+++ b/paddle/function/GemmFunctor.cpp
--- a/paddle/function/GemmFunctor.h
+++ b/paddle/function/GemmFunctor.h
--- a/paddle/function/GruFunctor.h
+++ b/paddle/function/GruFunctor.h
--- a/paddle/function/Im2Col.h
+++ b/paddle/function/Im2Col.h
--- a/paddle/function/Im2ColOp.cpp
+++ b/paddle/function/Im2ColOp.cpp
--- a/paddle/function/Im2ColOpGpu.cu
+++ b/paddle/function/Im2ColOpGpu.cu
--- a/paddle/function/Im2ColTest.cpp
+++ b/paddle/function/Im2ColTest.cpp
--- a/paddle/function/MulOp.cpp
+++ b/paddle/function/MulOp.cpp
--- a/paddle/function/MulOpGpu.cu
+++ b/paddle/function/MulOpGpu.cu
--- a/paddle/function/MulOpTest.cpp
+++ b/paddle/function/MulOpTest.cpp
--- a/paddle/function/NaiveConvOp.cpp
+++ b/paddle/function/NaiveConvOp.cpp
--- a/paddle/function/PadOp.cpp
+++ b/paddle/function/PadOp.cpp
--- a/paddle/function/PadOpGpu.cu
+++ b/paddle/function/PadOpGpu.cu
--- a/paddle/function/PadOpTest.cpp
+++ b/paddle/function/PadOpTest.cpp
--- a/paddle/function/RowConvOp.cpp
+++ b/paddle/function/RowConvOp.cpp
--- a/paddle/function/RowConvOp.h
+++ b/paddle/function/RowConvOp.h
--- a/paddle/function/RowConvOpGpu.cu
+++ b/paddle/function/RowConvOpGpu.cu
--- a/paddle/function/RowConvOpTest.cpp
+++ b/paddle/function/RowConvOpTest.cpp
--- a/paddle/function/ScaleSubRegionOp.cpp
+++ b/paddle/function/ScaleSubRegionOp.cpp
--- a/paddle/function/ScaleSubRegionOp.h
+++ b/paddle/function/ScaleSubRegionOp.h
--- a/paddle/function/ScaleSubRegionOpGpu.cu
+++ b/paddle/function/ScaleSubRegionOpGpu.cu
--- a/paddle/function/ScaleSubRegionOpTest.cpp
+++ b/paddle/function/ScaleSubRegionOpTest.cpp
--- a/paddle/function/SwitchOp.cpp
+++ b/paddle/function/SwitchOp.cpp
--- a/paddle/function/SwitchOp.h
+++ b/paddle/function/SwitchOp.h
--- a/paddle/function/SwitchOpGpu.cu
+++ b/paddle/function/SwitchOpGpu.cu
--- a/paddle/function/SwitchOpTest.cpp
+++ b/paddle/function/SwitchOpTest.cpp
--- a/paddle/function/TensorShapeTest.cpp
+++ b/paddle/function/TensorShapeTest.cpp
--- a/paddle/function/TensorTypeTest.cpp
+++ b/paddle/function/TensorTypeTest.cpp
--- a/paddle/function/neon/NeonDepthwiseConv.cpp
+++ b/paddle/function/neon/NeonDepthwiseConv.cpp
--- a/paddle/function/neon/NeonDepthwiseConv.h
+++ b/paddle/function/neon/NeonDepthwiseConv.h
--- a/paddle/function/neon/NeonDepthwiseConvTranspose.cpp
+++ b/paddle/function/neon/NeonDepthwiseConvTranspose.cpp
--- a/paddle/function/neon/neon_util.h
+++ b/paddle/function/neon/neon_util.h
--- a/paddle/function/nnpack/NNPACKConvOp.cpp
+++ b/paddle/function/nnpack/NNPACKConvOp.cpp
--- a/paddle/function/nnpack/NNPACKConvOpTest.cpp
+++ b/paddle/function/nnpack/NNPACKConvOpTest.cpp
--- a/paddle/gserver/CMakeLists.txt
+++ b/paddle/gserver/CMakeLists.txt
--- a/paddle/gserver/activations/ActivationFunction.cpp
+++ b/paddle/gserver/activations/ActivationFunction.cpp
--- a/paddle/gserver/activations/MKLDNNActivation.cpp
+++ b/paddle/gserver/activations/MKLDNNActivation.cpp
--- a/paddle/gserver/activations/MKLDNNActivation.h
+++ b/paddle/gserver/activations/MKLDNNActivation.h
--- a/paddle/gserver/dataproviders/DataProvider.cpp
+++ b/paddle/gserver/dataproviders/DataProvider.cpp
--- a/paddle/gserver/dataproviders/DataProvider.h
+++ b/paddle/gserver/dataproviders/DataProvider.h
--- a/paddle/gserver/dataproviders/ProtoDataProvider.cpp
+++ b/paddle/gserver/dataproviders/ProtoDataProvider.cpp
--- a/paddle/gserver/dataproviders/ProtoDataProvider.h
+++ b/paddle/gserver/dataproviders/ProtoDataProvider.h
--- a/paddle/gserver/evaluators/CTCErrorEvaluator.cpp
+++ b/paddle/gserver/evaluators/CTCErrorEvaluator.cpp
--- a/paddle/gserver/evaluators/ChunkEvaluator.cpp
+++ b/paddle/gserver/evaluators/ChunkEvaluator.cpp
--- a/paddle/gserver/evaluators/DetectionMAPEvaluator.cpp
+++ b/paddle/gserver/evaluators/DetectionMAPEvaluator.cpp
--- a/paddle/gserver/evaluators/Evaluator.cpp
+++ b/paddle/gserver/evaluators/Evaluator.cpp
--- a/paddle/gserver/evaluators/Evaluator.h
+++ b/paddle/gserver/evaluators/Evaluator.h
--- a/paddle/gserver/gradientmachines/GradientMachine.cpp
+++ b/paddle/gserver/gradientmachines/GradientMachine.cpp
--- a/paddle/gserver/gradientmachines/GradientMachine.h
+++ b/paddle/gserver/gradientmachines/GradientMachine.h
--- a/paddle/gserver/gradientmachines/MultiGradientMachine.cpp
+++ b/paddle/gserver/gradientmachines/MultiGradientMachine.cpp
--- a/paddle/gserver/gradientmachines/MultiGradientMachine.h
+++ b/paddle/gserver/gradientmachines/MultiGradientMachine.h
--- a/paddle/gserver/gradientmachines/NeuralNetwork.cpp
+++ b/paddle/gserver/gradientmachines/NeuralNetwork.cpp
--- a/paddle/gserver/gradientmachines/NeuralNetwork.h
+++ b/paddle/gserver/gradientmachines/NeuralNetwork.h
--- a/paddle/gserver/gradientmachines/RecurrentGradientMachine.cpp
+++ b/paddle/gserver/gradientmachines/RecurrentGradientMachine.cpp
--- a/paddle/gserver/gradientmachines/RecurrentGradientMachine.h
+++ b/paddle/gserver/gradientmachines/RecurrentGradientMachine.h
--- a/paddle/gserver/layers/AgentLayer.cpp
+++ b/paddle/gserver/layers/AgentLayer.cpp
--- a/paddle/gserver/layers/AgentLayer.h
+++ b/paddle/gserver/layers/AgentLayer.h
--- a/paddle/gserver/layers/AverageLayer.h
+++ b/paddle/gserver/layers/AverageLayer.h
--- a/paddle/gserver/layers/BatchNormBaseLayer.cpp
+++ b/paddle/gserver/layers/BatchNormBaseLayer.cpp
--- a/paddle/gserver/layers/BatchNormBaseLayer.h
+++ b/paddle/gserver/layers/BatchNormBaseLayer.h
--- a/paddle/gserver/layers/BatchNormalizationLayer.cpp
+++ b/paddle/gserver/layers/BatchNormalizationLayer.cpp
--- a/paddle/gserver/layers/BatchNormalizationLayer.h
+++ b/paddle/gserver/layers/BatchNormalizationLayer.h
--- a/paddle/gserver/layers/BlockExpandLayer.cpp
+++ b/paddle/gserver/layers/BlockExpandLayer.cpp
--- a/paddle/gserver/layers/BlockExpandLayer.h
+++ b/paddle/gserver/layers/BlockExpandLayer.h
--- a/paddle/gserver/layers/CRFLayer.cpp
+++ b/paddle/gserver/layers/CRFLayer.cpp
--- a/paddle/gserver/layers/ClipLayer.cpp
+++ b/paddle/gserver/layers/ClipLayer.cpp
--- a/paddle/gserver/layers/Conv3DLayer.cpp
+++ b/paddle/gserver/layers/Conv3DLayer.cpp
--- a/paddle/gserver/layers/Conv3DLayer.h
+++ b/paddle/gserver/layers/Conv3DLayer.h
--- a/paddle/gserver/layers/ConvBaseLayer.cpp
+++ b/paddle/gserver/layers/ConvBaseLayer.cpp
--- a/paddle/gserver/layers/ConvBaseLayer.h
+++ b/paddle/gserver/layers/ConvBaseLayer.h
--- a/paddle/gserver/layers/ConvBaseOperator.cpp
+++ b/paddle/gserver/layers/ConvBaseOperator.cpp
--- a/paddle/gserver/layers/ConvBaseProjection.cpp
+++ b/paddle/gserver/layers/ConvBaseProjection.cpp
--- a/paddle/gserver/layers/ConvBaseProjection.h
+++ b/paddle/gserver/layers/ConvBaseProjection.h
--- a/paddle/gserver/layers/ConvProjection.cpp
+++ b/paddle/gserver/layers/ConvProjection.cpp
--- a/paddle/gserver/layers/ConvTransProjection.cpp
+++ b/paddle/gserver/layers/ConvTransProjection.cpp
--- a/paddle/gserver/layers/CostLayer.cpp
+++ b/paddle/gserver/layers/CostLayer.cpp
--- a/paddle/gserver/layers/CostLayer.h
+++ b/paddle/gserver/layers/CostLayer.h
--- a/paddle/gserver/layers/CropLayer.cpp
+++ b/paddle/gserver/layers/CropLayer.cpp
--- a/paddle/gserver/layers/CropLayer.h
+++ b/paddle/gserver/layers/CropLayer.h
--- a/paddle/gserver/layers/CrossChannelNormLayer.cpp
+++ b/paddle/gserver/layers/CrossChannelNormLayer.cpp
--- a/paddle/gserver/layers/CrossEntropyOverBeam.cpp
+++ b/paddle/gserver/layers/CrossEntropyOverBeam.cpp
--- a/paddle/gserver/layers/CrossEntropyOverBeam.h
+++ b/paddle/gserver/layers/CrossEntropyOverBeam.h
--- a/paddle/gserver/layers/CudnnBatchNormLayer.cpp
+++ b/paddle/gserver/layers/CudnnBatchNormLayer.cpp
--- a/paddle/gserver/layers/CudnnBatchNormLayer.h
+++ b/paddle/gserver/layers/CudnnBatchNormLayer.h
--- a/paddle/gserver/layers/CudnnConvBaseLayer.cpp
+++ b/paddle/gserver/layers/CudnnConvBaseLayer.cpp
--- a/paddle/gserver/layers/CudnnPoolLayer.cpp
+++ b/paddle/gserver/layers/CudnnPoolLayer.cpp
--- a/paddle/gserver/layers/DeConv3DLayer.cpp
+++ b/paddle/gserver/layers/DeConv3DLayer.cpp
--- a/paddle/gserver/layers/DeConv3DLayer.h
+++ b/paddle/gserver/layers/DeConv3DLayer.h
--- a/paddle/gserver/layers/DetectionOutputLayer.cpp
+++ b/paddle/gserver/layers/DetectionOutputLayer.cpp
--- a/paddle/gserver/layers/DetectionOutputLayer.h
+++ b/paddle/gserver/layers/DetectionOutputLayer.h
--- a/paddle/gserver/layers/DetectionUtil.cpp
+++ b/paddle/gserver/layers/DetectionUtil.cpp
--- a/paddle/gserver/layers/DetectionUtil.h
+++ b/paddle/gserver/layers/DetectionUtil.h
--- a/paddle/gserver/layers/DotProdLayer.cpp
+++ b/paddle/gserver/layers/DotProdLayer.cpp
--- a/paddle/gserver/layers/ExpandConvBaseLayer.cpp
+++ b/paddle/gserver/layers/ExpandConvBaseLayer.cpp
--- a/paddle/gserver/layers/ExpandConvBaseLayer.h
+++ b/paddle/gserver/layers/ExpandConvBaseLayer.h
--- a/paddle/gserver/layers/ExpandConvLayer.cpp
+++ b/paddle/gserver/layers/ExpandConvLayer.cpp
--- a/paddle/gserver/layers/ExpandConvLayer.h
+++ b/paddle/gserver/layers/ExpandConvLayer.h
--- a/paddle/gserver/layers/ExpandConvTransLayer.cpp
+++ b/paddle/gserver/layers/ExpandConvTransLayer.cpp
--- a/paddle/gserver/layers/ExpandConvTransLayer.h
+++ b/paddle/gserver/layers/ExpandConvTransLayer.h
--- a/paddle/gserver/layers/FactorizationMachineLayer.cpp
+++ b/paddle/gserver/layers/FactorizationMachineLayer.cpp
--- a/paddle/gserver/layers/FactorizationMachineLayer.h
+++ b/paddle/gserver/layers/FactorizationMachineLayer.h
--- a/paddle/gserver/layers/FeatureMapExpandLayer.cpp
+++ b/paddle/gserver/layers/FeatureMapExpandLayer.cpp
--- a/paddle/gserver/layers/GruCompute.cpp
+++ b/paddle/gserver/layers/GruCompute.cpp
--- a/paddle/gserver/layers/GruCompute.cu
+++ b/paddle/gserver/layers/GruCompute.cu
--- a/paddle/gserver/layers/HierarchicalSigmoidLayer.cpp
+++ b/paddle/gserver/layers/HierarchicalSigmoidLayer.cpp
--- a/paddle/gserver/layers/HierarchicalSigmoidLayer.h
+++ b/paddle/gserver/layers/HierarchicalSigmoidLayer.h
--- a/paddle/gserver/layers/KmaxSeqScoreLayer.cpp
+++ b/paddle/gserver/layers/KmaxSeqScoreLayer.cpp
--- a/paddle/gserver/layers/L2DistanceLayer.cpp
+++ b/paddle/gserver/layers/L2DistanceLayer.cpp
--- a/paddle/gserver/layers/L2DistanceLayer.h
+++ b/paddle/gserver/layers/L2DistanceLayer.h
--- a/paddle/gserver/layers/Layer.cpp
+++ b/paddle/gserver/layers/Layer.cpp
--- a/paddle/gserver/layers/Layer.h
+++ b/paddle/gserver/layers/Layer.h
--- a/paddle/gserver/layers/LinearChainCRF.cpp
+++ b/paddle/gserver/layers/LinearChainCRF.cpp
--- a/paddle/gserver/layers/LstmCompute.cu
+++ b/paddle/gserver/layers/LstmCompute.cu
--- a/paddle/gserver/layers/MKLDNNAddtoLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNAddtoLayer.cpp
--- a/paddle/gserver/layers/MKLDNNAddtoLayer.h
+++ b/paddle/gserver/layers/MKLDNNAddtoLayer.h
--- a/paddle/gserver/layers/MKLDNNBase.h
+++ b/paddle/gserver/layers/MKLDNNBase.h
--- a/paddle/gserver/layers/MKLDNNBatchNormLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNBatchNormLayer.cpp
--- a/paddle/gserver/layers/MKLDNNBatchNormLayer.h
+++ b/paddle/gserver/layers/MKLDNNBatchNormLayer.h
--- a/paddle/gserver/layers/MKLDNNConcatLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNConcatLayer.cpp
--- a/paddle/gserver/layers/MKLDNNConcatLayer.h
+++ b/paddle/gserver/layers/MKLDNNConcatLayer.h
--- a/paddle/gserver/layers/MKLDNNConvLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNConvLayer.cpp
--- a/paddle/gserver/layers/MKLDNNConvLayer.h
+++ b/paddle/gserver/layers/MKLDNNConvLayer.h
--- a/paddle/gserver/layers/MKLDNNFcLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNFcLayer.cpp
--- a/paddle/gserver/layers/MKLDNNFcLayer.h
+++ b/paddle/gserver/layers/MKLDNNFcLayer.h
--- a/paddle/gserver/layers/MKLDNNLRNLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNLRNLayer.cpp
--- a/paddle/gserver/layers/MKLDNNLRNLayer.h
+++ b/paddle/gserver/layers/MKLDNNLRNLayer.h
--- a/paddle/gserver/layers/MKLDNNLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNLayer.cpp
--- a/paddle/gserver/layers/MKLDNNLayer.h
+++ b/paddle/gserver/layers/MKLDNNLayer.h
--- a/paddle/gserver/layers/MKLDNNPoolLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNPoolLayer.cpp
--- a/paddle/gserver/layers/MKLDNNPoolLayer.h
+++ b/paddle/gserver/layers/MKLDNNPoolLayer.h
--- a/paddle/gserver/layers/MaxLayer.h
+++ b/paddle/gserver/layers/MaxLayer.h
--- a/paddle/gserver/layers/MaxPoolWithMaskLayer.cpp
+++ b/paddle/gserver/layers/MaxPoolWithMaskLayer.cpp
--- a/paddle/gserver/layers/MaxPoolWithMaskLayer.h
+++ b/paddle/gserver/layers/MaxPoolWithMaskLayer.h
--- a/paddle/gserver/layers/MultiBoxLossLayer.cpp
+++ b/paddle/gserver/layers/MultiBoxLossLayer.cpp
--- a/paddle/gserver/layers/MultiBoxLossLayer.h
+++ b/paddle/gserver/layers/MultiBoxLossLayer.h
--- a/paddle/gserver/layers/NormLayer.cpp
+++ b/paddle/gserver/layers/NormLayer.cpp
--- a/paddle/gserver/layers/Pool3DLayer.cpp
+++ b/paddle/gserver/layers/Pool3DLayer.cpp
--- a/paddle/gserver/layers/Pool3DLayer.h
+++ b/paddle/gserver/layers/Pool3DLayer.h
--- a/paddle/gserver/layers/PoolLayer.cpp
+++ b/paddle/gserver/layers/PoolLayer.cpp
--- a/paddle/gserver/layers/PoolLayer.h
+++ b/paddle/gserver/layers/PoolLayer.h
--- a/paddle/gserver/layers/PoolProjection.cpp
+++ b/paddle/gserver/layers/PoolProjection.cpp
--- a/paddle/gserver/layers/PoolProjection.h
+++ b/paddle/gserver/layers/PoolProjection.h
--- a/paddle/gserver/layers/PrintLayer.cpp
+++ b/paddle/gserver/layers/PrintLayer.cpp
--- a/paddle/gserver/layers/ROIPoolLayer.cpp
+++ b/paddle/gserver/layers/ROIPoolLayer.cpp
--- a/paddle/gserver/layers/ROIPoolLayer.h
+++ b/paddle/gserver/layers/ROIPoolLayer.h
--- a/paddle/gserver/layers/RotateLayer.h
+++ b/paddle/gserver/layers/RotateLayer.h
--- a/paddle/gserver/layers/RowConvLayer.cpp
+++ b/paddle/gserver/layers/RowConvLayer.cpp
--- a/paddle/gserver/layers/RowConvLayer.h
+++ b/paddle/gserver/layers/RowConvLayer.h
--- a/paddle/gserver/layers/RowL2NormLayer.cpp
+++ b/paddle/gserver/layers/RowL2NormLayer.cpp
--- a/paddle/gserver/layers/ScaleShiftLayer.cpp
+++ b/paddle/gserver/layers/ScaleShiftLayer.cpp
--- a/paddle/gserver/layers/ScaleSubRegionLayer.cpp
+++ b/paddle/gserver/layers/ScaleSubRegionLayer.cpp
--- a/paddle/gserver/layers/ScaleSubRegionLayer.h
+++ b/paddle/gserver/layers/ScaleSubRegionLayer.h
--- a/paddle/gserver/layers/SequenceLastInstanceLayer.cpp
+++ b/paddle/gserver/layers/SequenceLastInstanceLayer.cpp
--- a/paddle/gserver/layers/SequencePoolLayer.cpp
+++ b/paddle/gserver/layers/SequencePoolLayer.cpp
--- a/paddle/gserver/layers/SequencePoolLayer.h
+++ b/paddle/gserver/layers/SequencePoolLayer.h
--- a/paddle/gserver/layers/SequenceReshapeLayer.cpp
+++ b/paddle/gserver/layers/SequenceReshapeLayer.cpp
--- a/paddle/gserver/layers/SequenceSliceLayer.cpp
+++ b/paddle/gserver/layers/SequenceSliceLayer.cpp
--- a/paddle/gserver/layers/SliceProjection.cpp
+++ b/paddle/gserver/layers/SliceProjection.cpp
--- a/paddle/gserver/layers/SubNestedSequenceLayer.cpp
+++ b/paddle/gserver/layers/SubNestedSequenceLayer.cpp
--- a/paddle/gserver/layers/SubSequenceLayer.cpp
+++ b/paddle/gserver/layers/SubSequenceLayer.cpp
--- a/paddle/gserver/layers/SwitchOrderLayer.cpp
+++ b/paddle/gserver/layers/SwitchOrderLayer.cpp
--- a/paddle/gserver/layers/SwitchOrderLayer.h
+++ b/paddle/gserver/layers/SwitchOrderLayer.h
--- a/paddle/gserver/layers/TransLayer.cpp
+++ b/paddle/gserver/layers/TransLayer.cpp
--- a/paddle/gserver/tests/CMakeLists.txt
+++ b/paddle/gserver/tests/CMakeLists.txt
--- a/paddle/gserver/tests/LayerGradUtil.cpp
+++ b/paddle/gserver/tests/LayerGradUtil.cpp
--- a/paddle/gserver/tests/LayerGradUtil.h
+++ b/paddle/gserver/tests/LayerGradUtil.h
--- a/paddle/gserver/tests/MKLDNNTester.cpp
+++ b/paddle/gserver/tests/MKLDNNTester.cpp
--- a/paddle/gserver/tests/MKLDNNTester.h
+++ b/paddle/gserver/tests/MKLDNNTester.h
--- a/paddle/gserver/tests/concat_slice_a.conf
+++ b/paddle/gserver/tests/concat_slice_a.conf
--- a/paddle/gserver/tests/concat_slice_b.conf
+++ b/paddle/gserver/tests/concat_slice_b.conf
--- a/paddle/gserver/tests/img_conv_cudnn.py
+++ b/paddle/gserver/tests/img_conv_cudnn.py
--- a/paddle/gserver/tests/img_conv_exconv.py
+++ b/paddle/gserver/tests/img_conv_exconv.py
--- a/paddle/gserver/tests/mkldnn_branch_net.conf
+++ b/paddle/gserver/tests/mkldnn_branch_net.conf
--- a/paddle/gserver/tests/mkldnn_simple_net.conf
+++ b/paddle/gserver/tests/mkldnn_simple_net.conf
--- a/paddle/gserver/tests/proto_files.txt
+++ b/paddle/gserver/tests/proto_files.txt
--- a/paddle/gserver/tests/proto_files_compressed.txt
+++ b/paddle/gserver/tests/proto_files_compressed.txt
--- a/paddle/gserver/tests/rnn_data_provider.py
+++ b/paddle/gserver/tests/rnn_data_provider.py
--- a/paddle/gserver/tests/sequence_lstm.conf
+++ b/paddle/gserver/tests/sequence_lstm.conf
--- a/paddle/gserver/tests/sequence_nest_layer_group.conf
+++ b/paddle/gserver/tests/sequence_nest_layer_group.conf
--- a/paddle/gserver/tests/sequence_nest_rnn_multi_input.conf
+++ b/paddle/gserver/tests/sequence_nest_rnn_multi_input.conf
--- a/paddle/gserver/tests/sequence_recurrent.py
+++ b/paddle/gserver/tests/sequence_recurrent.py
--- a/paddle/gserver/tests/sequence_recurrent_group.py
+++ b/paddle/gserver/tests/sequence_recurrent_group.py
--- a/paddle/gserver/tests/sequence_rnn_matched_inputs.py
+++ b/paddle/gserver/tests/sequence_rnn_matched_inputs.py
--- a/paddle/gserver/tests/sequence_rnn_mixed_inputs.py
+++ b/paddle/gserver/tests/sequence_rnn_mixed_inputs.py
--- a/paddle/gserver/tests/sequence_rnn_multi_input.conf
+++ b/paddle/gserver/tests/sequence_rnn_multi_input.conf
--- a/paddle/gserver/tests/test_ActivationGrad.cpp
+++ b/paddle/gserver/tests/test_ActivationGrad.cpp
--- a/paddle/gserver/tests/test_BatchNorm.cpp
+++ b/paddle/gserver/tests/test_BatchNorm.cpp
--- a/paddle/gserver/tests/test_CRFLayerGrad.cpp
+++ b/paddle/gserver/tests/test_CRFLayerGrad.cpp
--- a/paddle/gserver/tests/test_CompareSparse.cpp
+++ b/paddle/gserver/tests/test_CompareSparse.cpp
--- a/paddle/gserver/tests/test_CompareTwoNets.cpp
+++ b/paddle/gserver/tests/test_CompareTwoNets.cpp
--- a/paddle/gserver/tests/test_ConvTrans.cpp
+++ b/paddle/gserver/tests/test_ConvTrans.cpp
--- a/paddle/gserver/tests/test_ConvUnify.cpp
+++ b/paddle/gserver/tests/test_ConvUnify.cpp
--- a/paddle/gserver/tests/test_CrossEntropyOverBeamGrad.cpp
+++ b/paddle/gserver/tests/test_CrossEntropyOverBeamGrad.cpp
--- a/paddle/gserver/tests/test_DetectionOutput.cpp
+++ b/paddle/gserver/tests/test_DetectionOutput.cpp
--- a/paddle/gserver/tests/test_Evaluator.cpp
+++ b/paddle/gserver/tests/test_Evaluator.cpp
--- a/paddle/gserver/tests/test_Expand.cpp
+++ b/paddle/gserver/tests/test_Expand.cpp
--- a/paddle/gserver/tests/test_KmaxSeqScore.cpp
+++ b/paddle/gserver/tests/test_KmaxSeqScore.cpp
--- a/paddle/gserver/tests/test_LayerGrad.cpp
+++ b/paddle/gserver/tests/test_LayerGrad.cpp
--- a/paddle/gserver/tests/test_MKLDNN.cpp
+++ b/paddle/gserver/tests/test_MKLDNN.cpp
--- a/paddle/gserver/tests/test_MaxPoolingWithMaskOutput.cpp
+++ b/paddle/gserver/tests/test_MaxPoolingWithMaskOutput.cpp
--- a/paddle/gserver/tests/test_NetworkCompare.cpp
+++ b/paddle/gserver/tests/test_NetworkCompare.cpp
--- a/paddle/gserver/tests/test_PriorBox.cpp
+++ b/paddle/gserver/tests/test_PriorBox.cpp
--- a/paddle/gserver/tests/test_ProtoDataProvider.cpp
+++ b/paddle/gserver/tests/test_ProtoDataProvider.cpp
--- a/paddle/gserver/tests/test_PyDataProvider.cpp
+++ b/paddle/gserver/tests/test_PyDataProvider.cpp
--- a/paddle/gserver/tests/test_PyDataProvider2.py
+++ b/paddle/gserver/tests/test_PyDataProvider2.py
--- a/paddle/gserver/tests/test_RecurrentGradientMachine.cpp
+++ b/paddle/gserver/tests/test_RecurrentGradientMachine.cpp
--- a/paddle/gserver/tests/test_SelectiveFCLayer.cpp
+++ b/paddle/gserver/tests/test_SelectiveFCLayer.cpp
--- a/paddle/gserver/tests/test_SeqSliceLayerGrad.cpp
+++ b/paddle/gserver/tests/test_SeqSliceLayerGrad.cpp
--- a/paddle/gserver/tests/test_WarpCTCLayer.cpp
+++ b/paddle/gserver/tests/test_WarpCTCLayer.cpp
--- a/paddle/math/Allocator.h
+++ b/paddle/math/Allocator.h
--- a/paddle/math/BaseMatrix.cu
+++ b/paddle/math/BaseMatrix.cu
--- a/paddle/math/BaseMatrix.h
+++ b/paddle/math/BaseMatrix.h
--- a/paddle/math/CMakeLists.txt
+++ b/paddle/math/CMakeLists.txt
--- a/paddle/math/CpuSparseMatrix.cpp
+++ b/paddle/math/CpuSparseMatrix.cpp
--- a/paddle/math/CpuSparseMatrix.h
+++ b/paddle/math/CpuSparseMatrix.h
--- a/paddle/math/MKLDNNMatrix.cpp
+++ b/paddle/math/MKLDNNMatrix.cpp
--- a/paddle/math/MKLDNNMatrix.h
+++ b/paddle/math/MKLDNNMatrix.h
--- a/paddle/math/MathFunctions.cpp
+++ b/paddle/math/MathFunctions.cpp
--- a/paddle/math/MathFunctions.h
+++ b/paddle/math/MathFunctions.h
--- a/paddle/math/MathUtils.cpp
+++ b/paddle/math/MathUtils.cpp
--- a/paddle/math/Matrix.cpp
+++ b/paddle/math/Matrix.cpp
--- a/paddle/math/Matrix.h
+++ b/paddle/math/Matrix.h
--- a/paddle/math/NEONFunctions.cpp
+++ b/paddle/math/NEONFunctions.cpp
--- a/paddle/math/NEONFunctions.h
+++ b/paddle/math/NEONFunctions.h
--- a/paddle/math/RowBuffer.h
+++ b/paddle/math/RowBuffer.h
--- a/paddle/math/SIMDFunctions.h
+++ b/paddle/math/SIMDFunctions.h
--- a/paddle/math/SparseMatrix.cpp
+++ b/paddle/math/SparseMatrix.cpp
--- a/paddle/math/SparseMatrix.h
+++ b/paddle/math/SparseMatrix.h
--- a/paddle/math/SparseRowMatrix.cpp
+++ b/paddle/math/SparseRowMatrix.cpp
--- a/paddle/math/SparseRowMatrix.h
+++ b/paddle/math/SparseRowMatrix.h
--- a/paddle/math/Storage.cpp
+++ b/paddle/math/Storage.cpp
--- a/paddle/math/TensorEvaluate.h
+++ b/paddle/math/TensorEvaluate.h
--- a/paddle/math/TrainingAlgorithmOp.cu
+++ b/paddle/math/TrainingAlgorithmOp.cu
--- a/paddle/math/Vector.cpp
+++ b/paddle/math/Vector.cpp
--- a/paddle/math/Vector.h
+++ b/paddle/math/Vector.h
--- a/paddle/math/float16.h
+++ b/paddle/math/float16.h
--- a/paddle/math/tests/CMakeLists.txt
+++ b/paddle/math/tests/CMakeLists.txt
--- a/paddle/math/tests/TensorCheck.h
+++ b/paddle/math/tests/TensorCheck.h
--- a/paddle/math/tests/TestUtils.h
+++ b/paddle/math/tests/TestUtils.h
--- a/paddle/math/tests/test_Allocator.cpp
+++ b/paddle/math/tests/test_Allocator.cpp
--- a/paddle/math/tests/test_BaseMatrix.cpp
+++ b/paddle/math/tests/test_BaseMatrix.cpp
--- a/paddle/math/tests/test_CpuGpuVector.cpp
+++ b/paddle/math/tests/test_CpuGpuVector.cpp
--- a/paddle/math/tests/test_ExecViaCpu.cpp
+++ b/paddle/math/tests/test_ExecViaCpu.cpp
--- a/paddle/math/tests/test_GpuProfiler.cpp
+++ b/paddle/math/tests/test_GpuProfiler.cpp
--- a/paddle/math/tests/test_Matrix.cpp
+++ b/paddle/math/tests/test_Matrix.cpp
--- a/paddle/math/tests/test_SparseMatrix.cpp
+++ b/paddle/math/tests/test_SparseMatrix.cpp
--- a/paddle/math/tests/test_Tensor.cu
+++ b/paddle/math/tests/test_Tensor.cu
--- a/paddle/math/tests/test_TrainingAlgorithm.cpp
+++ b/paddle/math/tests/test_TrainingAlgorithm.cpp
--- a/paddle/math/tests/test_batchTranspose.cpp
+++ b/paddle/math/tests/test_batchTranspose.cpp
--- a/paddle/math/tests/test_float16.cpp
+++ b/paddle/math/tests/test_float16.cpp
--- a/paddle/math/tests/test_float16.cu
+++ b/paddle/math/tests/test_float16.cu
--- a/paddle/math/tests/test_lazyAssign.cu
+++ b/paddle/math/tests/test_lazyAssign.cu
--- a/paddle/math/tests/test_matrixCompare.cpp
+++ b/paddle/math/tests/test_matrixCompare.cpp
--- a/paddle/math/tests/test_perturbation.cpp
+++ b/paddle/math/tests/test_perturbation.cpp
--- a/paddle/math/tests/test_sparseMatrixCompare.cpp
+++ b/paddle/math/tests/test_sparseMatrixCompare.cpp
--- a/paddle/memory/.clang-format
+++ b/paddle/memory/.clang-format
--- a/paddle/memory/CMakeLists.txt
+++ b/paddle/memory/CMakeLists.txt
--- a/paddle/memory/README.md
+++ b/paddle/memory/README.md
--- a/paddle/memory/detail/CMakeLists.txt
+++ b/paddle/memory/detail/CMakeLists.txt
--- a/paddle/memory/detail/buddy_allocator.cc
+++ b/paddle/memory/detail/buddy_allocator.cc
--- a/paddle/memory/detail/buddy_allocator.h
+++ b/paddle/memory/detail/buddy_allocator.h
--- a/paddle/memory/detail/memory_block.cc
+++ b/paddle/memory/detail/memory_block.cc
--- a/paddle/memory/detail/memory_block.h
+++ b/paddle/memory/detail/memory_block.h
--- a/paddle/memory/detail/meta_cache.cc
+++ b/paddle/memory/detail/meta_cache.cc
--- a/paddle/memory/detail/meta_cache.h
+++ b/paddle/memory/detail/meta_cache.h
--- a/paddle/memory/detail/meta_data.cc
+++ b/paddle/memory/detail/meta_data.cc
--- a/paddle/memory/detail/meta_data.h
+++ b/paddle/memory/detail/meta_data.h
--- a/paddle/memory/detail/system_allocator.cc
+++ b/paddle/memory/detail/system_allocator.cc
--- a/paddle/memory/detail/system_allocator.h
+++ b/paddle/memory/detail/system_allocator.h
--- a/paddle/memory/detail/system_allocator_test.cc
+++ b/paddle/memory/detail/system_allocator_test.cc
--- a/paddle/memory/memcpy.cc
+++ b/paddle/memory/memcpy.cc
--- a/paddle/memory/memcpy.h
+++ b/paddle/memory/memcpy.h
--- a/paddle/memory/memory.cc
+++ b/paddle/memory/memory.cc
--- a/paddle/memory/memory.h
+++ b/paddle/memory/memory.h
--- a/paddle/memory/memory_test.cc
+++ b/paddle/memory/memory_test.cc
--- a/paddle/operators/.clang-format
+++ b/paddle/operators/.clang-format
--- a/paddle/operators/CMakeLists.txt
+++ b/paddle/operators/CMakeLists.txt
--- a/paddle/operators/accuracy_op.cc
+++ b/paddle/operators/accuracy_op.cc
--- a/paddle/operators/accuracy_op.cu
+++ b/paddle/operators/accuracy_op.cu
--- a/paddle/operators/accuracy_op.h
+++ b/paddle/operators/accuracy_op.h
--- a/paddle/operators/activation_op.cc
+++ b/paddle/operators/activation_op.cc
--- a/paddle/operators/activation_op.cu
+++ b/paddle/operators/activation_op.cu
--- a/paddle/operators/activation_op.h
+++ b/paddle/operators/activation_op.h
--- a/paddle/operators/adadelta_op.cc
+++ b/paddle/operators/adadelta_op.cc
--- a/paddle/operators/adadelta_op.cu
+++ b/paddle/operators/adadelta_op.cu
--- a/paddle/operators/adadelta_op.h
+++ b/paddle/operators/adadelta_op.h
--- a/paddle/operators/adagrad_op.cc
+++ b/paddle/operators/adagrad_op.cc
--- a/paddle/operators/adagrad_op.cu
+++ b/paddle/operators/adagrad_op.cu
--- a/paddle/operators/adagrad_op.h
+++ b/paddle/operators/adagrad_op.h
--- a/paddle/operators/adam_op.cc
+++ b/paddle/operators/adam_op.cc
--- a/paddle/operators/adam_op.cu
+++ b/paddle/operators/adam_op.cu
--- a/paddle/operators/adam_op.h
+++ b/paddle/operators/adam_op.h
--- a/paddle/operators/adamax_op.cc
+++ b/paddle/operators/adamax_op.cc
--- a/paddle/operators/adamax_op.cu
+++ b/paddle/operators/adamax_op.cu
--- a/paddle/operators/adamax_op.h
+++ b/paddle/operators/adamax_op.h