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

.pre-commit-config.yaml

@@ -3,8 +3,8 @@
     hooks:
     -   id: remove-crlf
         files: (?!.*third_party)^.*$ | (?!.*book)^.*$
--   repo: https://github.com/reyoung/mirrors-yapf.git
-    sha: v0.13.2
+-   repo: https://github.com/PaddlePaddle/mirrors-yapf.git
+    sha: 0d79c0c469bab64f7229c9aca2b1186ef47f0e37
     hooks:
     -   id: yapf
         files: (.*\.(py|bzl)|BUILD|.*\.BUILD|WORKSPACE)$

README.md

@@ -70,7 +70,7 @@ before looking into the
 We provide [English](http://www.paddlepaddle.org/develop/doc/) and
 [Chinese](http://www.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.
 

cmake/cudnn.cmake

@@ -11,11 +11,23 @@ find_path(CUDNN_INCLUDE_DIR cudnn.h
 
 get_filename_component(__libpath_hist ${CUDA_CUDART_LIBRARY} PATH)
 
+if(NOT ${CMAKE_HOST_SYSTEM_PROCESSOR})
+    execute_process(
+        COMMAND uname -m COMMAND tr -d '\n'
+        OUTPUT_VARIABLE HOST_ARCH
+        RESULT_VARIABLE UNAME_RESULT)
+    if(${UNAME_RESULT})
+        set(HOST_ARCH "x86_64")
+    endif(${UNAME_RESULT})
+else(NOT ${CMAKE_HOST_SYSTEM_PROCESSOR})
+    set(HOST_ARCH ${CMAKE_HOST_SYSTEM_PROCESSOR})
+endif(NOT ${CMAKE_HOST_SYSTEM_PROCESSOR})
+
 list(APPEND CUDNN_CHECK_LIBRARY_DIRS
     ${CUDNN_ROOT}
     ${CUDNN_ROOT}/lib64
     ${CUDNN_ROOT}/lib
-    ${CUDNN_ROOT}/lib/x86_64-linux-gnu
+    ${CUDNN_ROOT}/lib/${HOST_ARCH}-linux-gnu
     $ENV{CUDNN_ROOT}
     $ENV{CUDNN_ROOT}/lib64
     $ENV{CUDNN_ROOT}/lib

cmake/external/any.cmake

@@ -18,3 +18,4 @@ ExternalProject_Add(
 )
 
 add_definitions(-DANY_IMPL_ANY_CAST_MOVEABLE)
+LIST(APPEND external_project_dependencies linb_any)

cmake/external/gflags.cmake

@@ -26,7 +26,7 @@ ENDIF(WIN32)
 INCLUDE_DIRECTORIES(${GFLAGS_INCLUDE_DIR})
 
 ExternalProject_Add(
-    gflags
+    extern_gflags
     ${EXTERNAL_PROJECT_LOG_ARGS}
     GIT_REPOSITORY  "https://github.com/gflags/gflags.git"
     PREFIX          ${GFLAGS_SOURCES_DIR}
@@ -44,4 +44,8 @@ ExternalProject_Add(
                      -DCMAKE_BUILD_TYPE:STRING=Release
 )
 
+ADD_LIBRARY(gflags STATIC IMPORTED)
+SET_PROPERTY(TARGET gflags PROPERTY IMPORTED_LOCATION ${GFLAGS_LIBRARIES})
+ADD_DEPENDENCIES(gflags extern_gflags)
+
 LIST(APPEND external_project_dependencies gflags)

cmake/external/glog.cmake

@@ -27,7 +27,7 @@ ENDIF(WIN32)
 INCLUDE_DIRECTORIES(${GLOG_INCLUDE_DIR})
 
 ExternalProject_Add(
-    glog
+    extern_glog
     ${EXTERNAL_PROJECT_LOG_ARGS}
     DEPENDS gflags
     GIT_REPOSITORY  "https://github.com/google/glog.git"
@@ -48,4 +48,8 @@ ExternalProject_Add(
                      -DCMAKE_BUILD_TYPE:STRING=Release
 )
 
+ADD_LIBRARY(glog STATIC IMPORTED)
+SET_PROPERTY(TARGET glog PROPERTY IMPORTED_LOCATION ${GLOG_LIBRARIES})
+ADD_DEPENDENCIES(glog extern_glog)
+
 LIST(APPEND external_project_dependencies glog)

cmake/external/gtest.cmake

@@ -35,7 +35,7 @@ IF(WITH_TESTING)
     ENDIF(WIN32)
 
     ExternalProject_Add(
-        gtest
+        extern_gtest
         ${EXTERNAL_PROJECT_LOG_ARGS}
         GIT_REPOSITORY  "https://github.com/google/googletest.git"
         GIT_TAG         "release-1.8.0"
@@ -55,5 +55,14 @@ IF(WITH_TESTING)
                          -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
                          -DCMAKE_BUILD_TYPE:STRING=Release
     )
-    LIST(APPEND external_project_dependencies gtest)
+
+    ADD_LIBRARY(gtest STATIC IMPORTED)
+    SET_PROPERTY(TARGET gtest PROPERTY IMPORTED_LOCATION ${GTEST_LIBRARIES})
+    ADD_DEPENDENCIES(gtest extern_gtest)
+
+    ADD_LIBRARY(gtest_main STATIC IMPORTED)
+    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)

cmake/external/openblas.cmake

@@ -37,11 +37,11 @@ IF(NOT ${CBLAS_FOUND})
         SET(OPTIONAL_ARGS HOSTCC=${HOST_C_COMPILER} TARGET=ARMV7 USE_THREAD=0 libs)
     ELSE()
         SET(OPENBLAS_COMMIT "v0.2.19")
-        SET(OPENBLAS_ARGS DYNAMIC_ARCH=1 libs)
+        SET(OPTIONAL_ARGS DYNAMIC_ARCH=1 libs NUM_THREADS=64)
     ENDIF()
 
     ExternalProject_Add(
-        openblas
+        extern_openblas
         ${EXTERNAL_PROJECT_LOG_ARGS}
         GIT_REPOSITORY      https://github.com/xianyi/OpenBLAS.git
         GIT_TAG             ${OPENBLAS_COMMIT}
@@ -53,8 +53,14 @@ IF(NOT ${CBLAS_FOUND})
         UPDATE_COMMAND      ""
         CONFIGURE_COMMAND   ""
     )
-    LIST(APPEND external_project_dependencies openblas)
 ENDIF(NOT ${CBLAS_FOUND})
 
 MESSAGE(STATUS "BLAS library: ${CBLAS_LIBRARIES}")
 INCLUDE_DIRECTORIES(${CBLAS_INC_DIR})
+
+ADD_LIBRARY(cblas STATIC IMPORTED)
+SET_PROPERTY(TARGET cblas PROPERTY IMPORTED_LOCATION ${CBLAS_LIBRARIES})
+IF(NOT ${CBLAS_FOUND})
+    ADD_DEPENDENCIES(cblas extern_openblas)
+    LIST(APPEND external_project_dependencies cblas)
+ENDIF(NOT ${CBLAS_FOUND})

cmake/external/protobuf.cmake

@@ -17,9 +17,14 @@ INCLUDE(ExternalProject)
 macro(PROMPT_PROTOBUF_LIB)
     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})
     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}")
+endmacro()
 
 set(PROTOBUF_ROOT "" CACHE PATH "Folder contains protobuf")
 if (NOT "${PROTOBUF_ROOT}" STREQUAL "")
@@ -30,6 +35,7 @@ if (NOT "${PROTOBUF_ROOT}" STREQUAL "")
     find_program(PROTOBUF_PROTOC_EXECUTABLE protoc PATHS ${PROTOBUF_ROOT}/bin)
     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}.")
@@ -100,8 +106,7 @@ IF(NOT CMAKE_CROSSCOMPILING)
     FIND_PACKAGE(Protobuf ${PROTOBUF_VERSION})
 
     IF(PROTOBUF_FOUND)
-        EXEC_PROGRAM(${PROTOBUF_PROTOC_EXECUTABLE} ARGS --version OUTPUT_VARIABLE PROTOBUF_VERSION)
-        STRING(REGEX MATCH "[0-9]+.[0-9]+" PROTOBUF_VERSION "${PROTOBUF_VERSION}")
+        SET_PROTOBUF_VERSION()
         IF("${PROTOBUF_VERSION}" VERSION_LESS "3.1.0")
             SET(PROTOBUF_FOUND OFF)
         ENDIF()

cmake/external/warpctc.cmake

@@ -43,7 +43,7 @@ ELSE()
 ENDIF()
 
 ExternalProject_Add(
-    warpctc
+    extern_warpctc
     ${EXTERNAL_PROJECT_LOG_ARGS}
     GIT_REPOSITORY  "https://github.com/gangliao/warp-ctc.git"
     PREFIX          ${WARPCTC_SOURCES_DIR}
@@ -65,4 +65,8 @@ ExternalProject_Add(
                      -DCMAKE_INSTALL_PREFIX:PATH=${WARPCTC_INSTALL_DIR}
 )
 
+ADD_LIBRARY(warpctc STATIC IMPORTED)
+SET_PROPERTY(TARGET warpctc PROPERTY IMPORTED_LOCATION ${WARPCTC_LIBRARIES})
+ADD_DEPENDENCIES(warpctc extern_warpctc)
+
 LIST(APPEND external_project_dependencies warpctc)

cmake/generic.cmake

@@ -29,6 +29,11 @@
 # https://cmake.org/cmake/help/v3.0/module/CMakeParseArguments.html
 #
 
+if(NOT APPLE)
+    find_package(Threads REQUIRED)
+    link_libraries(${CMAKE_THREAD_LIBS_INIT})
+endif(NOT APPLE)
+
 # cc_library parses tensor.cc and figures out that target also depend on tensor.h.
 # cc_library(tensor
 #   SRCS
@@ -45,7 +50,9 @@ function(cc_library TARGET_NAME)
   else()
     add_library(${TARGET_NAME} STATIC ${cc_library_SRCS})
   endif()
-  add_dependencies(${TARGET_NAME} ${cc_library_DEPS} ${external_project_dependencies})
+  if (cc_library_DEPS)
+    add_dependencies(${TARGET_NAME} ${cc_library_DEPS})
+  endif()
 endfunction(cc_library)
 
 # cc_binary parses tensor.cc and figures out that target also depend on tensor.h.
@@ -58,7 +65,6 @@ function(cc_binary TARGET_NAME)
   set(multiValueArgs SRCS DEPS)
   cmake_parse_arguments(cc_binary "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
   add_executable(${TARGET_NAME} ${cc_binary_SRCS})
-  link_paddle_exe(${TARGET_NAME})
   if(cc_binary_DEPS)
     target_link_libraries(${TARGET_NAME} ${cc_binary_DEPS})
     add_dependencies(${TARGET_NAME} ${cc_binary_DEPS})
@@ -73,17 +79,16 @@ endfunction(cc_binary)
 #   DEPS
 #   tensor)
 function(cc_test TARGET_NAME)
+  if(WITH_TESTING)
     set(options "")
     set(oneValueArgs "")
     set(multiValueArgs SRCS DEPS)
     cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
     add_executable(${TARGET_NAME} ${cc_test_SRCS})
-  link_paddle_test(${TARGET_NAME})
-  if(cc_test_DEPS)
-    target_link_libraries(${TARGET_NAME} ${cc_test_DEPS})
-    add_dependencies(${TARGET_NAME} ${cc_test_DEPS})
-  endif()
+    target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} gtest gtest_main)
+    add_dependencies(${TARGET_NAME} ${cc_test_DEPS} gtest gtest_main)
     add_test(${TARGET_NAME} ${TARGET_NAME})
+  endif()
 endfunction(cc_test)
 
 # Suppose that ops.cu includes global functions that take Tensor as
@@ -95,6 +100,7 @@ endfunction(cc_test)
 #   DEPS
 #   tensor)
 function(nv_library TARGET_NAME)
+  if (WITH_GPU)
     set(options OPTIONAL)
     set(oneValueArgs "")
     set(multiValueArgs SRCS DEPS)
@@ -104,20 +110,24 @@ function(nv_library TARGET_NAME)
     else()
       cuda_add_library(${TARGET_NAME} STATIC ${nv_library_SRCS})
     endif()
-  add_dependencies(${TARGET_NAME} ${nv_library_DEPS} ${external_project_dependencies})
+    if (nv_library_DEPS)
+      add_dependencies(${TARGET_NAME} ${nv_library_DEPS})
+    endif()
+  endif()
 endfunction(nv_library)
 
 function(nv_binary TARGET_NAME)
+  if (WITH_GPU)
     set(options "")
     set(oneValueArgs "")
     set(multiValueArgs SRCS DEPS)
     cmake_parse_arguments(nv_binary "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
     cuda_add_executable(${TARGET_NAME} ${nv_binary_SRCS})
-  link_paddle_exe(${TARGET_NAME})  
     if(nv_binary_DEPS)
       target_link_libraries(${TARGET_NAME} ${nv_binary_DEPS})
       add_dependencies(${TARGET_NAME} ${nv_binary_DEPS})
     endif()
+  endif()
 endfunction(nv_binary)
 
 # The dependency to target tensor implies that if any of
@@ -128,17 +138,16 @@ endfunction(nv_binary)
 #   DEPS
 #   ops)
 function(nv_test TARGET_NAME)
+  if (WITH_GPU AND WITH_TESTING)
     set(options "")
     set(oneValueArgs "")
     set(multiValueArgs SRCS DEPS)
     cmake_parse_arguments(nv_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
     cuda_add_executable(${TARGET_NAME} ${nv_test_SRCS})
-  link_paddle_test(${TARGET_NAME})  
-  if(nv_test_DEPS)
-    target_link_libraries(${TARGET_NAME} ${nv_test_DEPS})
-    add_dependencies(${TARGET_NAME} ${nv_test_DEPS})
-  endif()
+    target_link_libraries(${TARGET_NAME} ${nv_test_DEPS} gtest gtest_main)
+    add_dependencies(${TARGET_NAME} ${nv_test_DEPS} gtest gtest_main)
     add_test(${TARGET_NAME} ${TARGET_NAME})
+  endif()
 endfunction(nv_test)
 
 set(GOPATH "${CMAKE_CURRENT_BINARY_DIR}/go")

doc/design/cluster_train/pserver_client.md

@@ -136,6 +136,9 @@ int paddle_send_grads(paddle_pserver_client* client, const paddle_gradient* grad
 /**
  * @brief paddle_get_params gets parameters from parameter servers.
  *
+ * paddle_get_params will block until parameters are initialized on
+ * the parameter servers.
+ *
  * @param names the array of names of the parameters to get.
  * @param dst the destination array of parameters to save to.
  * @param len the length of the names array and the paddle_parameter

doc/howto/deep_model/rnn/hierarchical_layer_cn.rst

@@ -28,17 +28,17 @@ pooling 的使用示例如下，详细见 :ref:`api_v2.layer_pooling` 配置API
 
         seq_pool = pooling(input=layer,
                            pooling_type=pooling.Max(),
-                           agg_level=AggregateLevel.EACH_SEQUENCE)
+                           agg_level=AggregateLevel.TO_SEQUENCE)
         
 - `pooling_type` 目前支持两种，分别是：pooling.Max()和pooling.Avg()。
 
-- `agg_level=AggregateLevel.EACH_TIMESTEP` 时（默认值）：
+- `agg_level=AggregateLevel.TO_NO_SEQUENCE` 时（默认值）：
 
   - 作用：双层序列经过运算变成一个0层序列，或单层序列经过运算变成一个0层序列
   - 输入：一个双层序列，或一个单层序列
   - 输出：一个0层序列，即整个输入序列（单层或双层）的平均值（或最大值）
 
-- `agg_level=AggregateLevel.EACH_SEQUENCE` 时：
+- `agg_level=AggregateLevel.TO_SEQUENCE` 时：
 
   - 作用：一个双层序列经过运算变成一个单层序列
   - 输入：必须是一个双层序列
@@ -52,15 +52,15 @@ last_seq 的使用示例如下（ :ref:`api_v2.layer_first_seq` 类似），详
 ..	code-block:: bash
 
         last = last_seq(input=layer,
-                        agg_level=AggregateLevel.EACH_SEQUENCE)
+                        agg_level=AggregateLevel.TO_SEQUENCE)
         
-- `agg_level=AggregateLevel.EACH_TIMESTEP` 时（默认值）：
+- `agg_level=AggregateLevel.TO_NO_SEQUENCE` 时（默认值）：
 
   - 作用：一个双层序列经过运算变成一个0层序列，或一个单层序列经过运算变成一个0层序列
   - 输入：一个双层序列或一个单层序列
   - 输出：一个0层序列，即整个输入序列（双层或者单层）最后一个，或第一个元素。
 
-- `agg_level=AggregateLevel.EACH_SEQUENCE` 时：
+- `agg_level=AggregateLevel.TO_SEQUENCE` 时：
   - 作用：一个双层序列经过运算变成一个单层序列
   - 输入：必须是一个双层序列
   - 输出：一个单层序列，其中每个元素是双层序列中每个subseq最后一个（或第一个）元素。
@@ -74,9 +74,9 @@ expand 的使用示例如下，详细见 :ref:`api_v2.layer_expand` 配置API。
 
         ex = expand(input=layer1,
                     expand_as=layer2,
-                    expand_level=ExpandLevel.FROM_TIMESTEP)
+                    expand_level=ExpandLevel.FROM_NO_SEQUENCE)
         
-- `expand_level=ExpandLevel.FROM_TIMESTEP` 时（默认值）：
+- `expand_level=ExpandLevel.FROM_NO_SEQUENCE` 时（默认值）：
 
   - 作用：一个0层序列经过运算扩展成一个单层序列，或者一个双层序列
   - 输入：layer1必须是一个0层序列，是待扩展的数据；layer2 可以是一个单层序列，或者是一个双层序列，提供扩展的长度信息

doc/howto/deep_model/rnn/hrnn_rnn_api_compare_cn.rst

@@ -81,7 +81,7 @@
   
   * 在本例中，我们将原始数据的每一组，通过\ :code:`recurrent_group`\ 进行拆解，拆解成的每一句话再通过一个LSTM网络。这和单层RNN的配置是等价的。
 
-* 与单层RNN的配置类似，我们只需要使用LSTM encode成的最后一个向量。所以对\ :code:`recurrent_group`\ 进行了\ :code:`last_seq`\ 操作。但和单层RNN不同，我们是对每一个子序列取最后一个元素，因此\ :code:`agg_level=AggregateLevel.EACH_SEQUENCE`\ 。
+* 与单层RNN的配置类似，我们只需要使用LSTM encode成的最后一个向量。所以对\ :code:`recurrent_group`\ 进行了\ :code:`last_seq`\ 操作。但和单层RNN不同，我们是对每一个子序列取最后一个元素，因此\ :code:`agg_level=AggregateLevel.TO_SEQUENCE`\ 。
 
 * 至此，\ :code:`lstm_last`\ 便和单层RNN配置中的\ :code:`lstm_last`\ 具有相同的结果了。
 

go/cmake/CMakeDetermineGoCompiler.cmake

@@ -38,7 +38,7 @@ endif()
 
 mark_as_advanced(CMAKE_Go_COMPILER)
 
-configure_file(${CMAKE_CURRENT_SOURCE_DIR}/cmake/CMakeGoCompiler.cmake.in
+configure_file(${CMAKE_MODULE_PATH}/CMakeGoCompiler.cmake.in
   ${CMAKE_PLATFORM_INFO_DIR}/CMakeGoCompiler.cmake @ONLY)
 
 set(CMAKE_Go_COMPILER_ENV_VAR "GO_COMPILER")

go/cmake/golang.cmake

@@ -0,0 +1,50 @@
+set(GOPATH "${CMAKE_CURRENT_BINARY_DIR}/go")
+file(MAKE_DIRECTORY ${GOPATH})
+set(PADDLE_IN_GOPATH "${GOPATH}/src/github.com/PaddlePaddle")
+file(MAKE_DIRECTORY ${PADDLE_IN_GOPATH})
+
+function(GO_LIBRARY NAME BUILD_TYPE)
+  if(BUILD_TYPE STREQUAL "STATIC")
+    set(BUILD_MODE -buildmode=c-archive)
+    set(LIB_NAME "lib${NAME}.a")
+  else()
+    set(BUILD_MODE -buildmode=c-shared)
+    if(APPLE)
+      set(LIB_NAME "lib${NAME}.dylib")
+    else()
+      set(LIB_NAME "lib${NAME}.so")
+    endif()
+  endif()
+
+  file(GLOB GO_SOURCE RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}" "*.go")
+  file(RELATIVE_PATH rel ${CMAKE_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR})
+
+  # find Paddle directory.
+  get_filename_component(PARENT_DIR ${CMAKE_CURRENT_SOURCE_DIR} DIRECTORY)
+  get_filename_component(PARENT_DIR ${PARENT_DIR} DIRECTORY)
+  get_filename_component(PADDLE_DIR ${PARENT_DIR} DIRECTORY)
+
+  # automatically get all dependencies specified in the source code
+  # for given target.
+  add_custom_target(goGet env GOPATH=${GOPATH} ${CMAKE_Go_COMPILER} get -d ${rel}/...)
+
+  # make a symlink that references Paddle inside $GOPATH, so go get
+  # will use the local changes in Paddle rather than checkout Paddle
+  # in github.
+  add_custom_target(copyPaddle
+    COMMAND ln -sf ${PADDLE_DIR} ${PADDLE_IN_GOPATH})
+  add_dependencies(goGet copyPaddle)
+
+  add_custom_command(OUTPUT ${OUTPUT_DIR}/.timestamp
+    COMMAND env GOPATH=${GOPATH} ${CMAKE_Go_COMPILER} build ${BUILD_MODE}
+    -o "${CMAKE_CURRENT_BINARY_DIR}/${LIB_NAME}"
+    ${CMAKE_GO_FLAGS} ${GO_SOURCE}
+    WORKING_DIRECTORY ${CMAKE_CURRENT_LIST_DIR})
+
+  add_custom_target(${NAME} ALL DEPENDS ${OUTPUT_DIR}/.timestamp ${ARGN})
+  add_dependencies(${NAME} goGet)
+
+  if(NOT BUILD_TYPE STREQUAL "STATIC")
+    install(PROGRAMS ${CMAKE_CURRENT_BINARY_DIR}/${LIB_NAME} DESTINATION bin)
+  endif()
+endfunction(GO_LIBRARY)

go/cmd/master/master.go

@@ -13,8 +13,8 @@ import (
 
 	"github.com/namsral/flag"
 
-	"github.com/PaddlePaddle/Paddle/paddle/go/master"
-	"github.com/PaddlePaddle/Paddle/paddle/go/recordio"
+	"github.com/PaddlePaddle/Paddle/go/master"
+	"github.com/PaddlePaddle/Paddle/go/recordio"
 )
 
 func main() {

go/cmd/pserver/pserver.go

@@ -8,7 +8,7 @@ import (
 
 	"github.com/namsral/flag"
 
-	"github.com/PaddlePaddle/Paddle/paddle/go/pserver"
+	"github.com/PaddlePaddle/Paddle/go/pserver"
 )
 
 func main() {

go/master/service.go

@@ -6,7 +6,7 @@ import (
 	"sync"
 	"time"
 
-	"github.com/PaddlePaddle/Paddle/paddle/go/recordio"
+	"github.com/PaddlePaddle/Paddle/go/recordio"
 )
 
 const (