Merge remote-tracking branch 'baidu/develop' into feature/sppnet

Author: QiJune

.travis.yml

@@ -38,10 +38,21 @@ addons:
       - curl
       - lcov
       - graphviz
+      - swig
 before_install:
+  - |
+    if [ ${JOB} == "BUILD_AND_TEST" ]; then
+      if [ "$TRAVIS_PULL_REQUEST" != "false" ]; then
+        TRAVIS_COMMIT_RANGE="FETCH_HEAD...$TRAVIS_BRANCH"
+      fi
+      git diff --name-only $TRAVIS_COMMIT_RANGE | grep -qvE '(\.md$)' || {
+        echo "Only markdown docs were updated, stopping build process."
+        exit
+      }
+    fi
   - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then sudo paddle/scripts/travis/before_install.linux.sh; fi
   - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then paddle/scripts/travis/before_install.osx.sh; fi
-  - pip install wheel protobuf sphinx breathe recommonmark
+  - pip install wheel protobuf sphinx breathe recommonmark virtualenv numpy
 script:
   - paddle/scripts/travis/main.sh
 notifications:

CMakeLists.txt

@@ -95,11 +95,26 @@ if(NOT WITH_GPU)
     add_definitions(-DHPPL_STUB_FUNC)
     list(APPEND CMAKE_CXX_SOURCE_FILE_EXTENSIONS cu)
 else()
+    if(${CUDA_VERSION_MAJOR} GREATER 6)
+        if(COMPILER_SUPPORT_CXX11)
+            LIST(APPEND CUDA_NVCC_FLAGS -std=c++11)
+        endif()
+    endif()
+
     # TODO(yuyang18): Change it to remove std=c++11 in cuda compile.
     set(CUDA_PROPAGATE_HOST_FLAGS OFF)
     if(NOT CUDNN_FOUND)
         message(FATAL_ERROR "Paddle need cudnn to compile")
     endif()
+    set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-g -O3 --use_fast_math")
+
+    if(WITH_AVX)
+        if(AVX_FOUND)
+            set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler -mavx")
+        endif(AVX_FOUND)
+    else(WITH_AVX)
+        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler -msse3")
+    endif(WITH_AVX)
 
     if(WITH_DSO)
         set(CUDA_LIBRARIES "")

README.md

@@ -14,7 +14,7 @@ developed by Baidu scientists and engineers for the purpose of applying deep
 learning to many products at Baidu.
 
 Our vision is to enable deep learning for everyone via PaddlePaddle.
-Please refer to our [release log](https://github.com/baidu/Paddle/releases) to track the latest feature of PaddlePaddle. 
+Please refer to our [release announcement](https://github.com/baidu/Paddle/releases) to track the latest feature of PaddlePaddle. 
 
 ## Features
 

cmake/util.cmake

@@ -188,14 +188,6 @@ macro(add_simple_unittest TARGET_NAME)
     add_unittest(${TARGET_NAME} ${TARGET_NAME}.cpp)
 endmacro()
 
-macro(add_paddle_culib TARGET_NAME)
-    set(NVCC_FLAG ${CUDA_NVCC_FLAGS})
-    set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};--use_fast_math)
-    cuda_add_library(${TARGET_NAME} STATIC ${ARGN})
-    set(CUDA_NVCC_FLAGS ${NVCC_FLAG})
-endmacro()
-
-
 # Creates C resources file from files in given resource file
 function(create_resources res_file output)
     # Create empty output file

demo/quick_start/preprocess.sh

@@ -21,14 +21,21 @@
 set -e
 
 export LC_ALL=C
+UNAME_STR=`uname`
+
+if [[ ${UNAME_STR} == 'Linux' ]]; then
+  SHUF_PROG='shuf'
+else
+  SHUF_PROG='gshuf'
+fi
 
 mkdir -p data/tmp
 python preprocess.py -i data/reviews_Electronics_5.json.gz
 # uniq and shuffle
 cd data/tmp
 echo 'uniq and shuffle...'
-cat pos_*|sort|uniq|shuf> pos.shuffed
-cat neg_*|sort|uniq|shuf> neg.shuffed
+cat pos_*|sort|uniq|${SHUF_PROG}> pos.shuffed
+cat neg_*|sort|uniq|${SHUF_PROG}> neg.shuffed
 
 min_len=`sed -n '$=' neg.shuffed`
 test_num=$((min_len/10))
@@ -42,8 +49,8 @@ head -n$train_num neg.shuffed >train.neg
 tail -n$test_num pos.shuffed >test.pos
 tail -n$test_num neg.shuffed >test.neg
 
-cat train.pos train.neg|shuf>../train.txt
-cat test.pos test.neg|shuf>../test.txt
+cat train.pos train.neg | ${SHUF_PROG} >../train.txt
+cat test.pos test.neg | ${SHUF_PROG} >../test.txt
 
 cd -
 echo 'data/train.txt' > data/train.list

doc_cn/demo/quick_start/index.md

@@ -134,9 +134,8 @@ define_py_data_sources2(train_list='data/train.list',
 * obj="process": 指定生成数据的函数
 * args={"dictionary": word_dict}: 额外的参数，这里指定词典
 
-更详细用例请参考文档<a href = "../../../doc/ui/data_provider/python_case.html">Python Use Case</a>，
-数据格式和详细文档请参考<a href = "../../../doc/ui/data_provider/pydataprovider2.html">
-PyDataProviderWrapper</a>。
+更详细数据格式和用例请参考<a href = "../../ui/data_provider/pydataprovider2.html">
+PyDataProvider2</a>。
 
 ## 网络结构(Network Architecture)
 本节我们将专注于网络结构的介绍。

doc_cn/faq/index.rst

@@ -177,3 +177,40 @@ PaddlePaddle的参数使用名字 :code:`name` 作为参数的ID，相同名字
 
     pip install --upgrade pip
 
+8.  python相关的单元测试都过不了
+--------------------------------
+
+如果出现以下python相关的单元测试都过不了的情况：
+
+..  code-block:: bash
+
+    24 - test_PyDataProvider (Failed)
+    26 - test_RecurrentGradientMachine (Failed)
+    27 - test_NetworkCompare (Failed)
+    28 - test_PyDataProvider2 (Failed)
+    32 - test_Prediction (Failed)
+    33 - test_Compare (Failed)
+    34 - test_Trainer (Failed)
+    35 - test_TrainerOnePass (Failed)
+    36 - test_CompareTwoNets (Failed)
+    37 - test_CompareTwoOpts (Failed)
+    38 - test_CompareSparse (Failed)
+    39 - test_recurrent_machine_generation (Failed)
+    40 - test_PyDataProviderWrapper (Failed)
+    41 - test_config_parser (Failed)
+    42 - test_swig_api (Failed)
+    43 - layers_test (Failed)
+    
+并且查询PaddlePaddle单元测试的日志，提示：
+
+..  code-block:: bash
+    
+    paddle package is already in your PYTHONPATH. But unittest need a clean environment.
+    Please uninstall paddle package before start unittest. Try to 'pip uninstall paddle'.
+    
+解决办法是：卸载paddle包 :code:`pip uninstall paddle`。
+
+原因是：单元测试使用了一个旧版本的python包，而没有测试到代码中实际修改的python包。即单元测试需要一个干净的环境：
+
+* 如果paddle包已经在python的site-packages里面了，那么单元测试时使用的paddle包，就是site-packages里面的python包，而不是源码目录里 :code:`/python` 目录下的python包。
+* 即便设置了 :code:`PYTHONPATH` 到 :code:`/python` 也没用，因为python的搜索路径是优先已经安装的python包。

paddle/api/paddle_api_config.py.in

@@ -1,6 +1,7 @@
 PADDLE_BUILD_DIR="@CMAKE_CURRENT_BINARY_DIR@/../"
 WITH_GPU="@WITH_GPU@"
 PROTOBUF_LIB="@PROTOBUF_LIBRARY@"
+ZLIB_LIB="@ZLIB_LIBRARIES@"
 CMAKE_THREAD_LIB="@CMAKE_THREAD_LIBS_INIT@"
 CMAKE_DL_LIBS="@CMAKE_DL_LIBS@"
 
@@ -15,3 +16,4 @@ GFLAGS_LOCATION="@GFLAGS_LOCATION@"
 CBLAS_LIBRARIES="@CBLAS_LIBS@"
 
 CUDA_LIBRARIES="@CUDA_LIBRARIES@"
+WITH_COVERALLS="@ON_COVERALLS@"

paddle/api/paddle_ld_flags.py

@@ -38,6 +38,7 @@ def __init__(self):
             self.paddle_build_dir = os.path.abspath(self.paddle_build_dir)
             self.with_gpu = PaddleLDFlag.cmake_bool(WITH_GPU)
             self.protolib = PROTOBUF_LIB
+            self.zlib = ZLIB_LIB
             self.thread = CMAKE_THREAD_LIB
             self.dl_libs = CMAKE_DL_LIBS
             self.with_python = PaddleLDFlag.cmake_bool(WITH_PYTHON)
@@ -47,6 +48,7 @@ def __init__(self):
             self.glog_libs = LIBGLOG_LIBRARY
 
             self.with_gflags = PaddleLDFlag.cmake_bool(WITH_GFLAGS)
+            self.with_coverage = PaddleLDFlag.cmake_bool(WITH_COVERALLS)
             self.gflags_libs = GFLAGS_LIBRARIES
             self.gflags_location = GFLAGS_LOCATION
             self.cblas_libs = CBLAS_LIBRARIES
@@ -64,7 +66,7 @@ def libs_dir_str(self):
 
         def parent_dir_str(self):
             libdirs = PARENT_LIB_DIRS
-            return " ".join(map(lambda x: "-L" + os.path.join(self.paddle_build_dir, '..', x), 
+            return " ".join(map(lambda x: "-L" + os.path.join(self.paddle_build_dir, '..', x),
                 libdirs))
 
         def libs_str(self):
@@ -82,6 +84,7 @@ def libs_str(self):
                 "-lpaddle_cuda",
                 "-lpaddle_api",
                 self.normalize_flag(self.protolib),
+                self.normalize_flag(self.zlib),
                 self.normalize_flag(self.thread),
                 self.normalize_flag(self.dl_libs),
                 self.normalize_flag(self.cblas_libs),
@@ -95,6 +98,8 @@ def libs_str(self):
                 libs.append(self.normalize_flag(self.gflags_libs))
             if self.with_gpu:
                 libs.append(self.normalize_flag(self.curt))
+            if self.with_coverage:
+                libs.append("-fprofile-arcs")
             return " ".join(filter(lambda l: len(l) != 0, libs))
 
         def normalize_flag(self, cmake_flag):
@@ -131,8 +136,14 @@ def cmake_bool(cmake_str):
                 return False
             else:
                 return True
-
+        def c_flag(self):
+            if self.with_coverage:
+                return ["-fprofile-arcs", "-ftest-coverage", "-O0", "-g"]
+            else:
+                return None
 except ImportError:
     class PaddleLDFlag(object):
         def ldflag_str(self):
             pass
+        def c_flag(self):
+            pass

paddle/cuda/include/hl_base.h

@@ -209,6 +209,15 @@ typedef struct {
 #define HL_FLOAT_MIN        2.2250738585072014e-308
 #endif
 
+
+/**
+ * The maximum input value for exp, used to avoid overflow problem.
+ *
+ * Currently only used for tanh function.
+ */
+#define EXP_MAX_INPUT       40.0
+
+
 /**
  * @brief DIVUP(x, y) is similar to ceil(x / y).
  * @note  For CUDA, DIVUP will be used to specify