Merge branch 'develop' of upstream into deconv_group

7 years ago · adbf97b42d
parent 7a557040e7 01c0a81b4d
commit adbf97b42d
104 changed files with 2228 additions and 4986 deletions
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@ -100,6 +100,9 @@ endif()
 set(THIRD_PARTY_PATH "${CMAKE_BINARY_DIR}/third_party" CACHE STRING
  "A path setting third party libraries download & build directories.")

+set(FLUID_INSTALL_DIR "${CMAKE_BINARY_DIR}/fluid_install_dir" CACHE STRING
+  "A path setting fluid shared and static libraries")
+
 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 "
@ -117,13 +120,14 @@ else()
 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)
+if (NOT DEFINED WITH_MKLDNN)
+    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()
 endif()
-
 ########################################################################################

 include(external/mklml)     # download mklml package
--- a/2
+++ b/2
@ -70,7 +70,7 @@ RUN localedef -i en_US -f UTF-8 en_US.UTF-8
 # specify sphinx version as 1.5.6 and remove -U option for [pip install -U
 # sphinx-rtd-theme] since -U option will cause sphinx being updated to newest
 # version(1.7.1 for now), which causes building documentation failed.
-RUN pip install --upgrade pip==9.0.3 && \
+RUN easy_install -U pip && \
    pip install -U wheel && \
    pip install -U docopt PyYAML sphinx==1.5.6 && \
    pip install sphinx-rtd-theme==0.1.9 recommonmark
--- a/benchmark/fluid/mnist.py
+++ b/benchmark/fluid/mnist.py
@ -159,6 +159,7 @@ def run_benchmark(model, args):
        paddle.dataset.mnist.train(), batch_size=args.batch_size)

    accuracy = fluid.metrics.Accuracy()
+    train_exe = fluid.ParallelExecutor(use_cuda=True, loss_name=avg_cost.name)
    iters, num_samples, start_time = 0, 0, time.time()
    for pass_id in range(args.pass_num):
        accuracy.reset()
@ -175,17 +176,20 @@ def run_benchmark(model, args):
            y_data = np.array(map(lambda x: x[1], data)).astype("int64")
            y_data = y_data.reshape([len(y_data), 1])

-            outs = exe.run(
-                fluid.default_main_program(),
+            outs = train_exe.run(
                feed={"pixel": img_data,
                      "label": y_data},
-                fetch_list=[avg_cost, batch_acc, batch_size_tensor]
+                fetch_list=[
+                    avg_cost.name, batch_acc.name, batch_size_tensor.name
+                ]
            )  # The accuracy is the accumulation of batches, but not the current batch.
-            accuracy.update(value=outs[1], weight=outs[2])
+            accuracy.update(
+                value=np.array(np.mean(outs[1])),
+                weight=np.mean(np.array(outs[2])))
            iters += 1
            num_samples += len(y_data)
-            loss = np.array(outs[0])
-            acc = np.array(outs[1])
+            loss = np.mean(np.array(outs[0]))
+            acc = np.mean(np.array(outs[1]))
            train_losses.append(loss)
            train_accs.append(acc)
            print("Pass: %d, Iter: %d, Loss: %f, Accuracy: %f" %
--- a/benchmark/fluid/resnet.py
+++ b/benchmark/fluid/resnet.py
@ -241,6 +241,7 @@ def run_benchmark(model, args):
    exe = fluid.Executor(place)
    exe.run(fluid.default_startup_program())
    accuracy = fluid.average.WeightedAverage()
+    train_exe = fluid.ParallelExecutor(use_cuda=True, loss_name=avg_cost.name)
    if args.use_fake_data:
        data = train_reader().next()
        image = np.array(map(lambda x: x[0].reshape(dshape), data)).astype(
@ -264,14 +265,17 @@ def run_benchmark(model, args):
                                     data)).astype('float32')
                label = np.array(map(lambda x: x[1], data)).astype('int64')
                label = label.reshape([-1, 1])
-            loss, acc, weight = exe.run(
-                fluid.default_main_program(),
+            loss, acc, weight = train_exe.run(
                feed={'data': image,
                      'label': label},
-                fetch_list=[avg_cost, batch_acc, batch_size_tensor])
+                fetch_list=[
+                    avg_cost.name, batch_acc.name, batch_size_tensor.name
+                ])
            iters += 1
            num_samples += len(label)
-            accuracy.add(value=acc, weight=weight)
+            accuracy.add(value=np.array(np.mean(acc)), weight=np.mean(weight))
+            loss = np.mean(np.array(loss))
+            acc = np.mean(np.array(acc))
            train_losses.append(loss)
            train_accs.append(acc)
            print("Pass: %d, Iter: %d, Loss: %f, Accuracy: %f" %
--- a/benchmark/fluid/vgg.py
+++ b/benchmark/fluid/vgg.py
@ -169,6 +169,7 @@ def main():

    iters, num_samples, start_time = 0, 0, time.time()
    accuracy = fluid.average.WeightedAverage()
+    train_exe = fluid.ParallelExecutor(use_cuda=True, loss_name=avg_cost.name)
    for pass_id in range(args.pass_num):
        accuracy.reset()
        train_accs = []
@ -184,14 +185,17 @@ def main():
            y_data = np.array(map(lambda x: x[1], data)).astype("int64")
            y_data = y_data.reshape([-1, 1])

-            loss, acc, weight = exe.run(
-                fluid.default_main_program(),
+            loss, acc, weight = train_exe.run(
                feed={"pixel": img_data,
                      "label": y_data},
-                fetch_list=[avg_cost, batch_acc, batch_size_tensor])
-            accuracy.add(value=acc, weight=weight)
+                fetch_list=[
+                    avg_cost.name, batch_acc.name, batch_size_tensor.name
+                ])
+            accuracy.add(value=np.array(np.mean(acc)), weight=np.mean(weight))
            iters += 1
            num_samples += len(y_data)
+            loss = np.mean(np.array(loss))
+            acc = np.mean(np.array(acc))
            print(
                "Pass = %d, Iter = %d, Loss = %f, Accuracy = %f" %
                (pass_id, iters, loss, acc)
--- a/cmake/external/boost.cmake
+++ b/cmake/external/boost.cmake
@ -23,8 +23,12 @@ set(BOOST_PROJECT       "extern_boost")
 # checked that the devtools package of CentOS 6 installs boost 1.41.0.
 # So we use 1.41.0 here.
 set(BOOST_VER           "1.41.0")
-set(BOOST_TAR           "boost_1_41_0")
-set(BOOST_URL           "http://paddlepaddledeps.bj.bcebos.com/${BOOST_TAR}.tar.gz")
+if((NOT DEFINED BOOST_TAR) OR (NOT DEFINED BOOST_URL))
+    message(STATUS "use pre defined download url")
+    set(BOOST_TAR "boost_1_41_0" CACHE STRING "" FORCE)
+    set(BOOST_URL "http://paddlepaddledeps.bj.bcebos.com/${BOOST_TAR}.tar.gz" CACHE STRING "" FORCE)
+endif()
+MESSAGE(STATUS "BOOST_TAR: ${BOOST_TAR}, BOOST_URL: ${BOOST_URL}")
 set(BOOST_SOURCES_DIR ${THIRD_PARTY_PATH}/boost)
 set(BOOST_DOWNLOAD_DIR  "${BOOST_SOURCES_DIR}/src/${BOOST_PROJECT}")
 set(BOOST_INCLUDE_DIR "${BOOST_DOWNLOAD_DIR}/${BOOST_TAR}" CACHE PATH "boost include directory." FORCE)
--- a/cmake/external/eigen.cmake
+++ b/cmake/external/eigen.cmake
@ -21,11 +21,12 @@ else()
    ExternalProject_Add(
        extern_eigen3
        ${EXTERNAL_PROJECT_LOG_ARGS}
-        GIT_REPOSITORY  "https://github.com/RLovelett/eigen.git"
+        GIT_REPOSITORY  "https://github.com/eigenteam/eigen-git-mirror"
        # eigen on cuda9.1 missing header of math_funtions.hpp
        # https://stackoverflow.com/questions/43113508/math-functions-hpp-not-found-when-using-cuda-with-eigen
        GIT_TAG         917060c364181f33a735dc023818d5a54f60e54c
        PREFIX          ${EIGEN_SOURCE_DIR}
+        DOWNLOAD_NAME   "eigen"
        UPDATE_COMMAND  ""
        CONFIGURE_COMMAND ""
        BUILD_COMMAND     ""
--- a/cmake/external/mkldnn.cmake
+++ b/cmake/external/mkldnn.cmake
@ -53,11 +53,9 @@ ExternalProject_Add(
    ${EXTERNAL_PROJECT_LOG_ARGS}
    DEPENDS             ${MKLDNN_DEPENDS}
    GIT_REPOSITORY      "https://github.com/01org/mkl-dnn.git"
-    GIT_TAG             "v0.14"
+    GIT_TAG             "db3424ad44901513c03a1ea31ccaacdf633fbe9f"
    PREFIX              ${MKLDNN_SOURCES_DIR}
    UPDATE_COMMAND      ""
-    # Patch MKLDNN to compile with gcc 4.8, the related issue is in intel/mkl-dnn#237.
-    PATCH_COMMAND       ${CMAKE_COMMAND} -E copy_if_different ${CMAKE_CURRENT_SOURCE_DIR}/patches/mkldnn.hpp ${MKLDNN_SOURCES_DIR}/src/extern_mkldnn/include/mkldnn.hpp
    CMAKE_ARGS          -DCMAKE_INSTALL_PREFIX=${MKLDNN_INSTALL_DIR}
    CMAKE_ARGS          -DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE} 
    CMAKE_ARGS          -DMKLROOT=${MKLML_ROOT}
--- a/cmake/external/mklml.cmake
+++ b/cmake/external/mklml.cmake
@ -27,8 +27,12 @@ ENDIF()
 INCLUDE(ExternalProject)

 SET(MKLML_PROJECT       "extern_mklml")
-SET(MKLML_VER           "mklml_lnx_2018.0.3.20180406")
-SET(MKLML_URL           "http://paddlepaddledeps.bj.bcebos.com/${MKLML_VER}.tgz")
+IF((NOT DEFINED MKLML_VER) OR (NOT DEFINED MKLML_URL))
+  MESSAGE(STATUS "use pre defined download url")
+  SET(MKLML_VER "mklml_lnx_2018.0.3.20180406" CACHE STRING "" FORCE)
+  SET(MKLML_URL "http://paddlepaddledeps.bj.bcebos.com/${MKLML_VER}.tgz" CACHE STRING "" FORCE)
+ENDIF()
+MESSAGE(STATUS "MKLML_VER: ${MKLML_VER}, MKLML_URL: ${MKLML_URL}")
 SET(MKLML_SOURCE_DIR    "${THIRD_PARTY_PATH}/mklml")
 SET(MKLML_DOWNLOAD_DIR  "${MKLML_SOURCE_DIR}/src/${MKLML_PROJECT}")
 SET(MKLML_DST_DIR       "mklml")
--- a/cmake/generic.cmake
+++ b/cmake/generic.cmake
@ -231,7 +231,7 @@ endfunction(cc_binary)

 function(cc_test TARGET_NAME)
  if(WITH_TESTING)
-    set(options "")
+    set(options SERIAL)
    set(oneValueArgs "")
    set(multiValueArgs SRCS DEPS ARGS)
    cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
@ -241,6 +241,9 @@ function(cc_test TARGET_NAME)
    add_test(NAME ${TARGET_NAME}
             COMMAND ${TARGET_NAME} ${cc_test_ARGS}
             WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
+    if (${cc_test_SERIAL})
+        set_property(TEST ${TARGET_NAME} PROPERTY SERIAL 1)
+    endif()
  endif()
 endfunction(cc_test)

@ -295,7 +298,7 @@ endfunction(nv_binary)

 function(nv_test TARGET_NAME)
  if (WITH_GPU AND WITH_TESTING)
-    set(options "")
+    set(options SERIAL)
    set(oneValueArgs "")
    set(multiValueArgs SRCS DEPS)
    cmake_parse_arguments(nv_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
@ -303,6 +306,9 @@ function(nv_test TARGET_NAME)
    target_link_libraries(${TARGET_NAME} ${nv_test_DEPS} paddle_gtest_main memory gtest gflags glog)
    add_dependencies(${TARGET_NAME} ${nv_test_DEPS} paddle_gtest_main memory gtest gflags glog)
    add_test(${TARGET_NAME} ${TARGET_NAME})
+    if (nv_test_SERIAL)
+        set_property(TEST ${TARGET_NAME} PROPERTY SERIAL 1)
+    endif()
  endif()
 endfunction(nv_test)

--- a/cmake/inference_lib.cmake
+++ b/cmake/inference_lib.cmake
@ -52,32 +52,32 @@ function(copy TARGET)
 endfunction()

 # third party
-set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/eigen3")
+set(dst_dir "${FLUID_INSTALL_DIR}/third_party/eigen3")
 copy(eigen3_lib
  SRCS ${EIGEN_INCLUDE_DIR}/Eigen/Core ${EIGEN_INCLUDE_DIR}/Eigen/src ${EIGEN_INCLUDE_DIR}/unsupported/Eigen
  DSTS ${dst_dir}/Eigen ${dst_dir}/Eigen ${dst_dir}/unsupported
 )

-set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/gflags")
+set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/gflags")
 copy(gflags_lib
  SRCS ${GFLAGS_INCLUDE_DIR} ${GFLAGS_LIBRARIES}
  DSTS ${dst_dir} ${dst_dir}/lib
 )

-set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/glog")
+set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/glog")
 copy(glog_lib
  SRCS ${GLOG_INCLUDE_DIR} ${GLOG_LIBRARIES}
  DSTS ${dst_dir} ${dst_dir}/lib
 )

-set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/boost/")
+set(dst_dir "${FLUID_INSTALL_DIR}/third_party/boost/")
 copy(boost_lib
  SRCS ${BOOST_INCLUDE_DIR}/boost
  DSTS ${dst_dir}
 )

 if(NOT PROTOBUF_FOUND)
-    set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/protobuf")
+    set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/protobuf")
    copy(protobuf_lib
      SRCS ${PROTOBUF_INCLUDE_DIR} ${PROTOBUF_LIBRARY}
      DSTS ${dst_dir} ${dst_dir}/lib
@ -85,31 +85,39 @@ if(NOT PROTOBUF_FOUND)
 endif()

 if(NOT CBLAS_FOUND)
-    set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/openblas")
+    set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/openblas")
    copy(openblas_lib
      SRCS ${CBLAS_INSTALL_DIR}/lib ${CBLAS_INSTALL_DIR}/include
      DSTS ${dst_dir} ${dst_dir}
    )
 elseif (WITH_MKLML)
-    set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/mklml")
+    set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/mklml")
    copy(mklml_lib
      SRCS ${MKLML_LIB} ${MKLML_IOMP_LIB} ${MKLML_INC_DIR}
      DSTS ${dst_dir}/lib ${dst_dir}/lib ${dst_dir}
    )
 endif()

+if(WITH_MKLDNN)
+  set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/mkldnn")
+  copy(mkldnn_lib
+    SRCS ${MKLDNN_INC_DIR} ${MKLDNN_SHARED_LIB}
+    DSTS ${dst_dir} ${dst_dir}/lib
+  )
+endif()
+
 if(NOT MOBILE_INFERENCE AND NOT RPI)
-  set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/snappy")
+  set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/snappy")
  copy(snappy_lib
    SRCS ${SNAPPY_INCLUDE_DIR} ${SNAPPY_LIBRARIES}
    DSTS ${dst_dir} ${dst_dir}/lib)

-  set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/snappystream")
+  set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/snappystream")
  copy(snappystream_lib
    SRCS ${SNAPPYSTREAM_INCLUDE_DIR} ${SNAPPYSTREAM_LIBRARIES}
    DSTS ${dst_dir} ${dst_dir}/lib)

-  set(dst_dir "${CMAKE_INSTALL_PREFIX}/third_party/install/zlib")
+  set(dst_dir "${FLUID_INSTALL_DIR}/third_party/install/zlib")
  copy(zlib_lib
    SRCS ${ZLIB_INCLUDE_DIR} ${ZLIB_LIBRARIES}
    DSTS ${dst_dir} ${dst_dir}/lib)
@ -117,7 +125,7 @@ endif()

 # paddle fluid module
 set(src_dir "${PADDLE_SOURCE_DIR}/paddle/fluid")
-set(dst_dir "${CMAKE_INSTALL_PREFIX}/paddle/fluid")
+set(dst_dir "${FLUID_INSTALL_DIR}/paddle/fluid")
 set(module "framework")
 copy(framework_lib DEPS framework_py_proto 
  SRCS ${src_dir}/${module}/*.h ${src_dir}/${module}/details/*.h ${PADDLE_BINARY_DIR}/paddle/fluid/framework/framework.pb.h
@ -148,4 +156,30 @@ copy(string_lib
  DSTS ${dst_dir}/${module} ${dst_dir}/${module}/tinyformat
 )

+set(module "pybind")
+copy(pybind_lib
+  SRCS ${CMAKE_CURRENT_BINARY_DIR}/paddle/fluid/${module}/pybind.h
+  DSTS ${dst_dir}/${module}
+)
+
+# CMakeCache Info
+copy(cmake_cache
+  SRCS ${CMAKE_CURRENT_BINARY_DIR}/CMakeCache.txt
+  DSTS ${FLUID_INSTALL_DIR})
+
 add_custom_target(inference_lib_dist DEPENDS ${inference_lib_dist_dep}) 
+
+# paddle fluid version
+execute_process(
+  COMMAND ${GIT_EXECUTABLE} log --pretty=format:%H -1
+  OUTPUT_VARIABLE PADDLE_GIT_COMMIT)
+set(version_file ${FLUID_INSTALL_DIR}/version.txt)
+file(WRITE ${version_file}
+  "GIT COMMIT ID: ${PADDLE_GIT_COMMIT}\n"
+  "WITH_MKL: ${WITH_MKL}\n"
+  "WITH_GPU: ${WITH_GPU}\n")
+if(WITH_GPU)
+  file(APPEND ${version_file}
+    "CUDA version: ${CUDA_VERSION}\n"
+    "CUDNN version: v${CUDNN_MAJOR_VERSION}\n")
+endif()
--- a/doc/fluid/api/layers.rst
+++ b/doc/fluid/api/layers.rst
@ -485,7 +485,7 @@ roi_pool
 ..  autofunction:: paddle.fluid.layers.roi_pool
    :noindex:

-    
+
 ops
 ===

@ -828,4 +828,10 @@ topk
 ..  autofunction:: paddle.fluid.layers.topk
    :noindex:

+dice_loss
+----
+
+..  autofunction:: paddle.fluid.layers.dice_loss
+    :noindex:
+

--- a/doc/fluid/design/concepts/functions_operators_layers.md
+++ b/doc/fluid/design/concepts/functions_operators_layers.md
@ -40,7 +40,7 @@ template <typename T>
 class FCOp : public OperatorBase {
 public:
  void Run(...) {
-    add(mul(Input<T>("X"), Input<T>("W")), Input<T>("b");
+    add(mul(Input<T>("X"), Input<T>("W")), Input<T>("b"));
  }
 };
 REGISTER_OP(FCOp, "fc");
--- a/doc/fluid/howto/optimization/cpu_profiling_cn.md
+++ b/doc/fluid/howto/optimization/cpu_profiling_cn.md
@ -1,3 +1,5 @@
+# CPU性能调优
+
 此教程会介绍如何使用Python的cProfile包、Python库yep、Google perftools来进行性能分析 (profiling) 与调优（performance tuning）。

 Profling 指发现性能瓶颈。系统中的瓶颈可能和程序员开发过程中想象的瓶颈相去甚远。Tuning 指消除瓶颈。性能优化的过程通常是不断重复地 profiling 和 tuning。
@ -8,7 +10,7 @@ PaddlePaddle 用户一般通过调用 Python API 编写深度学习程序。大
 * Python 与 C++ 混合代码的性能分析


-# Python代码的性能分析
+## Python代码的性能分析

 ### 生成性能分析文件

--- a/doc/fluid/howto/optimization/cpu_profiling_en.md
+++ b/doc/fluid/howto/optimization/cpu_profiling_en.md
@ -1,3 +1,5 @@
+# Tune CPU performance
+
 This tutorial introduces techniques we use to profile and tune the
 CPU performance of PaddlePaddle.  We will use Python packages
 `cProfile` and `yep`, and Google's `perftools`.
@ -14,7 +16,7 @@ the profiling and tuning of
 1. the Python code and
 1. the mixture of Python and C++ code.

-# Profiling the Python Code
+## Profiling the Python Code

 ### Generate the Performance Profiling File

--- a/doc/v2/build_and_install/build_from_source_cn.rst
+++ b/doc/v2/build_and_install/build_from_source_cn.rst
@ -35,7 +35,7 @@ PaddlePaddle需要使用Docker环境完成编译，这样可以免去单独安
   # 2. 可选步骤：源码中构建用于编译PaddlePaddle的Docker镜像
   docker build -t paddle:dev .
   # 3. 执行下面的命令编译CPU-Only的二进制
-   docker run -it -v $PWD:/paddle -e "WITH_GPU=OFF" -e "WITH_TESTING=OFF" paddlepaddle/paddle_manylinux_devel:cuda8.0_cudnn5 bash -x /paddle/paddle/scripts/docker/build.sh
+   docker run -it -v $PWD:/paddle -e "WITH_GPU=OFF" -e "WITH_TESTING=OFF" paddlepaddle/paddle_manylinux_devel:cuda8.0_cudnn5 bash -x /paddle/paddle/scripts/paddle_build.sh build
   # 4. 或者也可以使用为上述可选步骤构建的镜像（必须先执行第2步）
   docker run -it -v $PWD:/paddle -e "WITH_GPU=OFF" -e "WITH_TESTING=OFF" paddle:dev

--- a/doc/v2/build_and_install/build_from_source_en.rst
+++ b/doc/v2/build_and_install/build_from_source_en.rst
@ -34,7 +34,7 @@ Or you can build your own image from source as the optional step below:
   # 2. Optional: build development docker image from source
   docker build -t paddle:dev .
   # 3. Run the following command to build a CPU-Only binaries
-   docker run -it -v $PWD:/paddle -e "WITH_GPU=OFF" -e "WITH_TESTING=OFF" paddlepaddle/paddle_manylinux_devel:cuda8.0_cudnn5 bash -x /paddle/paddle/scripts/docker/build.sh
+   docker run -it -v $PWD:/paddle -e "WITH_GPU=OFF" -e "WITH_TESTING=OFF" paddlepaddle/paddle_manylinux_devel:cuda8.0_cudnn5 bash -x /paddle/paddle/scripts/paddle_build.sh build
   # 4. Or, use your built Docker image to build PaddlePaddle (must run step 2)
   docker run -it -v $PWD:/paddle -e "WITH_GPU=OFF" -e "WITH_TESTING=OFF" paddle:dev

--- a/doc/v2/build_and_install/pip_install_cn.rst
+++ b/doc/v2/build_and_install/pip_install_cn.rst
@ -10,20 +10,38 @@ PaddlePaddle可以使用常用的Python包管理工具
 使用pip安装
 ------------------------------

-
-执行下面的命令即可在当前机器上安装PaddlePaddle的运行时环境，并自动下载安装依赖软件，版本为cpu_avx_openblas。
+执行下面的命令即可在当前机器上安装PaddlePaddle的运行时环境，并自动下载安装依赖软件。

  .. code-block:: bash

     pip install paddlepaddle

+当前的默认版本为0.12.0，cpu_avx_openblas，您可以通过指定版本号来安装其它版本，例如:
+
+  .. code-block:: bash
+
+      pip install paddlepaddle==0.11.0
+

-如果需要安装支持GPU的版本（cuda7.5_cudnn5_avx_openblas），需要执行：
+如果需要安装支持GPU的版本（cuda8.0_cudnn5_avx_openblas），需要执行：

  .. code-block:: bash

     pip install paddlepaddle-gpu

+当前的默认版本也是0.12.0，PaddlePaddle针对不同需求提供了更多版本的安装包，部分列表如下：
+
+=================================   ========================================
+版本号                               版本说明
+=================================   ========================================
+paddlepaddle-gpu==0.12.0            使用CUDA 8.0和cuDNN 5编译的0.12.0版本
+paddlepaddle-gpu==0.11.0.post87     使用CUDA 8.0和cuDNN 7编译的0.11.0版本
+paddlepaddle-gpu==0.11.0.post8      使用CUDA 8.0和cuDNN 5编译的0.11.0版本
+paddlepaddle-gpu==0.11.0            使用CUDA 7.5和cuDNN 5编译的0.11.0版本
+=================================   ========================================
+
+您可以在 `Release History <https://pypi.org/project/paddlepaddle-gpu/#history>`_ 中找到paddlepaddle-gpu的各个发行版本。
+
 如果需要获取并安装最新的（开发分支）PaddlePaddle，可以从我们的CI系统中下载最新的whl安装包和c-api开发包并安装，
 您可以从下面的表格中找到需要的版本：

@ -37,12 +55,11 @@ PaddlePaddle可以使用常用的Python包管理工具
    :header: "版本说明", "cp27-cp27mu", "cp27-cp27m"
    :widths: 1, 3, 3

-    "cpu_avx_mkl", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cpu_avx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cpu_noavx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cuda7.5_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
+    "cpu_avx_mkl", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cpu_avx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cpu_noavx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"

 .. _pip_dependency:

@ -69,7 +86,7 @@ PaddlePaddle发布的安装包会尽量对齐 `manylinux1 <https://www.python.or
 ------------------------------

 - paddlepaddle*.whl is not a supported wheel on this platform.
-  
+
  出现这个问题的主要原因是，没有找到和当前系统匹配的paddlepaddle安装包。请检查Python版本是否为2.7系列。另外最新的pip官方源中的安装包默认是manylinux1标准，需要使用最新的pip (>9.0.0) 才可以安装。可以使用下面的命令更新您的pip：

    .. code-block:: bash
--- a/doc/v2/build_and_install/pip_install_en.rst
+++ b/doc/v2/build_and_install/pip_install_en.rst
@ -12,20 +12,38 @@ Install using pip
 ------------------------------

 Run the following command to install PaddlePaddle on the current
-machine, it will also download requirements, the version is cpu_avx_openblas.
+machine, it will also download requirements.

  .. code-block:: bash

     pip install paddlepaddle

+the default version is 0.12.0, cpu_avx_openblas, you can specify the versions to satisfy your demands, like:

-If you wish to install GPU version (cuda7.5_cudnn5_avx_openblas), just run:
+  .. code-block:: bash
+
+      pip install paddlepaddle==0.11.0
+
+If you need to install a GPU-enabled version (cuda8.0_cudnn5_avx_openblas), you need to run:

  .. code-block:: bash

     pip install paddlepaddle-gpu

-If you wish to install the latest develop branch PaddlePaddle, 
+The default version is also 0.12.0, PaddlePaddle provides several versions of packages for different needs, as shown in the table:
+
+=================================   ========================================
+版本号                               版本说明
+=================================   ========================================
+paddlepaddle-gpu==0.12.0            0.12.0 built with CUDA 8.0 and cuDNN 5
+paddlepaddle-gpu==0.11.0.post87     0.11.0 built with CUDA 8.0 and cuDNN 7
+paddlepaddle-gpu==0.11.0.post8      0.11.0 built with CUDA 8.0 and cuDNN 5
+paddlepaddle-gpu==0.11.0            0.11.0 built with CUDA 7.5 and cuDNN 5
+=================================   ========================================
+
+You can find all versions released of paddlepaddle-gpu in `Release History <https://pypi.org/project/paddlepaddle-gpu/#history>`_ .
+
+If you wish to install the latest develop branch PaddlePaddle,
 you can download the latest whl package from our CI system. Access
 the below links, log in as guest, then click at the "Artifact"
 tab, you'll find the download link of whl packages.
@ -40,12 +58,11 @@ If the links below shows up the login form, just click "Log in as guest" to star
    :header: "version", "cp27-cp27mu", "cp27-cp27m"
    :widths: 1, 3, 3

-    "cpu_avx_mkl", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cpu_avx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cpu_noavx_openblas", "`paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cuda7.5_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda75cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
-    "cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-0.11.0-cp27-cp27m-linux_x86_64.whl>`_"
+    "cpu_avx_mkl", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxCp27cp27mu/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cpu_avx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuAvxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cpu_noavx_openblas", "`paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_CpuNoavxOpenblas/.lastSuccessful/paddlepaddle-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cuda8.0_cudnn5_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda80cudnn5cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"
+    "cuda8.0_cudnn7_avx_mkl", "`paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27mu-linux_x86_64.whl>`_", "`paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl <https://guest:@paddleci.ngrok.io/repository/download/Manylinux1_Cuda8cudnn7cp27cp27mu/.lastSuccessful/paddlepaddle_gpu-latest-cp27-cp27m-linux_x86_64.whl>`_"

 .. _pip_dependency:

@ -79,7 +96,7 @@ FAQ
 ------------------------------

 - paddlepaddle*.whl is not a supported wheel on this platform.
-  
+
  The main cause of this issue is that your current platform is
  not supported. Please check that you are using Python 2.7 series.
  Besides, pypi only supports manylinux1 standard, you'll need to
--- a/paddle/contrib/float16/.gitignore
+++ b/paddle/contrib/float16/.gitignore
--- a/paddle/contrib/float16/README.md
+++ b/paddle/contrib/float16/README.md
--- a/paddle/contrib/float16/float16_benchmark.md
+++ b/paddle/contrib/float16/float16_benchmark.md
--- a/paddle/contrib/float16/float16_inference_demo.py
+++ b/paddle/contrib/float16/float16_inference_demo.py
--- a/paddle/contrib/float16/float16_transpiler.py
+++ b/paddle/contrib/float16/float16_transpiler.py
--- a/paddle/contrib/float16/run_float16_demo.sh
+++ b/paddle/contrib/float16/run_float16_demo.sh
--- a/Show More
+++ b/Show More