Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into image_v2

.travis.yml

@@ -48,7 +48,7 @@ before_install:
   - if [[ "$JOB" == "PRE_COMMIT" ]]; then sudo ln -s /usr/bin/clang-format-3.8 /usr/bin/clang-format; fi
   # Paddle is using protobuf 3.1 currently. Protobuf 3.2 breaks the compatibility. So we specify the python 
   # protobuf version.
-  - pip install numpy wheel 'protobuf==3.1' sphinx recommonmark sphinx-rtd-theme==0.1.9 virtualenv pre-commit requests==2.9.2 LinkChecker
+  - pip install numpy wheel 'protobuf==3.1' sphinx==1.5.6 recommonmark sphinx-rtd-theme==0.1.9 virtualenv pre-commit requests==2.9.2 LinkChecker
   - |
     function timeout() { perl -e 'alarm shift; exec @ARGV' "$@"; }
 script:

CMakeLists.txt

@@ -92,6 +92,7 @@ include(external/swig)      # download, build, install swig
 include(external/warpctc)   # download, build, install warpctc
 include(external/any)       # download libn::any
 
+include(generic)            # simplify cmake module
 include(package)            # set paddle packages
 include(cpplint)            # set paddle c++ style
 include(ccache)             # set ccache for compilation

cmake/generic.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.
+
+
+# To simplify the build process of PaddlePaddle, we defined couple of
+# fundamental abstractions, e.g., how to build library, binary and
+# test in C++, CUDA and Go.
+# 
+# -------------------------------------------
+#    C++	      CUDA C++	      Go
+# -------------------------------------------
+# cc_library	 nv_library	  go_library
+# cc_binary  	 nv_binary	  go_binary
+# cc_test        nv_test	  go_test
+# -------------------------------------------
+#
+# cmake_parse_arguments can help us to achieve this goal.
+# https://cmake.org/cmake/help/v3.0/module/CMakeParseArguments.html
+
+# cc_library parses tensor.cc and figures out that target also depend on tensor.h.
+# cc_library(tensor
+#   SRCS
+#   tensor.cc
+#   DEPS
+#   variant)
+function(cc_library TARGET_NAME)
+  set(options OPTIONAL)
+  set(oneValueArgs "")
+  set(multiValueArgs SRCS DEPS)
+  cmake_parse_arguments(cc_library "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+  if (${cc_library_OPTIONAL} STREQUAL "SHARED")
+    add_library(${TARGET_NAME} SHARED ${cc_library_SRCS})
+  else()
+    add_library(${TARGET_NAME} STATIC ${cc_library_SRCS})
+  endif()
+  add_dependencies(${TARGET_NAME} ${cc_library_DEPS} ${external_project_dependencies})
+endfunction(cc_library)
+
+# cc_binary parses tensor.cc and figures out that target also depend on tensor.h.
+# cc_binary(tensor
+#   SRCS
+#   tensor.cc)
+function(cc_binary TARGET_NAME)
+  set(options OPTIONAL)
+  set(oneValueArgs "")
+  set(multiValueArgs SRCS DEPS)
+  cmake_parse_arguments(cc_binary "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+  add_executable(${TARGET_NAME} ${cc_binary_SRCS})
+  add_dependencies(${TARGET_NAME} ${cc_binary_DEPS} ${external_project_dependencies})
+  target_link_libraries(${TARGET_NAME} ${cc_binary_DEPS})
+endfunction(cc_binary)
+
+# The dependency to target tensor implies that if any of
+# tensor{.h,.cc,_test.cc} is changed, tensor_test need to be re-built.
+# cc_test(tensor_test
+#   SRCS
+#   tensor_test.cc
+#   DEPS
+#   tensor)
+function(cc_test TARGET_NAME)
+  set(options "")
+  set(oneValueArgs "")
+  set(multiValueArgs SRCS DEPS)
+  cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+  add_executable(${TARGET_NAME} ${cc_test_SRCS})
+  add_dependencies(${TARGET_NAME} ${cc_test_DEPS} ${external_project_dependencies})
+  target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} ${GTEST_MAIN_LIBRARIES} ${GTEST_LIBRARIES})
+  add_test(${TARGET_NAME} ${TARGET_NAME})
+endfunction(cc_test)
+
+# Suppose that ops.cu includes global functions that take Tensor as
+# their parameters, so ops depend on tensor. This implies that if
+# any of tensor.{h.cc}, ops.{h,cu} is changed, ops need to be re-built.
+# nv_library(ops
+#   SRCS
+#   ops.cu
+#   DEPS
+#   tensor)
+function(nv_library TARGET_NAME)
+  set(options OPTIONAL)
+  set(oneValueArgs "")
+  set(multiValueArgs SRCS DEPS)
+  cmake_parse_arguments(nv_library "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+  if (${nv_library_OPTIONAL} STREQUAL "SHARED")
+    cuda_add_library(${TARGET_NAME} SHARED ${nv_library_SRCS})
+  else()
+    cuda_add_library(${TARGET_NAME} STATIC ${nv_library_SRCS})
+  endif()
+  add_dependencies(${TARGET_NAME} ${nv_library_DEPS} ${external_project_dependencies})
+endfunction(nv_library)
+
+function(nv_binary TARGET_NAME)
+  set(options "")
+  set(oneValueArgs "")
+  set(multiValueArgs SRCS DEPS)
+  cmake_parse_arguments(nv_binary "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+  cuda_add_executable(${TARGET_NAME} ${nv_binary_SRCS})
+  add_dependencies(${TARGET_NAME} ${nv_binary_DEPS} ${external_project_dependencies})
+  target_link_libraries(${TARGET_NAME} ${nv_binary_DEPS})
+endfunction(nv_binary)
+
+# The dependency to target tensor implies that if any of
+# ops{.h,.cu,_test.cu} is changed, ops_test need to be re-built.
+# nv_test(ops_test
+#   SRCS
+#   ops_test.cu
+#   DEPS
+#   ops)
+function(nv_test TARGET_NAME)
+  set(options "")
+  set(oneValueArgs "")
+  set(multiValueArgs SRCS DEPS)
+  cmake_parse_arguments(nv_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+  cuda_add_executable(${TARGET_NAME} ${nv_test_SRCS})
+  add_dependencies(${TARGET_NAME} ${nv_test_DEPS} ${external_project_dependencies})
+  target_link_libraries(${TARGET_NAME} ${nv_test_DEPS} ${GTEST_MAIN_LIBRARIES} ${GTEST_LIBRARIES})
+  add_test(${TARGET_NAME} ${TARGET_NAME})
+endfunction(nv_test)

demo/semantic_role_labeling/api_train_v2.py

-import sys
 import math
 import numpy as np
-import paddle.v2 as paddle
+import gzip
+import logging
 import paddle.v2.dataset.conll05 as conll05
+import paddle.v2.evaluator as evaluator
+import paddle.v2 as paddle
 
+logger = logging.getLogger('paddle')
 
-def db_lstm():
-    word_dict, verb_dict, label_dict = conll05.get_dict()
-    word_dict_len = len(word_dict)
-    label_dict_len = len(label_dict)
-    pred_len = len(verb_dict)
+word_dict, verb_dict, label_dict = conll05.get_dict()
+word_dict_len = len(word_dict)
+label_dict_len = len(label_dict)
+pred_len = len(verb_dict)
 
-    mark_dict_len = 2
-    word_dim = 32
-    mark_dim = 5
-    hidden_dim = 512
-    depth = 8
+mark_dict_len = 2
+word_dim = 32
+mark_dim = 5
+hidden_dim = 512
+depth = 8
+default_std = 1 / math.sqrt(hidden_dim) / 3.0
+mix_hidden_lr = 1e-3
 
-    #8 features
-    def d_type(size):
-        return paddle.data_type.integer_value_sequence(size)
 
+def d_type(size):
+    return paddle.data_type.integer_value_sequence(size)
+
+
+def db_lstm():
+    #8 features
     word = paddle.layer.data(name='word_data', type=d_type(word_dict_len))
     predicate = paddle.layer.data(name='verb_data', type=d_type(pred_len))
 
@@ -31,11 +38,7 @@ def db_lstm():
     ctx_p2 = paddle.layer.data(name='ctx_p2_data', type=d_type(word_dict_len))
     mark = paddle.layer.data(name='mark_data', type=d_type(mark_dict_len))
 
-    target = paddle.layer.data(name='target', type=d_type(label_dict_len))
-
-    default_std = 1 / math.sqrt(hidden_dim) / 3.0
-
-    emb_para = paddle.attr.Param(name='emb', initial_std=0., learning_rate=0.)
+    emb_para = paddle.attr.Param(name='emb', initial_std=0., is_static=True)
     std_0 = paddle.attr.Param(initial_std=0.)
     std_default = paddle.attr.Param(initial_std=default_std)
 
@@ -63,7 +66,6 @@ def db_lstm():
                 input=emb, param_attr=std_default) for emb in emb_layers
         ])
 
-    mix_hidden_lr = 1e-3
     lstm_para_attr = paddle.attr.Param(initial_std=0.0, learning_rate=1.0)
     hidden_para_attr = paddle.attr.Param(
         initial_std=default_std, learning_rate=mix_hidden_lr)
@@ -111,6 +113,21 @@ def db_lstm():
                 input=input_tmp[1], param_attr=lstm_para_attr)
         ], )
 
+    return feature_out
+
+
+def load_parameter(file_name, h, w):
+    with open(file_name, 'rb') as f:
+        f.read(16)  # skip header.
+        return np.fromfile(f, dtype=np.float32).reshape(h, w)
+
+
+def train():
+    paddle.init(use_gpu=False, trainer_count=1)
+
+    # define network topology
+    feature_out = db_lstm()
+    target = paddle.layer.data(name='target', type=d_type(label_dict_len))
     crf_cost = paddle.layer.crf(size=label_dict_len,
                                 input=feature_out,
                                 label=target,
@@ -120,29 +137,15 @@ def db_lstm():
                                     learning_rate=mix_hidden_lr))
 
     crf_dec = paddle.layer.crf_decoding(
-        name='crf_dec_l',
         size=label_dict_len,
         input=feature_out,
         label=target,
         param_attr=paddle.attr.Param(name='crfw'))
-
-    return crf_cost, crf_dec
-
-
-def load_parameter(file_name, h, w):
-    with open(file_name, 'rb') as f:
-        f.read(16)  # skip header.
-        return np.fromfile(f, dtype=np.float32).reshape(h, w)
-
-
-def main():
-    paddle.init(use_gpu=False, trainer_count=1)
-
-    # define network topology
-    crf_cost, crf_dec = db_lstm()
+    evaluator.sum(input=crf_dec)
 
     # create parameters
-    parameters = paddle.parameters.create([crf_cost, crf_dec])
+    parameters = paddle.parameters.create(crf_cost)
+    parameters.set('emb', load_parameter(conll05.get_embedding(), 44068, 32))
 
     # create optimizer
     optimizer = paddle.optimizer.Momentum(
@@ -152,18 +155,12 @@ def main():
         model_average=paddle.optimizer.ModelAverage(
             average_window=0.5, max_average_window=10000), )
 
-    def event_handler(event):
-        if isinstance(event, paddle.event.EndIteration):
-            if event.batch_id % 100 == 0:
-                print "Pass %d, Batch %d, Cost %f, %s" % (
-                    event.pass_id, event.batch_id, event.cost, event.metrics)
-
     trainer = paddle.trainer.SGD(cost=crf_cost,
                                  parameters=parameters,
-                                 update_equation=optimizer)
-    parameters.set('emb', load_parameter(conll05.get_embedding(), 44068, 32))
+                                 update_equation=optimizer,
+                                 extra_layers=crf_dec)
 
-    trn_reader = paddle.batch(
+    reader = paddle.batch(
         paddle.reader.shuffle(
             conll05.test(), buf_size=8192), batch_size=10)
 
@@ -179,12 +176,102 @@ def main():
         'target': 8
     }
 
+    def event_handler(event):
+        if isinstance(event, paddle.event.EndIteration):
+            if event.batch_id % 100 == 0:
+                logger.info("Pass %d, Batch %d, Cost %f, %s" % (
+                    event.pass_id, event.batch_id, event.cost, event.metrics))
+            if event.batch_id and event.batch_id % 1000 == 0:
+                result = trainer.test(reader=reader, feeding=feeding)
+                logger.info("\nTest with Pass %d, Batch %d, %s" %
+                            (event.pass_id, event.batch_id, result.metrics))
+
+        if isinstance(event, paddle.event.EndPass):
+            # save parameters
+            with gzip.open('params_pass_%d.tar.gz' % event.pass_id, 'w') as f:
+                parameters.to_tar(f)
+
+            result = trainer.test(reader=reader, feeding=feeding)
+            logger.info("\nTest with Pass %d, %s" %
+                        (event.pass_id, result.metrics))
+
     trainer.train(
-        reader=trn_reader,
+        reader=reader,
         event_handler=event_handler,
-        num_passes=10000,
+        num_passes=10,
         feeding=feeding)
 
 
+def infer_a_batch(inferer, test_data, word_dict, pred_dict, label_dict):
+    probs = inferer.infer(input=test_data, field='id')
+    assert len(probs) == sum(len(x[0]) for x in test_data)
+
+    for idx, test_sample in enumerate(test_data):
+        start_id = 0
+        pred_str = "%s\t" % (pred_dict[test_sample[6][0]])
+
+        for w, tag in zip(test_sample[0],
+                          probs[start_id:start_id + len(test_sample[0])]):
+            pred_str += "%s[%s] " % (word_dict[w], label_dict[tag])
+        print(pred_str.strip())
+        start_id += len(test_sample[0])
+
+
+def infer():
+    label_dict_reverse = dict((value, key)
+                              for key, value in label_dict.iteritems())
+    word_dict_reverse = dict((value, key)
+                             for key, value in word_dict.iteritems())
+    pred_dict_reverse = dict((value, key)
+                             for key, value in verb_dict.iteritems())
+
+    test_creator = paddle.dataset.conll05.test()
+
+    paddle.init(use_gpu=False, trainer_count=1)
+
+    # define network topology
+    feature_out = db_lstm()
+    predict = paddle.layer.crf_decoding(
+        size=label_dict_len,
+        input=feature_out,
+        param_attr=paddle.attr.Param(name='crfw'))
+
+    test_pass = 0
+    with gzip.open('params_pass_%d.tar.gz' % (test_pass)) as f:
+        parameters = paddle.parameters.Parameters.from_tar(f)
+        inferer = paddle.inference.Inference(
+            output_layer=predict, parameters=parameters)
+
+        # prepare test data
+        test_data = []
+        test_batch_size = 50
+
+        for idx, item in enumerate(test_creator()):
+            test_data.append(item[0:8])
+
+            if idx and (not idx % test_batch_size):
+                infer_a_batch(
+                    inferer,
+                    test_data,
+                    word_dict_reverse,
+                    pred_dict_reverse,
+                    label_dict_reverse, )
+                test_data = []
+        infer_a_batch(
+            inferer,
+            test_data,
+            word_dict_reverse,
+            pred_dict_reverse,
+            label_dict_reverse, )
+        test_data = []
+
+
+def main(is_inferring=False):
+    if is_inferring:
+        infer()
+    else:
+        train()
+
+
 if __name__ == '__main__':
-    main()
+    main(is_inferring=False)

doc/howto/cross_compiling/cross_compiling_for_android_cn.md

+# 构建Android平台上的PaddlePaddle库
+
+用户可通过交叉编译的方式，在用户熟悉的开发平台（Linux，Mac OS X和Windows）上编译Android平台上适用的PaddlePaddle库。
+本文档将以Linux x86-64平台为例，介绍交叉编译Android平台上适用的PaddlePaddle库的方法和步骤。
+
+## 准备交叉编译环境
+
+从源码交叉编译PaddlePaddle，用户需要提前准备好交叉编译环境。Android平台上使用的C/C++交叉编译工具链为[Android NDK](https://developer.android.com/ndk/downloads/index.html?hl=zh-cn)，用户可自行前往下载预编译好的版本，也可通过以下命令获取：
+
+```bash
+wget -q https://dl.google.com/android/repository/android-ndk-r14b-linux-x86_64.zip
+unzip -q android-ndk-r14b-linux-x86_64.zip
+```
+
+Android NDK中包含了所有Android API级别、所有架构（arm/arm64/x86/mips）需要用到的编译工具和系统库。用户可根据自己的编译目标架构、所需支持的最低Android API级别，构建[独立工具链](https://developer.android.google.cn/ndk/guides/standalone_toolchain.html?hl=zh-cn)。
+比如：
+
+```bash
+your/path/to/android-ndk-r14b-linux-x86_64/build/tools/make-standalone-toolchain.sh \
+        --arch=arm --platform=android-21 --install-dir=your/path/to/my_standalone_toolchain
+```
+
+此命令将在your/path/to/my_standalone_toolchain目录生成一套编译工具链，面向架构为32位ARM架构，支持的最小的Android API级别为21，使用的编译器为arm-linux-androideabi-gcc (GCC) 4.9。
+
+注意：**PaddlePaddle要求使用的编译工具链所支持的Andoid API级别不小于21**。
+
+## 配置交叉编译参数
+
+CMake系统对交叉编译提供了支持[cmake-toolchains](https://cmake.org/cmake/help/v3.0/manual/cmake-toolchains.7.html#cross-compiling)。为了简化cmake配置，PaddlePaddle为交叉编译提供了工具链配置文档[cmake/cross_compiling/android.cmake](https://github.com/PaddlePaddle/Paddle/blob/develop/cmake/cross_compiling/android.cmake)，以提供一些默认的编译器和编译参数相关配置。注意，从CMake 3.7版本开始，CMake官方对Android平台的交叉编译提供了通用的支持。PaddlePaddle若检测到用户使用的CMake版本不低于3.7时，将会将用户传进来的配置参数传递CMake系统，交由CMake系统本身来处理。有关参数配置的详细说明见[cmake-toolchains](https://cmake.org/cmake/help/v3.7/manual/cmake-toolchains.7.html#cross-compiling)。
+
+交叉编译Android版本的PaddlePaddle库时，有一些必须配置的参数：
+- `CMAKE_SYSTEM_NAME`，CMake编译的目标平台，必须设置为`Android`。在设置`CMAKE_SYSTEM_NAME=Android`后，PaddlePaddle的CMake系统才认为是在交叉编译Android系统的版本，并自动编译宿主机版protoc可执行文件、目标机版protobuf库、以及Android所需`arm_soft_fp_abi`分支的目标机版OpenBLAS库。此外，还会强制设置一些PaddlePaddle参数的值（`WITH_GPU=OFF`、`WITH_AVX=OFF`、`WITH_PYTHON=OFF`、`WITH_RDMA=OFF`）。
+- `WITH_C_API`，必须设置为`ON`。在Android平台上只支持使用C-API来预测。
+- `WITH_SWIG_PY`，必须设置为`OFF`。在Android平台上不支持通过swig调用来训练或者预测。
+
+Android平台可选配置参数：
+
+- `ANDROID_STANDALONE_TOOLCHAIN`，独立工具链所在的绝对路径，或者相对于构建目录的相对路径。PaddlePaddle的CMake系统将根据该值自动推导和设置需要使用的交叉编译器、sysroot、以及Android API级别；否则，用户需要在cmake时手动设置这些值。无默认值。
+- `ANDROID_ABI`，目标架构ABI。目前只支持`armeabi-v7a`，默认值为`armeabi-v7a`。
+- `ANDROID_NATIVE_API_LEVEL`，工具链的Android API级别。若没有显式设置，PaddlePaddle将根据`ANDROID_STANDALONE_TOOLCHAIN`的值自动推导得到。
+- `ANROID_ARM_MODE`，是否使用ARM模式。可设置`ON/OFF`，默认值为`ON`。
+- `ANDROID_ARM_NEON`，是否使用NEON指令。目前必须设置成`ON`，默认值为`ON`。
+
+其他配置参数：
+
+- `HOST_C/CXX_COMPILER`，宿主机的C/C++编译器。在编译宿主机版protoc可执行文件和目标机版OpenBLAS库时需要用到。默认设置成环境变量`CC`的值；若环境变量`CC`没有设置，则设置成`cc`编译器。
+
+一种常用的cmake配置如下：
+
+```bash
+cmake -DCMAKE_SYSTEM_NAME=Android \
+      -DANDROID_STANDALONE_TOOLCHAIN=your/path/to/my_standalone_toolchain \
+      -DANDROID_ABI=armeabi-v7a \
+      -DANDROID_ARM_NEON=ON \
+      -DANDROID_ARM_MODE=ON \
+      -DCMAKE_INSTALL_PREFIX=your/path/to/install \
+      -DWITH_C_API=ON \
+      -DWITH_SWIG_PY=OFF \
+      ..
+```
+
+用户还可根据自己的需求设置其他编译参数。比如希望最小化生成的库的大小，可以设置`CMAKE_BUILD_TYPE`为`MinSizeRel`；若希望最快的执行速度，则可设置`CMAKE_BUILD_TYPE`为`Release`。亦可以通过手动设置`CMAKE_C/CXX_FLAGS_MINSIZEREL/RELEASE`来影响PaddlePaddle的编译过程。
+
+## 编译和安装
+
+CMake配置完成后，执行以下命令，PaddlePaddle将自动下载和编译所有第三方依赖库、编译和安装PaddlePaddle预测库。
+
+```bash
+make
+make install
+```
+
+注意：如果你曾经在源码目录下编译过其他平台的PaddlePaddle库，请先使用`rm -rf`命令删除`third_party`目录和`build`目录，以确保所有的第三方依赖库和PaddlePaddle代码都是针对新的CMake配置重新编译的。
+
+执行完安装命令后，`your/path/to/install`目录中会包含`include`和`lib`目录，其中`include`中包含C-API的头文件，`lib`中包含一个Android版本的库。自此，PaddlePaddle的已经安装完成，用户可将`your/path/to/install`目录下的生成文件用于深度学习相关Android App中，调用方法见C-API文档。

doc/howto/cross_compiling/cross_compiling_for_raspberry_cn.md

+# 构建Raspberry Pi平台上的PaddlePaddle库
+
+对于Rasspberry Pi系统，用户可通过ssh等方式登录到Raspberry Pi系统上，按照[源码编译PaddlePaddle](http://www.paddlepaddle.org/doc_cn/getstarted/build_and_install/cmake/build_from_source_cn.html)相关文档所述，直接编译Raspberry Pi平台上适用的PaddlePaddle库。
+
+用户也可以在自己熟悉的开发平台上，通过交叉编译的方式来编译。这篇文档将以Linux x86-64平台为例，介绍交叉编译Raspberry Pi平台上适用的PaddlePaddle的方法和步骤。
+
+## 准备交叉编译环境
+
+从源码交叉编译PaddlePaddle，用户需要提前准备好交叉编译环境。用户可自行前往[github](https://github.com/raspberrypi/tools)下载Raspberry Pi平台使用的C/C++交叉编译工具链，也可通过以下命令获取：
+
+```bash
+git clone https://github.com/raspberrypi/tools.git
+```
+
+该github仓库中包含若干个预编译好的、针对不同平台的编译工具。宿主机是Linux x86-64环境，则需选用`arm-bcm2708/gcc-linaro-arm-linux-gnueabihf-raspbian-x64`下的作为编译工具，所使用的编译器为arm-linux-gnueabihf-gcc 4.8.3。
+
+注意，该编译工具链需要系统glibc支持2.14以上。
+
+## 配置交叉编译参数
+
+CMake系统对交叉编译提供了支持[cmake-toolchains](https://cmake.org/cmake/help/v3.0/manual/cmake-toolchains.7.html#cross-compiling)。为了简化cmake配置，PaddlePaddle为交叉编译提供了工具链配置文档[cmake/cross_compiling/raspberry_pi.cmake](https://github.com/PaddlePaddle/Paddle/blob/develop/cmake/cross_compiling/raspberry_pi.cmake)，以提供一些默认的编译器和编译参数相关配置。
+
+交叉编译Raspberry Pi版本PaddlePaddle库时，有一些必须配置的参数：
+
+- `CMAKE_SYSTEM_NAME`，CMake编译的目标平台，必须配置为`RPi`。在设置`CMAKE_SYSTEM_NAME=RPi`后，PaddlePaddle的CMake系统才认为在是在交叉编译Raspberry Pi系统的版本，并自动编译宿主机版protoc可执行文件、目标机版protobuf库、以及目标机版OpenBLAS库。
+
+Raspberry Pi平台可选配置参数：
+
+- `RPI_TOOLCHAIN`，编译工具链所在的绝对路径，或者相对于构建目录的相对路径。PaddlePaddle的CMake系统将根据该值自动设置需要使用的交叉编译器；否则，用户需要在cmake时手动设置这些值。无默认值。
+- `RPI_ARM_NEON`，是否使用NEON指令。目前必须设置成`ON`，默认值为`ON`。
+
+其他配置参数：
+
+- `HOST_C/CXX_COMPILER`，宿主机的C/C++编译器。在编译宿主机版protoc可执行文件和目标机版OpenBLAS库时需要用到。默认设置成环境变量`CC`的值；若环境变量`CC`没有设置，则设置成`cc`编译器。
+
+cmake参数如下；
+
+```
+cmake -DCMAKE_SYSTEM_NAME=RPi \
+      -DRPI_TOOLCHAIN=your/path/to/arm-bcm2708/gcc-linaro-arm-linux-gnueabihf-raspbian-x64 \
+      -DRPI_ARM_NEON=ON \
+      -DCMAKE_INSTALL_PREFIX=your/path/to/install \
+      -DWITH_GPU=OFF \
+      -DWITH_C_API=ON \
+      -DWITH_PYTHON=OFF \
+      -DWITH_SWIG_PY=OFF \
+      ..
+```
+
+用户还可根据自己的需求设置其他编译参数。比如希望最小化生成的库的大小，可以设置`CMAKE_BUILD_TYPE`为`MinSizeRel`；若希望最快的执行速度，则可设置`CMAKE_BUILD_TYPE`为`Release`。亦可以通过手动设置`CMAKE_C/CXX_FLAGS_MINSIZEREL/RELEASE`来影响PaddlePaddle的编译过程。
+
+## 编译和安装
+
+CMake配置完成后，执行以下命令，PaddlePaddle将自动下载和编译所有第三方依赖库、编译和安装PaddlePaddle。
+
+```bash
+make
+make install
+```
+
+注意：如果你曾经在源码目录下编译过其他平台的PaddlePaddle库，请先使用`rm -rf`命令删除`third_party`目录和`build`目录，以确保所有的第三方依赖库和PaddlePaddle代码都是针对新的CMake配置重新编译的。
+
+执行完安装命令后，由于上一步cmake配置中`WITH_C_API`设置为`ON`，`your/path/to/install`目录中会包含`include`和`lib`目录，其中`include`中包含C-API的头文件，`lib`中包含一个Raspberry Pi版本的库。
+
+更多的编译配置见[源码编译PaddlePaddle](http://www.paddlepaddle.org/doc_cn/getstarted/build_and_install/cmake/build_from_source_cn.html)相关文档。

paddle/api/CMakeLists.txt

@@ -26,7 +26,7 @@ FILE(GLOB PY_PADDLE_PYTHON_FILES ${PROJ_ROOT}/paddle/py_paddle/*.py)
 SET_SOURCE_FILES_PROPERTIES(Paddle.i PROPERTIES CPLUSPLUS ON)
 
 SET(CMAKE_SWIG_OUTDIR ${CMAKE_CURRENT_BINARY_DIR})
-SET(CMAKE_CXX_FLAGS "-std=c++11 -fPIC -Wall")
+SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -fPIC -Wall")
 IF(WITH_COVERAGE)
     SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -g -O0 -fprofile-arcs -ftest-coverage")
 ENDIF(WITH_COVERAGE)

paddle/majel/CMakeLists.txt

-cmake_minimum_required(VERSION 3.0)
-
-if(${CMAKE_CURRENT_SOURCE_DIR} STREQUAL ${CMAKE_SOURCE_DIR})
-    # find #include <majel/xx.h>
-    get_filename_component(PARENT_DIR ${CMAKE_CURRENT_SOURCE_DIR} DIRECTORY)
-    include_directories(${PARENT_DIR})
-
-    # find cmake directory modules
-    get_filename_component(PARENT_DIR ${PARENT_DIR} DIRECTORY)
-    set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${PARENT_DIR}/cmake")
-
-    # enable c++11
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
-
-    # enable gtest
-    set(THIRD_PARTY_PATH ${CMAKE_CURRENT_SOURCE_DIR}/third_party)
-    set(WITH_TESTING ON)
-    include(external/gtest)
-else()
-    message("-- Found gtest (include: ${GTEST_INCLUDE_DIR}, library: ${GTEST_LIBRARIES})")
-endif()
-
-########################### Build Majel #############################
-set(MAJEL_CXX_FILES place.cc)
-set(MAJEL_CUDA_FILES "")
-
-if(CUDA_FOUND)
-    cuda_add_library(majel ${MAJEL_CUDA_FILES} ${MAJEL_CXX_FILES})
-else()
-    add_library(majel ${MAJEL_CXX_FILES})
-endif()
-add_dependencies(majel ${external_project_dependencies})
-#####################################################################
+cc_library(majel SRCS place.cc)
 
 if(WITH_TESTING)
     add_subdirectory(test)

paddle/majel/README.md

@@ -93,6 +93,19 @@ typedef boost::variant<
 
 Because `variant` may be thought of as "multi-type, single value", we can utilize it to implement unified interfaces for PaddlePaddle.
 
+`DDim` plays two kinds of roles in Majel. First, it is used to indicate the size of a tensor. For example, we can construct a new `DArray` by following way:
+ 
+ ```c++
+ DArray arr = make_darray(make_ddim({2,3}), 0.0f);
+ ```
+ It means that `arr` will be a two-dimension tensor, or a matrix. The size of its first dimension is 2 and the second is 3. All the element value of `arr` will be initialized as 0.0 .
+ 
+ The second meaning of `DDim` is tensor index. For example, if we want to access the value in the 1st row and 2nd column of `arr` and set it to 1.0, we can do like this:
+
+ ```c++
+ arr[make_ddim({0, 1})] = 1.0；
+ ```
+
 ## implement Tensor in Paddle
 
 Before writing code, please make sure you already look through Majel Source Code and grabbed the design philosophy of `DArray` in Majel.
@@ -113,7 +126,7 @@ To assign subtasks to our colleagues, we have to discuss how to divide it to ind
 
 - [ ] 3. Re-implement `Dim`.
 
-    `Dim` is an excellent implementation in Majel. 
+    `Dim` is an excellent implementation in Majel.
 
 > ???
 

paddle/majel/test/CMakeLists.txt

-file(GLOB_RECURSE ALL_TEST_FILES RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}" "*.cc")
+cc_test(place_test
+    SRCS place_test.cc
+    DEPS majel)
 
-add_executable(majel_tests ${ALL_TEST_FILES})
-add_dependencies(majel_tests majel)
-target_link_libraries(majel_tests     
-                      ${Boost_LIBRARIES}
-                      ${GTEST_LIBRARIES}
-                      ${GTEST_MAIN_LIBRARIES}
-                      majel
-                     )
-add_test(majel_tests majel_tests)
+if(WITH_GPU)
+    nv_test(cuda_test SRCS cuda_test.cu)
+endif()

paddle/majel/test/cuda_test.cu

+#include <cuda_runtime.h>
+#include <stdio.h>
+#include "gtest/gtest.h"
+
+#define CHECK_ERR(x)                 \
+  if (x != cudaSuccess) {            \
+    fprintf(stderr,                  \
+            "%s in %s at line %d\n", \
+            cudaGetErrorString(err), \
+            __FILE__,                \
+            __LINE__);               \
+    exit(-1);                        \
+  }
+
+__global__ void vecAdd(float *d_A, float *d_B, float *d_C, int n) {
+  int i = blockDim.x * blockIdx.x + threadIdx.x;
+  if (i < n) {
+    d_C[i] = d_A[i] + d_B[i];
+  }
+}
+
+TEST(Cuda, Equality) {
+  int n = 10;
+  // Memory allocation for h_A, h_B and h_C (in the host)
+  float h_A[10] = {1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 0.0};
+  float h_B[10] = {0.0, 9.0, 8.0, 7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0};
+  float h_C[10];
+  float *d_A, *d_B, *d_C;
+  cudaError_t err;
+  // Memory allocation for d_A, d_B and d_C (in the device)
+  err = cudaMalloc((void **)&d_A, sizeof(float) * n);
+  CHECK_ERR(err);
+
+  err = cudaMalloc((void **)&d_B, sizeof(float) * n);
+  CHECK_ERR(err);
+
+  err = cudaMalloc((void **)&d_C, sizeof(float) * n);
+  CHECK_ERR(err);
+
+  // Copying memory to device
+  err = cudaMemcpy(d_A, h_A, sizeof(float) * n, cudaMemcpyHostToDevice);
+  CHECK_ERR(err);
+
+  err = cudaMemcpy(d_B, h_B, sizeof(float) * n, cudaMemcpyHostToDevice);
+  CHECK_ERR(err);
+
+  // Calling the kernel
+  vecAdd<<<ceil(n / 256.0), 256>>>(d_A, d_B, d_C, n);
+
+  // Copying results back to host
+  err = cudaMemcpy(h_C, d_C, sizeof(float) * n, cudaMemcpyDeviceToHost);
+  CHECK_ERR(err);
+
+  EXPECT_EQ(h_C[0], 1.0);
+  for (int i = 1; i < n - 1; ++i) {
+    EXPECT_EQ(h_C[i], 11.0);
+  }
+  EXPECT_EQ(h_C[9], 1.0);
+}

paddle/scripts/travis/docs.sh

@@ -60,7 +60,6 @@ function deploy_docs() {
 
 deploy_docs "master" "." 
 deploy_docs "develop" "./develop/"
-deploy_docs "release/0.10.0" "./release/0.10.0/"
 
 # Check is there anything changed.
 set +e

paddle/utils/CpuId.cpp

@@ -19,19 +19,22 @@ limitations under the License. */
 /// for MSVC
 #define CPUID(info, x) __cpuidex(info, x, 0)
 
-#elif !defined(__ANDROID__)
+#else
 
+#if !defined(__arm__)
 #include <cpuid.h>
-
 /// for GCC/Clang
 #define CPUID(info, x) __cpuid_count(x, 0, info[0], info[1], info[2], info[3])
+#endif
 
 #endif
 
 namespace paddle {
 
 SIMDFlags::SIMDFlags() {
-#if !defined(__ANDROID__)
+#if defined(__arm__)
+  simd_flags_ = SIMD_NEON;
+#else
   unsigned int cpuInfo[4];
   // CPUID: https://en.wikipedia.org/wiki/CPUID
   // clang-format off
@@ -52,8 +55,6 @@ SIMDFlags::SIMDFlags() {
   CPUID(cpuInfo, 0x80000001);
   simd_flags_ |= cpuInfo[2] & (1 << 16) ? SIMD_FMA4  : SIMD_NONE;
   // clang-fotmat on
-#else
-  simd_flags_ = SIMD_NEON;
 #endif
 }
 

python/paddle/trainer/config_parser.py

@@ -2320,6 +2320,9 @@ def Memory(name,
         memory_name = name + "+delay1"
     agent_name = memory_name
     if is_sequence:
+        config_assert(
+            boot_layer is not None,
+            "there must be boot_layer in network when is_sequence = True")
         agent_layer = SequenceAgentLayer(agent_name, size)
     else:
         agent_layer = AgentLayer(agent_name, size)