Merge remote-tracking branch 'baidu/develop' into port_blas

CMakeLists.txt

@@ -164,10 +164,12 @@ if(WITH_GOLANG)
     add_subdirectory(go)
 endif(WITH_GOLANG)
 
+set(PADDLE_PYTHON_BUILD_DIR "${CMAKE_CURRENT_BINARY_DIR}/python/build")
 add_subdirectory(paddle)
 if(WITH_PYTHON)
   add_subdirectory(python)
 endif()
+
 if(WITH_DOC)
     add_subdirectory(doc)
 endif()

Dockerfile

@@ -64,7 +64,7 @@ RUN pip install --upgrade pip && \
     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' && \
-    pip install rarfile
+    pip install opencv-python rarfile 'scipy>=0.19.0' 'nltk>=3.2.2'
 
 # 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

cmake/generic.cmake

@@ -411,7 +411,7 @@ function(py_test TARGET_NAME)
     set(multiValueArgs SRCS DEPS)
     cmake_parse_arguments(py_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})  
     add_test(NAME ${TARGET_NAME}
-             COMMAND env PYTHONPATH=${PADDLE_PYTHON_PACKAGE_DIR}
+             COMMAND env PYTHONPATH=${PADDLE_PYTHON_BUILD_DIR}/lib-python
              python2 ${py_test_SRCS}
              WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
   endif()

paddle/framework/CMakeLists.txt

@@ -35,6 +35,11 @@ py_proto_compile(framework_py_proto SRCS attribute.proto op_proto.proto op_desc.
 # Generate an empty __init__.py to make framework_py_proto as a valid python module.
 add_custom_target(framework_py_proto_init ALL COMMAND ${CMAKE_COMMAND} -E touch __init__.py)
 add_dependencies(framework_py_proto framework_py_proto_init)
+add_custom_command(TARGET framework_py_proto POST_BUILD
+    COMMAND ${CMAKE_COMMAND} -E make_directory ${PROJ_ROOT}/python/paddle/v2/framework/proto
+    COMMAND cp *.py ${PROJ_ROOT}/python/paddle/v2/framework/proto/
+    COMMENT "Copy generated python proto into directory paddle/v2/framework/proto."
+    WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
 
 cc_library(backward SRCS backward.cc DEPS net_op)
 cc_test(backward_test SRCS backward_test.cc DEPS backward)

paddle/framework/attribute.h

@@ -14,7 +14,6 @@ limitations under the License. */
 
 #pragma once
 
-#include <boost/variant.hpp>
 #include <functional>
 #include <string>
 #include <unordered_map>
@@ -24,6 +23,7 @@ limitations under the License. */
 #include "paddle/framework/attribute.pb.h"
 #include "paddle/framework/op_desc.pb.h"
 #include "paddle/platform/enforce.h"
+#include "paddle/platform/variant.h"
 
 namespace paddle {
 namespace framework {

paddle/framework/ddim.h

@@ -14,12 +14,12 @@ limitations under the License. */
 
 #pragma once
 
-#include <boost/variant.hpp>
 #include <initializer_list>
 #include <stdexcept>
 #include <vector>
 #include "paddle/framework/dim.h"
 #include "paddle/platform/enforce.h"
+#include "paddle/platform/variant.h"
 #include "unsupported/Eigen/CXX11/Tensor"
 
 namespace paddle {

paddle/framework/operator.h

@@ -15,7 +15,6 @@ limitations under the License. */
 #pragma once
 
 #include <algorithm>
-#include <boost/variant.hpp>
 #include <string>
 #include <unordered_map>
 #include <vector>
@@ -27,6 +26,7 @@ limitations under the License. */
 #include "paddle/framework/tensor.h"
 #include "paddle/platform/device_context.h"
 #include "paddle/platform/place.h"
+#include "paddle/platform/variant.h"
 #include "paddle/utils/Error.h"
 
 namespace paddle {

paddle/framework/tensor.h

@@ -79,11 +79,11 @@ class Tensor {
   inline const DDim& dims() const;
 
   /*! Resize the dimensions of the memory block. */
-  inline void Resize(const DDim& dims);
+  inline Tensor& Resize(const DDim& dims);
 
   /*! The internal of two tensors share the same memory block. */
   template <typename T>
-  inline void ShareDataWith(const Tensor& src);
+  inline Tensor& ShareDataWith(const Tensor& src);
 
   /**
    * @brief   Copy the content of external tensor to a new place.

paddle/framework/tensor_impl.h

@@ -23,9 +23,11 @@ template <typename T>
 inline void Tensor::check_memory_size() const {
   PADDLE_ENFORCE_NOT_NULL(
       holder_, "Tenosr holds no memory. Call Tensor::mutable_data first.");
-  PADDLE_ENFORCE_GE(holder_->size(), product(dims_) * sizeof(T) + offset_,
+  PADDLE_ENFORCE_GE(
+      holder_->size(), product(dims_) * sizeof(T) + offset_,
       "Tensor's dims_ is out of bound. Call Tensor::mutable_data "
-                    "first to re-allocate memory.");
+      "first to re-allocate memory.\n"
+      "or maybe the required data-type mismatches the data already stored.");
 }
 
 template <typename T>
@@ -78,9 +80,10 @@ inline T* Tensor::mutable_data(platform::Place place) {
 }
 
 template <typename T>
-inline void Tensor::ShareDataWith(const Tensor& src) {
+inline Tensor& Tensor::ShareDataWith(const Tensor& src) {
   src.check_memory_size<T>();
   *this = src;
+  return *this;
 }
 
 template <typename T>
@@ -136,7 +139,10 @@ inline Tensor Tensor::Slice(const int& begin_idx, const int& end_idx) const {
   return dst;
 }
 
-inline void Tensor::Resize(const DDim& dims) { dims_ = dims; }
+inline Tensor& Tensor::Resize(const DDim& dims) {
+  dims_ = dims;
+  return *this;
+}
 
 inline const DDim& Tensor::dims() const { return dims_; }
 

paddle/operators/gaussian_random_op.cu

@@ -40,8 +40,8 @@ class GaussianRandomKernel : public framework::OpKernel {
         &g, CURAND_RNG_PSEUDO_DEFAULT));
     PADDLE_ENFORCE(
         platform::dynload::curandSetPseudoRandomGeneratorSeed(g, seed));
-    curandGenerateNormal(g, data, framework::product(tensor->dims()), mean,
-                         std);
+    platform::dynload::curandGenerateNormal(
+        g, data, framework::product(tensor->dims()), mean, std);
   }
 };
 

paddle/platform/dynload/cublas.h

@@ -50,8 +50,8 @@ extern void *cublas_dso_handle;
 #else
 #define DECLARE_DYNAMIC_LOAD_CUBLAS_WRAP(__name)     \
   struct DynLoad__##__name {                         \
-    inline template <typename... Args>           \
-    cublasStatus_t operator()(Args... args) {    \
+    template <typename... Args>                      \
+    inline cublasStatus_t operator()(Args... args) { \
       return __name(args...);                        \
     }                                                \
   };                                                 \

paddle/platform/dynload/curand.h

@@ -55,6 +55,7 @@ extern void *curand_dso_handle;
   __macro(curandSetPseudoRandomGeneratorSeed); \
   __macro(curandGenerateUniform);              \
   __macro(curandGenerateUniformDouble);        \
+  __macro(curandGenerateNormal);               \
   __macro(curandDestroyGenerator);
 
 CURAND_RAND_ROUTINE_EACH(DECLARE_DYNAMIC_LOAD_CURAND_WRAP);

paddle/platform/place.h

@@ -14,8 +14,8 @@ limitations under the License. */
 
 #pragma once
 
-#include <boost/variant.hpp>
 #include <iostream>
+#include "paddle/platform/variant.h"
 
 namespace paddle {
 namespace platform {

paddle/platform/variant.h

+/* 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. */
+
+#pragma once
+
+#include <boost/config.hpp>
+
+#ifndef PADDLE_ONLY_CPU
+
+// Because boost's variadic templates has bug on nvcc, boost will disable
+// variadic template support when GPU enabled on nvcc.
+// Define BOOST_NO_CXX11_VARIADIC_TEMPLATES on gcc/clang to generate same
+// function symbols.
+//
+// https://github.com/PaddlePaddle/Paddle/issues/3386
+#ifndef BOOST_NO_CXX11_VARIADIC_TEMPLATES
+#define BOOST_NO_CXX11_VARIADIC_TEMPLATES
+#endif
+#endif
+
+#include <boost/variant.hpp>

paddle/scripts/docker/build.sh

@@ -74,11 +74,11 @@ cat <<EOF
 Running unit tests ...
 ========================================
 EOF
+    ctest --output-on-failure
     # make install should also be test when unittest
     make install -j `nproc`
     pip install /usr/local/opt/paddle/share/wheels/*.whl
     paddle version
-    ctest --output-on-failure
 fi
 
 

python/CMakeLists.txt

-set(OUTPUT_DIR
-    "${CMAKE_CURRENT_BINARY_DIR}/build")
 
 file(GLOB TRAINER_PY_FILES . ./paddle/trainer/*.py)
 file(GLOB HELPERS_PY_FILES . ./paddle/trainer_config_helpers/*.py)
@@ -33,13 +31,15 @@ add_custom_command(OUTPUT ${PROJ_ROOT}/python/paddle/v2/framework/core.so
 add_custom_target(copy_paddle_pybind ALL DEPENDS ${PROJ_ROOT}/python/paddle/v2/framework/core.so)
 
 
-add_custom_command(OUTPUT ${OUTPUT_DIR}/.timestamp
+add_custom_command(OUTPUT ${PADDLE_PYTHON_BUILD_DIR}/.timestamp
     COMMAND env ${py_env} ${PYTHON_EXECUTABLE} setup.py bdist_wheel
-    COMMAND ${CMAKE_COMMAND} -E touch ${OUTPUT_DIR}/.timestamp
+    COMMAND ${CMAKE_COMMAND} -E touch ${PADDLE_PYTHON_BUILD_DIR}/.timestamp
+    COMMAND ${CMAKE_COMMAND} -E remove_directory ${PADDLE_PYTHON_BUILD_DIR}/lib-python
+    COMMAND ${CMAKE_COMMAND} -E copy_directory ${PADDLE_PYTHON_BUILD_DIR}/lib* ${PADDLE_PYTHON_BUILD_DIR}/lib-python
     DEPENDS gen_proto_py copy_paddle_pybind framework_py_proto ${PY_FILES} ${external_project_dependencies} ${COPY_PADDLE_MASTER})
 
 add_custom_target(paddle_python ALL DEPENDS
-    ${OUTPUT_DIR}/.timestamp paddle_pserver_main paddle_trainer paddle_merge_model python_api_wheel)
+    ${PADDLE_PYTHON_BUILD_DIR}/.timestamp paddle_pserver_main paddle_trainer paddle_merge_model python_api_wheel)
 
 set(PADDLE_PYTHON_PACKAGE_DIR ${CMAKE_CURRENT_BINARY_DIR}/dist/)
 

python/paddle/v2/framework/.gitignore

+proto

python/paddle/v2/framework/tests/CMakeLists.txt

@@ -23,6 +23,5 @@ py_test(test_rowwise_add_op SRCS test_rowwise_add_op.py)
 py_test(test_default_scope_funcs SRCS test_default_scope_funcs.py)
 
 py_test(test_operator SRCS test_operator.py)
-
-py_test(test_gaussian_random_op SRCS test_gaussian_random_op.py)
+# py_test(test_gaussian_random_op SRCS test_gaussian_random_op.py)
 py_test(test_uniform_random_op SRCS test_uniform_random_op.py)