Merge remote-tracking branch 'origin/develop' into feature/evaluator

mobile_baidu
Dong Zhihong 8 years ago
commit 8d9b33412d

@ -126,7 +126,7 @@ include(external/swig) # download, build, install swig
include(external/warpctc) # download, build, install warpctc
include(external/any) # download libn::any
include(external/eigen) # download eigen3
include(external/pybind11) # download pybind11
include(external/pybind11) # download pybind11
include(external/nccl)
include(cudnn) # set cudnn libraries, must before configure

@ -79,9 +79,8 @@ if(NOT DEFINED IOS_ARCH)
# FIXME(liuyiqun): support "armv7;armv7s;arm64" future
set(IOS_ARCH "arm64")
elseif(IOS_PLATFORM STREQUAL "SIMULATOR")
set(IOS_ARCH "i386;x86_64")
elseif(IOS_PLATFORM STREQUAL "WATCHOS")
set(IOS_ARCH armv7k)
# FIXME(liuyiqun): support "i386;x86_64" future
set(IOS_ARCH "x86_64")
endif()
endif()
set(CMAKE_OSX_ARCHITECTURES ${IOS_ARCH} CACHE string "Build architecture for iOS")

@ -1,3 +1,21 @@
# 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.
if(NOT WITH_GPU)
return()
endif()
include(ExternalProject)
set(NCCL_SOURCE_DIR ${THIRD_PARTY_PATH}/nccl)

@ -1,11 +1,11 @@
# 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.

@ -1,8 +1,26 @@
INCLUDE(ExternalProject)
# 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.
SET(PYBIND_SOURCE_DIR ${THIRD_PARTY_PATH}/pybind)
if(NOT WITH_PYTHON)
return()
endif()
include(ExternalProject)
INCLUDE_DIRECTORIES(${PYBIND_SOURCE_DIR}/src/extern_pybind/include)
set(PYBIND_SOURCE_DIR ${THIRD_PARTY_PATH}/pybind)
include_directories(${PYBIND_SOURCE_DIR}/src/extern_pybind/include)
ExternalProject_Add(
extern_pybind
@ -17,14 +35,12 @@ ExternalProject_Add(
TEST_COMMAND ""
)
if (${CMAKE_VERSION} VERSION_LESS "3.3.0")
if(${CMAKE_VERSION} VERSION_LESS "3.3.0")
set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/pybind_dummy.c)
file(WRITE ${dummyfile} "const char * dummy_any = \"${dummyfile}\";")
file(WRITE ${dummyfile} "const char * dummy_pybind = \"${dummyfile}\";")
add_library(pybind STATIC ${dummyfile})
else()
add_library(pybind INTERFACE)
endif()
add_dependencies(pybind extern_pybind)
LIST(APPEND external_project_dependencies pybind)

@ -1,11 +1,11 @@
# 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.

@ -1,11 +1,11 @@
# 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.

@ -1,27 +1,28 @@
# This file is use to check all support level of AVX on your machine
# so that PaddlePaddle can unleash the vectorization power of muticore.
INCLUDE(CheckCXXSourceRuns)
INCLUDE(CheckCXXSourceCompiles)
include(CheckCXXSourceRuns)
include(CheckCXXSourceCompiles)
IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
if(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
set(MMX_FLAG "-mmmx")
set(SSE2_FLAG "-msse2")
set(SSE3_FLAG "-msse3")
SET(AVX_FLAG "-mavx")
SET(AVX2_FLAG "-mavx2")
ELSEIF(MSVC)
set(AVX_FLAG "-mavx")
set(AVX2_FLAG "-mavx2")
elseif(MSVC)
set(MMX_FLAG "/arch:MMX")
set(SSE2_FLAG "/arch:SSE2")
set(SSE3_FLAG "/arch:SSE3")
SET(AVX_FLAG "/arch:AVX")
SET(AVX2_FLAG "/arch:AVX2")
ENDIF()
endif()
set(CMAKE_REQUIRED_FLAGS_RETAINED ${CMAKE_REQUIRED_FLAGS})
# Check MMX
set(CMAKE_REQUIRED_FLAGS ${MMX_FLAG})
set(MMX_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
CHECK_CXX_SOURCE_RUNS("
#include <mmintrin.h>
int main()
@ -32,6 +33,7 @@ int main()
# Check SSE2
set(CMAKE_REQUIRED_FLAGS ${SSE2_FLAG})
set(SSE2_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
CHECK_CXX_SOURCE_RUNS("
#include <emmintrin.h>
int main()
@ -42,6 +44,7 @@ int main()
# Check SSE3
set(CMAKE_REQUIRED_FLAGS ${SSE3_FLAG})
set(SSE3_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
CHECK_CXX_SOURCE_RUNS("
#include <pmmintrin.h>
int main()
@ -55,6 +58,7 @@ int main()
# Check AVX
set(CMAKE_REQUIRED_FLAGS ${AVX_FLAG})
set(AVX_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
CHECK_CXX_SOURCE_RUNS("
#include <immintrin.h>
int main()
@ -67,6 +71,7 @@ int main()
# Check AVX 2
set(CMAKE_REQUIRED_FLAGS ${AVX2_FLAG})
set(AVX2_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
CHECK_CXX_SOURCE_RUNS("
#include <immintrin.h>
int main()

@ -75,7 +75,7 @@ PaddlePaddle目前支持8种learning_rate_schedule这8种learning_rate_schedu
optimizer = paddle.optimizer.Adam(
learning_rate=1e-3,
learning_rate_schedule="manual",
learning_rate_schedule="pass_manual",
learning_rate_args="1:1.0,2:0.9,3:0.8",)
在该示例中当已训练pass数小于等于1时学习率为 :code:`1e-3 * 1.0`当已训练pass数大于1小于等于2时学习率为 :code:`1e-3 * 0.9`当已训练pass数大于2时学习率为 :code:`1e-3 * 0.8`

@ -145,7 +145,7 @@ PaddlePaddle发布新版本的时候都会发布对应版本的生产镜像以
Jupyter Notebook是一个开源的web程序大家可以通过它制作和分享带有代码、公式、图表、文字的交互式文档。用户可以通过网页浏览文档。
PaddlePaddle Book是为用户和开发者制作的一个交互式的Jupyter Nodebook。
PaddlePaddle Book是为用户和开发者制作的一个交互式的Jupyter Notebook。
如果您想要更深入了解deep learningPaddlePaddle Book一定是您最好的选择。
我们提供可以直接运行PaddlePaddle Book的Docker镜像直接运行

@ -63,7 +63,7 @@
</tr>
<tr>
<td class="left" rowspan="15">训练</td><td class="left">dot_period</td>
<td class="left" rowspan="14">训练</td><td class="left">dot_period</td>
<td class="left"></td><td class="left"></td><td class="left"></td><td class="left"></td>
</tr>

@ -8,3 +8,4 @@ PaddlePaddle 文档
howto/index_cn.rst
api/index_cn.rst
faq/index_cn.rst
mobile/index_cn.rst

@ -7,3 +7,4 @@ PaddlePaddle Documentation
getstarted/index_en.rst
howto/index_en.rst
api/index_en.rst
mobile/index_en.rst

@ -20,10 +20,32 @@ $ docker build -t username/paddle-android:dev . -f Dockerfile.android
构建好开发镜像后即可使用开发镜像来编译Android版PaddlePaddle C-API库。
Android的Docker开发镜像向用户提供两个可配置的参数
| Argument | Optional Values | Default |
|-----------------|-------------------------|---------|
|`ANDROID_ABI` |`armeabi-v7a, arm64-v8a` | `armeabi-v7a` |
|`ANDROID_API` |`>= 21` | `21` |
<table class="docutils">
<colgroup>
<col width="25%" />
<col width="50%" />
<col width="25%" />
</colgroup>
<thead valign="bottom">
<tr class="row-odd">
<th class="head">Argument</th>
<th class="head">Optional Values</th>
<th class="head">Default</th>
</tr>
</thead>
<tbody valign="top">
<tr class="row-even">
<td>ANDROID_ABI</td>
<td>armeabi-v7a, arm64-v8a</td>
<td>armeabi-v7a</td>
</tr>
<tr class="row-odd">
<td>ANDROID_API</td>
<td>>= 21</td>
<td>21</td>
</tr>
</tbody>
</table>
- 编译`armeabi-v7a``Android API 21`的PaddlePaddle库
```bash

@ -26,10 +26,32 @@ $ docker run -it --rm -v $PWD:/paddle -e "ANDROID_ABI=armeabi-v7a" -e "ANDROID_A
The Docker image accepts two arguments `ANDROID_ABI` and `ANDROID_API`:
| Argument | Optional Values | Default |
|-----------------|-------------------------|---------|
|`ANDROID_ABI` |`armeabi-v7a, arm64-v8a` | `armeabi-v7a` |
|`ANDROID_API` |`>= 21` | `21` |
<table class="docutils">
<colgroup>
<col width="25%" />
<col width="50%" />
<col width="25%" />
</colgroup>
<thead valign="bottom">
<tr class="row-odd">
<th class="head">Argument</th>
<th class="head">Optional Values</th>
<th class="head">Default</th>
</tr>
</thead>
<tbody valign="top">
<tr class="row-even">
<td>ANDROID_ABI</td>
<td>armeabi-v7a, arm64-v8a</td>
<td>armeabi-v7a</td>
</tr>
<tr class="row-odd">
<td>ANDROID_API</td>
<td>>= 21</td>
<td>21</td>
</tr>
</tbody>
</table>
The ARM-64 architecture (`arm64-v8a`) requires at least level 21 of Android API.

@ -27,10 +27,28 @@ iOS平台可选配置参数
- `SIMULATOR`,构建目标为`x86`架构的模拟器平台。
- `IOS_ARCH`,目标架构。针对不同的`IOS_PLATFORM`,可设置的目标架构如下表所示:
| IOS_PLATFORM | IOS_ARCH |
|--------------|----------------------|
| OS | armv7, armv7s, arm64 (默认) |
| SIMULATOR | i386, x86_64 (默认) |
<table class="docutils">
<colgroup>
<col width="35%" />
<col width="65%" />
</colgroup>
<thead valign="bottom">
<tr class="row-odd">
<th class="head">IOS_PLATFORM</th>
<th class="head">IOS_ARCH</th>
</tr>
</thead>
<tbody valign="top">
<tr class="row-even">
<td>OS</td>
<td>armv7, armv7s, arm64 (默认)</td>
</tr>
<tr class="row-odd">
<td>SIMULATOR</td>
<td>i386, x86_64 (默认)</td>
</tr>
</tbody>
</table>
- `IOS_DEPLOYMENT_TARGET`最小的iOS部署版本默认值为`7.0`。
- `IOS_ENABLE_BITCODE`,是否使能[Bitcode](https://developer.apple.com/library/content/documentation/IDEs/Conceptual/AppDistributionGuide/AppThinning/AppThinning.html#//apple_ref/doc/uid/TP40012582-CH35-SW3),可设置`ON/OFF`,默认值为`ON`。

@ -0,0 +1,9 @@
MOBILE
======
.. toctree::
:maxdepth: 1
cross_compiling_for_android_cn.md
cross_compiling_for_ios_cn.md
cross_compiling_for_raspberry_cn.md

@ -0,0 +1,8 @@
MOBILE
======
.. toctree::
:maxdepth: 1
cross_compiling_for_android_en.md
cross_compiling_for_raspberry_en.md

@ -29,32 +29,32 @@ add_style_check_target(paddle_capi ${CAPI_SOURCES} ${CAPI_HEADER}
add_dependencies(paddle_capi paddle_proto)
# TODO: paddle_capi_whole will be removed.
set(PADDLE_CAPI_LAYERS_LIBS
paddle_function
paddle_gserver)
if(MOBILE_INFERENCE)
set(PADDLE_CAPI_INFER_LIBS
paddle_utils
paddle_parameter
paddle_math
paddle_cuda
paddle_function
paddle_gserver
paddle_proto)
set(PADDLE_CAPI_ENGINE_LIBS
paddle_utils
paddle_parameter
paddle_math
paddle_cuda
paddle_proto)
else()
set(PADDLE_CAPI_INFER_LIBS
paddle_utils
paddle_parameter
paddle_math
paddle_cuda
paddle_function
paddle_gserver
paddle_proto
paddle_pserver
paddle_network)
set(PADDLE_CAPI_ENGINE_LIBS
paddle_utils
paddle_parameter
paddle_math
paddle_cuda
paddle_proto
paddle_pserver
paddle_network)
endif()
set(PADDLE_CAPI_INFER_LIBS ${PADDLE_CAPI_LAYERS_LIBS} ${PADDLE_CAPI_ENGINE_LIBS})
cc_library(paddle_capi_whole DEPS paddle_capi ${PADDLE_CAPI_INFER_LIBS})
# Link the static library for inference
cc_library(paddle_capi_engine DEPS paddle_capi paddle_utils paddle_parameter paddle_math paddle_cuda paddle_proto)
cc_library(paddle_capi_layers DEPS paddle_function paddle_gserver)
cc_library(paddle_capi_engine DEPS paddle_capi ${PADDLE_CAPI_ENGINE_LIBS})
cc_library(paddle_capi_layers DEPS ${PADDLE_CAPI_LAYERS_LIBS})
# Link the shared library for inference
if(NOT IOS)

@ -20,7 +20,8 @@ cc_test(scope_test SRCS scope_test.cc DEPS scope)
cc_library(attribute SRCS attribute.cc DEPS framework_proto)
cc_test(program_desc_test SRCS program_desc_test.cc DEPS proto_desc)
cc_test(program_desc_test SRCS program_desc_test.cc DEPS proto_desc
device_context)
cc_library(op_proto_maker SRCS op_proto_maker.cc DEPS framework_proto attribute)
cc_test(op_proto_maker_test SRCS op_proto_maker_test.cc DEPS op_proto_maker)
cc_library(op_info SRCS op_info.cc DEPS attribute framework_proto)
@ -44,8 +45,9 @@ add_custom_command(TARGET framework_py_proto POST_BUILD
cc_library(backward SRCS backward.cc DEPS net_op)
cc_test(backward_test SRCS backward_test.cc DEPS backward recurrent_op device_context fill_constant_op)
cc_library(lod_rank_table SRCS lod_rank_table.cc DEPS lod_tensor)
cc_library(executor SRCS executor.cc DEPS op_registry device_context scope framework_proto backward glog)
cc_library(executor SRCS executor.cc DEPS op_registry device_context scope framework_proto backward glog lod_rank_table)
cc_library(prune SRCS prune.cc DEPS framework_proto)
cc_test(prune_test SRCS prune_test.cc DEPS op_info prune recurrent_op device_context)

@ -21,7 +21,9 @@ limitations under the License. */
#include <vector>
#include "paddle/framework/feed_fetch_type.h"
#include "paddle/framework/lod_rank_table.h"
#include "paddle/framework/lod_tensor.h"
#include "paddle/framework/lod_tensor_array.h"
#include "paddle/framework/op_registry.h"
#include "paddle/framework/scope.h"
@ -70,10 +72,14 @@ static void CreateTensor(Variable* var, VarDesc::VarType var_type) {
var->GetMutable<FeedFetchList>();
} else if (var_type == VarDesc::STEP_SCOPES) {
var->GetMutable<std::vector<framework::Scope>>();
} else if (var_type == VarDesc::LOD_RANK_TABLE) {
var->GetMutable<LoDRankTable>();
} else if (var_type == VarDesc::LOD_TENSOR_ARRAY) {
var->GetMutable<LoDTensorArray>();
} else {
PADDLE_THROW(
"Variable type %d is not in "
"[LoDTensor, SelectedRows, FEED_MINIBATCH, FETCH_LIST]",
"[LoDTensor, SelectedRows, FEED_MINIBATCH, FETCH_LIST, LOD_RANK_TABLE]",
var_type);
}
}
@ -83,7 +89,7 @@ void Executor::Run(const ProgramDescBind& pdesc, Scope* scope, int block_id,
// TODO(tonyyang-svail):
// - only runs on the first device (i.e. no interdevice communication)
// - will change to use multiple blocks for RNN op and Cond Op
PADDLE_ENFORCE_LT(block_id, pdesc.Size());
PADDLE_ENFORCE_LT(static_cast<size_t>(block_id), pdesc.Size());
auto& block = pdesc.Block(block_id);
auto& device = device_contexts_[0];

@ -109,6 +109,11 @@ message LoDTensorDesc {
optional int32 lod_level = 2 [ default = 0 ];
}
message LoDTensorArrayDesc {
required TensorDesc tensor = 1;
optional int32 lod_level = 2 [ default = 0 ];
}
message VarDesc {
enum VarType {
LOD_TENSOR = 1;
@ -116,11 +121,14 @@ message VarDesc {
FEED_MINIBATCH = 3;
FETCH_LIST = 4;
STEP_SCOPES = 5;
LOD_RANK_TABLE = 6;
LOD_TENSOR_ARRAY = 7;
}
required string name = 1;
required VarType type = 2;
optional LoDTensorDesc lod_tensor = 3;
optional TensorDesc selected_rows = 4;
optional LoDTensorArrayDesc tensor_array = 6;
optional bool persistable = 5 [ default = false ];
}

@ -0,0 +1,48 @@
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
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. */
#include "paddle/framework/lod_rank_table.h"
namespace paddle {
namespace framework {
void LoDRankTable::Reset(const LoD& lod, size_t level) {
this->coarse_lod_.clear();
this->items_.clear();
PADDLE_ENFORCE(level < lod.size(),
"Cannot rank lod since the level %d is less than lod size %d",
level, lod.size());
coarse_lod_.reserve(level);
for (size_t i = 0; i < level; ++i) {
coarse_lod_.push_back(lod[i]);
}
auto& vec = lod[level];
for (size_t i = 0; i < vec.size() - 1; ++i) {
TableItem item;
item.index = i;
item.length = vec[i + 1] - vec[i];
items_.emplace_back(item);
}
// NOTE(yuyang18):
//
// The time complexity of stable_sort is O(N*log(N)) if additional memory is
// available. It is easy to debug and unit test when using `stable_sort`
// instead of `sort`. Also, the items of a rank table will not be too large.
std::stable_sort(items_.begin(), items_.end(),
[](const TableItem& a, const TableItem& b) {
return a.length > b.length;
});
}
} // namespace framework
} // namespace paddle

Some files were not shown because too many files have changed in this diff Show More

Loading…
Cancel
Save