Skip to content

P
PaddleDetection

PaddlePaddle / PaddleDetection
大约 1 年 前同步成功

通知 694
Star 11112
Fork 2696
P
PaddleDetection

体验新版 GitCode，发现更多精彩内容 >>

提交 988a4a66 编写于 作者: D dzhwinter
浏览文件
操作

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

上级 ddc25873 5369adf1
展开全部 隐藏空白更改
内联 并排
Showing with 57758 addition and 17446 deletion

要显示的变更太多。

To preserve performance only 1000 of 1000+ files are displayed.
文本差异 电子邮件补丁
.clang-format
浏览文件 @ 988a4a66
......@@ -25,4 +25,3 @@ AllowAllParametersOfDeclarationOnNextLine: true
BinPackParameters: false
BinPackArguments: false
...
.gitignore
浏览文件 @ 988a4a66
......@@ -21,10 +21,11 @@ third_party/
cmake-build-*
# generated while compiling
python/paddle/v2/framework/core.so
python/paddle/v2/fluid/core.so
paddle/pybind/pybind.h
CMakeFiles
cmake_install.cmake
paddle/.timestamp
python/paddlepaddle.egg-info/
paddle/pybind/pybind.h
python/paddle/version.py
.pre-commit-config.yaml
浏览文件 @ 988a4a66
......@@ -31,6 +31,3 @@
- id: go-fmt
types:
- go
- id: gometalinter
types:
- go
.travis.yml
浏览文件 @ 988a4a66
......@@ -30,6 +30,7 @@ addons:
- automake
- libtool
- ccache
ssh_known_hosts: 52.76.173.135
before_install:
- if [[ "$JOB" == "check_style" ]]; then sudo ln -s /usr/bin/clang-format-3.8 /usr/bin/clang-format; fi
# Paddle is using protobuf 3.1 currently. Protobuf 3.2 breaks the compatibility. So we specify the python
......@@ -41,7 +42,15 @@ before_install:
script:
- |
timeout 2580 paddle/scripts/travis/${JOB}.sh # 43min timeout
RESULT=$?; if [ $RESULT -eq 0 ] || [ $RESULT -eq 142 ]; then true; else false; fi;
RESULT=$?; if [ $RESULT -eq 0 ] || [ $RESULT -eq 142 ]; then true ;else exit 1; fi;
- |
if [[ "$JOB" != "build_doc" ]]; then exit 0; fi;
if [[ "$TRAVIS_PULL_REQUEST" != "false" ]]; then exit 0; fi;
if [[ "$TRAVIS_BRANCH" != "develop" && ! "$TRAVIS_BRANCH" =~ ^v[[:digit:]]+\.[[:digit:]]+(\.[[:digit:]]+)?(-\S*)?$ ]]; then exit 0; fi;
export DEPLOY_DOCS_SH=https://raw.githubusercontent.com/PaddlePaddle/PaddlePaddle.org/master/scripts/deploy/deploy_docs.sh
export DOCS_DIR=`pwd`
cd ..
curl $DEPLOY_DOCS_SH | bash -s $CONTENT_DEC_PASSWD $TRAVIS_BRANCH $DOCS_DIR $DOCS_DIR/build/doc
notifications:
email:
on_success: change
......
CMakeLists.txt
浏览文件 @ 988a4a66
......@@ -16,10 +16,14 @@ cmake_minimum_required(VERSION 3.0)
set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/cmake")
set(PADDLE_SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR})
set(PADDLE_BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR})
SET(CMAKE_CXX_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
SET(CMAKE_C_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
include(system)
project(paddle CXX C Go)
message(STATUS "CXX compiler: " ${CMAKE_CXX_COMPILER} ", version: " ${CMAKE_CXX_COMPILER_VERSION})
message(STATUS "C compiler: " ${CMAKE_C_COMPILER} ", version: " ${CMAKE_C_COMPILER_VERSION})
find_package(Sphinx)
if(NOT CMAKE_CROSSCOMPILING)
......@@ -36,8 +40,7 @@ include(simd)
################################ Configurations #######################################
option(WITH_GPU "Compile PaddlePaddle with NVIDIA GPU" ${CUDA_FOUND})
option(WITH_AVX "Compile PaddlePaddle with AVX intrinsics" ${AVX_FOUND})
option(WITH_MKLDNN "Compile PaddlePaddle with mkl-dnn support." ${AVX_FOUND})
option(WITH_MKLML "Compile PaddlePaddle with mklml package." ${AVX_FOUND})
option(WITH_MKL "Compile PaddlePaddle with MKL support." ${AVX_FOUND})
option(WITH_DSO "Compile PaddlePaddle with dynamic linked CUDA" ON)
option(WITH_TESTING "Compile PaddlePaddle with unit testing" ON)
option(WITH_SWIG_PY "Compile PaddlePaddle with inference api" ON)
......@@ -55,7 +58,9 @@ option(WITH_C_API "Compile PaddlePaddle with C-API(Prediction)" OFF)
option(WITH_GOLANG "Compile PaddlePaddle with GOLANG" OFF)
option(GLIDE_INSTALL "Download and install go dependencies " ON)
option(USE_NNPACK "Compile PaddlePaddle with NNPACK library" OFF)
option(WITH_DISTRIBUTE "Compile with grpc distributed support" OFF)
option(USE_EIGEN_FOR_BLAS "Use matrix multiplication in Eigen" OFF)
option(WITH_ARM_FP16 "Use half precision support on armv8.2-a cpu" OFF)
# CMAKE_BUILD_TYPE
if(NOT CMAKE_BUILD_TYPE)
......@@ -68,9 +73,6 @@ if(ANDROID OR IOS)
if(ANDROID)
if(${CMAKE_SYSTEM_VERSION} VERSION_LESS "16")
message(FATAL_ERROR "Unsupport standalone toolchains with Android API level lower than 16")
elseif(${CMAKE_SYSTEM_VERSION} VERSION_LESS "21")
# TODO: support glog for Android api 16 ~ 19 in the future
message(WARNING "Using the unofficial git repository <https://github.com/Xreki/glog.git> instead")
endif()
endif()
......@@ -82,10 +84,18 @@ if(ANDROID OR IOS)
"Disable PYTHON when cross-compiling for Android and iOS" FORCE)
set(WITH_RDMA OFF CACHE STRING
"Disable RDMA when cross-compiling for Android and iOS" FORCE)
set(WITH_MKLDNN OFF CACHE STRING
"Disable MKLDNN when cross-compiling for Android and iOS" FORCE)
set(WITH_MKLML OFF CACHE STRING
"Disable MKLML package when cross-compiling for Android and iOS" FORCE)
set(WITH_MKL OFF CACHE STRING
"Disable MKL when cross-compiling for Android and iOS" FORCE)
set(WITH_GOLANG OFF CACHE STRING
"Disable golang when cross-compiling for Android and iOS" FORCE)
# Compile PaddlePaddle mobile inference library
if (NOT WITH_C_API)
set(WITH_C_API ON CACHE STRING
"Always compile the C_API when cross-compiling for Android and iOS" FORCE)
endif()
set(MOBILE_INFERENCE ON)
add_definitions(-DPADDLE_MOBILE_INFERENCE)
endif()
set(THIRD_PARTY_PATH "${CMAKE_BINARY_DIR}/third_party" CACHE STRING
......@@ -97,6 +107,20 @@ if (WITH_C_API AND WITH_PYTHON)
"different Python interpreter from compiling.")
endif()
if(MOBILE_INFERENCE)
set(THIRD_PARTY_BUILD_TYPE MinSizeRel)
else()
set(THIRD_PARTY_BUILD_TYPE Release)
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)
endif()
########################################################################################
include(external/mklml) # download mklml package
......@@ -112,7 +136,10 @@ 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(external/cares)
include(external/grpc)
include(cudnn) # set cudnn libraries, must before configure
include(configure) # add paddle env configuration
......@@ -143,14 +170,15 @@ set(EXTERNAL_LIBS
)
if(WITH_GPU)
list(APPEND EXTERNAL_LIBS ${CUDA_LIBRARIES} ${CUDA_rt_LIBRARY})
if(NOT WITH_DSO)
list(APPEND EXTERNAL_LIBS ${CUDNN_LIBRARY} ${CUDA_CUBLAS_LIBRARIES} ${CUDA_curand_LIBRARY})
endif(NOT WITH_DSO)
include(cuda)
endif(WITH_GPU)
if(WITH_MKLML)
list(APPEND EXTERNAL_LIBS ${MKLML_IOMP_LIB})
endif()
if(WITH_MKLDNN)
list(APPEND EXTERNAL_LIBS ${MKLDNN_LIB} ${MKLDNN_IOMP_LIB})
list(APPEND EXTERNAL_LIBS ${MKLDNN_LIB})
endif()
if(USE_NNPACK)
......@@ -160,9 +188,11 @@ endif(USE_NNPACK)
add_subdirectory(proto)
# "add_subdirectory(go)" should be placed after the following loine,
# because it depends on paddle/optimizer.
add_subdirectory(paddle/optimizer)
if(NOT MOBILE_INFERENCE)
# "add_subdirectory(go)" should be placed after the following loine,
# because it depends on paddle/optimizer.
add_subdirectory(paddle/optimizer)
endif()
# "add_subdirectory(paddle)" and "add_subdirectory(python)" should be
# placed after this block, because they depends on it.
......
CONTRIBUTING.md
浏览文件 @ 988a4a66
./doc/howto/dev/contribute_to_paddle_en.md
# Contribute Code
We sincerely appreciate your contribution. This document explains our workflow and work style.
## Workflow
PaddlePaddle uses this [Git branching model](http://nvie.com/posts/a-successful-git-branching-model/). The following steps guide usual contributions.
1. Fork
Our development community has been growing fastly; it doesn't make sense for everyone to write into the official repo. So, please file Pull Requests from your fork. To make a fork, just head over to the GitHub page and click the ["Fork" button](https://help.github.com/articles/fork-a-repo/).
1. Clone
To make a copy of your fork to your local computers, please run
```bash
git clone https://github.com/your-github-account/paddle
cd paddle
```
1. Create the local feature branch
For daily works like adding a new feature or fixing a bug, please open your feature branch before coding:
```bash
git checkout -b my-cool-stuff
```
1. Commit
Before issuing your first `git commit` command, please install [`pre-commit`](http://pre-commit.com/) by running the following commands:
```bash
pip install pre-commit
pre-commit install
```
Our pre-commit configuration requires clang-format 3.8 for auto-formating C/C++ code and yapf for Python.
Once installed, `pre-commit` checks the style of code and documentation in every commit. We will see something like the following when you run `git commit`:
```
➜ git commit
CRLF end-lines remover...............................(no files to check)Skipped
yapf.................................................(no files to check)Skipped
Check for added large files..............................................Passed
Check for merge conflicts................................................Passed
Check for broken symlinks................................................Passed
Detect Private Key...................................(no files to check)Skipped
Fix End of Files.....................................(no files to check)Skipped
clang-formater.......................................(no files to check)Skipped
[my-cool-stuff c703c041] add test file
1 file changed, 0 insertions(+), 0 deletions(-)
create mode 100644 233
```
1. Build and test
Users can build PaddlePaddle natively on Linux and Mac OS X. But to unify the building environment and to make it easy for debugging, the recommended way is [using Docker](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/howto/dev/build_en.md).
1. Keep pulling
An experienced Git user pulls from the official repo often -- daily or even hourly, so they notice conflicts with others work early, and it's easier to resolve smaller conflicts.
```bash
git remote add upstream https://github.com/PaddlePaddle/Paddle
git pull upstream develop
```
1. Push and file a pull request
You can "push" your local work into your forked repo:
```bash
git push origin my-cool-stuff
```
The push allows you to create a pull request, requesting owners of this [official repo](https://github.com/PaddlePaddle/Paddle) to pull your change into the official one.
To create a pull request, please follow [these steps](https://help.github.com/articles/creating-a-pull-request/).
If your change is for fixing an issue, please write ["Fixes <issue-URL>"](https://help.github.com/articles/closing-issues-using-keywords/) in the description section of your pull request. Github would close the issue when the owners merge your pull request.
Please remember to specify some reviewers for your pull request. If you don't know who are the right ones, please follow Github's recommendation.
1. Delete local and remote branches
To keep your local workspace and your fork clean, you might want to remove merged branches:
```bash
git push origin :my-cool-stuff
git checkout develop
git pull upstream develop
git branch -d my-cool-stuff
```
### Code Review
- Please feel free to ping your reviewers by sending them the URL of your pull request via IM or email. Please do this after your pull request passes the CI.
- Please answer reviewers' every comment. If you are to follow the comment, please write "Done"; please give a reason otherwise.
- If you don't want your reviewers to get overwhelmed by email notifications, you might reply their comments by [in a batch](https://help.github.com/articles/reviewing-proposed-changes-in-a-pull-request/).
- Reduce the unnecessary commits. Some developers commit often. It is recommended to append a sequence of small changes into one commit by running `git commit --amend` instead of `git commit`.
## Coding Standard
### Code Style
Our C/C++ code follows the [Google style guide](http://google.github.io/styleguide/cppguide.html).
Our Python code follows the [PEP8 style guide](https://www.python.org/dev/peps/pep-0008/).
Our build process helps to check the code style. In [`build.sh`](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/paddle/scripts/docker/build.sh#L42), the entry point of our [builder Docker image](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/Dockerfile#L88), the CMake argument `WITH_STYLE_CHECK` is set to `ON` by default. This flag is on
Please install pre-commit, which automatically reformat the changes to C/C++ and Python code whenever we run `git commit`. To check the whole codebase, we can run the command `pre-commit run -a`, as in the [`check_style.sh` file](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/paddle/scripts/travis/check_style.sh#L30), which is invoked by [our Travis CI configuration](https://github.com/PaddlePaddle/Paddle/blob/b84e8226514b8bb4405c3c28e54aa5077193d179/.travis.yml#L43).
### Unit Tests
Please remember to add related unit tests.
- For C/C++ code, please follow [`google-test` Primer](https://github.com/google/googletest/blob/master/googletest/docs/Primer.md).
- For Python code, please use [Python's standard `unittest` package](http://pythontesting.net/framework/unittest/unittest-introduction/).
### Writing Logs
We use [glog](https://github.com/google/glog) for logging in our C/C++ code.
For general information, please use `LOG`. For debug information, please use [`VLOG`](http://htmlpreview.github.io/?https://github.com/google/glog/blob/master/doc/glog.html#verbose). The reason is at [here](https://groups.google.com/a/chromium.org/d/msg/chromium-dev/3NDNd1KzXeY/AZKMMx37fdQJ).
`VLOG` requires a *verbose level* parameter. For example:
```c++
VLOG(3) << "Operator FC is taking " << num_inputs << "inputs."
```
When we run a PaddlePaddle application or test, we can specify a verbose threshold. For example:
```bash
GLOG_vmodule=buddy_allocator=2 \
GLOG_v=10 \
python \
../python/paddle/v2/framework/tests/test_recurrent_op.py
```
This will enable VLOG messages generated by `buddy_allocator.{h,cc}` and in the verbose range of 0 to 3, so you will see above example VLOG message, which is in level 3. This suggests that we output overall messages in lower verbose levels, so they display with higher probability. When coding C++, please follow the verbose level convention as follows:
- verbose level 1: [framework](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/framework)
- verbose level 3: [operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators)
- verbose level 5: [memory](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/memory), [platform](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/platform)
- verbose level 7: [math](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/math)
Dockerfile
浏览文件 @ 988a4a66
......@@ -22,14 +22,14 @@ COPY ./paddle/scripts/docker/root/ /root/
RUN apt-get update && \
apt-get install -y \
git python-pip python-dev openssh-server bison \
git python-pip python-dev openssh-server bison libnccl-dev \
wget unzip unrar tar xz-utils bzip2 gzip coreutils ntp \
curl sed grep graphviz libjpeg-dev zlib1g-dev \
python-matplotlib gcc-4.8 g++-4.8 \
automake locales clang-format swig doxygen cmake \
liblapack-dev liblapacke-dev libboost-dev \
clang-3.8 llvm-3.8 libclang-3.8-dev \
net-tools && \
net-tools libtool && \
apt-get clean -y
# Install Go and glide
......
RELEASE.cn.md
浏览文件 @ 988a4a66
# v0.11.0版本
## PaddlePaddle Fluid
- PaddlePaddle发布版本v0.11.0包含一个新的特性*PaddlePaddle Fluid*. Fluid 是设计用来让用户像Pytorch和Tensorflow Eager Execution一样执行程序。在这些系统中，不再有*模型*这个概念，应用也不再包含一个用于描述Operator图或者一系列层的符号描述，而是像通用程序那样描述训练或者预测的过程。而Fluid与PyTorch或Eager Execution的区别在于Fluid不依赖Python提供的控制流，例如 if-else-then或者for，而是提供了基于C++实现的控制流并暴露了对应的用with语法实现的Python接口。例如：
https://github.com/PaddlePaddle/Paddle/blob/3df78ed2a98d37f7ae6725894cc7514effd5664b/python/paddle/v2/fluid/tests/test_while_op.py#L36-L44
- 在v0.11.0版本中，我们提供了一个C++类`Executor`用于运行一个Fluid程序。Executor类似一个解释器。在未来的版本中，我们将提升和优化Executor成为一个调试器，就像GDB。并可能提供一些编译器，这个编译器会读取一个上文所描述的应用然后编译成一个等价的
源代码，这个源代码可以被nvcc编译成可以使用CUDA的二进制，或者被icc编译成可以充分利用Intel CPU的二进制。
## 新特点
* 发布 `PaddlePaddle Fluid`。
* 增加了用于模型预测的C-API。
* 用Fluid API实现了一个简单的GAN的例子。
* 增加了关于性能调优的文档。
* 为`paddle.v2.dataset`下载数据集提供了重试机制.
* C++中使用protobuf-lite替换protobuf减少了二进制的大小。
* 发布了新特性 [Elastic Deep Learning (EDL)](https://github.com/PaddlePaddle/cloud/tree/develop/doc/autoscale/experiment).
* 基于Bazel API利用cmake实现了一个的新的构建系统函数库。
* 当使用编译选项`WITH_MKL=ON`时自动下载和编译Intel® [MKLML](https://github.com/01org/mkl-dnn/releases/download/v0.11/mklml_lnx_2018.0.1.20171007.tgz) 函数库.
* [Intel® MKL-DNN on PaddlePaddle](https://github.com/PaddlePaddle/Paddle/tree/develop/doc/design/mkldnn):
- 完成了 11个 MKL-DNN 层: Convolution, Fully connectivity, Pooling, ReLU, Tanh, ELU, Softmax, BatchNorm, AddTo, Concat, LRN。
- 完成了 3个 MKL-DNN 网络: VGG-19, ResNet-50, GoogleNet
- 基于Intel Skylake 6148 CPU的[性能测试](https://github.com/PaddlePaddle/Paddle/blob/develop/benchmark/IntelOptimizedPaddle.md) : 相对于MKLML有2~3倍的训练加速。
* 增加 [softsign activation](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/activation.html#softsign)
* 增加 [dot product layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#dot-prod)
* 增加 [L2 distance layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#l2-distance)
* 增加 [sub-nested sequence layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#sub-nested-seq)
* 增加 [kmax sequence score layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#kmax-sequence-score)
* 增加 [sequence slice layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#seq-slice)
* 增加 [row convolution layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#row-conv)
* 增加移动端友好的网页
## 改进
* 使用一个Python`whl`包即可安装.
* [V2 API可以实现用户定制化评估](https://github.com/PaddlePaddle/models/tree/develop/ltr#训练过程中输出自定义评估指标)。
* 将 `PADDLE_ONLY_CPU` 改为 `PADDLE_WITH_GPU`, 因为我们会支持多种设备。
* 删除了有一些bug的BarrierStat。
* 清理和删除了paddle::Parameter中未使用的函数。
* 删除了ProtoDataProvider。
* Huber loss同时支持回归和分类。
* 为sequence pooling 层增加`stride`参数。
* v2 API自动使用cudnn batch normalization。
* 可以使用一个固定的参数名共享BN层的参数。
* 2D convolution operation支持variable-dimension input特性。
* 重构cmake中关于CUDA的部分并实现自动检测GPU架构的功能。
* 优化网页导航。
## 错误修复
* 修复ROI pooling的Bug. cc9a761
* 修复当label是dense vector是AUC变成0的问题. #5274
* 修复WarpCTC 层的Bug.
# v0.10.0版本
我们非常高兴发布了PaddlePaddle V0.10.0版，并开发了新的[Python API](http://research.baidu.com/paddlepaddles-new-api-simplifies-deep-learning-programs/)。
......
RELEASE.md
浏览文件 @ 988a4a66
# Release v0.11.0
## PaddlePaddle Fluid
- Release 0.11.0 includes a new feature *PaddlePaddle Fluid*. Fluid is
designed to allow users to program like PyTorch and TensorFlow Eager Execution.
In these systems, there is no longer the concept *model* and applications
do not include a symbolic description of a graph of operators nor a sequence
of layers. Instead, applications look exactly like a usual program that
describes a process of training or inference. The difference between
Fluid and PyTorch or Eager Execution is that Fluid doesn't rely on Python's
control-flow, `if-then-else` nor `for`. Instead, Fluid provides its
C++ implementations and their Python binding using the `with` statement. For an example
https://github.com/PaddlePaddle/Paddle/blob/3df78ed2a98d37f7ae6725894cc7514effd5664b/python/paddle/v2/fluid/tests/test_while_op.py#L36-L44
- In 0.11.0, we provides a C++ class `Executor` to run a Fluid program.
Executor works like an interpreter. In future version, we will improve
`Executor` into a debugger like GDB, and we might provide some compilers,
which, for example, takes an application like the above one, and outputs
an equivalent C++ source program, which can be compiled using
[`nvcc`](http://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html)
to generate binaries that use CUDA, or using
[`icc`](https://software.intel.com/en-us/c-compilers) to generate binaries
that make full use of Intel CPUs.
## New Features
* Release `PaddlePaddle Fluid`.
* Add C-API for model inference
* Use fluid API to create a simple GAN demo.
* Add develop guide about performance tunning.
* Add retry when download `paddle.v2.dataset`.
* Linking protobuf-lite not protobuf in C++. Reduce the binary size.
* Feature [Elastic Deep Learning (EDL)](https://github.com/PaddlePaddle/cloud/tree/develop/doc/autoscale/experiment) released.
* A new style cmake functions for Paddle. It is based on Bazel API.
* Automatically download and compile with Intel® [MKLML](https://github.com/01org/mkl-dnn/releases/download/v0.11/mklml_lnx_2018.0.1.20171007.tgz) library as CBLAS when build `WITH_MKL=ON`.
* [Intel® MKL-DNN on PaddlePaddle](https://github.com/PaddlePaddle/Paddle/tree/develop/doc/design/mkldnn):
- Complete 11 MKL-DNN layers: Convolution, Fully connectivity, Pooling, ReLU, Tanh, ELU, Softmax, BatchNorm, AddTo, Concat, LRN.
- Complete 3 MKL-DNN networks: VGG-19, ResNet-50, GoogleNet
- [Benchmark](https://github.com/PaddlePaddle/Paddle/blob/develop/benchmark/IntelOptimizedPaddle.md) on Intel Skylake 6148 CPU: 2~3x training speedup compared with MKLML.
* Add the [`softsign` activation](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/activation.html#softsign).
* Add the [dot product layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#dot-prod).
* Add the [L2 distance layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#l2-distance).
* Add the [sub-nested sequence layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#sub-nested-seq).
* Add the [kmax sequence score layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#kmax-sequence-score).
* Add the [sequence slice layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#seq-slice).
* Add the [row convolution layer](http://www.paddlepaddle.org/docs/develop/documentation/zh/api/v2/config/layer.html#row-conv)
* Add mobile friendly webpages.
## Improvements
* Build and install using a single `whl` package.
* [Custom evaluating in V2 API](https://github.com/PaddlePaddle/models/tree/develop/ltr#训练过程中输出自定义评估指标).
* Change `PADDLE_ONLY_CPU` to `PADDLE_WITH_GPU`, since we will support many kinds of devices.
* Remove buggy BarrierStat.
* Clean and remove unused functions in paddle::Parameter.
* Remove ProtoDataProvider.
* Huber loss supports both regression and classification.
* Add the `stride` parameter for sequence pooling layers.
* Enable v2 API use cudnn batch normalization automatically.
* The BN layer's parameter can be shared by a fixed the parameter name.
* Support variable-dimension input feature for 2D convolution operation.
* Refine cmake about CUDA to automatically detect GPU architecture.
* Improved website navigation.
## Bug Fixes
* Fix bug in ROI pooling. cc9a761
* Fix AUC is zero when label is dense vector. #5274
* Fix bug in WarpCTC layer.
# Release v0.10.0
We are glad to release version 0.10.0. In this version, we are happy to release the new
......
benchmark/IntelOptimizedPaddle.md 0 → 100644
浏览文件 @ 988a4a66
# Benchmark
Machine:
- Server: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz, 2 Sockets, 20 Cores per socket
- Laptop: TBD
System: CentOS release 6.3 (Final), Docker 1.12.1.
PaddlePaddle: (TODO: will rerun after 0.11.0)
- paddlepaddle/paddle:latest (for MKLML and MKL-DNN)
- MKL-DNN tag v0.11
- MKLML 2018.0.1.20171007
- paddlepaddle/paddle:latest-openblas (for OpenBLAS)
- OpenBLAS v0.2.20
On each machine, we will test and compare the performance of training on single node using MKL-DNN / MKLML / OpenBLAS respectively.
## Benchmark Model
### Server
Test on batch size 64, 128, 256 on Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Input image size - 3 * 224 * 224, Time: images/second
- VGG-19
| BatchSize | 64 | 128 | 256 |
|--------------|-------| -----| --------|
| OpenBLAS | 7.80 | 9.00 | 10.80 |
| MKLML | 12.12 | 13.70 | 16.18 |
| MKL-DNN | 28.46 | 29.83 | 30.44 |
<img src="figs/vgg-cpu-train.png" width="500">
- ResNet-50
| BatchSize | 64 | 128 | 256 |
|--------------|-------| ------| -------|
| OpenBLAS | 25.22 | 25.68 | 27.12 |
| MKLML | 32.52 | 31.89 | 33.12 |
| MKL-DNN | 81.69 | 82.35 | 84.08 |
<img src="figs/resnet-cpu-train.png" width="500">
- GoogLeNet
| BatchSize | 64 | 128 | 256 |
|--------------|-------| ------| -------|
| OpenBLAS | 89.52 | 96.97 | 108.25 |
| MKLML | 128.46| 137.89| 158.63 |
| MKL-DNN     | 250.46| 264.83| 269.50 |
<img src="figs/googlenet-cpu-train.png" width="500">
### Laptop
TBD
benchmark/paddle/image/googlenet.py
浏览文件 @ 988a4a66
......@@ -5,10 +5,22 @@ height = 224
width = 224
num_class = 1000
batch_size = get_config_arg('batch_size', int, 128)
args = {'height': height, 'width': width, 'color': True, 'num_class': num_class}
use_gpu = get_config_arg('use_gpu', bool, True)
is_infer = get_config_arg("is_infer", bool, False)
args = {
'height': height,
'width': width,
'color': True,
'num_class': num_class,
'is_infer': is_infer
}
define_py_data_sources2(
"train.list", None, module="provider", obj="process", args=args)
"train.list" if not is_infer else None,
"test.list" if is_infer else None,
module="provider",
obj="process",
args=args)
settings(
batch_size=batch_size,
......@@ -16,6 +28,8 @@ settings(
learning_method=MomentumOptimizer(0.9),
regularization=L2Regularization(0.0005 * batch_size))
conv_projection = conv_projection if use_gpu else img_conv_layer
def inception2(name, input, channels, \
filter1,
filter3R, filter3,
......@@ -138,12 +152,11 @@ def inception(name, input, channels, \
cat = concat_layer(
name=name,
input=[cov1, cov3, cov5, covprj],
bias_attr=True,
bias_attr=True if use_gpu else False,
act=ReluActivation())
return cat
lab = data_layer(name="label", size=1000)
data = data_layer(name="input", size=3 * height * width)
# stage 1
......@@ -221,6 +234,10 @@ pool5 = img_pool_layer(
dropout = dropout_layer(name="dropout", input=pool5, dropout_rate=0.4)
out3 = fc_layer(
name="output3", input=dropout, size=1000, act=SoftmaxActivation())
loss3 = cross_entropy(name='loss3', input=out3, label=lab)
outputs(loss3)
if is_infer:
outputs(out3)
else:
lab = data_layer(name="label", size=num_class)
loss3 = cross_entropy(name='loss3', input=out3, label=lab)
outputs(loss3)
benchmark/paddle/image/provider.py
浏览文件 @ 988a4a66
......@@ -13,14 +13,20 @@ def initHook(settings, height, width, color, num_class, **kwargs):
settings.data_size = settings.height * settings.width * 3
else:
settings.data_size = settings.height * settings.width
settings.slots = [dense_vector(settings.data_size), integer_value(1)]
settings.is_infer = kwargs.get('is_infer', False)
if settings.is_infer:
settings.slots = [dense_vector(settings.data_size)]
else:
settings.slots = [dense_vector(settings.data_size), integer_value(1)]
@provider(
init_hook=initHook, min_pool_size=-1, cache=CacheType.CACHE_PASS_IN_MEM)
def process(settings, file_list):
for i in xrange(1024):
for i in xrange(2560 if settings.is_infer else 1024):
img = np.random.rand(1, settings.data_size).reshape(-1, 1).flatten()
lab = random.randint(0, settings.num_class - 1)
yield img.astype('float32'), int(lab)
if settings.is_infer:
yield img.astype('float32')
else:
lab = random.randint(0, settings.num_class - 1)
yield img.astype('float32'), int(lab)
benchmark/paddle/image/resnet.py 0 → 100644
浏览文件 @ 988a4a66
#!/usr/bin/env python
from paddle.trainer_config_helpers import *
height = 224
width = 224
num_class = 1000
batch_size = get_config_arg('batch_size', int, 64)
layer_num = get_config_arg("layer_num", int, 50)
is_infer = get_config_arg("is_infer", bool, False)
args = {
'height': height,
'width': width,
'color': True,
'num_class': num_class,
'is_infer': is_infer
}
define_py_data_sources2(
"train.list" if not is_infer else None,
"test.list" if is_infer else None,
module="provider",
obj="process",
args=args)
settings(
batch_size=batch_size,
learning_rate=0.01 / batch_size,
learning_method=MomentumOptimizer(0.9),
regularization=L2Regularization(0.0005 * batch_size))
#######################Network Configuration #############
def conv_bn_layer(name,
input,
filter_size,
num_filters,
stride,
padding,
channels=None,
active_type=ReluActivation()):
"""
A wrapper for conv layer with batch normalization layers.
Note:
conv layer has no activation.
"""
tmp = img_conv_layer(
name=name + "_conv",
input=input,
filter_size=filter_size,
num_channels=channels,
num_filters=num_filters,
stride=stride,
padding=padding,
act=LinearActivation(),
bias_attr=False)
return batch_norm_layer(
name=name + "_bn",
input=tmp,
act=active_type,
use_global_stats=is_infer)
def bottleneck_block(name, input, num_filters1, num_filters2):
"""
A wrapper for bottlenect building block in ResNet.
Last conv_bn_layer has no activation.
Addto layer has activation of relu.
"""
last_name = conv_bn_layer(
name=name + '_branch2a',
input=input,
filter_size=1,
num_filters=num_filters1,
stride=1,
padding=0)
last_name = conv_bn_layer(
name=name + '_branch2b',
input=last_name,
filter_size=3,
num_filters=num_filters1,
stride=1,
padding=1)
last_name = conv_bn_layer(
name=name + '_branch2c',
input=last_name,
filter_size=1,
num_filters=num_filters2,
stride=1,
padding=0,
active_type=LinearActivation())
return addto_layer(
name=name + "_addto", input=[input, last_name], act=ReluActivation())
def mid_projection(name, input, num_filters1, num_filters2, stride=2):
"""
A wrapper for middile projection in ResNet.
projection shortcuts are used for increasing dimensions,
and other shortcuts are identity
branch1: projection shortcuts are used for increasing
dimensions, has no activation.
branch2x: bottleneck building block, shortcuts are identity.
"""
# stride = 2
branch1 = conv_bn_layer(
name=name + '_branch1',
input=input,
filter_size=1,
num_filters=num_filters2,
stride=stride,
padding=0,
active_type=LinearActivation())
last_name = conv_bn_layer(
name=name + '_branch2a',
input=input,
filter_size=1,
num_filters=num_filters1,
stride=stride,
padding=0)
last_name = conv_bn_layer(
name=name + '_branch2b',
input=last_name,
filter_size=3,
num_filters=num_filters1,
stride=1,
padding=1)
last_name = conv_bn_layer(
name=name + '_branch2c',
input=last_name,
filter_size=1,
num_filters=num_filters2,
stride=1,
padding=0,
active_type=LinearActivation())
return addto_layer(
name=name + "_addto", input=[branch1, last_name], act=ReluActivation())
img = data_layer(name='image', size=height * width * 3)
def deep_res_net(res2_num=3, res3_num=4, res4_num=6, res5_num=3):
"""
A wrapper for 50,101,152 layers of ResNet.
res2_num: number of blocks stacked in conv2_x
res3_num: number of blocks stacked in conv3_x
res4_num: number of blocks stacked in conv4_x
res5_num: number of blocks stacked in conv5_x
"""
# For ImageNet
# conv1: 112x112
tmp = conv_bn_layer(
"conv1",
input=img,
filter_size=7,
channels=3,
num_filters=64,
stride=2,
padding=3)
tmp = img_pool_layer(name="pool1", input=tmp, pool_size=3, stride=2)
# conv2_x: 56x56
tmp = mid_projection(
name="res2_1", input=tmp, num_filters1=64, num_filters2=256, stride=1)
for i in xrange(2, res2_num + 1, 1):
tmp = bottleneck_block(
name="res2_" + str(i), input=tmp, num_filters1=64, num_filters2=256)
# conv3_x: 28x28
tmp = mid_projection(
name="res3_1", input=tmp, num_filters1=128, num_filters2=512)
for i in xrange(2, res3_num + 1, 1):
tmp = bottleneck_block(
name="res3_" + str(i),
input=tmp,
num_filters1=128,
num_filters2=512)
# conv4_x: 14x14
tmp = mid_projection(
name="res4_1", input=tmp, num_filters1=256, num_filters2=1024)
for i in xrange(2, res4_num + 1, 1):
tmp = bottleneck_block(
name="res4_" + str(i),
input=tmp,
num_filters1=256,
num_filters2=1024)
# conv5_x: 7x7
tmp = mid_projection(
name="res5_1", input=tmp, num_filters1=512, num_filters2=2048)
for i in xrange(2, res5_num + 1, 1):
tmp = bottleneck_block(
name="res5_" + str(i),
input=tmp,
num_filters1=512,
num_filters2=2048)
tmp = img_pool_layer(
name='avgpool',
input=tmp,
pool_size=7,
stride=1,
pool_type=AvgPooling())
return fc_layer(input=tmp, size=num_class, act=SoftmaxActivation())
if layer_num == 50:
resnet = deep_res_net(3, 4, 6, 3)
elif layer_num == 101:
resnet = deep_res_net(3, 4, 23, 3)
elif layer_num == 152:
resnet = deep_res_net(3, 8, 36, 3)
else:
print("Wrong layer number.")
if is_infer:
outputs(resnet)
else:
lbl = data_layer(name="label", size=num_class)
loss = cross_entropy(name='loss', input=resnet, label=lbl)
outputs(loss)
benchmark/paddle/image/run_mkldnn_infer.sh 0 → 100755
浏览文件 @ 988a4a66
set -e
function clock_to_seconds() {
hours=`echo $1 | awk -F ':' '{print $1}'`
mins=`echo $1 | awk -F ':' '{print $2}'`
secs=`echo $1 | awk -F ':' '{print $3}'`
echo `awk 'BEGIN{printf "%.2f",('$secs' + '$mins' * 60 + '$hours' * 3600)}'`
}
function infer() {
unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
topology=$1
layer_num=$2
bs=$3
use_mkldnn=$4
if [ $4 == "True" ]; then
thread=1
log="logs/infer-${topology}-${layer_num}-mkldnn-${bs}.log"
elif [ $4 == "False" ]; then
thread=`nproc`
if [ $thread -gt $bs ]; then
thread=$bs
fi
log="logs/infer-${topology}-${layer_num}-${thread}mklml-${bs}.log"
else
echo "Wrong input $4, use True or False."
exit 0
fi
models_in="models/${topology}-${layer_num}/pass-00000/"
if [ ! -d $models_in ]; then
echo "Training model ${topology}_${layer_num}"
paddle train --job=train \
--config="${topology}.py" \
--use_mkldnn=True \
--use_gpu=False \
--trainer_count=1 \
--num_passes=1 \
--save_dir="models/${topology}-${layer_num}" \
--config_args="batch_size=128,layer_num=${layer_num}" \
> /dev/null 2>&1
echo "Done"
fi
log_period=$((256 / bs))
paddle train --job=test \
--config="${topology}.py" \
--use_mkldnn=$use_mkldnn \
--use_gpu=False \
--trainer_count=$thread \
--log_period=$log_period \
--config_args="batch_size=${bs},layer_num=${layer_num},is_infer=True" \
--init_model_path=$models_in \
2>&1 | tee ${log}
# calculate the last 5 logs period time of 1280 samples,
# the time before are burning time.
start=`tail ${log} -n 7 | head -n 1 | awk -F ' ' '{print $2}' | xargs`
end=`tail ${log} -n 2 | head -n 1 | awk -F ' ' '{print $2}' | xargs`
start_sec=`clock_to_seconds $start`
end_sec=`clock_to_seconds $end`
fps=`awk 'BEGIN{printf "%.2f",(1280 / ('$end_sec' - '$start_sec'))}'`
echo "Last 1280 samples start: ${start}(${start_sec} sec), end: ${end}(${end_sec} sec;" >> ${log}
echo "FPS: $fps images/sec" 2>&1 | tee -a ${log}
}
if [ ! -f "train.list" ]; then
echo " " > train.list
fi
if [ ! -f "test.list" ]; then
echo " " > test.list
fi
if [ ! -d "logs" ]; then
mkdir logs
fi
if [ ! -d "models" ]; then
mkdir -p models
fi
# inference benchmark
for use_mkldnn in True False; do
for batchsize in 1 2 4 8 16; do
infer googlenet v1 $batchsize $use_mkldnn
infer resnet 50 $batchsize $use_mkldnn
infer vgg 19 $batchsize $use_mkldnn
done
done
benchmark/paddle/image/run_mkldnn.sh → benchmark/paddle/image/run_mkldnn_train.sh
浏览文件 @ 988a4a66
set -e
function train() {
unset OMP_NUM_THREADS MKL_NUM_THREADS
export OMP_DYNAMIC="FALSE"
export KMP_AFFINITY="granularity=fine,compact,0,0"
unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
topology=$1
bs=$2
use_mkldnn=$3
if [ $3 == "True" ]; then
layer_num=$2
bs=$3
use_mkldnn=$4
if [ $4 == "True" ]; then
thread=1
log="logs/${topology}-mkldnn-${bs}.log"
elif [ $3 == "False" ]; then
log="logs/train-${topology}-${layer_num}-mkldnn-${bs}.log"
elif [ $4 == "False" ]; then
thread=`nproc`
# each trainer_count use only 1 core to avoid conflict
export OMP_NUM_THREADS=1
export MKL_NUM_THREADS=1
log="logs/${topology}-${thread}mklml-${bs}.log"
log="logs/train-${topology}-${layer_num}-${thread}mklml-${bs}.log"
else
echo "Wrong input $3, use True or False."
echo "Wrong input $4, use True or False."
exit 0
fi
args="batch_size=${bs}"
args="batch_size=${bs},layer_num=${layer_num}"
config="${topology}.py"
paddle train --job=time \
--config=$config \
......@@ -33,19 +30,18 @@ function train() {
2>&1 | tee ${log}
}
if [ ! -d "train.list" ]; then
if [ ! -f "train.list" ]; then
echo " " > train.list
fi
if [ ! -d "logs" ]; then
mkdir logs
fi
#========== mkldnn ==========#
train vgg 64 True
train vgg 128 True
train vgg 256 True
#========== mklml ===========#
train vgg 64 False
train vgg 128 False
train vgg 256 False
# training benchmark
for use_mkldnn in True False; do
for batchsize in 64 128 256; do
train vgg 19 $batchsize $use_mkldnn
train resnet 50 $batchsize $use_mkldnn
train googlenet v1 $batchsize $use_mkldnn
done
done
benchmark/paddle/image/vgg.py
浏览文件 @ 988a4a66
......@@ -6,14 +6,25 @@ width = 224
num_class = 1000
batch_size = get_config_arg('batch_size', int, 64)
layer_num = get_config_arg('layer_num', int, 19)
is_infer = get_config_arg("is_infer", bool, False)
args = {'height': height, 'width': width, 'color': True, 'num_class': num_class}
args = {
'height': height,
'width': width,
'color': True,
'num_class': num_class,
'is_infer': is_infer
}
define_py_data_sources2(
"train.list", None, module="provider", obj="process", args=args)
"train.list" if not is_infer else None,
"test.list" if is_infer else None,
module="provider",
obj="process",
args=args)
settings(
batch_size=batch_size,
learning_rate=0.01 / batch_size,
learning_rate=0.001 / batch_size,
learning_method=MomentumOptimizer(0.9),
regularization=L2Regularization(0.0005 * batch_size))
......@@ -98,6 +109,9 @@ elif layer_num == 19:
else:
print("Wrong layer number.")
lab = data_layer('label', num_class)
loss = cross_entropy(input=vgg, label=lab)
outputs(loss)
if is_infer:
outputs(vgg)
else:
lab = data_layer('label', num_class)
loss = cross_entropy(input=vgg, label=lab)
outputs(loss)
cmake/cblas.cmake
浏览文件 @ 988a4a66
# Find the CBlas and lapack libraries
#
# It will search MKL, atlas, OpenBlas, reference-cblas in order.
# It will search MKLML, atlas, OpenBlas, reference-cblas in order.
#
# If any cblas implementation found, the following variable will be set.
# CBLAS_PROVIDER # one of MKL, ATLAS, OPENBLAS, REFERENCE
# CBLAS_PROVIDER # one of MKLML, ATLAS, OPENBLAS, REFERENCE
# CBLAS_INC_DIR # the include directory for cblas.
# CBLAS_LIBS # a list of libraries should be linked by paddle.
# # Each library should be full path to object file.
#
# User should set one of MKL_ROOT, ATLAS_ROOT, OPENBLAS_ROOT, REFERENCE_CBLAS_ROOT
# during cmake. If none of them set, it will try to find cblas implementation in
# system paths.
#
set(CBLAS_FOUND OFF)
......@@ -30,44 +25,6 @@ if(WITH_MKLML AND MKLML_INC_DIR AND MKLML_LIB)
return()
endif()
## Then find MKL.
set(INTEL_MKL_ROOT "/opt/intel/mkl" CACHE PATH "Folder contains intel mkl libs")
set(MKL_ROOT $ENV{MKL_ROOT} CACHE PATH "Folder contains env MKL")
set(MKL_INCLUDE_SEARCH_PATHS
${MKL_ROOT}/include
${INTEL_MKL_ROOT}/include)
set(MKL_LIB_SEARCH_PATHS
${MKL_ROOT}/lib
${MKL_ROOT}/lib/intel64
${INTEL_MKL_ROOT}/lib
${INTEL_MKL_ROOT}/lib/intel64)
find_path(MKL_INC_DIR mkl.h PATHS
${MKL_INCLUDE_SEARCH_PATHS})
find_path(MKL_LAPACK_INC_DIR mkl_lapacke.h PATHS
${MKL_INCLUDE_SEARCH_PATHS})
find_library(MKL_CORE_LIB NAMES mkl_core PATHS
${MKL_LIB_SEARCH_PATHS})
find_library(MKL_SEQUENTIAL_LIB NAMES mkl_sequential PATHS
${MKL_LIB_SEARCH_PATHS})
find_library(MKL_INTEL_LP64 NAMES mkl_intel_lp64 PATHS
${MKL_LIB_SEARCH_PATHS})
if(MKL_LAPACK_INC_DIR AND MKL_INC_DIR AND MKL_CORE_LIB AND MKL_SEQUENTIAL_LIB AND MKL_INTEL_LP64)
set(CBLAS_FOUND ON)
set(CBLAS_PROVIDER MKL)
set(CBLAS_INC_DIR ${MKL_INC_DIR} ${MKL_LAPACK_INC_DIR})
set(CBLAS_LIBRARIES ${MKL_INTEL_LP64} ${MKL_SEQUENTIAL_LIB} ${MKL_CORE_LIB})
add_definitions(-DPADDLE_USE_MKL)
add_definitions(-DLAPACK_FOUND)
message(STATUS "Found MKL (include: ${MKL_INC_DIR}, library: ${CBLAS_LIBRARIES})")
message(STATUS "Found lapack in MKL (include: ${MKL_LAPACK_INC_DIR})")
return()
endif()
## Then find atlas.
set(ATLAS_ROOT $ENV{ATLAS_ROOT} CACHE PATH "Folder contains Atlas")
set(ATLAS_INCLUDE_SEARCH_PATHS
......
cmake/configure.cmake
浏览文件 @ 988a4a66
......@@ -24,6 +24,15 @@ if(WITH_DOUBLE)
add_definitions(-DPADDLE_TYPE_DOUBLE)
endif(WITH_DOUBLE)
if(WITH_ARM_FP16)
add_definitions(-DPADDLE_ARM_FP16)
add_definitions("-march=armv8.2-a+fp16+simd")
endif(WITH_ARM_FP16)
if(WITH_TESTING)
add_definitions(-DPADDLE_WITH_TESTING)
endif(WITH_TESTING)
if(NOT WITH_TIMER)
add_definitions(-DPADDLE_DISABLE_TIMER)
endif(NOT WITH_TIMER)
......@@ -49,19 +58,20 @@ if(NOT WITH_GOLANG)
endif(NOT WITH_GOLANG)
if(NOT WITH_GPU)
add_definitions(-DPADDLE_ONLY_CPU)
add_definitions(-DHPPL_STUB_FUNC)
list(APPEND CMAKE_CXX_SOURCE_FILE_EXTENSIONS cu)
else()
add_definitions(-DPADDLE_WITH_CUDA)
FIND_PACKAGE(CUDA REQUIRED)
if(${CUDA_VERSION_MAJOR} VERSION_LESS 7)
message(FATAL_ERROR "Paddle need CUDA >= 7.0 to compile")
message(FATAL_ERROR "Paddle needs CUDA >= 7.0 to compile")
endif()
if(NOT CUDNN_FOUND)
message(FATAL_ERROR "Paddle need cudnn to compile")
message(FATAL_ERROR "Paddle needs cudnn to compile")
endif()
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} "-Xcompiler ${SIMD_FLAG}")
......@@ -71,27 +81,14 @@ else()
include_directories(${CUDA_TOOLKIT_INCLUDE})
endif(NOT WITH_GPU)
if(WITH_MKLDNN)
add_definitions(-DPADDLE_USE_MKLDNN)
if (WITH_MKLML AND MKLDNN_IOMP_DIR)
message(STATUS "Enable Intel OpenMP at ${MKLDNN_IOMP_DIR}")
set(OPENMP_FLAGS "-fopenmp")
set(CMAKE_C_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
set(CMAKE_CXX_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OPENMP_FLAGS}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OPENMP_FLAGS}")
else()
find_package(OpenMP)
if(OPENMP_FOUND)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
else()
message(WARNING "Can not find OpenMP."
"Some performance features in MKLDNN may not be available")
endif()
endif()
endif(WITH_MKLDNN)
if (WITH_MKLML AND MKLML_IOMP_LIB)
message(STATUS "Enable Intel OpenMP with ${MKLML_IOMP_LIB}")
set(OPENMP_FLAGS "-fopenmp")
set(CMAKE_C_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
set(CMAKE_CXX_CREATE_SHARED_LIBRARY_FORBIDDEN_FLAGS ${OPENMP_FLAGS})
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OPENMP_FLAGS}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OPENMP_FLAGS}")
endif()
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${SIMD_FLAG}")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${SIMD_FLAG}")
......
cmake/cross_compiling/ios.cmake
浏览文件 @ 988a4a66
......@@ -76,12 +76,9 @@ set(IOS_PLATFORM ${IOS_PLATFORM} CACHE STRING "Type of iOS Platform")
# Set the architecture for iOS
if(NOT DEFINED IOS_ARCH)
if(IOS_PLATFORM STREQUAL "OS")
# FIXME(liuyiqun): support "armv7;armv7s;arm64" future
set(IOS_ARCH "arm64")
set(IOS_ARCH "armv7;armv7s;arm64")
elseif(IOS_PLATFORM STREQUAL "SIMULATOR")
set(IOS_ARCH "i386;x86_64")
elseif(IOS_PLATFORM STREQUAL "WATCHOS")
set(IOS_ARCH armv7k)
endif()
endif()
set(CMAKE_OSX_ARCHITECTURES ${IOS_ARCH} CACHE string "Build architecture for iOS")
......@@ -249,7 +246,7 @@ set(IOS_COMPILER_FLAGS "${XCODE_IOS_PLATFORM_VERSION_FLAGS} ${XCODE_IOS_BITCODE_
# Hidden visibilty is required for cxx on iOS
set(CMAKE_C_FLAGS "${IOS_COMPILER_FLAGS} ${CMAKE_C_FLAGS}" CACHE STRING "C flags")
set(CMAKE_CXX_FLAGS "${IOS_COMPILER_FLAGS} -fvisibility-inlines-hidden ${CMAKE_CXX_FLAGS}" CACHE STRING "CXX flags")
set(CMAKE_CXX_FLAGS "${IOS_COMPILER_FLAGS} -fvisibility=hidden -fvisibility-inlines-hidden ${CMAKE_CXX_FLAGS}" CACHE STRING "CXX flags")
set(IOS_LINK_FLAGS "${XCODE_IOS_PLATFORM_VERSION_FLAGS} -Wl,-search_paths_first")
......
cmake/cuda.cmake 0 → 100644
浏览文件 @ 988a4a66
if(NOT WITH_GPU)
return()
endif()
set(paddle_known_gpu_archs "30 35 50 52 60 61 70")
set(paddle_known_gpu_archs7 "30 35 50 52")
set(paddle_known_gpu_archs8 "30 35 50 52 60 61")
######################################################################################
# A function for automatic detection of GPUs installed (if autodetection is enabled)
# Usage:
# detect_installed_gpus(out_variable)
function(detect_installed_gpus out_variable)
if(NOT CUDA_gpu_detect_output)
set(cufile ${PROJECT_BINARY_DIR}/detect_cuda_archs.cu)
file(WRITE ${cufile} ""
"#include <cstdio>\n"
"int main() {\n"
" int count = 0;\n"
" if (cudaSuccess != cudaGetDeviceCount(&count)) return -1;\n"
" if (count == 0) return -1;\n"
" for (int device = 0; device < count; ++device) {\n"
" cudaDeviceProp prop;\n"
" if (cudaSuccess == cudaGetDeviceProperties(&prop, device))\n"
" std::printf(\"%d.%d \", prop.major, prop.minor);\n"
" }\n"
" return 0;\n"
"}\n")
execute_process(COMMAND "${CUDA_NVCC_EXECUTABLE}" "-ccbin=${CUDA_HOST_COMPILER}"
"--run" "${cufile}"
WORKING_DIRECTORY "${PROJECT_BINARY_DIR}/CMakeFiles/"
RESULT_VARIABLE nvcc_res OUTPUT_VARIABLE nvcc_out
ERROR_QUIET OUTPUT_STRIP_TRAILING_WHITESPACE)
if(nvcc_res EQUAL 0)
# only keep the last line of nvcc_out
STRING(REGEX REPLACE ";" "\\\\;" nvcc_out "${nvcc_out}")
STRING(REGEX REPLACE "\n" ";" nvcc_out "${nvcc_out}")
list(GET nvcc_out -1 nvcc_out)
string(REPLACE "2.1" "2.1(2.0)" nvcc_out "${nvcc_out}")
set(CUDA_gpu_detect_output ${nvcc_out} CACHE INTERNAL "Returned GPU architetures from detect_installed_gpus tool" FORCE)
endif()
endif()
if(NOT CUDA_gpu_detect_output)
message(STATUS "Automatic GPU detection failed. Building for all known architectures.")
set(${out_variable} ${paddle_known_gpu_archs} PARENT_SCOPE)
else()
set(${out_variable} ${CUDA_gpu_detect_output} PARENT_SCOPE)
endif()
endfunction()
########################################################################
# Function for selecting GPU arch flags for nvcc based on CUDA_ARCH_NAME
# Usage:
# select_nvcc_arch_flags(out_variable)
function(select_nvcc_arch_flags out_variable)
# List of arch names
set(archs_names "Kepler" "Maxwell" "Pascal" "All" "Manual")
set(archs_name_default "All")
if(NOT CMAKE_CROSSCOMPILING)
list(APPEND archs_names "Auto")
endif()
# set CUDA_ARCH_NAME strings (so it will be seen as dropbox in CMake-Gui)
set(CUDA_ARCH_NAME ${archs_name_default} CACHE STRING "Select target NVIDIA GPU achitecture.")
set_property( CACHE CUDA_ARCH_NAME PROPERTY STRINGS "" ${archs_names} )
mark_as_advanced(CUDA_ARCH_NAME)
# verify CUDA_ARCH_NAME value
if(NOT ";${archs_names};" MATCHES ";${CUDA_ARCH_NAME};")
string(REPLACE ";" ", " archs_names "${archs_names}")
message(FATAL_ERROR "Only ${archs_names} architeture names are supported.")
endif()
if(${CUDA_ARCH_NAME} STREQUAL "Manual")
set(CUDA_ARCH_BIN ${paddle_known_gpu_archs} CACHE STRING "Specify 'real' GPU architectures to build binaries for, BIN(PTX) format is supported")
set(CUDA_ARCH_PTX "50" CACHE STRING "Specify 'virtual' PTX architectures to build PTX intermediate code for")
mark_as_advanced(CUDA_ARCH_BIN CUDA_ARCH_PTX)
else()
unset(CUDA_ARCH_BIN CACHE)
unset(CUDA_ARCH_PTX CACHE)
endif()
if(${CUDA_ARCH_NAME} STREQUAL "Kepler")
set(cuda_arch_bin "30 35")
elseif(${CUDA_ARCH_NAME} STREQUAL "Maxwell")
set(cuda_arch_bin "50")
elseif(${CUDA_ARCH_NAME} STREQUAL "Pascal")
set(cuda_arch_bin "60 61")
elseif(${CUDA_ARCH_NAME} STREQUAL "Volta")
set(cuda_arch_bin "70")
elseif(${CUDA_ARCH_NAME} STREQUAL "All")
set(cuda_arch_bin ${paddle_known_gpu_archs})
elseif(${CUDA_ARCH_NAME} STREQUAL "Auto")
detect_installed_gpus(cuda_arch_bin)
else() # (${CUDA_ARCH_NAME} STREQUAL "Manual")
set(cuda_arch_bin ${CUDA_ARCH_BIN})
endif()
# remove dots and convert to lists
string(REGEX REPLACE "\\." "" cuda_arch_bin "${cuda_arch_bin}")
string(REGEX REPLACE "\\." "" cuda_arch_ptx "${CUDA_ARCH_PTX}")
string(REGEX MATCHALL "[0-9()]+" cuda_arch_bin "${cuda_arch_bin}")
string(REGEX MATCHALL "[0-9]+" cuda_arch_ptx "${cuda_arch_ptx}")
list(REMOVE_DUPLICATES cuda_arch_bin)
list(REMOVE_DUPLICATES cuda_arch_ptx)
set(nvcc_flags "")
set(nvcc_archs_readable "")
# Tell NVCC to add binaries for the specified GPUs
foreach(arch ${cuda_arch_bin})
if(arch MATCHES "([0-9]+)\\(([0-9]+)\\)")
# User explicitly specified PTX for the concrete BIN
list(APPEND nvcc_flags -gencode arch=compute_${CMAKE_MATCH_2},code=sm_${CMAKE_MATCH_1})
list(APPEND nvcc_archs_readable sm_${CMAKE_MATCH_1})
else()
# User didn't explicitly specify PTX for the concrete BIN, we assume PTX=BIN
list(APPEND nvcc_flags -gencode arch=compute_${arch},code=sm_${arch})
list(APPEND nvcc_archs_readable sm_${arch})
endif()
endforeach()
# Tell NVCC to add PTX intermediate code for the specified architectures
foreach(arch ${cuda_arch_ptx})
list(APPEND nvcc_flags -gencode arch=compute_${arch},code=compute_${arch})
list(APPEND nvcc_archs_readable compute_${arch})
endforeach()
string(REPLACE ";" " " nvcc_archs_readable "${nvcc_archs_readable}")
set(${out_variable} ${nvcc_flags} PARENT_SCOPE)
set(${out_variable}_readable ${nvcc_archs_readable} PARENT_SCOPE)
endfunction()
message(STATUS "CUDA detected: " ${CUDA_VERSION})
if (${CUDA_VERSION} LESS 7.0)
set(paddle_known_gpu_archs ${paddle_known_gpu_archs})
elseif (${CUDA_VERSION} LESS 8.0) # CUDA 7.x
set(paddle_known_gpu_archs ${paddle_known_gpu_archs7})
list(APPEND CUDA_NVCC_FLAGS "-D_MWAITXINTRIN_H_INCLUDED")
list(APPEND CUDA_NVCC_FLAGS "-D__STRICT_ANSI__")
elseif (${CUDA_VERSION} LESS 9.0) # CUDA 8.x
set(paddle_known_gpu_archs ${paddle_known_gpu_archs8})
list(APPEND CUDA_NVCC_FLAGS "-D_MWAITXINTRIN_H_INCLUDED")
list(APPEND CUDA_NVCC_FLAGS "-D__STRICT_ANSI__")
# CUDA 8 may complain that sm_20 is no longer supported. Suppress the
# warning for now.
list(APPEND CUDA_NVCC_FLAGS "-Wno-deprecated-gpu-targets")
endif()
include_directories(${CUDA_INCLUDE_DIRS})
list(APPEND EXTERNAL_LIBS ${CUDA_LIBRARIES} ${CUDA_rt_LIBRARY})
if(NOT WITH_DSO)
list(APPEND EXTERNAL_LIBS ${CUDNN_LIBRARY} ${CUDA_CUBLAS_LIBRARIES} ${CUDA_curand_LIBRARY} ${NCCL_LIBRARY})
endif(NOT WITH_DSO)
# setting nvcc arch flags
select_nvcc_arch_flags(NVCC_FLAGS_EXTRA)
list(APPEND CUDA_NVCC_FLAGS ${NVCC_FLAGS_EXTRA})
message(STATUS "Added CUDA NVCC flags for: ${NVCC_FLAGS_EXTRA_readable}")
# Set C++11 support
set(CUDA_PROPAGATE_HOST_FLAGS OFF)
# Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
# So, don't set these flags here.
list(APPEND CUDA_NVCC_FLAGS "-std=c++11")
list(APPEND CUDA_NVCC_FLAGS "--use_fast_math")
list(APPEND CUDA_NVCC_FLAGS "-Xcompiler -fPIC")
# Set :expt-relaxed-constexpr to suppress Eigen warnings
list(APPEND CUDA_NVCC_FLAGS "--expt-relaxed-constexpr")
if(CMAKE_BUILD_TYPE STREQUAL "Debug")
list(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_DEBUG})
elseif(CMAKE_BUILD_TYPE STREQUAL "Release")
list(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_RELEASE})
elseif(CMAKE_BUILD_TYPE STREQUAL "RelWithDebInfo")
list(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_RELWITHDEBINFO})
elseif(CMAKE_BUILD_TYPE STREQUAL "MinSizeRel")
list(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_MINSIZEREL})
endif()
mark_as_advanced(CUDA_BUILD_CUBIN CUDA_BUILD_EMULATION CUDA_VERBOSE_BUILD)
mark_as_advanced(CUDA_SDK_ROOT_DIR CUDA_SEPARABLE_COMPILATION)
cmake/external/cares.cmake 0 → 100644
浏览文件 @ 988a4a66
# 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(MOBILE_INFERENCE OR NOT WITH_DISTRIBUTE)
return()
ENDIF()
include (ExternalProject)
# NOTE: c-ares is needed when linking with grpc.
SET(CARES_SOURCES_DIR ${THIRD_PARTY_PATH}/cares)
SET(CARES_INSTALL_DIR ${THIRD_PARTY_PATH}/install/cares)
SET(CARES_INCLUDE_DIR "${CARES_INSTALL_DIR}/include/" CACHE PATH "cares include directory." FORCE)
ExternalProject_Add(
extern_cares
GIT_REPOSITORY "https://github.com/c-ares/c-ares.git"
GIT_TAG "cares-1_13_0"
PREFIX ${CARES_SOURCES_DIR}
UPDATE_COMMAND ""
CONFIGURE_COMMAND ./buildconf && ./configure --disable-shared --prefix=${CARES_INSTALL_DIR}
BUILD_IN_SOURCE 1
BUILD_COMMAND make -j8
INSTALL_COMMAND make install
)
ADD_LIBRARY(cares STATIC IMPORTED GLOBAL)
SET_PROPERTY(TARGET cares PROPERTY IMPORTED_LOCATION
"${CARES_INSTALL_DIR}/lib/libcares.a")
include_directories(${CARES_INCLUDE_DIR})
ADD_DEPENDENCIES(cares extern_cares)
cmake/external/eigen.cmake
浏览文件 @ 988a4a66
......@@ -8,7 +8,7 @@ ExternalProject_Add(
extern_eigen3
${EXTERNAL_PROJECT_LOG_ARGS}
GIT_REPOSITORY "https://github.com/RLovelett/eigen.git"
GIT_TAG "master"
GIT_TAG 70661066beef694cadf6c304d0d07e0758825c10
PREFIX ${EIGEN_SOURCE_DIR}
UPDATE_COMMAND ""
CONFIGURE_COMMAND ""
......
cmake/external/gflags.cmake
浏览文件 @ 988a4a66
......@@ -28,14 +28,8 @@ INCLUDE_DIRECTORIES(${GFLAGS_INCLUDE_DIR})
ExternalProject_Add(
extern_gflags
${EXTERNAL_PROJECT_LOG_ARGS}
# TODO(yiwang): The annoying warnings mentioned in
# https://github.com/PaddlePaddle/Paddle/issues/3277 are caused by
# gflags. I fired a PR https://github.com/gflags/gflags/pull/230
# to fix it. Before it gets accepted by the gflags team, we use
# my personal fork, which contains above fix, temporarily. Let's
# change this back to the official Github repo once my PR is
# merged.
GIT_REPOSITORY "https://github.com/wangkuiyi/gflags.git"
GIT_REPOSITORY "https://github.com/gflags/gflags.git"
GIT_TAG 77592648e3f3be87d6c7123eb81cbad75f9aef5a
PREFIX ${GFLAGS_SOURCES_DIR}
UPDATE_COMMAND ""
CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
......@@ -45,11 +39,11 @@ ExternalProject_Add(
-DCMAKE_INSTALL_PREFIX=${GFLAGS_INSTALL_DIR}
-DCMAKE_POSITION_INDEPENDENT_CODE=ON
-DBUILD_TESTING=OFF
-DCMAKE_BUILD_TYPE=Release
-DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
${EXTERNAL_OPTIONAL_ARGS}
CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GFLAGS_INSTALL_DIR}
-DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-DCMAKE_BUILD_TYPE:STRING=Release
-DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
)
ADD_LIBRARY(gflags STATIC IMPORTED GLOBAL)
......
cmake/external/glog.cmake
浏览文件 @ 988a4a66
......@@ -26,11 +26,21 @@ ENDIF(WIN32)
INCLUDE_DIRECTORIES(${GLOG_INCLUDE_DIR})
IF(ANDROID AND ${CMAKE_SYSTEM_VERSION} VERSION_LESS "21")
# Using the unofficial glog for Android API < 21
SET(GLOG_REPOSITORY "https://github.com/Xreki/glog.git")
SET(GLOG_TAG "8a547150548b284382ccb6582408e9140ff2bea8")
ELSE()
SET(GLOG_REPOSITORY "https://github.com/google/glog.git")
SET(GLOG_TAG "v0.3.5")
ENDIF()
ExternalProject_Add(
extern_glog
${EXTERNAL_PROJECT_LOG_ARGS}
DEPENDS gflags
GIT_REPOSITORY "https://github.com/google/glog.git"
GIT_REPOSITORY ${GLOG_REPOSITORY}
GIT_TAG ${GLOG_TAG}
PREFIX ${GLOG_SOURCES_DIR}
UPDATE_COMMAND ""
CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
......@@ -43,12 +53,12 @@ ExternalProject_Add(
-DWITH_GFLAGS=ON
-Dgflags_DIR=${GFLAGS_INSTALL_DIR}/lib/cmake/gflags
-DBUILD_TESTING=OFF
-DCMAKE_BUILD_TYPE=Release
-DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
${EXTERNAL_OPTIONAL_ARGS}
CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GLOG_INSTALL_DIR}
-DCMAKE_INSTALL_LIBDIR:PATH=${GLOG_INSTALL_DIR}/lib
-DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-DCMAKE_BUILD_TYPE:STRING=Release
-DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
)
ADD_LIBRARY(glog STATIC IMPORTED GLOBAL)
......
cmake/external/grpc.cmake 0 → 100644
浏览文件 @ 988a4a66
# 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(MOBILE_INFERENCE OR NOT WITH_DISTRIBUTE)
return()
ENDIF()
include (ExternalProject)
SET(GRPC_SOURCES_DIR ${THIRD_PARTY_PATH}/grpc)
SET(GRPC_INSTALL_DIR ${THIRD_PARTY_PATH}/install/grpc)
SET(GRPC_INCLUDE_DIR "${GRPC_INSTALL_DIR}/include/" CACHE PATH "grpc include directory." FORCE)
SET(GRPC_CPP_PLUGIN "${GRPC_INSTALL_DIR}/bin/grpc_cpp_plugin" CACHE FILEPATH "GRPC_CPP_PLUGIN" FORCE)
IF(APPLE)
SET(BUILD_CMD make -n HAS_SYSTEM_PROTOBUF=false -s -j8 static grpc_cpp_plugin | sed "s/-Werror//g" | sh)
ELSE()
SET(BUILD_CMD make HAS_SYSTEM_PROTOBUF=false -s -j8 static grpc_cpp_plugin)
ENDIF()
ExternalProject_Add(
extern_grpc
DEPENDS protobuf zlib
GIT_REPOSITORY "https://github.com/grpc/grpc.git"
GIT_TAG "v1.7.x"
PREFIX ${GRPC_SOURCES_DIR}
UPDATE_COMMAND ""
CONFIGURE_COMMAND ""
BUILD_IN_SOURCE 1
# NOTE(yuyang18):
# Disable -Werror, otherwise the compile will fail in MacOS.
# It seems that we cannot configure that by make command.
# Just dry run make command and remove `-Werror`, then use a shell to run make commands
BUILD_COMMAND ${BUILD_CMD}
INSTALL_COMMAND make prefix=${GRPC_INSTALL_DIR} install
)
# FIXME(typhoonzero): hack to get static lib path, try a better way like merge them.
ADD_LIBRARY(grpc++_unsecure STATIC IMPORTED GLOBAL)
SET_PROPERTY(TARGET grpc++_unsecure PROPERTY IMPORTED_LOCATION
"${GRPC_INSTALL_DIR}/lib/libgrpc++_unsecure.a")
ADD_LIBRARY(grpc++ STATIC IMPORTED GLOBAL)
SET_PROPERTY(TARGET grpc++ PROPERTY IMPORTED_LOCATION
"${GRPC_INSTALL_DIR}/lib/libgrpc++.a")
ADD_LIBRARY(gpr STATIC IMPORTED GLOBAL)
SET_PROPERTY(TARGET gpr PROPERTY IMPORTED_LOCATION
"${GRPC_INSTALL_DIR}/lib/libgpr.a")
ADD_LIBRARY(grpc_unsecure STATIC IMPORTED GLOBAL)
SET_PROPERTY(TARGET grpc_unsecure PROPERTY IMPORTED_LOCATION
"${GRPC_INSTALL_DIR}/lib/libgrpc_unsecure.a")
include_directories(${GRPC_INCLUDE_DIR})
ADD_DEPENDENCIES(grpc++_unsecure extern_grpc)
cmake/external/gtest.cmake
浏览文件 @ 988a4a66
......@@ -56,11 +56,11 @@ IF(WITH_TESTING)
-DBUILD_GMOCK=ON
-Dgtest_disable_pthreads=ON
-Dgtest_force_shared_crt=ON
-DCMAKE_BUILD_TYPE=Release
-DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
${EXTERNAL_OPTIONAL_ARGS}
CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${GTEST_INSTALL_DIR}
-DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-DCMAKE_BUILD_TYPE:STRING=Release
-DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
)
ADD_LIBRARY(gtest STATIC IMPORTED GLOBAL)
......
cmake/external/mkldnn.cmake
浏览文件 @ 988a4a66
......@@ -40,28 +40,32 @@ INCLUDE_DIRECTORIES(${MKLDNN_INC_DIR})
IF(${CBLAS_PROVIDER} STREQUAL "MKLML")
SET(MKLDNN_DEPENDS ${MKLML_PROJECT})
SET(MKLDNN_MKLROOT ${MKLML_ROOT})
SET(MKLDNN_IOMP_LIB ${MKLML_IOMP_LIB})
SET(MKLDNN_IOMP_DIR ${MKLML_LIB_DIR})
MESSAGE(STATUS "Build MKLDNN with ${MKLDNN_MKLROOT}")
MESSAGE(STATUS "Build MKLDNN with MKLML ${MKLML_ROOT}")
ELSE()
MESSAGE(FATAL_ERROR "Should enable MKLML when build MKLDNN")
ENDIF()
SET(MKLDNN_CFLAG "${CMAKE_C_FLAGS} -Wno-error=strict-overflow")
SET(MKLDNN_CXXFLAG "${CMAKE_CXX_FLAGS} -Wno-error=strict-overflow")
ExternalProject_Add(
${MKLDNN_PROJECT}
${EXTERNAL_PROJECT_LOG_ARGS}
DEPENDS ${MKLDNN_DEPENDS}
GIT_REPOSITORY "https://github.com/01org/mkl-dnn.git"
GIT_TAG "v0.10"
GIT_TAG "v0.11"
PREFIX ${MKLDNN_SOURCES_DIR}
UPDATE_COMMAND ""
CMAKE_ARGS -DCMAKE_INSTALL_PREFIX=${MKLDNN_INSTALL_DIR}
CMAKE_ARGS -DMKLROOT=${MKLDNN_MKLROOT}
CMAKE_ARGS -DMKLROOT=${MKLML_ROOT}
CMAKE_ARGS -DCMAKE_C_FLAGS=${MKLDNN_CFLAG}
CMAKE_ARGS -DCMAKE_CXX_FLAGS=${MKLDNN_CXXFLAG}
CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${MKLDNN_INSTALL_DIR}
-DMKLROOT:PATH=${MKLDNN_MKLROOT}
-DMKLROOT:PATH=${MKLML_ROOT}
)
ADD_LIBRARY(mkldnn SHARED IMPORTED GLOBAL)
SET_PROPERTY(TARGET mkldnn PROPERTY IMPORTED_LOCATION ${MKLDNN_LIB})
ADD_DEPENDENCIES(mkldnn ${MKLDNN_PROJECT})
MESSAGE(STATUS "Mkldnn library: ${MKLDNN_LIB}")
MESSAGE(STATUS "MKLDNN library: ${MKLDNN_LIB}")
add_definitions(-DPADDLE_USE_MKLDNN)
LIST(APPEND external_project_dependencies mkldnn)
cmake/external/mklml.cmake
浏览文件 @ 988a4a66
......@@ -27,8 +27,8 @@ ENDIF()
INCLUDE(ExternalProject)
SET(MKLML_PROJECT "extern_mklml")
SET(MKLML_VER "mklml_lnx_2018.0.20170720")
SET(MKLML_URL "https://github.com/01org/mkl-dnn/releases/download/v0.10/${MKLML_VER}.tgz")
SET(MKLML_VER "mklml_lnx_2018.0.1.20171007")
SET(MKLML_URL "https://github.com/01org/mkl-dnn/releases/download/v0.11/${MKLML_VER}.tgz")
SET(MKLML_SOURCE_DIR "${THIRD_PARTY_PATH}/mklml")
SET(MKLML_DOWNLOAD_DIR "${MKLML_SOURCE_DIR}/src/${MKLML_PROJECT}")
SET(MKLML_DST_DIR "mklml")
......
cmake/external/nccl.cmake 0 → 100644
浏览文件 @ 988a4a66
# 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 ${THIRD_PARTY_PATH}/nccl)
include_directories(${NCCL_SOURCE_DIR}/src/extern_nccl/src)
if(WITH_DSO)
# If we use DSO, we do not build nccl, just download the dependencies
set(NCCL_BUILD_COMMAND "")
set(NCCL_INSTALL_COMMAND "")
set(NCCL_INSTALL_DIR "")
else()
# otherwise, we build nccl and link it.
set(NCCL_INSTALL_DIR ${THIRD_PARTY_PATH}/install/nccl)
# Note: cuda 8.0 is needed to make nccl
# When cuda is not installed on the system directory, need to set CUDA_HOME to your cuda root
set(NCCL_BUILD_COMMAND "make -j 8")
set(NCCL_INSTALL_COMMAND "make install PREFIX=${NCCL_INSTALL_DIR}")
endif()
ExternalProject_Add(
extern_nccl
${EXTERNAL_PROJECT_LOG_ARGS}
GIT_REPOSITORY "https://github.com/NVIDIA/nccl.git"
GIT_TAG "v1.3.4-1"
PREFIX "${NCCL_SOURCE_DIR}"
UPDATE_COMMAND ""
CONFIGURE_COMMAND ""
BUILD_COMMAND "${NCCL_BUILD_COMMAND}"
INSTALL_COMMAND "${NCCL_INSTALL_COMMAND}"
INSTALL_DIR "${NCCL_INSTALL_DIR}"
TEST_COMMAND ""
)
if(WITH_DSO)
if(${CMAKE_VERSION} VERSION_LESS "3.3.0")
set(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/lib_nccl_dummy.c)
file(WRITE ${dummyfile} "const char * dummy_nccl = \"${dummyfile}\";")
add_library(nccl STATIC ${dummyfile})
else()
add_library(nccl INTERFACE)
endif()
else()
add_library(nccl STATIC IMPORTED GLOBAL)
set_property(TARGET nccl PROPERTY IMPORTED_LOCATION
${NCCL_INSTALL_DIR}/lib/libnccl_static.a)
endif()
add_dependencies(nccl extern_nccl)
cmake/external/openblas.cmake
浏览文件 @ 988a4a66
# 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.
......@@ -29,7 +29,7 @@ IF(NOT ${CBLAS_FOUND})
"${CBLAS_INSTALL_DIR}/lib/${CMAKE_STATIC_LIBRARY_PREFIX}openblas${CMAKE_STATIC_LIBRARY_SUFFIX}"
CACHE FILEPATH "openblas library." FORCE)
SET(OPENBLAS_CC "${CMAKE_C_COMPILER}")
SET(OPENBLAS_CC "${CMAKE_C_COMPILER} -Wno-unused-but-set-variable -Wno-unused-variable")
IF(CMAKE_CROSSCOMPILING)
SET(OPTIONAL_ARGS HOSTCC=${HOST_C_COMPILER})
......@@ -45,15 +45,14 @@ IF(NOT ${CBLAS_FOUND})
SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} TARGET=ARMV8 BINARY=64 USE_THREAD=0)
ENDIF()
ELSEIF(IOS)
# FIXME(liuyiqun): support multiple architectures
SET(OPENBLAS_COMMIT "b5c96fcfcdc82945502a2303116a64d89985daf5")
SET(OPENBLAS_CC "${OPENBLAS_CC} ${CMAKE_C_FLAGS} -isysroot ${CMAKE_OSX_SYSROOT}")
IF(CMAKE_OSX_ARCHITECTURES MATCHES "armv7")
SET(OPENBLAS_CC "${OPENBLAS_CC} -arch armv7")
SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} TARGET=ARMV7 ARM_SOFTFP_ABI=1 USE_THREAD=0)
ELSEIF(CMAKE_OSX_ARCHITECTURES MATCHES "arm64")
IF(CMAKE_OSX_ARCHITECTURES MATCHES "arm64")
SET(OPENBLAS_COMMIT "b5c96fcfcdc82945502a2303116a64d89985daf5")
SET(OPENBLAS_CC "${OPENBLAS_CC} ${CMAKE_C_FLAGS} -isysroot ${CMAKE_OSX_SYSROOT}")
SET(OPENBLAS_CC "${OPENBLAS_CC} -arch arm64")
SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} TARGET=ARMV8 BINARY=64 USE_THREAD=0 CROSS_SUFFIX=${CROSS_SUFFIX})
ELSE()
MESSAGE(FATAL_ERROR "OpenBLAS only support arm64 architectures on iOS. "
"You can set IOS_USE_VECLIB_FOR_BLAS=ON or USE_EIGEN_FOR_BLAS=ON to use other blas library instead.")
ENDIF()
ELSEIF(RPI)
# use hardfp
......@@ -86,7 +85,7 @@ IF(NOT ${CBLAS_FOUND})
UPDATE_COMMAND ""
CONFIGURE_COMMAND ""
)
SET(CBLAS_PROVIDER openblas)
IF(WITH_C_API)
INSTALL(DIRECTORY ${CBLAS_INC_DIR} DESTINATION third_party/openblas)
# Because libopenblas.a is a symbolic link of another library, thus need to
......@@ -98,7 +97,7 @@ IF(NOT ${CBLAS_FOUND})
ENDIF()
INSTALL(CODE "execute_process(
COMMAND ${CMAKE_COMMAND} -E copy_directory ${CBLAS_INSTALL_DIR}/lib
destination ${CMAKE_INSTALL_PREFIX}/${TMP_INSTALL_DIR}
${CMAKE_INSTALL_PREFIX}/${TMP_INSTALL_DIR}
)"
)
INSTALL(CODE "MESSAGE(STATUS \"Installing: \"
......@@ -115,11 +114,7 @@ INCLUDE_DIRECTORIES(${CBLAS_INC_DIR})
# linear algebra libraries for cc_library(xxx SRCS xxx.c DEPS cblas)
SET(dummyfile ${CMAKE_CURRENT_BINARY_DIR}/cblas_dummy.c)
FILE(WRITE ${dummyfile} "const char * dummy = \"${dummyfile}\";")
IF(${CBLAS_PROVIDER} MATCHES MKL)
ADD_LIBRARY(cblas SHARED ${dummyfile})
ELSE()
ADD_LIBRARY(cblas STATIC ${dummyfile})
ENDIF()
ADD_LIBRARY(cblas STATIC ${dummyfile})
TARGET_LINK_LIBRARIES(cblas ${CBLAS_LIBRARIES})
IF(NOT ${CBLAS_FOUND})
......
cmake/external/protobuf.cmake
浏览文件 @ 988a4a66
......@@ -15,7 +15,18 @@
INCLUDE(ExternalProject)
# Always invoke `FIND_PACKAGE(Protobuf)` for importing function protobuf_generate_cpp
FIND_PACKAGE(Protobuf QUIET)
SET(PROTOBUF_FOUND "OFF")
macro(UNSET_VAR VAR_NAME)
UNSET(${VAR_NAME} CACHE)
UNSET(${VAR_NAME})
endmacro()
UNSET_VAR(PROTOBUF_INCLUDE_DIR)
UNSET_VAR(PROTOBUF_FOUND)
UNSET_VAR(PROTOBUF_PROTOC_EXECUTABLE)
UNSET_VAR(PROTOBUF_PROTOC_LIBRARY)
UNSET_VAR(PROTOBUF_LITE_LIBRARY)
UNSET_VAR(PROTOBUF_LIBRARY)
UNSET_VAR(PROTOBUF_INCLUDE_DIR)
UNSET_VAR(Protobuf_PROTOC_EXECUTABLE)
if(NOT COMMAND protobuf_generate_python) # before cmake 3.4, protobuf_genrerate_python is not defined.
function(protobuf_generate_python SRCS)
......@@ -110,7 +121,6 @@ macro(PROMPT_PROTOBUF_LIB)
# FIND_Protobuf.cmake uses `Protobuf_PROTOC_EXECUTABLE`.
# make `protobuf_generate_cpp` happy.
SET(Protobuf_PROTOC_EXECUTABLE ${PROTOBUF_PROTOC_EXECUTABLE})
FOREACH(dep ${protobuf_DEPS})
ADD_DEPENDENCIES(protobuf ${dep})
ADD_DEPENDENCIES(protobuf_lite ${dep})
......@@ -128,11 +138,11 @@ endmacro()
set(PROTOBUF_ROOT "" CACHE PATH "Folder contains protobuf")
if (NOT "${PROTOBUF_ROOT}" STREQUAL "")
find_path(PROTOBUF_INCLUDE_DIR google/protobuf/message.h PATHS ${PROTOBUF_ROOT}/include)
find_library(PROTOBUF_LIBRARY protobuf PATHS ${PROTOBUF_ROOT}/lib)
find_library(PROTOBUF_LITE_LIBRARY protobuf-lite PATHS ${PROTOBUF_ROOT}/lib)
find_library(PROTOBUF_PROTOC_LIBRARY protoc PATHS ${PROTOBUF_ROOT}/lib)
find_program(PROTOBUF_PROTOC_EXECUTABLE protoc PATHS ${PROTOBUF_ROOT}/bin)
find_path(PROTOBUF_INCLUDE_DIR google/protobuf/message.h PATHS ${PROTOBUF_ROOT}/include NO_DEFAULT_PATH)
find_library(PROTOBUF_LIBRARY protobuf PATHS ${PROTOBUF_ROOT}/lib NO_DEFAULT_PATH)
find_library(PROTOBUF_LITE_LIBRARY protobuf-lite PATHS ${PROTOBUF_ROOT}/lib NO_DEFAULT_PATH)
find_library(PROTOBUF_PROTOC_LIBRARY protoc PATHS ${PROTOBUF_ROOT}/lib NO_DEFAULT_PATH)
find_program(PROTOBUF_PROTOC_EXECUTABLE protoc PATHS ${PROTOBUF_ROOT}/bin NO_DEFAULT_PATH)
if (PROTOBUF_INCLUDE_DIR AND PROTOBUF_LIBRARY AND PROTOBUF_LITE_LIBRARY AND PROTOBUF_PROTOC_LIBRARY AND PROTOBUF_PROTOC_EXECUTABLE)
message(STATUS "Using custom protobuf library in ${PROTOBUF_ROOT}.")
SET_PROTOBUF_VERSION()
......@@ -178,32 +188,48 @@ FUNCTION(build_protobuf TARGET_NAME BUILD_FOR_HOST)
SET(OPTIONAL_CACHE_ARGS "-DZLIB_ROOT:STRING=${ZLIB_ROOT}")
ENDIF()
SET(PROTOBUF_REPO "https://github.com/google/protobuf.git")
SET(PROTOBUF_TAG "9f75c5aa851cd877fb0d93ccc31b8567a6706546")
IF(MOBILE_INFERENCE)
# The reason why the official version is not used is described in
# https://github.com/PaddlePaddle/Paddle/issues/6114
SET(PROTOBUF_REPO "https://github.com/qingqing01/protobuf.git")
SET(PROTOBUF_TAG "v3.2.0")
IF(NOT BUILD_FOR_HOST)
SET(OPTIONAL_ARGS ${OPTIONAL_ARGS} "-Dprotobuf_BUILD_PROTOC_BINARIES=OFF")
ENDIF()
ENDIF()
ExternalProject_Add(
${TARGET_NAME}
${EXTERNAL_PROJECT_LOG_ARGS}
PREFIX ${PROTOBUF_SOURCES_DIR}
UPDATE_COMMAND ""
DEPENDS zlib
GIT_REPOSITORY "https://github.com/google/protobuf.git"
GIT_TAG "9f75c5aa851cd877fb0d93ccc31b8567a6706546"
GIT_REPOSITORY ${PROTOBUF_REPO}
GIT_TAG ${PROTOBUF_TAG}
CONFIGURE_COMMAND
${CMAKE_COMMAND} ${PROTOBUF_SOURCES_DIR}/src/${TARGET_NAME}/cmake
${OPTIONAL_ARGS}
-Dprotobuf_BUILD_TESTS=OFF
-DCMAKE_POSITION_INDEPENDENT_CODE=ON
-DCMAKE_BUILD_TYPE=Release
-DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
-DCMAKE_INSTALL_PREFIX=${PROTOBUF_INSTALL_DIR}
-DCMAKE_INSTALL_LIBDIR=lib
CMAKE_CACHE_ARGS
-DCMAKE_INSTALL_PREFIX:PATH=${PROTOBUF_INSTALL_DIR}
-DCMAKE_BUILD_TYPE:STRING=Release
-DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
-DCMAKE_VERBOSE_MAKEFILE:BOOL=OFF
-DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
${OPTIONAL_CACHE_ARGS}
)
ENDFUNCTION()
SET(PROTOBUF_VERSION 3.1)
IF(NOT MOBILE_INFERENCE)
SET(PROTOBUF_VERSION 3.1)
ELSE()
SET(PROTOBUF_VERSION 3.2)
ENDIF()
IF(CMAKE_CROSSCOMPILING)
build_protobuf(protobuf_host TRUE)
LIST(APPEND external_project_dependencies protobuf_host)
......
cmake/external/pybind11.cmake
浏览文件 @ 988a4a66
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)
cmake/external/swig.cmake
浏览文件 @ 988a4a66
# 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.
......
cmake/external/warpctc.cmake
浏览文件 @ 988a4a66
......@@ -12,6 +12,10 @@
# See the License for the specific language governing permissions and
# limitations under the License.
IF(MOBILE_INFERENCE)
return()
ENDIF()
INCLUDE(ExternalProject)
SET(WARPCTC_SOURCES_DIR ${THIRD_PARTY_PATH}/warpctc)
......@@ -35,6 +39,7 @@ ExternalProject_Add(
extern_warpctc
${EXTERNAL_PROJECT_LOG_ARGS}
GIT_REPOSITORY "https://github.com/gangliao/warp-ctc.git"
GIT_TAG b63a0644654a3e0ed624c85a1767bc8193aead09
PREFIX ${WARPCTC_SOURCES_DIR}
UPDATE_COMMAND ""
CMAKE_ARGS -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
......@@ -48,9 +53,9 @@ ExternalProject_Add(
-DCMAKE_DISABLE_FIND_PACKAGE_Torch=ON
-DBUILD_SHARED=ON
-DCMAKE_POSITION_INDEPENDENT_CODE=ON
-DCMAKE_BUILD_TYPE=Release
-DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
${EXTERNAL_OPTIONAL_ARGS}
CMAKE_CACHE_ARGS -DCMAKE_BUILD_TYPE:STRING=Release
CMAKE_CACHE_ARGS -DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
-DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-DCMAKE_INSTALL_PREFIX:PATH=${WARPCTC_INSTALL_DIR}
)
......
cmake/external/zlib.cmake
浏览文件 @ 988a4a66
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
#
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#
# http://www.apache.org/licenses/LICENSE-2.0
#
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
......@@ -42,14 +42,16 @@ ExternalProject_Add(
-DBUILD_SHARED_LIBS=OFF
-DCMAKE_POSITION_INDEPENDENT_CODE=ON
-DCMAKE_MACOSX_RPATH=ON
-DCMAKE_BUILD_TYPE=Release
-DCMAKE_BUILD_TYPE=${THIRD_PARTY_BUILD_TYPE}
${EXTERNAL_OPTIONAL_ARGS}
CMAKE_CACHE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=${ZLIB_INSTALL_DIR}
-DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
-DCMAKE_BUILD_TYPE:STRING=Release
-DCMAKE_BUILD_TYPE:STRING=${THIRD_PARTY_BUILD_TYPE}
)
LIST(APPEND external_project_dependencies zlib)
ADD_LIBRARY(zlib_target STATIC IMPORTED GLOBAL)
SET_PROPERTY(TARGET zlib_target PROPERTY IMPORTED_LOCATION ${ZLIB_LIBRARIES})
IF(WITH_C_API)
INSTALL(DIRECTORY ${ZLIB_INCLUDE_DIR} DESTINATION third_party/zlib)
......
cmake/flags.cmake
浏览文件 @ 988a4a66
......@@ -111,6 +111,8 @@ set(COMMON_FLAGS
-Wno-error=sign-compare
-Wno-error=unused-local-typedefs
-Wno-error=parentheses-equality # Warnings in pybind11
-Wno-error=ignored-attributes # Warnings in Eigen, gcc 6.3
-Wno-error=terminate # Warning in PADDLE_ENFORCE
)
set(GPU_COMMON_FLAGS
......@@ -149,58 +151,3 @@ endforeach()
foreach(flag ${GPU_COMMON_FLAGS})
safe_set_nvflag(${flag})
endforeach()
set(CUDA_PROPAGATE_HOST_FLAGS OFF)
# Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
# So, don't set these flags here.
LIST(APPEND CUDA_NVCC_FLAGS -std=c++11)
LIST(APPEND CUDA_NVCC_FLAGS --use_fast_math)
if(CMAKE_BUILD_TYPE STREQUAL "Debug")
LIST(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_DEBUG})
elseif(CMAKE_BUILD_TYPE STREQUAL "Release")
LIST(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_RELEASE})
elseif(CMAKE_BUILD_TYPE STREQUAL "RelWithDebInfo")
LIST(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_RELWITHDEBINFO})
elseif(CMAKE_BUILD_TYPE STREQUAL "MinSizeRel")
LIST(APPEND CUDA_NVCC_FLAGS ${CMAKE_CXX_FLAGS_MINSIZEREL})
endif()
function(specify_cuda_arch cuda_version cuda_arch)
if(${cuda_version} VERSION_GREATER "8.0")
foreach(capability 61 62)
if(${cuda_arch} STREQUAL ${capability})
list(APPEND __arch_flags " -gencode arch=compute_${cuda_arch},code=sm_${cuda_arch}")
endif()
endforeach()
elseif(${cuda_version} VERSION_GREATER "7.0" and ${cuda_arch} STREQUAL "53")
list(APPEND __arch_flags " -gencode arch=compute_${cuda_arch},code=sm_${cuda_arch}")
endif()
endfunction()
# Common gpu architectures: Kepler, Maxwell
foreach(capability 30 35 50)
list(APPEND __arch_flags " -gencode arch=compute_${capability},code=sm_${capability}")
endforeach()
if (CUDA_VERSION VERSION_GREATER "7.0" OR CUDA_VERSION VERSION_EQUAL "7.0")
list(APPEND __arch_flags " -gencode arch=compute_52,code=sm_52")
endif()
# Modern gpu architectures: Pascal
if (CUDA_VERSION VERSION_GREATER "8.0" OR CUDA_VERSION VERSION_EQUAL "8.0")
list(APPEND __arch_flags " -gencode arch=compute_60,code=sm_60")
list(APPEND CUDA_NVCC_FLAGS --expt-relaxed-constexpr)
endif()
# Custom gpu architecture
set(CUDA_ARCH)
if(CUDA_ARCH)
specify_cuda_arch(${CUDA_VERSION} ${CUDA_ARCH})
endif()
set(CUDA_NVCC_FLAGS ${__arch_flags} ${CUDA_NVCC_FLAGS})
cmake/generic.cmake
浏览文件 @ 988a4a66
......@@ -93,7 +93,7 @@ include_directories(${CMAKE_CURRENT_BINARY_DIR})
if(NOT APPLE AND NOT ANDROID)
find_package(Threads REQUIRED)
link_libraries(${CMAKE_THREAD_LIBS_INIT})
set(CMAKE_CXX_LINK_EXECUTABLE "${CMAKE_CXX_LINK_EXECUTABLE} -ldl -lrt")
set(CMAKE_CXX_LINK_EXECUTABLE "${CMAKE_CXX_LINK_EXECUTABLE} -pthread -ldl -lrt")
endif(NOT APPLE AND NOT ANDROID)
function(merge_static_libs TARGET_NAME)
......@@ -227,8 +227,8 @@ function(cc_test TARGET_NAME)
set(multiValueArgs SRCS DEPS)
cmake_parse_arguments(cc_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
add_executable(${TARGET_NAME} ${cc_test_SRCS})
target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} gtest gtest_main)
add_dependencies(${TARGET_NAME} ${cc_test_DEPS} gtest gtest_main)
target_link_libraries(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main paddle_memory gtest gflags)
add_dependencies(${TARGET_NAME} ${cc_test_DEPS} paddle_gtest_main paddle_memory gtest gflags)
add_test(NAME ${TARGET_NAME} COMMAND ${TARGET_NAME} WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
endif()
endfunction(cc_test)
......@@ -288,8 +288,8 @@ function(nv_test TARGET_NAME)
set(multiValueArgs SRCS DEPS)
cmake_parse_arguments(nv_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
cuda_add_executable(${TARGET_NAME} ${nv_test_SRCS})
target_link_libraries(${TARGET_NAME} ${nv_test_DEPS} gtest gtest_main)
add_dependencies(${TARGET_NAME} ${nv_test_DEPS} gtest gtest_main)
target_link_libraries(${TARGET_NAME} ${nv_test_DEPS} paddle_gtest_main paddle_memory gtest gflags)
add_dependencies(${TARGET_NAME} ${nv_test_DEPS} paddle_gtest_main paddle_memory gtest gflags)
add_test(${TARGET_NAME} ${TARGET_NAME})
endif()
endfunction(nv_test)
......@@ -389,13 +389,60 @@ function(go_test TARGET_NAME)
WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
endfunction(go_test)
# Modification of standard 'protobuf_generate_cpp()' with protobuf-lite support
# Usage:
# paddle_protobuf_generate_cpp(<proto_srcs> <proto_hdrs> <proto_files>)
function(paddle_protobuf_generate_cpp SRCS HDRS)
if(NOT ARGN)
message(SEND_ERROR "Error: paddle_protobuf_generate_cpp() called without any proto files")
return()
endif()
set(${SRCS})
set(${HDRS})
if (MOBILE_INFERENCE)
set(EXTRA_FLAG "lite:")
else()
set(EXTRA_FLAG "")
endif()
foreach(FIL ${ARGN})
get_filename_component(ABS_FIL ${FIL} ABSOLUTE)
get_filename_component(FIL_WE ${FIL} NAME_WE)
set(_protobuf_protoc_src "${CMAKE_CURRENT_BINARY_DIR}/${FIL_WE}.pb.cc")
set(_protobuf_protoc_hdr "${CMAKE_CURRENT_BINARY_DIR}/${FIL_WE}.pb.h")
list(APPEND ${SRCS} "${_protobuf_protoc_src}")
list(APPEND ${HDRS} "${_protobuf_protoc_hdr}")
add_custom_command(
OUTPUT "${_protobuf_protoc_src}"
"${_protobuf_protoc_hdr}"
COMMAND ${CMAKE_COMMAND} -E make_directory "${CMAKE_CURRENT_BINARY_DIR}"
COMMAND ${PROTOBUF_PROTOC_EXECUTABLE}
-I${CMAKE_CURRENT_SOURCE_DIR}
--cpp_out "${EXTRA_FLAG}${CMAKE_CURRENT_BINARY_DIR}" ${ABS_FIL}
DEPENDS ${ABS_FIL} protoc
COMMENT "Running C++ protocol buffer compiler on ${FIL}"
VERBATIM )
endforeach()
set_source_files_properties(${${SRCS}} ${${HDRS}} PROPERTIES GENERATED TRUE)
set(${SRCS} ${${SRCS}} PARENT_SCOPE)
set(${HDRS} ${${HDRS}} PARENT_SCOPE)
endfunction()
function(proto_library TARGET_NAME)
set(oneValueArgs "")
set(multiValueArgs SRCS DEPS)
cmake_parse_arguments(proto_library "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
set(proto_srcs)
set(proto_hdrs)
protobuf_generate_cpp(proto_srcs proto_hdrs ${proto_library_SRCS})
paddle_protobuf_generate_cpp(proto_srcs proto_hdrs ${proto_library_SRCS})
cc_library(${TARGET_NAME} SRCS ${proto_srcs} DEPS ${proto_library_DEPS} protobuf)
endfunction()
......@@ -412,11 +459,58 @@ function(py_test TARGET_NAME)
if(WITH_TESTING)
set(options STATIC static SHARED shared)
set(oneValueArgs "")
set(multiValueArgs SRCS DEPS)
cmake_parse_arguments(py_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
set(multiValueArgs SRCS DEPS ARGS)
cmake_parse_arguments(py_test "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
add_test(NAME ${TARGET_NAME}
COMMAND env PYTHONPATH=${PADDLE_PYTHON_BUILD_DIR}/lib-python
python2 ${py_test_SRCS}
${PYTHON_EXECUTABLE} -u ${py_test_SRCS} ${py_test_ARGS}
WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
endif()
endfunction()
# grpc_library generate grpc code using grpc_cpp_plugin and protoc
# then build the generated protobuf code and grpc code with your
# implementation source codes together. Use SRCS argument for your
# implementation source files and PROTO argument for your .proto
# files.
#
# Usage: grpc_library(my_target SRCS my_client.cc PROTO my_target.proto DEPS my_dep)
function(grpc_library TARGET_NAME)
set(oneValueArgs PROTO)
set(multiValueArgs SRCS DEPS)
set(options "")
cmake_parse_arguments(grpc_library "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
message(STATUS "generating grpc ${grpc_library_PROTO}")
get_filename_component(ABS_PROTO ${grpc_library_PROTO} ABSOLUTE)
get_filename_component(PROTO_WE ${grpc_library_PROTO} NAME_WE)
get_filename_component(PROTO_PATH ${ABS_PROTO} PATH)
protobuf_generate_cpp(grpc_proto_srcs grpc_proto_hdrs "${ABS_PROTO}")
set(grpc_grpc_srcs "${CMAKE_CURRENT_BINARY_DIR}/${PROTO_WE}.grpc.pb.cc")
set(grpc_grpc_hdrs "${CMAKE_CURRENT_BINARY_DIR}/${PROTO_WE}.grpc.pb.h")
cc_library("${TARGET_NAME}_proto" SRCS "${grpc_proto_srcs}")
add_custom_command(
OUTPUT "${grpc_grpc_srcs}" "${grpc_grpc_hdrs}"
COMMAND ${PROTOBUF_PROTOC_EXECUTABLE}
ARGS --grpc_out "${CMAKE_CURRENT_BINARY_DIR}" -I "${PROTO_PATH}"
--plugin=protoc-gen-grpc="${GRPC_CPP_PLUGIN}" "${ABS_PROTO}"
DEPENDS "${ABS_PROTO}" ${PROTOBUF_PROTOC_EXECUTABLE} extern_grpc)
# FIXME(typhoonzero): grpc generated code do not generate virtual-dtor, mark it
# as compiler warnings instead of error. Should try remove the warnings also.
set_source_files_properties(
${grpc_grpc_srcs}
PROPERTIES
COMPILE_FLAGS "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor")
cc_library("${TARGET_NAME}_grpc" SRCS "${grpc_grpc_srcs}")
set_source_files_properties(
${grpc_library_SRCS}
PROPERTIES
COMPILE_FLAGS "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor")
cc_library("${TARGET_NAME}" SRCS "${grpc_library_SRCS}" DEPS "${TARGET_NAME}_grpc" "${TARGET_NAME}_proto" "${grpc_library_DEPS}")
endfunction()
cmake/simd.cmake
浏览文件 @ 988a4a66
# 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 ${AVX2_FLAG})
set(AVX2_FOUND_EXITCODE 1 CACHE STRING "Result from TRY_RUN" FORCE)
CHECK_CXX_SOURCE_RUNS("
#include <immintrin.h>
int main()
......
cmake/util.cmake
浏览文件 @ 988a4a66
......@@ -73,32 +73,50 @@ function(link_paddle_exe TARGET_NAME)
generate_rdma_links()
endif()
target_circle_link_libraries(${TARGET_NAME}
ARCHIVE_START
paddle_gserver
paddle_function
ARCHIVE_END
paddle_pserver
paddle_trainer_lib
paddle_network
paddle_math
paddle_utils
paddle_parameter
paddle_proto
paddle_cuda
paddle_optimizer
${EXTERNAL_LIBS}
${CMAKE_THREAD_LIBS_INIT}
${CMAKE_DL_LIBS}
${RDMA_LD_FLAGS}
${RDMA_LIBS})
if(MOBILE_INFERENCE)
target_circle_link_libraries(${TARGET_NAME}
ARCHIVE_START
paddle_gserver
paddle_function
ARCHIVE_END
paddle_math
paddle_utils
paddle_parameter
paddle_proto
paddle_cuda
${EXTERNAL_LIBS}
${CMAKE_THREAD_LIBS_INIT}
${CMAKE_DL_LIBS}
${RDMA_LD_FLAGS}
${RDMA_LIBS})
else()
target_circle_link_libraries(${TARGET_NAME}
ARCHIVE_START
paddle_gserver
paddle_function
ARCHIVE_END
paddle_pserver
paddle_trainer_lib
paddle_network
paddle_math
paddle_utils
paddle_parameter
paddle_proto
paddle_cuda
paddle_optimizer
${EXTERNAL_LIBS}
${CMAKE_THREAD_LIBS_INIT}
${CMAKE_DL_LIBS}
${RDMA_LD_FLAGS}
${RDMA_LIBS})
endif()
if(ANDROID)
target_link_libraries(${TARGET_NAME} log)
endif(ANDROID)
if(WITH_MKLDNN AND WITH_MKLML AND MKLDNN_IOMP_DIR)
target_link_libraries(${TARGET_NAME} "-L${MKLDNN_IOMP_DIR} -liomp5 -Wl,--as-needed")
if(WITH_MKLML AND MKLML_LIB_DIR AND MKLML_IOMP_LIB)
target_link_libraries(${TARGET_NAME} "-L${MKLML_LIB_DIR} -liomp5 -Wl,--as-needed")
endif()
add_dependencies(${TARGET_NAME} ${external_project_dependencies})
......@@ -150,17 +168,3 @@ function(create_resources res_file output_file)
COMMAND python ARGS ${PADDLE_SOURCE_DIR}/cmake/make_resource.py ${res_file} ${output_file}
DEPENDS ${res_file} ${PADDLE_SOURCE_DIR}/cmake/make_resource.py)
endfunction()
# Create a python unittest using run_python_tests.sh,
# which takes care of making correct running environment
function(add_python_test TEST_NAME)
foreach(arg ${ARGN})
get_filename_component(py_fn ${arg} NAME_WE)
set(TRG_NAME ${TEST_NAME}_${py_fn})
add_test(NAME ${TRG_NAME}
COMMAND env PYTHONPATH=${PADDLE_PYTHON_PACKAGE_DIR}
python2 ${arg}
WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR})
endforeach()
endfunction()
doc/api/index_en.rst
浏览文件 @ 988a4a66
......@@ -7,3 +7,4 @@ API
v2/model_configs.rst
v2/data.rst
v2/run_logic.rst
v2/fluid.rst
doc/api/v2/config/activation.rst
浏览文件 @ 988a4a66
......@@ -99,3 +99,10 @@ STanh
.. automodule:: paddle.v2.activation
:members: STanh
:noindex:
SoftSign
========
.. automodule:: paddle.v2.activation
:members: SoftSign
:noindex:
doc/api/v2/config/layer.rst
浏览文件 @ 988a4a66
......@@ -54,7 +54,7 @@ img_conv
.. _api_v2.layer_context_projection:
context_projection
context_projection
------------------
.. autoclass:: paddle.v2.layer.context_projection
:noindex:
......@@ -70,7 +70,7 @@ Image Pooling Layer
img_pool
--------
.. autoclass:: paddle.v2.layer.img_pool
:noindex:
:noindex:
spp
---
......@@ -82,6 +82,11 @@ maxout
.. autoclass:: paddle.v2.layer.maxout
:noindex:
roi_pool
--------
.. autoclass:: paddle.v2.layer.roi_pool
:noindex:
Norm Layer
==========
......@@ -99,7 +104,7 @@ sum_to_one_norm
---------------
.. autoclass:: paddle.v2.layer.sum_to_one_norm
:noindex:
cross_channel_norm
------------------
.. autoclass:: paddle.v2.layer.cross_channel_norm
......@@ -109,7 +114,7 @@ row_l2_norm
-----------
.. autoclass:: paddle.v2.layer.row_l2_norm
:noindex:
Recurrent Layers
================
......@@ -330,6 +335,16 @@ bilinear_interp
.. autoclass:: paddle.v2.layer.bilinear_interp
:noindex:
dot_prod
---------
.. autoclass:: paddle.v2.layer.dot_prod
:noindex:
out_prod
--------
.. autoclass:: paddle.v2.layer.out_prod
:noindex:
power
-----
.. autoclass:: paddle.v2.layer.power
......@@ -367,6 +382,11 @@ cos_sim
.. autoclass:: paddle.v2.layer.cos_sim
:noindex:
l2_distance
-----------
.. autoclass:: paddle.v2.layer.l2_distance
:noindex:
trans
-----
.. autoclass:: paddle.v2.layer.trans
......@@ -395,6 +415,13 @@ multiplex
.. autoclass:: paddle.v2.layer.multiplex
:noindex:
Factorization Machine Layer
============================
factorization_machine
---------------------
.. autoclass:: paddle.v2.layer.factorization_machine
:noindex:
Slicing and Joining Layers
==========================
......
doc/api/v2/config/networks.rst
浏览文件 @ 988a4a66
......@@ -125,3 +125,8 @@ simple_attention
:members: simple_attention
:noindex:
dot_product_attention
---------------------
.. automodule:: paddle.v2.networks
:members: dot_product_attention
:noindex:
doc/api/v2/data.rst
浏览文件 @ 988a4a66
......@@ -2,112 +2,9 @@
Data Reader Interface and DataSets
==================================
.. toctree::
:maxdepth: 1
DataTypes
=========
.. automodule:: paddle.v2.data_type
:members:
:noindex:
DataFeeder
==========
.. automodule:: paddle.v2.data_feeder
:members:
:noindex:
Reader
======
.. automodule:: paddle.v2.reader
:members:
:noindex:
.. automodule:: paddle.v2.reader.creator
:members:
:noindex:
minibatch
=========
.. automodule:: paddle.v2.minibatch
:members:
:noindex:
Dataset
=======
.. automodule:: paddle.v2.dataset
:members:
:noindex:
mnist
+++++
.. automodule:: paddle.v2.dataset.mnist
:members:
:noindex:
cifar
+++++
.. automodule:: paddle.v2.dataset.cifar
:members:
:noindex:
conll05
+++++++
.. automodule:: paddle.v2.dataset.conll05
:members: get_dict,get_embedding,test
:noindex:
imdb
++++
.. automodule:: paddle.v2.dataset.imdb
:members:
:noindex:
imikolov
++++++++
.. automodule:: paddle.v2.dataset.imikolov
:members:
:noindex:
movielens
+++++++++
.. automodule:: paddle.v2.dataset.movielens
:members:
:noindex:
.. autoclass:: paddle.v2.dataset.movielens.MovieInfo
:noindex:
.. autoclass:: paddle.v2.dataset.movielens.UserInfo
:noindex:
sentiment
+++++++++
.. automodule:: paddle.v2.dataset.sentiment
:members:
:noindex:
uci_housing
+++++++++++
.. automodule:: paddle.v2.dataset.uci_housing
:members:
:noindex:
wmt14
+++++
.. automodule:: paddle.v2.dataset.wmt14
:members:
:noindex:
data/data_reader.rst
data/image.rst
data/dataset.rst
doc/api/v2/data/data_reader.rst 0 → 100644
浏览文件 @ 988a4a66
=====================
Data Reader Interface
=====================
DataTypes
=========
.. automodule:: paddle.v2.data_type
:members:
:noindex:
DataFeeder
==========
.. automodule:: paddle.v2.data_feeder
:members:
:noindex:
Reader
======
.. automodule:: paddle.v2.reader
:members:
:noindex:
.. automodule:: paddle.v2.reader.creator
:members:
:noindex:
minibatch
=========
.. automodule:: paddle.v2.minibatch
:members:
:noindex:
doc/api/v2/data/dataset.rst 0 → 100644
浏览文件 @ 988a4a66
Dataset
=======
.. automodule:: paddle.v2.dataset
:members:
:noindex:
mnist
+++++
.. automodule:: paddle.v2.dataset.mnist
:members:
:noindex:
cifar
+++++
.. automodule:: paddle.v2.dataset.cifar
:members:
:noindex:
conll05
+++++++
.. automodule:: paddle.v2.dataset.conll05
:members: get_dict,get_embedding,test
:noindex:
imdb
++++
.. automodule:: paddle.v2.dataset.imdb
:members:
:noindex:
imikolov
++++++++
.. automodule:: paddle.v2.dataset.imikolov
:members:
:noindex:
movielens
+++++++++
.. automodule:: paddle.v2.dataset.movielens
:members:
:noindex:
.. autoclass:: paddle.v2.dataset.movielens.MovieInfo
:noindex:
.. autoclass:: paddle.v2.dataset.movielens.UserInfo
:noindex:
sentiment
+++++++++
.. automodule:: paddle.v2.dataset.sentiment
:members:
:noindex:
uci_housing
+++++++++++
.. automodule:: paddle.v2.dataset.uci_housing
:members:
:noindex:
wmt14
+++++
.. automodule:: paddle.v2.dataset.wmt14
:members:
:noindex:
doc/api/v2/data/image.rst 0 → 100644
浏览文件 @ 988a4a66
Image Interface
===============
.. automodule:: paddle.v2.image
:members:
doc/api/v2/fluid.rst 0 → 100644
浏览文件 @ 988a4a66
======================
Fluid
======================
.. toctree::
:maxdepth: 1
fluid/layers.rst
fluid/data_feeder.rst
fluid/executor.rst
fluid/initializer.rst
fluid/evaluator.rst
fluid/nets.rst
fluid/optimizer.rst
fluid/param_attr.rst
fluid/profiler.rst
fluid/regularizer.rst
doc/api/v2/fluid/data_feeder.rst 0 → 100644
浏览文件 @ 988a4a66
===========
DataFeeder
===========
DataFeeder
-----------
.. automodule:: paddle.v2.fluid.data_feeder
:members: DataFeeder
:noindex:
doc/api/v2/fluid/evaluator.rst 0 → 100644
浏览文件 @ 988a4a66
===========
Evaluator
===========
Evaluator
-----------
.. automodule:: paddle.v2.fluid.evaluator
:members: Evaluator
:noindex:
doc/api/v2/fluid/executor.rst 0 → 100644
浏览文件 @ 988a4a66
===========
Executor
===========
Executor
-----------
.. automodule:: paddle.v2.fluid.executor
:members: Executor
:noindex:
doc/api/v2/fluid/initializer.rst 0 → 100644
浏览文件 @ 988a4a66
===========
Initializer
===========
Initializer
-----------
.. automodule:: paddle.v2.fluid.initializer
:members: Initializer
:noindex:
ConstantInitializer
-------------------
.. automodule:: paddle.v2.fluid.initializer
:members: ConstantInitializer
:noindex:
UniformInitializer
------------------
.. automodule:: paddle.v2.fluid.initializer
:members: UniformInitializer
:noindex:
NormalInitializer
-----------------
.. automodule:: paddle.v2.fluid.initializer
:members: NormalInitializer
:noindex:
XavierInitializer
-----------------
.. automodule:: paddle.v2.fluid.initializer
:members: XavierInitializer
:noindex:
MSRAInitializer
---------------
.. automodule:: paddle.v2.fluid.initializer
:members: MSRAInitializer
:noindex:
doc/api/v2/fluid/layers.rst 0 → 100644
浏览文件 @ 988a4a66
==========
Layers
==========
fc
---
.. autofunction:: paddle.v2.fluid.layers.fc
:noindex:
embedding
---------
.. autofunction:: paddle.v2.fluid.layers.embedding
:noindex:
dynamic_lstm
------------
.. autofunction:: paddle.v2.fluid.layers.dynamic_lstm
:noindex:
data
---------
.. autofunction:: paddle.v2.fluid.layers.data
:noindex:
mean
---------
.. autofunction:: paddle.v2.fluid.layers.mean
:noindex:
mul
---------
.. autofunction:: paddle.v2.fluid.layers.mul
:noindex:
elementwise_add
---------------
.. autofunction:: paddle.v2.fluid.layers.elementwise_add
:noindex:
elementwise_div
---------------
.. autofunction:: paddle.v2.fluid.layers.elementwise_div
:noindex:
dropout
---------
.. autofunction:: paddle.v2.fluid.layers.dropout
:noindex:
reshape
---------
.. autofunction:: paddle.v2.fluid.layers.reshape
:noindex:
sigmoid
---------
.. autofunction:: paddle.v2.fluid.layers.sigmoid
:noindex:
scale
---------
.. autofunction:: paddle.v2.fluid.layers.scale
:noindex:
reshape
---------
.. autofunction:: paddle.v2.fluid.layers.reshape
:noindex:
transpose
---------
.. autofunction:: paddle.v2.fluid.layers.transpose
:noindex:
sigmoid_cross_entropy_with_logits
---------
.. autofunction:: paddle.v2.fluid.layers.esigmoid_cross_entropy_with_logits
:noindex:
cast
---------
.. autofunction:: paddle.v2.fluid.layers.cast
:noindex:
concat
---------
.. autofunction:: paddle.v2.fluid.layers.concat
:noindex:
sums
---------
.. autofunction:: paddle.v2.fluid.layers.sums
:noindex:
linear_chain_crf
---------
.. autofunction:: paddle.v2.fluid.layers.linear_chain_crf
:noindex:
assign
---------
.. autofunction:: paddle.v2.fluid.layers.embedding
:noindex:
split_lod_tensor
---------
.. autofunction:: paddle.v2.fluid.layers.split_lod_tensor
:noindex:
merge_lod_tensor
---------
.. autofunction:: paddle.v2.fluid.layers.merge_lod_tensor
:noindex:
cos_sim
---------
.. autofunction:: paddle.v2.fluid.layers.cos_sim
:noindex:
cross_entropy
---------
.. autofunction:: paddle.v2.fluid.layers.cross_entropy
:noindex:
square_error_cost
---------
.. autofunction:: paddle.v2.fluid.layers.square_error_cost
:noindex:
accuracy
---------
.. autofunction:: paddle.v2.fluid.layers.accuracy
:noindex:
sequence_conv
---------
.. autofunction:: paddle.v2.fluid.layers.sequence_conv
:noindex:
conv2d
---------
.. autofunction:: paddle.v2.fluid.layers.conv2d
:noindex:
sequence_pool
---------
.. autofunction:: paddle.v2.fluid.layers.sequence_pool
:noindex:
pool2d
---------
.. autofunction:: paddle.v2.fluid.layers.pool2d
:noindex:
batch_norm
---------
.. autofunction:: paddle.v2.fluid.layers.batch_norm
:noindex:
beam_search_decode
---------
.. autofunction:: paddle.v2.fluid.layers.beam_search_decode
:noindex:
lstm
---------
.. autofunction:: paddle.v2.fluid.layers.lstm
:noindex:
lod_rank_table
---------
.. autofunction:: paddle.v2.fluid.layers.lod_rank_table
:noindex:
max_sequence_len
---------
.. autofunction:: paddle.v2.fluid.layers.max_sequence_len
:noindex:
topk
---------
.. autofunction:: paddle.v2.fluid.layers.topk
:noindex:
lod_tensor_to_array
---------
.. autofunction:: paddle.v2.fluid.layers.lod_tensor_to_array
:noindex:
array_to_lod_tensor
---------
.. autofunction:: paddle.v2.fluid.layers.array_to_lod_tensor
:noindex:
fill_constant
---------
.. autofunction:: paddle.v2.fluid.layers.fill_constant
:noindex:
fill_constant_batch_size_like
---------
.. autofunction:: paddle.v2.fluid.layers.fill_constant_batch_size_like
:noindex:
ones
---------
.. autofunction:: paddle.v2.fluid.layers.ones
:noindex:
zeros
---------
.. autofunction:: paddle.v2.fluid.layers.zeros
:noindex:
increment
---------
.. autofunction:: paddle.v2.fluid.layers.increment
:noindex:
array_write
---------
.. autofunction:: paddle.v2.fluid.layers.array_write
:noindex:
create_array
---------
.. autofunction:: paddle.v2.fluid.layers.create_array
:noindex:
less_than
---------
.. autofunction:: paddle.v2.fluid.layers.less_than
:noindex:
array_read
---------
.. autofunction:: paddle.v2.fluid.layers.array_read
:noindex:
shrink_memory
---------
.. autofunction:: paddle.v2.fluid.layers.shrink_memory
:noindex:
array_length
---------
.. autofunction:: paddle.v2.fluid.layers.array_length
:noindex:
conv2d_transpose
---------
.. autofunction:: paddle.v2.fluid.layers.conv2d_transpose
:noindex:
doc/api/v2/fluid/nets.rst 0 → 100644
浏览文件 @ 988a4a66
===========
Nets
===========
simple_img_conv_pool
-----------
.. autofunction:: paddle.v2.fluid.nets.simple_img_conv_pool
:noindex:
img_conv_group
-----------
.. autofunction:: paddle.v2.fluid.nets.img_conv_group
:noindex:
sequence_conv_pool
-----------
.. autofunction:: paddle.v2.fluid.nets.sequence_conv_pool
:noindex:
doc/api/v2/fluid/optimizer.rst 0 → 100644
浏览文件 @ 988a4a66
===========
Optimizer
===========
Optimizer
-----------
.. automodule:: paddle.v2.fluid.optimizer
:members: Optimizer
:noindex:
SGDOptimizer
-----------
.. automodule:: paddle.v2.fluid.optimizer
:members: SGDOptimizer
:noindex:
MomentumOptimizer
-----------
.. automodule:: paddle.v2.fluid.optimizer
:members: MomentumOptimizer
:noindex:
AdagradOptimizer
-----------
.. automodule:: paddle.v2.fluid.optimizer
:members: AdagradOptimizer
:noindex:
AdamOptimizer
-----------
.. automodule:: paddle.v2.fluid.optimizer
:members: AdamOptimizer
:noindex:
AdamaxOptimizer
-----------
.. automodule:: paddle.v2.fluid.optimizer
:members: AdamaxOptimizer
:noindex:
DecayedAdagradOptimizer
-----------
.. automodule:: paddle.v2.fluid.optimizer
:members: DecayedAdagradOptimizer
:noindex:
doc/api/v2/fluid/param_attr.rst 0 → 100644
浏览文件 @ 988a4a66
===========
ParamAttr
===========
ParamAttr
-----------
.. automodule:: paddle.v2.fluid.param_attr
:members: ParamAttr
:noindex:
doc/api/v2/fluid/profiler.rst 0 → 100644
浏览文件 @ 988a4a66
===========
Profiler
===========
Profiler
-----------
.. autofunction:: paddle.v2.fluid.profiler.cuda_profiler
:noindex:
doc/api/v2/fluid/regularizer.rst 0 → 100644
浏览文件 @ 988a4a66
===========
Regularizer
===========
WeightDecayRegularizer
-----------
.. automodule:: paddle.v2.fluid.regularizer
:members: WeightDecayRegularizer
:noindex:
L2DecayRegularizer
-----------
.. automodule:: paddle.v2.fluid.regularizer
:members: L2DecayRegularizer
:noindex:
L1DecayRegularizer
-----------
.. automodule:: paddle.v2.fluid.regularizer
:members: L1DecayRegularizer
doc/design/block.md
浏览文件 @ 988a4a66
......@@ -5,12 +5,12 @@
Both deep learning systems and programming languages help users describe computation procedures. These systems use various representations of computation:
- Caffe, Torch, and Paddle: sequences of layers.
- TensorFlow, Caffe2, Mxnet: graphs of operators.
- TensorFlow, Caffe2, Mxnet: graph of operators.
- PaddlePaddle: nested blocks, like C++ and Java programs.
## Block in Programming Languages and Deep Learning
In programming languages, a block is a pair of curly braces that includes local variables definitions and a sequence of instructions, or operators.
In programming languages, a block is a pair of curly braces that includes local variables definitions and a sequence of instructions or operators.
Blocks work with control flow structures like `if`, `else`, and `for`, which have equivalents in deep learning:
......@@ -24,14 +24,14 @@ A key difference is that a C++ program describes a one pass computation, whereas
## Stack Frames and the Scope Hierarchy
The existence of the backward makes the execution of a block of traditional programs and PaddlePaddle different to each other:
The existence of the backward pass makes the execution of a block of PaddlePaddle different from traditional programs:
| programming languages | PaddlePaddle |
|-----------------------|-------------------------------|
| stack | scope hierarchy |
| stack frame | scope |
| push at entering block| push at entering block |
| pop at leaving block | destroy at minibatch completes|
| programming languages | PaddlePaddle |
|-----------------------|---------------------------------|
| stack | scope hierarchy |
| stack frame | scope |
| push at entering block| push at entering block |
| pop at leaving block | destroy when minibatch completes|
1. In traditional programs:
......@@ -42,9 +42,9 @@ The existence of the backward makes the execution of a block of traditional prog
1. In PaddlePaddle
- When the execution enters a block, PaddlePaddle adds a new scope, where it realizes variables.
- PaddlePaddle doesn't pop a scope after the execution of the block because variables therein are to be used by the backward pass. So it has a stack forest known as a *scope hierarchy*.
- PaddlePaddle doesn't pop a scope after the execution of the block because variables therein are used by the backward pass. So it has a stack forest known as a *scope hierarchy*.
- The height of the highest tree is the maximum depth of nested blocks.
- After the process of a minibatch, PaddlePaddle destroys the scope hierarchy.
- After the processing of a minibatch, PaddlePaddle destroys the scope hierarchy.
## Use Blocks in C++ and PaddlePaddle Programs
......@@ -55,17 +55,23 @@ Let us consolidate the discussion by presenting some examples.
The following C++ programs shows how blocks are used with the `if-else` structure:
```c++
namespace pd = paddle;
int x = 10;
int y = 20;
int out;
int y = 1;
int z = 10;
bool cond = false;
int o1, o2;
if (cond) {
int z = x + y;
out = softmax(z);
o1 = z;
o2 = pd::layer::softmax(z);
} else {
int z = fc(x);
out = z;
int d = pd::layer::fc(z);
o1 = d;
o2 = d+1;
}
```
An equivalent PaddlePaddle program from the design doc of the [IfElseOp operator](./if_else_op.md) is as follows:
......@@ -73,57 +79,55 @@ An equivalent PaddlePaddle program from the design doc of the [IfElseOp operator
```python
import paddle as pd
x = var(10)
y = var(20)
cond = var(false)
ie = pd.create_ifelseop(inputs=[x], output_num=1)
x = minibatch([10, 20, 30]) # shape=[None, 1]
y = var(1) # shape=[1], value=1
z = minibatch([10, 20, 30]) # shape=[None, 1]
cond = larger_than(x, 15) # [false, true, true]
ie = pd.ifelse()
with ie.true_block():
x = ie.inputs(true, 0)
z = operator.add(x, y)
ie.set_output(true, 0, operator.softmax(z))
d = pd.layer.add_scalar(x, y)
ie.output(d, pd.layer.softmax(d))
with ie.false_block():
x = ie.inputs(false, 0)
z = layer.fc(x)
ie.set_output(true, 0, operator.softmax(z))
out = b(cond)
d = pd.layer.fc(z)
ie.output(d, d+1)
o1, o2 = ie(cond)
```
In both examples, the left branch computes `softmax(x+y)` and the right branch computes `fc(x)`.
In both examples, the left branch computes `x+y` and `softmax(x+y)`, the right branch computes `fc(x)` and `x+1` .
The difference is that variables in the C++ program contain scalar values, whereas those in the PaddlePaddle programs are mini-batches of instances.
A difference is that variables in the C++ program contain scalar values, whereas those in the PaddlePaddle programs are mini-batches of instances. The `ie.input(true, 0)` invocation returns instances in the 0-th input, `x`, that corresponds to true values in `cond` as the local variable `x`, where `ie.input(false, 0)` returns instances corresponding to false values.
### Blocks with `for` and `RNNOp`
The following RNN model from the [RNN design doc](./rnn.md)
The following RNN model in PaddlePaddle from the [RNN design doc](./rnn.md) :
```python
x = sequence([10, 20, 30])
m = var(0)
W = tensor()
U = tensor()
rnn = create_rnn(inputs=[input])
with rnn.stepnet() as net:
x = net.set_inputs(0)
h = net.add_memory(init=m)
fc_out = pd.matmul(W, x)
hidden_out = pd.matmul(U, h.pre(n=1))
sum = pd.add_two(fc_out, hidden_out)
act = pd.sigmoid(sum)
h.update(act) # update memory with act
net.set_outputs(0, act, hidden_out) # two outputs
x = sequence([10, 20, 30]) # shape=[None, 1]
m = var(0) # shape=[1]
W = var(0.314, param=true) # shape=[1]
U = var(0.375, param=true) # shape=[1]
rnn = pd.rnn()
with rnn.step():
h = rnn.memory(init = m)
h_prev = rnn.previous_memory(h)
a = layer.fc(W, x)
b = layer.fc(U, h_prev)
s = pd.add(a, b)
act = pd.sigmoid(s)
rnn.update_memory(h, act)
rnn.output(a, b)
o1, o2 = rnn()
print o1, o2
```
has its equivalent C++ program as follows
```c++
int* x = {10, 20, 30};
int m = 0;
int W = some_value();
int U = some_other_value();
int* m = {0};
int* W = {0.314};
int* U = {0.375};
int mem[sizeof(x) / sizeof(x[0]) + 1];
int o1[sizeof(x) / sizeof(x[0]) + 1];
......@@ -131,25 +135,21 @@ int o2[sizeof(x) / sizeof(x[0]) + 1];
for (int i = 1; i <= sizeof(x)/sizeof(x[0]); ++i) {
int x = x[i-1];
if (i == 1) mem[0] = m;
int fc_out = W * x;
int hidden_out = Y * mem[i-1];
int sum = fc_out + hidden_out;
int a = W * x;
int b = Y * mem[i-1];
int s = fc_out + hidden_out;
int act = sigmoid(sum);
mem[i] = act;
o1[i] = act;
o2[i] = hidden_out;
}
print_array(o1);
print_array(o2);
```
## Compilation and Execution
Like TensorFlow programs, a PaddlePaddle program is written in Python. The first part describes a neural network as a protobuf message, and the rest part executes the message for training or inference.
Like TensorFlow, a PaddlePaddle program is written in Python. The first part describes a neural network as a protobuf message, and the rest executes the message for training or inference.
The generation of this protobuf message is like what a compiler generates a binary executable file. The execution of the message that the OS executes the binary file.
The generation of this protobuf message is similar to how a compiler generates a binary executable file. The execution of the message is similar to how the OS executes the binary file.
## The "Binary Executable File Format"
......@@ -186,10 +186,10 @@ Also, the RNN operator in above example is serialized into a protobuf message of
```
OpDesc {
inputs = {0} // the index of x
outputs = {5, 3} // indices of act and hidden_out
inputs = {0} // the index of x in vars of BlockDesc above
outputs = {5, 3} // indices of act and hidden_out in vars of BlockDesc above
attrs {
"memories" : {1} // the index of h
"states" : {1} // the index of h
"step_net" : <above step net>
}
};
......@@ -203,32 +203,32 @@ This `OpDesc` value is in the `ops` field of the `BlockDesc` value representing
During the generation of the Protobuf message, the Block should store VarDesc (the Protobuf message which describes Variable) and OpDesc (the Protobuf message which describes Operator).
VarDesc in a block should have its name scope to avoid local variables affect parent block's name scope.
Child block's name scopes should inherit the parent's so that OpDesc in child block can reference a VarDesc that stored in parent block. For example
Child block's name scopes should inherit the parent's so that OpDesc in child block can reference a VarDesc that stored in parent block. For example:
```python
a = pd.Varaible(shape=[20, 20])
a = pd.Variable(shape=[20, 20])
b = pd.fc(a, params=["fc.w", "fc.b"])
rnn = pd.create_rnn()
with rnn.stepnet() as net:
x = net.set_inputs(a)
with rnn.stepnet():
x = a.as_step_input()
# reuse fc's parameter
fc_without_b = pd.get_variable("fc.w")
net.set_outputs(fc_without_b)
rnn.output(fc_without_b)
out = rnn()
```
the method `pd.get_variable` can help retrieve a Variable by a name, a Variable may store in a parent block, but might be retrieved in a child block, so block should have a variable scope that supports inheritance.
The method `pd.get_variable` can help retrieve a Variable by the name. The Variable may be stored in a parent block, but might be retrieved in a child block, so block should have a variable scope that supports inheritance.
In compiler design, the symbol table is a data structure created and maintained by compilers to store information about the occurrence of various entities such as variable names, function names, classes, etc.
To store the definition of variables and operators, we define a C++ class `SymbolTable`, like the one used in compilers.
`SymbolTable` can do the following stuff:
`SymbolTable` can do the following:
- store the definitions (some names and attributes) of variables and operators,
- to verify if a variable was declared,
- to make it possible to implement type checking (offer Protobuf message pointers to `InferShape` handlers).
- verify if a variable was declared,
- make it possible to implement type checking (offer Protobuf message pointers to `InferShape` handlers).
```c++
......@@ -240,19 +240,18 @@ class SymbolTable {
OpDesc* NewOp(const string& name="");
// TODO determine whether name is generated by python or C++
// currently assume that a unique name will be generated by C++ if the
// argument name left default.
VarDesc* NewVar(const string& name="");
// TODO determine whether name is generated by python or C++.
// Currently assume that a unique name will be generated by C++ if the
// argument name is left default.
VarDesc* Var(const string& name="");
// find a VarDesc by name, if recursive true, find parent's SymbolTable
// find a VarDesc by name, if recursive is true, find parent's SymbolTable
// recursively.
// this interface is introduced to support InferShape, find protobuf messages
// of variables and operators, pass pointers into InferShape.
// operator
//
// NOTE maybe some C++ classes such as VarDescBuilder and OpDescBuilder should
// be proposed and embedded into pybind to enable python operate on C++ pointers.
// be proposed and embedded into pybind to enable python operation on C++ pointers.
VarDesc* FindVar(const string& name, bool recursive=true);
OpDesc* FindOp(const string& name);
......@@ -270,7 +269,7 @@ class SymbolTable {
After all the description of variables and operators is added into SymbolTable,
the block has enough information to run.
The `Block` class takes a `BlockDesc` as input, and provide `Run` and `InferShape` functions.
The `Block` class takes a `BlockDesc` as input, and provides `Run` and `InferShape` functions.
```c++
......@@ -302,7 +301,7 @@ public:
void CreateVariables(const framework::Scope& scope);
void CreateOperators();
// some other necessary interfaces of NetOp are list below
// some other necessary interfaces of NetOp are listed below
// ...
private:
......@@ -316,15 +315,14 @@ private:
Block inherits from OperatorBase, which has a Run method.
Block's Run method will run its operators sequentially.
There is another important interface called `Eval`, which take some arguments called targets, and generate a minimal graph which takes targets as the end points and creates a new Block,
after `Run`, `Eval` will get the latest value and return the targets.
There is another important interface called `Eval`, which takes some arguments called targets and generates a minimal graph which treats targets as the end points and creates a new Block. After `Run`, `Eval` will get the latest value and return the targets.
The definition of Eval is as follows:
```c++
// clean a block description by targets using the corresponding dependency graph.
// return a new BlockDesc with minimal number of operators.
// NOTE not return a Block but the block's description so that this can be distributed
// NOTE: The return type is not a Block but the block's description so that this can be distributed
// to a cluster.
BlockDesc Prune(const BlockDesc& desc, vector<string> targets);
......
doc/design/evaluator.md 0 → 100644
浏览文件 @ 988a4a66
## Evaluator Design
### Problem Statement
During training or inference, we provide an evaluation function to measure the model performance, for example, accuracy, precision, etc. In the operator based framework design, the data passes through the network pipeline batch by batch. As a result, inside the operator, we only calculate the metrics for one minibatch. Thus, we need to provide a mechanism to calculate the metrics for each N pass/batch the user wants.
### Evaluator Design
Currently, every operation is expressed in the graph. We divide the evaluator process into three steps.
1. Initialize the metric state and add it into the block.
2. Calculate the concerned metrics for every mini-batch. The single evaluator operator is only responsible for calculating the necessary statistics for one mini-batch. For example, the accuracy operator only calculates the accuracy for a minibatch data if run once.
3. Merge the mini-batch statistics to form the evaluation result for multiple mini-batches. When it comes to distributed training/Multi-GPU training, aggregate the value from different devices.
### Implementation
This design is shown in the Python API.
Each metric operator needs to caculate the metric statistic and return the batch-aware states. Python side is responsible for accumulating the states for each pass.
```python
class Evaluator(object):
"""
Evaluator Base class.
"""
def __init__(self, name, **kwargs):
"""
Different evaluator may has different metric states. E.g, Accuracy need two variables, total and right sample counts.
Auc need four variables, `true_positives`,
`true_negatives`, `false_positives` and `false_negatives`. So every evaluator should create its needed variables and append to main_program
The initialization of Evaluator should be responsible for:
create metric states and append to the main_program
"""
pass
def _update_ops(self, input, label, **kwargs)
"""
Add mini-batch evaluator caculate operators to the main_program.
Add increment operator to accumulate the metric states.
"""
def reset(self, executor, reset_program=None):
"""
Reset metric states at the begin of each pass/user specified batch number.
Execute the reset_program to reset the states.
"""
def eval(self, executor, eval_program=None):
"""
Merge the mini-batch statistics to form the evaluation result for multiple mini-batches.
Execute the eval_program and return the result.
"""
return eval_result
```
doc/design/executor.md 0 → 100644
浏览文件 @ 988a4a66
# Executor Design Doc
## Motivation
We use executor to do the runtime evaluation of a `ProgramDesc`.
## Overview
An executor takes a `ProgramDesc`, a `block_id` and a `Scope`. The `ProgramDesc` is a list of blocks and each block contains the protobuf definition of all the parameters and operators. The `block_id` specifies the entrance block. And the `Scope` is the container of all the variable instance, which is persistent throughout different runs.
### What does executor do?
It evaluates all the operators in the `block_id`th block of a `ProgramDesc`.
### What does executor NOT do?
It does not do runtime optimization, meaning intelligently parse the dependency of each op a choose which one to be run and in which order they should be run.
It does not do graph partitioning, meaning dividing the `ProgramDesc` into several small pieces and executing them on different devices.
## Implementation
`Executor` evaluates a `ProgramDesc`. Essentially, it instantiates Variables and Operators, then run all the operators in sequence. [[code]](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/executor.cc)
doc/design/float16.md 0 → 100644
浏览文件 @ 988a4a66
# Design Doc: float16
## Why float16
Half precision (float16) is a binary floating-point format that occupies 16 bits in memory. float16 is half the size of traditional 32-bit single precision format (float) and has lower precision and smaller range.
When high precision computation is not required, using float16 data type could potentially
- reduce storage space, memory bandwidth, and power usages;
- increase the chance of data fitting into a smaller cache of lower latency;
- provide arithmetic speed up if supported by hardware.
## Survey of current float16 support
A brief survey of float16 support on different compilers, hardwares, and libraries can be found below. Interested readers can refer to [link1](https://github.com/PaddlePaddle/Paddle/issues/4853) and [link2](https://github.com/Xreki/Xreki.github.io/blob/master/multi_data_types_in_dl_framework/ppt/float16_and_quantized_type.md) for more info.
The goal of float16 is to serve as a key for the executor to find and run the correct version of compute method specialized for float16 in operator kernel. It should be compatible with various natively supported float16 implementations including `__half` for cuda, `float16_t` for ARM, and `Eigen::half` for Eigen to make writing customized float16 kernels easier.
### Compiler
- nvcc supports `__half` data type after CUDA 7.5.
- `__fp16` or `float16_t` is supported as storage type for gcc >= 6.1 and clang >= 3.4.
- `__fp16` or `float16_t` is supported as arithmetic type for gcc >= 7.1 and clang >= 3.9.
### Hardware
- `__half` is supported on GPU with compute capability >= 5.3.
- `__fp16` is supported as storage type for ARMv7-A, ARMv8-A, and above.
- `__fp16` is supported as arithmetic type after ARMv8.2-A (currently, the only microarchitecture implementing ARMv8.2-A is ARM Cortex-A75, which is announced in May 2017. There seems to be no application processors currently available on market that adopts this architecture. It is reported that Qualcomm Snapdragon 845 uses Cortex-A75 design and will be available in mobile devices in early 2018).
### Libraries
- [Eigen](https://github.com/RLovelett/eigen) >= 3.3 supports float16 calculation on both GPU and CPU using the `Eigen::half` class. It is mostly useful for Nvidia GPUs because of the overloaded arithmetic operators using cuda intrinsics. It falls back to using software emulation on CPU for calculation and there is no special treatment to ARM processors.
- [ARM compute library](https://github.com/ARM-software/ComputeLibrary) >= 17.02.01 supports NEON FP16 kernels (requires ARMv8.2-A CPU).
### CUDA version issue
There are currently three versions of CUDA that supports `__half` data type, namely, CUDA 7.5, 8.0, and 9.0.
CUDA 7.5 and 8.0 define `__half` as a simple struct that has a `uint16_t` data (see [`cuda_fp16.h`](https://github.com/ptillet/isaac/blob/9212ab5a3ddbe48f30ef373f9c1fb546804c7a8c/include/isaac/external/CUDA/cuda_fp16.h)) as follows:
```
typedef struct __align__(2) {
unsigned short x;
} __half;
typedef __half half;
```
This struct does not define any overloaded arithmetic operators. So you have to directly use `__hadd` instead of `+` to correctly add two half types:
```
__global__ void Add() {
half a, b, c;
c = __hadd(a, b); // correct
c = a + b; // compiler error: no operator "+" matches these operands
}
```
CUDA 9.0 provides a major update to the half data type. The related code can be found in the updated [`cuda_fp16.h`](https://github.com/ptillet/isaac/blob/master/include/isaac/external/CUDA/cuda_fp16.h) and the newly added [`cuda_fp16.hpp`](https://github.com/ptillet/isaac/blob/master/include/isaac/external/CUDA/cuda_fp16.hpp).
Essentially, CUDA 9.0 renames the original `__half` type in 7.5 and 8.0 as `__half_raw`, and defines a new `__half` class type that has constructors, conversion operators, and also provides overloaded arithmetic operators such as follows:
```
typedef struct __CUDA_ALIGN__(2) {
unsigned short x;
} __half_raw;
struct __CUDA_ALIGN__(2) __half {
protected:
unsigned short __x;
public:
// constructors and conversion operators from/to
// __half_raw and other built-in data types
}
typedef __half half;
__device__ __forceinline__
__half operator+(const __half &lh, const __half &rh) {
return __hadd(lh, rh);
}
// Other overloaded operators
```
This new design makes `c = a + b` work correctly for CUDA half data type.
## Implementation
The float16 class holds a 16-bit `uint16_t` data internally.
```
struct float16 {
uint16_t x;
};
```
float16 supports the following features:
- constructors / assignment operators that take input from primitive data types including bool, integers of various length, float, and double.
- constructors / assignment operators that take input from `__half` on cuda, `float16_t` on ARM, and `Eigen::half` on Eigen.
- conversion operators to primitive data types and half precision data types on cuda, ARM and Eigen.
- overloaded arithmetic operators for cuda, arm, and non-arm cpu, respectively. These operators will take advantage of the cuda and ARM intrinsics on the corresponding hardware.
To support the above features, two fundamental conversion functions are provided:
```
float16 float_to_half_rn(float f); // convert to half precision in round-to-nearest-even mode
float half_to_float(float16 h);
```
which provides one-to-one conversion between float32 and float16. These twos functions will do different conversion routines based on the current hardware. CUDA/ARM instrinsics will be used when the corresonding hardware is available. If the hardware or compiler level does not support float32 to float16 conversion, software emulation will be performed to do the conversion.
## To do
After float16 class is available, some of the future items are below:
- Update pybind/tensor_py.h to bind c++ float16 with numpy float16.
- Modify `GetKernelType()` method in `framework/operator.h` to make it compatible with float16.
- Create a type-casting operator that can convert the data type in tensor between float16 and other types.
doc/design/gan_api.md 0 → 100644
浏览文件 @ 988a4a66
# Design for GAN
GAN (General Adversarial Net [https://arxiv.org/abs/1406.2661]) is an important model for unsupervised learning and widely used in many areas.
It applies several important concepts in machine learning system design, including building and running subgraphs, dependency tracing, different optimizers in one executor and so forth.
In our GAN design, we wrap it as a user-friendly easily customized python API to design different models. We take the conditional DC-GAN (Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks [https://arxiv.org/abs/1511.06434]) as an example due to its good performance on image generation.
<p align="center">
<img src="./test.dot.png" width = "35%" align="center"/><br/>
Figure 1. The overall running logic of GAN. The black solid arrows indicate the forward pass; the green dashed arrows indicate the backward pass of generator training; the red dashed arrows indicate the backward pass of the discriminator training. The BP pass of the green (red) arrow should only update the parameters in the green (red) boxes. The diamonds indicate the data providers. d\_loss and g\_loss marked in red and green are the two targets we would like to run.
</p>
The operators, layers and functions required/optional to build a GAN demo is summarized in https://github.com/PaddlePaddle/Paddle/issues/4563.
<p align="center">
<img src="./dcgan.png" width = "90%" align="center"/><br/>
Figure 2. Photo borrowed from the original DC-GAN paper.
</p>
## The Conditional-GAN might be a class.
This design we adopt the popular open source design in https://github.com/carpedm20/DCGAN-tensorflow and https://github.com/rajathkmp/DCGAN. It contains following data structure:
- DCGAN(object): which contains everything required to build a GAN model. It provides following member functions methods as API:
- __init__(...): Initialize hyper-parameters (like conv dimension and so forth), and declare model parameters of discriminator and generator as well.
- generator(z, y=None): Generate a fake image from input noise z. If the label y is provided, the conditional GAN model will be chosen.
Returns a generated image.
- discriminator(image):
Given an image, decide if it is from a real source or a fake one.
Returns a 0/1 binary label.
- build_model(self):
build the whole GAN model, define training loss for both generator and discrimator.
## Discussion on Engine Functions required to build GAN
- Trace the tensor and variable dependency in the engine executor. (Very critical, otherwise GAN can'be be trained correctly)
- Different optimizers responsible for optimizing different loss.
To be more detailed, we introduce our design of DCGAN as following:
### Class member Function: Initializer
- Set up hyper-parameters, including condtional dimension, noise dimension, batch size and so forth.
- Declare and define all the model variables. All the discriminator parameters are included in the list self.theta_D and all the generator parameters are included in the list self.theta_G.
```python
class DCGAN(object):
def __init__(self, y_dim=None):
# hyper parameters
self.y_dim = y_dim # conditional gan or not
self.batch_size = 100
self.z_dim = z_dim # input noise dimension
# define parameters of discriminators
self.D_W0 = pd.Variable(shape=[3,3, 1, 128], data=pd.gaussian_normal_randomizer())
self.D_b0 = pd.Variable(np.zeros(128)) # variable also support initialization using a numpy data
self.D_W1 = pd.Variable(shape=[784, 128], data=pd.gaussian_normal_randomizer())
self.D_b1 = pd.Variable(np.zeros(128)) # variable also support initialization using a numpy data
self.D_W2 = pd.Varialble(np.random.rand(128, 1))
self.D_b2 = pd.Variable(np.zeros(128))
self.theta_D = [self.D_W0, self.D_b0, self.D_W1, self.D_b1, self.D_W2, self.D_b2]
# define parameters of generators
self.G_W0 = pd.Variable(shape=[784, 128], data=pd.gaussian_normal_randomizer())
self.G_b0 = pd.Variable(np.zeros(128)) # variable also support initialization using a numpy data
self.G_W1 = pd.Variable(shape=[784, 128], data=pd.gaussian_normal_randomizer())
self.G_b1 = pd.Variable(np.zeros(128)) # variable also support initialization using a numpy data
self.G_W2 = pd.Varialble(np.random.rand(128, 1))
self.G_b2 = pd.Variable(np.zeros(128))
self.theta_G = [self.G_W0, self.G_b0, self.G_W1, self.G_b1, self.G_W2, self.G_b2]
```
### Class member Function: Generator
- Given a noisy input z, returns a fake image.
- Concatenation, batch-norm, FC operations required;
- Deconv layer required, which is missing now...
```python
class DCGAN(object):
def generator(self, z, y = None):
# input z: the random noise
# input y: input data label (optional)
# output G_im: generated fake images
if not self.y_dim:
z = pd.layer.concat(1, [z, y])
G_h0 = pd.layer.fc(z, self.G_w0, self.G_b0)
G_h0_bn = pd.layer.batch_norm(G_h0)
G_h0_relu = pd.layer.relu(G_h0_bn)
G_h1 = pd.layer.deconv(G_h0_relu, self.G_w1, self.G_b1)
G_h1_bn = pd.layer.batch_norm(G_h1)
G_h1_relu = pd.layer.relu(G_h1_bn)
G_h2 = pd.layer.deconv(G_h1_relu, self.G_W2, self.G_b2))
G_im = pd.layer.tanh(G_im)
return G_im
```
### Class member function: Discriminator
- Given a noisy input z, returns a fake image.
- Concatenation, Convolution, batch-norm, FC, Leaky-ReLU operations required;
```python
class DCGAN(object):
def discriminator(self, image):
# input image: either generated images or real ones
# output D_h2: binary logit of the label
D_h0 = pd.layer.conv2d(image, w=self.D_w0, b=self.D_b0)
D_h0_bn = pd.layer.batchnorm(h0)
D_h0_relu = pd.layer.lrelu(h0_bn)
D_h1 = pd.layer.conv2d(D_h0_relu, w=self.D_w1, b=self.D_b1)
D_h1_bn = pd.layer.batchnorm(D_h1)
D_h1_relu = pd.layer.lrelu(D_h1_bn)
D_h2 = pd.layer.fc(D_h1_relu, w=self.D_w2, b=self.D_b2)
return D_h2
```
### Class member function: Build the model
- Define data readers as placeholders to hold the data;
- Build generator and discriminators;
- Define two training losses for discriminator and generator, respectively.
If we have execution dependency engine to back-trace all tensors, the module building our GAN model will be like this:
```python
class DCGAN(object):
def build_model(self):
if self.y_dim:
self.y = pd.data(pd.float32, [self.batch_size, self.y_dim])
self.images = pd.data(pd.float32, [self.batch_size, self.im_size, self.im_size])
self.faked_images = pd.data(pd.float32, [self.batch_size, self.im_size, self.im_size])
self.z = pd.data(tf.float32, [None, self.z_size])
# step 1: generate images by generator, classify real/fake images with discriminator
if self.y_dim: # if conditional GAN, includes label
self.G = self.generator(self.z, self.y)
self.D_t = self.discriminator(self.images)
# generated fake images
self.sampled = self.sampler(self.z, self.y)
self.D_f = self.discriminator(self.G)
else: # original version of GAN
self.G = self.generator(self.z)
self.D_t = self.discriminator(self.images)
# generate fake images
self.sampled = self.sampler(self.z)
self.D_f = self.discriminator(self.images)
# step 2: define the two losses
self.d_loss_real = pd.reduce_mean(pd.cross_entropy(self.D_t, np.ones(self.batch_size))
self.d_loss_fake = pd.reduce_mean(pd.cross_entropy(self.D_f, np.zeros(self.batch_size))
self.d_loss = self.d_loss_real + self.d_loss_fake
self.g_loss = pd.reduce_mean(pd.cross_entropy(self.D_f, np.ones(self.batch_szie))
```
If we do not have dependency engine but blocks, the module building our GAN model will be like this:
```python
class DCGAN(object):
def build_model(self, default_block):
# input data in the default block
if self.y_dim:
self.y = pd.data(pd.float32, [self.batch_size, self.y_dim])
self.images = pd.data(pd.float32, [self.batch_size, self.im_size, self.im_size])
# self.faked_images = pd.data(pd.float32, [self.batch_size, self.im_size, self.im_size])
self.z = pd.data(tf.float32, [None, self.z_size])
# step 1: generate images by generator, classify real/fake images with discriminator
with pd.default_block().g_block():
if self.y_dim: # if conditional GAN, includes label
self.G = self.generator(self.z, self.y)
self.D_g = self.discriminator(self.G, self.y)
else: # original version of GAN
self.G = self.generator(self.z)
self.D_g = self.discriminator(self.G, self.y)
self.g_loss = pd.reduce_mean(pd.cross_entropy(self.D_g, np.ones(self.batch_szie))
with pd.default_block().d_block():
if self.y_dim: # if conditional GAN, includes label
self.D_t = self.discriminator(self.images, self.y)
self.D_f = self.discriminator(self.G, self.y)
else: # original version of GAN
self.D_t = self.discriminator(self.images)
self.D_f = self.discriminator(self.G)
# step 2: define the two losses
self.d_loss_real = pd.reduce_mean(pd.cross_entropy(self.D_t, np.ones(self.batch_size))
self.d_loss_fake = pd.reduce_mean(pd.cross_entropy(self.D_f, np.zeros(self.batch_size))
self.d_loss = self.d_loss_real + self.d_loss_fake
```
Some small confusion and problems with this design:
- D\_g and D\_f are actually the same thing, but has to be written twice; i.e., if we want to run two sub-graphs conceptually, the same codes have to be written twice if they are shared by the graph.
- Requires ability to create a block anytime, rather than in if-else or rnn only;
## Main function for the demo:
Generally, the user of GAN just need to the following things:
- Define an object as DCGAN class;
- Build the DCGAN model;
- Specify two optimizers for two different losses with respect to different parameters.
```python
# pd for short, should be more concise.
from paddle.v2 as pd
import numpy as np
import logging
if __name__ == "__main__":
# dcgan class in the default graph/block
# if we use dependency engine as tensorflow
# the codes, will be slightly different like:
# dcgan = DCGAN()
# dcgan.build_model()
with pd.block() as def_block:
dcgan = DCGAN()
dcgan.build_model(def_block)
# load mnist data
data_X, data_y = self.load_mnist()
# Two subgraphs required!!!
with pd.block().d_block():
d_optim = pd.train.Adam(lr = .001, beta= .1)
d_step = d_optim.minimize(dcgan.d_loss, dcgan.theta_D)
with pd.block.g_block():
g_optim = pd.train.Adam(lr = .001, beta= .1)
g_step = pd.minimize(dcgan.g_loss, dcgan.theta_G)
# executor
sess = pd.executor()
# training
for epoch in xrange(10000):
for batch_id in range(N / batch_size):
idx = ...
# sample a batch
batch_im, batch_label = data_X[idx:idx+batch_size], data_y[idx:idx+batch_size]
# sample z
batch_z = np.random.uniform(-1., 1., [batch_size, z_dim])
if batch_id % 2 == 0:
sess.run(d_step,
feed_dict = {dcgan.images: batch_im,
dcgan.y: batch_label,
dcgan.z: batch_z})
else:
sess.run(g_step,
feed_dict = {dcgan.z: batch_z})
```
# More thinking about dependency engine v.s. block design:
- What if we just want to run an intermediate result? Do we need to run the whole block/graph?
- Should we call eval() to get the fake images in the first stage? And then train the discriminator in the second stage?
doc/design/graph_survey.md 0 → 100644
浏览文件 @ 988a4a66
## Survey on Graph
Neural network framework often provides symbolic API for users to write network topology conveniently. This doc manily focus on symbolic API in most popular neural network frameworks, and try to find out how to parse symbolic configuration to a portable file, such as protobuf or json.
### Mxnet
The core concept of symbolic API is `Symbol`. Mxnet implements `Symbol` class in C++, and export to Python using C-API. Please refer to the comments in Mxnet:
`Symbol` is help class used to represent the operator node in Graph.
`Symbol` acts as an interface for building graphs from different components like Variable, Functor and Group. `Symbol` is also exported to python front-end (while Graph is not) to enable quick test and deployment. Conceptually, symbol is the final operation of a graph and thus including all the information required (the graph) to evaluate its output value.
A simple network topology wrote by Symbol is as follows:
```python
def get_symbol(num_classes=10, **kwargs):
data = mx.symbol.Variable('data')
data = mx.symbol.Flatten(data=data)
fc1 = mx.symbol.FullyConnected(data = data, name='fc1', num_hidden=128)
act1 = mx.symbol.Activation(data = fc1, name='relu1', act_type="relu")
fc2 = mx.symbol.FullyConnected(data = act1, name = 'fc2', num_hidden = 64)
act2 = mx.symbol.Activation(data = fc2, name='relu2', act_type="relu")
fc3 = mx.symbol.FullyConnected(data = act2, name='fc3', num_hidden=num_classes)
mlp = mx.symbol.SoftmaxOutput(data = fc3, name = 'softmax')
return mlp
```
Varible here is actually a Symbol. Every basic Symbol will correspond to one Node, and every Node has its own NodeAttr. There is a op field in NodeAttr class, when a Symbol represents Variable(often input data), the op field is null.
Symbol contains a data member, std::vector<NodeEntry> outputs, and NodeEntry cantains a poniter to Node. We can follow the Node pointer to get all the Graph.
And Symbol can be saved to a Json file.
Here is a detailed example:
```
>>> import mxnet as mx
>>> data = mx.symbol.Variable('data')
>>> print data.debug_str()
Variable:data
>>> data = mx.symbol.Flatten(data=data)
>>> print data.debug_str()
Symbol Outputs:
output[0]=flatten0(0)
Variable:data
--------------------
Op:Flatten, Name=flatten0
Inputs:
arg[0]=data(0) version=0
>>> fc1 = mx.symbol.FullyConnected(data = data, name='fc1', num_hidden=128)
>>> print fc1.debug_str()
Symbol Outputs:
output[0]=fc1(0)
Variable:data
--------------------
Op:Flatten, Name=flatten0
Inputs:
arg[0]=data(0) version=0
Variable:fc1_weight
Variable:fc1_bias
--------------------
Op:FullyConnected, Name=fc1
Inputs:
arg[0]=flatten0(0)
arg[1]=fc1_weight(0) version=0
arg[2]=fc1_bias(0) version=0
Attrs:
num_hidden=128
```
### TensorFlow
The core concept of symbolic API is `Tensor`. Tensorflow defines `Tensor` in Python. Please refer to the comments in TensorFlow:
A `Tensor` is a symbolic handle to one of the outputs of an `Operation`. It does not hold the values of that operation's output, but instead provides a means of computing those values in a TensorFlow [Session](https://www.tensorflow.org/api_docs/python/tf/Session).
A simple example is as follows:
```python
# Build a dataflow graph.
c = tf.constant([[1.0, 2.0], [3.0, 4.0]])
d = tf.constant([[1.0, 1.0], [0.0, 1.0]])
e = tf.matmul(c, d)
# Construct a `Session` to execute the graph.
sess = tf.Session()
# Execute the graph and store the value that `e` represents in `result`.
result = sess.run(e)
```
The main method of `Tensor` is as follows:
```python
@property
def op(self):
"""The `Operation` that produces this tensor as an output."""
return self._op
@property
def dtype(self):
"""The `DType` of elements in this tensor."""
return self._dtype
@property
def graph(self):
"""The `Graph` that contains this tensor."""
return self._op.graph
@property
def name(self):
"""The string name of this tensor."""
if not self._op.name:
raise ValueError("Operation was not named: %s" % self._op)
return "%s:%d" % (self._op.name, self._value_index)
@property
def device(self):
"""The name of the device on which this tensor will be produced, or None."""
return self._op.device
```
Tensor can be taken as target to run by session. Tensor contains all the information of Graph, and tracks data dependency.
Here is a detailed example:
```
>>> import tensorflow as tf
>>> c = tf.constant([[1.0, 2.0], [3.0, 4.0]])
>>> print c.graph
<tensorflow.python.framework.ops.Graph object at 0x10f256d50>
>>> d = tf.constant([[1.0, 1.0], [0.0, 1.0]])
>>> print d.graph
<tensorflow.python.framework.ops.Graph object at 0x10f256d50>
>>> e = tf.matmul(c, d)
>>> print e.graph
<tensorflow.python.framework.ops.Graph object at 0x10f256d50>
```
### Dynet
The core concept of symbolic API is `Expression`, and Dynet defines `Expression` class in C++.
A simple example is as follows:
```cpp
ComputationGraph cg;
Expression W = parameter(cg, pW);
Expression in = input(cg, xs[i]);
Expression label = input(cg, ys[i]);
Expression pred = W * in;
Expression loss = square(pred - label);
```
The input data and parameter are also represented by Expression. Every basci Expression corresponds to a Node. And input data is also a Node.
Expression has a data member ComputationGraph, and ComputationGraph will be modified in users' configuring process. Expression can be a running target, beacuse Expression contains all dependency.
Here is a detailed example:
write topology in C++
```
ComputationGraph cg;
Expression W = parameter(cg, pW);
cg.print_graphviz();
Expression pred = W * xs[i];
cg.print_graphviz();
Expression loss = square(pred - ys[i]);
cg.print_graphviz();
```
compile and print
```
# first print
digraph G {
rankdir=LR;
nodesep=.05;
N0 [label="v0 = parameters({1}) @ 0x7ffe4de00110"];
}
# second print
digraph G {
rankdir=LR;
nodesep=.05;
N0 [label="v0 = parameters({1}) @ 0x7ffe4de00110"];
N1 [label="v1 = v0 * -0.98"];
N0 -> N1;
}
# third print
digraph G {
rankdir=LR;
nodesep=.05;
N0 [label="v0 = parameters({1}) @ 0x7ffe4de00110"];
N1 [label="v1 = v0 * -0.98"];
N0 -> N1;
N2 [label="v2 = -1.88387 - v1"];
N1 -> N2;
N3 [label="v3 = -v2"];
N2 -> N3;
N4 [label="v4 = square(v3)"];
N3 -> N4;
}
```
### Conclusion
Actually, Symbol/Tensor/Expression in Mxnet/TensorFlow/Dynet are the same level concepts. We use a unified name Expression here, this level concept has following features:
- Users wirte topoloy with symbolic API, and all return value is Expression, including input data and parameter.
- Expression corresponds with a global Graph, and Expression can also be composed.
- Expression tracks all dependency and can be taken as a run target
doc/design/if_else_op.md
浏览文件 @ 988a4a66
IfOp should have only one branch. An IfOp operator takes a `cond` variable whose value must be a vector of N boolean elements. Its return value has N instances. If cond[i] == True, input instance input[i] will go through true_block() and generate output[i]; otherwise it will produce output from false_bloack().
# The `IfElse` Operator
```python
import paddle as pd
PaddlePaddle's `IfElse` operator differs from TensorFlow's:
x = var()
y = var()
cond = var()
default_value = var()
b = pd.create_ifelseop(inputs=[x], output_num=1)
with b.true_block():
x = b.inputs(0)
z = operator.add(x, y)
b.set_output(0, operator.softmax(z))
with b.false_block():
x = b.inputs(0)
z = layer.fc(x)
b.set_output(0, operator.softmax(z))
out = b(cond)
```
- the TensorFlow version takes a scalar boolean value as the condition so that the whole mini-batch goes to either the true or the false branch, whereas
- the PaddlePaddle version takes a vector of boolean value as the condition, and instances corresponding to true values go to the true branch, those corresponding to false values go to the false branch.
## Example
The following PaddlePaddle program shows the usage of the IfElse operator:
If only true_block is set in an IfElseOp, a special case is that we can have a default value for false as:
```python
import paddle as pd
x = var()
y = var()
cond = var()
default_value = var()
b = pd.create_ifelseop(inputs=[x], output_num=1, default_value)
with b.true_block():
x = b.inputs(0)
z = operator.add(x, y)
b.set_output(0, operator.softmax(z))
x = minibatch([10, 20, 30]) # shape=[None, 1]
y = var(1) # shape=[1], value=1
z = minibatch([10, 20, 30]) # shape=[None, 1]
cond = larger_than(x, 15) # [false, true, true]
ie = pd.ifelse()
with ie.true_block():
d = pd.layer.add(x, y)
ie.output(d, pd.layer.softmax(d))
with ie.false_block():
d = pd.layer.fc(z)
ie.output(d, d+1)
o1, o2 = ie(cond)
```
out = b(cond)
A challenge to implement the `IfElse` operator is to infer those variables to be split, or, say, to identify the variable of the mini-batch or those derived from the mini-batch.
An equivalent C++ program is as follows:
```c++
namespace pd = paddle;
int x = 10;
int y = 1;
int z = 10;
bool cond = false;
int o1, o2;
if (cond) {
int d = x + y;
o1 = z;
o2 = pd::layer::softmax(z);
} else {
int d = pd::layer::fc(z);
o1 = d;
o2 = d+1;
}
```
where default_value is a list of vars for `cond` == False.
doc/design/images/graph_construction_example.dot
浏览文件 @ 988a4a66
......@@ -33,7 +33,6 @@ digraph ImageClassificationGraph {
cost -> MSE_Grad [color=red];
d_cost -> MSE_Grad [color=red];
x -> MSE_Grad [color=red];
l -> MSE_Grad [color=red];
y -> MSE_Grad -> d_y [color=red];
......
doc/design/infer_var_type.md 0 → 100644
浏览文件 @ 988a4a66
# Design Doc: InferVarType
## The Problem Posed
The variable in our design can hold variant types. Such as `LoDTensor` and `SelectedRows`. An operator should be able to inference the variable types of its output.
For example, a `lookup table` operator takes two `LoDTensor`; one is a float tensor as the embedding table, the other is an int tensor as word ID. The gradient operator of `lookup table` will generate a `SelectedRows` as its output. A `sum` operator can take both `LoDTensor` and `SelectedRows` as its inputs and will generate a `LoDTensor` if any of its inputs is `LoDTensor`, otherwise, the `sum` operator will generate `SelectedRows` as its output.
The variable type will be constant at runtime. Every variable's type can either be set by the user (input data and parameter) or be inferred by the operator in compile time.
## Proposed Solution
The `InferVarType` is a compile-time function which is registered to each operator. The inferface of that function is:
```c++
using InferVarTypeFN = std::function<
void (const OpDescBind& /*op_desc*/, BlockDescBind* /*block*/)>;
```
It takes an operator description as its input and will write the output variable type and store them in block description.
The `InferVarTypeFN` will be registered in `OpInfo`, to replace `infer_var_type_` field. The `OpInfo` should be
```cpp
struct OpInfo {
InferVarTypeFN infer_var_type_;
...
};
```
The default `InferVarType` will set output type as `LoDTensor`. It can be done by `GetInferVarType()`.
```cpp
void DefaultInferVarType(const OpDescBind& op_desc, BlockDescBind* block) {
// set the output type of variable as `LoDTensor`.
// ...
}
struct OpInfo {
InferVarTypeFN infer_var_type_;
InferVarTypeFN GetInferVarType() const {
if (infer_var_type_) {
return infer_var_type_;
} else {
return DefaultInferVarType;
}
}
};
```
## Register InferVarType
We provide a thin base class for registering an `InferVarTypeFN`. To use a base class will ease the implementation of registry since we can detect the registry entry is an `InferVarTypeFN` or not.
```cpp
class VarTypeInferer {
public:
virtual void operator()(const OpDescBind& op_desc, BlockDescBind* block) const = 0;
}
```
Operator developers can write the specialize `VarTypeInferer` as follow.
```cpp
class SpecialVarTypeInferer : public VarTypeInferer {
public:
virtual void operator()(const OpDescBind& op_desc, BlockDescBind* block) const {
// .. own logic
}
}
```
Then user can register the `InferVarType` just like `GradOpDescMaker` and `OpInfoMaker`.
```
REGISTER_OPERATOR(some_op, OpType, SpecialVarTypeInferer, ...);
```
doc/design/mkldnn/README.MD
浏览文件 @ 988a4a66
# Intel® MKL-DNN on PaddlePaddle: Design Doc
我们计划将Intel深度神经网络数学库(**MKL-DNN**\[[1](#references)\])集成到PaddlePaddle，充分展现英特尔平台的优势，有效提升PaddlePaddle在英特尔架构上的性能。
我们计划将英特尔深度神经网络数学库[Intel MKL-DNN](https://github.com/01org/mkl-dnn)
(Intel Math Kernel Library for Deep Neural Networks)集成到PaddlePaddle，
充分展现英特尔平台的优势，有效提升PaddlePaddle在英特尔架构上的性能。
我们短期内的基本目标是：
<div align="center">
<img src="image/overview.png"><br/>
Figure 1. PaddlePaddle on IA
</div>
- 完成常用layer的MKL-DNN实现。
近期目标
- 完成常用Layer的MKL-DNN实现。
- 完成常见深度神经网络VGG，GoogLeNet 和 ResNet的MKL-DNN实现。
目前的优化，主要针对PaddlePaddle在重构之前的代码框架以及V1的API。
具体的完成状态可以参见[这里](https://github.com/PaddlePaddle/Paddle/projects/21)。
## Contents
- [Overview](#overview)
- [Actions](#actions)
- [CMake](#cmake)
- [Matrix](#matrix)
- [Layers](#layers)
- [Activations](#activations)
- [Parameters](#parameters)
- [Gradients](#gradients)
- [Unit Tests](#unit-tests)
- [Protobuf Messages](#protobuf-messages)
- [Python API](#python-api)
- [Demos](#demos)
- [Benchmarking](#benchmarking)
- [Others](#others)
- [Design Concerns](#design-concerns)
## Overview
我们会把MKL-DNN作为第三方库集成进PaddlePaddle，整体框架图
我们会把MKL-DNN会作为第三方库集成进PaddlePaddle，与其他第三方库一样，会在编译PaddlePaddle的时候下载并编译MKL-DNN。
同时，为了进一步提升PaddlePaddle在基本数学运算的计算速度，我们也将MKLML即(MKL small library\[[1](#references)\])
作为另一个第三方库集成进PaddlePaddle，它只会包括生成好的动态库和头文件。
MKL，MKLML以及MKL-DNN三者关系如下表：
| Name | Open Source | License | Descriptions |
| :---------- | :--------------- | :---------- | :------------ |
| MKL | No | Proprietary | Accelerate math processing routines |
| MKLML | No | Proprietary | Small package of MKL, especially for Machine Learning |
| MKL-DNN | Yes | Apache 2.0 | Accelerate primitives processing routines especially for Deep Neural Networks |
MKLML可以与MKL-DNN共同使用，以此达到最好的性能。
<div align="center">
<img src="image/overview.png" width=350><br/>
Figure 1. PaddlePaddle on IA.
<img src="image/engine.png"><br/>
Figure 2. PaddlePaddle with MKL Engines
</div>
## Actions
我们把集成方案大致分为了如下几个方面。
添加的相关文件和目录结构如下：
```txt
PaddlePaddle/Paddle
├── ...
├── cmake/
│ ├── external/
│ │ ├── ...
│ │ ├── mkldnn.cmake
│ │ └── mklml.cmake
└── paddle/
├── ...
├── math/
│ ├── ...
│ └── MKLDNNMatrix.*
└── gserver/
├── ...
├── layers/
│ ├── ...
│ └── MKLDNN*Layer.*
├── activations/
│ ├── ...
│ └── MKLDNNActivations.*
└── tests/
├── ...
├── MKLDNNTester.*
└── test_MKLDNN.cpp
```
### CMake
我们会在`CMakeLists.txt`中会添加`WITH_MKLDNN`的选项，当设置这个值为`ON`的时候会启用编译MKL-DNN功能。同时会自动开启OpenMP用于提高MKL-DNN的性能。
在`CMakeLists.txt`中提供一个与MKL有关的总开关：`WITH_MKL`，它负责决定编译时是否使用MKLML和MKL-DNN
同时，我们会引入`WITH_MKLML`选项，用于选择是否使用MKL-DNN自带的MKLML安装包。这个安装包可以独立于MKL-DNN使用，但是建议在开启MKL-DNN的同时也打开MKLML的开关，这样才能发挥最好的性能。
- `WITH_MKLML` 控制是否使用MKLML库。
当打开`WITH_MKL`时，会自动使用MKLML库作为PaddlePaddle的CBLAS和LAPACK库，同时会开启Intel OpenMP用于提高MKLML的性能。
编译时会把对应的头文件和库放在`build/third_party/install/mklml/*`目录下对应的地方。
MKLML的库目前都是动态库，主要包括`libiomp5.so`和`libmklml_intel.so`。
- `WITH_MKLDNN` 控制是否使用MKL-DNN。
当开启`WITH_MKL`时，会自动根据硬件配置[[2](#references)]选择是否编译MKL-DNN。
编译时会把对应的头文件和库放在`build/third_party/install/mkldnn/*`目录下对应的地方。
MKL-DNN的库目前只有动态库`libmkldnn.so`。
所以，我们会在`cmake/external`目录新建`mkldnn.cmake`和`mklml.cmake`文件，它们会在编译PaddlePaddle的时候下载对应的软件包，并放到PaddlePaddle的third party目录中。
### Matrix
目前在PaddlePaddle中数据都是以`NCHW`的格式存储，但是在MKL-DNN中的排列方式不止这一种。
所以我们定义了一个`MKLDNNMatrix`用于管理MKL-DNN数据的不同格式以及相互之间的转换。
**备注**：当`WITH_MKLML=ON`的时候，会优先使用这个包作为PaddlePaddle的CBLAS和LAPACK库，所以会稍微改动`cmake/cblas.cmake`中的逻辑。
<div align="center">
<img src="image/matrix.png"><br/>
Figure 3. MKLDNNMatrix
</div>
### Layers
所有MKL-DNN相关的C++ layers，都会按照PaddlePaddle的目录结构存放在
`paddle/gserver/layers`中，并且文件名都会一以*Mkldnn*开头。
所有MKL-DNN的Layers都会继承于`MKLDNNLayer`，该类继承于PaddlePaddle的基类`Layer`。
在`MKLDNNLayer`中会提供一些必要的接口和函数，并且会写好`forward`和`backward`的基本逻辑，
子类只需要使用定义好的接口，实现具体的函数功能即可。
所有MKL-DNN的layers都会继承于一个叫做`MkldnnLayer`的父类，该父类继承于PaddlePaddle的基类`Layer`。
<div align="center">
<img src="image/layers.png"><br/>
Figure 4. MKLDNNLayer
</div>
### Activations
由于在PaddlePaddle中，激活函数是独立于layer概念的，所以会在`paddle/gserver/activations`目录下添加一个`MkldnnActivation.h`文件定义一些用于MKL-DNN的接口，实现方法还是会在`ActivationFunction.cpp`文件。
每个MKLDNNLayer都包含用于内部存储和外部存储的一系列MKLDNNMatrix：
### Unit Tests
会在`paddle/gserver/test`目录下添加`test_Mkldnn.cpp`和`MkldnnTester.*`用于MKL-DNN的测试。
- 内部存储（internel memory）：`inVal_`,`inGrad_`,`outVal_`和`outGrad_`，分别代表输入数据，输入梯度，输出数据和输出梯度。
- 外部存储（external memory）：都是以ext开头，比如`extInVal_`和`extInGrad_`，它们主要是用于，
当数据格式与PaddlePaddle默认的`NCHW`格式不匹配时，转换内存的工作。
需要注意的是，PaddlePaddle的activation会直接使用`output_.value`和`output_.grad`，
所以`extOutVal_`和`extOutGrad_`必须分别与`output_.value`和`output_.grad`共享内存，
如果不需要外部存储用于转换，那么对应的内部存储也会与它们共享内存。
- 转换函数（resetXXX）： 包括`resetInValue`，`resetInGrad`，`resetOutValue`和`resetOutGrad`，
表示对输入数据，输入梯度，输出数据和输出梯度的转换。
这些函数会根据输入参数重新设置内部和外部存储，当然这两者也可以相等，即表示不需要转换。
注意：每个`MKLDNNlayer`的子类只需要使用内部存储就可以了，所有外部的转换工作都会在reset系列函数中都准备好。
### Activations
在重构前的PaddlePaddle中，激活函数是独立于`Layer`的概念，并且输入输出都是共用一块内存，
所以添加了对应的`MKLDNNActivation`来实现，方式类似于`MKLDNNLayer`。
### Parameters
对于有参数的层，我们会保证`MKLDNNLayer`使用的参数与PaddlePaddle申请的buffer共用一块内存。
如果存在数据排列格式不一样的情况时，我们会在网络训练之前把格式转换为MKL-DNN希望的格式，
在训练结束的时候再保存为PaddlePaddle的格式，但是整个训练过程中不需要任何转换。
这样既使得最终保存的参数格式与PaddlePaddle一致，又可以避免不必要的转换。
### Gradients
由于MKL-DNN的操作都是直接覆盖的形式，也就是说输出的结果不会在原来的数据上累加，
这样带来的好处就是不需要一直清空memory，节省了不必要的操作。
但是注意的是，当网络出现分支且在`backward`的时候，需要累加不同Layer传过来的梯度。
所以在`MKLDNNlayer`中实现了一个merge的方法，此时每个小分支的`Input Gradient`
会先临时保存在`MKLDNNMatrix`中，由分支处的Layer负责求和，并把结果放到当前层的`output_.grad`中。
所以整体上，在实现每个子类的时候就不需要关心分支的事情了。
Activation的测试，计划在PaddlePaddle原有的测试文件上直接添加新的测试type。
<div align="center">
<img src="image/gradients.png"><br/>
Figure 5. Merge Gradients
</div>
### Protobuf Messages
根据具体layer的需求可能会在`proto/ModelConfig.proto`里面添加必要的选项。
### Unit Tests
我们会添加`test_MKLDNN.cpp`和`MKLDNNTester.*`用于MKL-DNN的测试。
测试分为每个Layer（或Activation）的单元测试和简单网络的整体测试。
每个测试会对比PaddlePaddle中CPU算出的结果与MKL-DNN的结果，小于某个比较小的阈值认为通过。
### Python API
目前只考虑**v1 API**。
......@@ -73,39 +172,40 @@ if use_mkldnn
self.layer_type = mkldnn_*
```
所有MKL-DNN的layer type会以*mkldnn_*开头，以示区分。
所有MKL-DNN的`layer_type`会以*mkldnn_*开头，这些会在`MKLDNN*Layer`注册layer的时候保证，以示区分。
并且可能在`python/paddle/trainer_config_helper`目录下的`activations.py `和`layers.py`里面添加必要的MKL-DNN的接口。
### Demos
会在`v1_api_demo`目录下添加一个`mkldnn`的文件夹，里面放入一些用于MKL-DNN测试的demo脚本。
同时,会在`paddle/utils.Flags`中添加一个`use_mkldnn`的flag，用于选择是否使用MKL-DNN的相关功能。
### Benchmarking
会考虑添加部分逻辑在`benchmark/paddle/image/run.sh`，添加使用MKL-DNN的测试。
会添加相应的脚本在[这里](https://github.com/PaddlePaddle/Paddle/tree/develop/benchmark/paddle/image)，用于测试和对比在使用MKL-DNN前后的CNN网络性能。
测试的性能对比结果会在[IntelOptimizedPaddle.md](https://github.com/PaddlePaddle/Paddle/blob/develop/benchmark/IntelOptimizedPaddle.md)
### Others
1. 如果在使用MKL-DNN的情况下，会把CPU的Buffer对齐为64。
1. 如果在使用MKL-DNN的情况下，会把CPU的Buffer对齐为4096，具体可以参考MKL-DNN中的[memory](https://github.com/01org/mkl-dnn/blob/master/include/mkldnn.hpp#L673)。
2. 深入PaddlePaddle，寻找有没有其他可以优化的可能，进一步优化。比如可能会用OpenMP改进SGD的更新性能。
## Design Concerns
为了更好的符合PaddlePaddle的代码风格\[[2](#references)\]，同时又尽可能少的牺牲MKL-DNN的性能\[[3](#references)\]。
为了更好的符合PaddlePaddle的代码风格\[[3](#references)\]，同时又尽可能少的牺牲MKL-DNN的性能\[[4](#references)\]。
我们总结出一些特别需要注意的点：
1. 使用**deviceId_**。为了尽可能少的在父类Layer中添加变量或者函数，我们决定使用已有的`deviceId_`变量来区分layer的属性，定义`-2`为`MkldnnLayer`特有的设备ID。
2. 重写父类Layer的**init**函数，修改`deviceId_`为`-2`，代表这个layer是用于跑在MKL-DNN的环境下。
3. 创建`MkldnnMatrix`，用于管理MKL-DNN会用到的相关memory函数、接口以及会用的到格式信息。
4. 创建`MkldnnBase`，定义一些除了layer和memory相关的类和函数。包括MKL-DNN会用到`MkldnnStream`和`CpuEngine`，和未来可能还会用到`FPGAEngine`等。
5. 在**Argument**里添加两个`MkldnnMatrixPtr`，取名为`mkldnnValue`和`mkldnnGrad`，用于存放`MkldnnLayer`会用到的memory buffer。 并且添加函数cvt(会修改为一个更加合适的函数名)，用于处理"CPU device"和"MKL-DNN device"之间memory的相互转化。
6. 在父类`Layer`中的`getOutput`函数中添加一段逻辑，用于判断`deviceId`，并针对device在MKL-DNN和CPU之间不统一的情况，做一个前期转换。 也就是调用`Argument`的cvt函数把output统一到需要的device上。
7. 在原来的`FLAGS`中添加一个`use_mkldnn`的flag，用于选择是否使用MKL-DNN的相关功能。
8. 关于MKLDNN参数的保存。由于MKLDNN参数的格式与PaddlePaddle原有的格式存在不一样的情况，所以需要在保存参数时同时保存该格式信息。目前准备扩展[Header](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/parameter/Parameter.h#L247)里面的`int32_t version`。这个值不管是在v1还是在v2里面，一直保存的是0，所以可以充分利用这个信息，定义一个枚举处理所有MKLDNN的参数格式，从而`MKLDNNLayer`就可以从输入的参数中获取需要的格式信息。
1. 使用**deviceId_**。为了尽可能少的在父类Layer中添加变量或者函数，
我们决定使用已有的`deviceId_`变量来区分layer的属性，定义`-2`为`MKLDNNLayer`特有的设备ID。
2. 重写父类Layer的**init**函数，修改`deviceId_`为`-2`，代表这个layer是用于跑在MKL-DNN的环境下。
3. 创建`MKLDNNBase`，定义一些除了layer和memory相关的类和函数。
包括MKL-DNN会用到`MKLDNNStream`和`CPUEngine`，和未来可能还会用到`FPGAEngine`等。
4. 如果MKL-DNN layer的后面接有cpu device，那么就会使`output_.value`与`extOutVal_`共享内存，
同时数据格式就是`NCHW`，这样下一个cpu device就能拿到正确的数据。
在有普通的CPU layer时， `extOutVal_`和`extOutGrad_`的格式始终是`NCHW`或者`NC`。
## References
1. [Intel Math Kernel Library for Deep Neural Networks (Intel MKL-DNN)](https://github.com/01org/mkl-dnn "Intel MKL-DNN")
2. [原来的方案](https://github.com/PaddlePaddle/Paddle/pull/3096)会引入**nextLayer**的信息。但是在PaddlePaddle中，无论是重构前的layer还是重构后的op，都不会想要知道next layer/op的信息。
3. MKL-DNN的高性能格式与PaddlePaddle原有的`NCHW`不同(PaddlePaddle中的CUDNN部分使用的也是`NCHW`，所以不存在这个问题)，所以需要引入一个转换方法，并且只需要在必要的时候转换这种格式，才能更好的发挥MKL-DNN的性能。
1. [MKL small library](https://github.com/01org/mkl-dnn#linking-your-application)是[Intel MKL](https://software.intel.com/en-us/mkl)的一个子集。
主要包括了深度学习相关的数学原语与操作，一般由MKL-DNN在发布[新版本](https://github.com/01org/mkl-dnn/releases)时一起更新。
2. [MKL-DNN System Requirements](https://github.com/01org/mkl-dnn#system-requirements)。
目前在PaddlePaddle中，仅会在支持AVX2指令集及以上的机器才使用MKL-DNN。
3. [原来的方案](https://github.com/PaddlePaddle/Paddle/pull/3096)会引入**nextLayer**的信息。
但是在PaddlePaddle中，无论是重构前的layer还是重构后的op，都不会想要知道next layer/op的信息。
4. MKL-DNN的高性能格式与PaddlePaddle原有的`NCHW`不同(PaddlePaddle中的cuDNN部分使用的也是`NCHW`，所以不存在这个问题)。
所以需要引入一个转换方法，并且只需要在必要的时候转换这种格式，才能更好的发挥MKL-DNN的性能。
doc/design/model_format.md 0 → 100644
浏览文件 @ 988a4a66
# Design Doc: Model Format
## Motivation
A model is an output of the training process. One complete model consists of two parts, the **topology** and the **parameters**. In order to support industrial deployment, the model format must be self-complete and must not expose any training source code.
As a result, In PaddlePaddle, the **topology** is represented as a [ProgramDesc](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/doc/design/program.md), which describes the model structure. The **parameters** contain all the trainable weights in the model. We must support large size parameters and efficient serialization/deserialization of parameters.
## Implementation
The topology is saved as a plain text in a detailed self-contain protobuf file.
The parameters are saved as a binary file. As we all know, the protobuf message has a limit of [64M size](https://developers.google.com/protocol-buffers/docs/reference/cpp/google.protobuf.io.coded_stream#CodedInputStream.SetTotalBytesLimit.details). We have done a [benchmark experiment](https://github.com/PaddlePaddle/Paddle/pull/4610), which shows that protobuf is not fit for the task.
As a result, we design a particular format for tensor serialization. By default, an arbitrary tensor in Paddle is a [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/lod_tensor.md), and has a description information proto of [LoDTensorDesc](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L99). We save the DescProto as the byte string header. It contains all the necessary information, such as the `dims`, and the `LoD` information in [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/paddle/framework/lod_tensor.md). A tensor stores values in a continuous memory buffer. For speed we dump the raw memory to disk and save it as the byte string content. So, the binary format of one tensor is,
The table below shows a tensor's byte view in detail. Note that all the signed values are written in the little-endian format.
|field name | type | description |
| --- | --- | --- |
| version | uint32_t | Version of saved file. Always 0 now. |
| tensor desc length | uint32_t | TensorDesc(Protobuf message) length in bytes. |
| tensor desc | void* | TensorDesc protobuf binary message |
| tensor data | void* | Tensor's data in binary format. The length of `tensor_data` is decided by `TensorDesc.dims()` and `TensorDesc.data_type()` |
| lod_level | uint64_t | Level of LoD |
| length of lod[0] | uint64_t | [Optional] length of lod[0] in bytes. |
| data of lod[0] | uint64_t* | [Optional] lod[0].data() |
| ... | ... | ... |
## Summary
- We introduce a model format.
- The model represented by its forward-pass computation procedure is saved in a **ProgramDesc** protobuf message.
- A bunch of specified format binary tensors describe the **parameters**.
doc/design/ops/images/2_level_rnn.dot
浏览文件 @ 988a4a66
digraph G {
rnn [label="1-th level RNN" shape=box]
rnn [label="1st level RNN" shape=box]
subgraph cluster0 {
label = "time step 0"
......@@ -8,7 +8,7 @@ digraph G {
sent0 [label="sentence"]
sent1 [label="sentence"]
rnn1 [label="2-th level RNN" shape=box]
rnn1 [label="2nd level RNN" shape=box]
sent0 -> rnn1
sent1 -> rnn1
......@@ -20,7 +20,7 @@ digraph G {
sent2 [label="sentence"]
sent3 [label="sentence"]
rnn2 [label="2-th level RNN" shape=box]
rnn2 [label="2nd level RNN" shape=box]
sent2 -> rnn2
sent3 -> rnn2
......@@ -32,7 +32,7 @@ digraph G {
sent4 [label="sentence"]
sent5 [label="sentence"]
rnn3 [label="2-th level RNN" shape=box]
rnn3 [label="2nd level RNN" shape=box]
sent4 -> rnn3
sent5 -> rnn3
......
doc/design/ops/rnn.md
浏览文件 @ 988a4a66
# RNNOp design
This document is about an RNN operator which requires that instances in a mini-batch have the same length. We will have a more flexible RNN operator.
This document describes the RNN (Recurrent Neural Network) operator and how it is implemented in PaddlePaddle. The RNN op requires that all instances in a mini-batch have the same length. We will have a more flexible dynamic RNN operator in the future.
## RNN Algorithm Implementation
<p aligh="center">
<p align="center">
<img src="./images/rnn.jpg"/>
</p>
The above diagram shows an RNN unrolled into a full network.
There are several important concepts:
There are several important concepts here:
- *step-net*: the sub-graph to run at each step,
- *memory*, $h_t$, the state of the current step,
- *ex-memory*, $h_{t-1}$, the state of the previous step,
- *initial memory value*, the ex-memory of the first step.
- *step-net*: the sub-graph that runs at each step.
- *memory*, $h_t$, the state of the current step.
- *ex-memory*, $h_{t-1}$, the state of the previous step.
- *initial memory value*, the memory of the first (initial) step.
### Step-scope
There could be local variables defined in step-nets. PaddlePaddle runtime realizes these variables in *step-scopes* -- scopes created for each step.
There could be local variables defined in each step-net. PaddlePaddle runtime realizes these variables in *step-scopes* which are created for each step.
<p aligh="center">
<p align="center">
<img src="./images/rnn.png"/><br/>
Figure 2 the RNN's data flow
Figure 2 illustrates the RNN's data flow
</p>
Please be aware that all steps run the same step-net. Each step
Please be aware that every step runs the same step-net. Each step does the following:
1. creates the step-scope,
2. realizes local variables, including step-outputs, in the step-scope, and
3. runs the step-net, which could use these variables.
1. Creates the step-scope.
2. Initializes the local variables including step-outputs, in the step-scope.
3. Runs the step-net, which uses the above mentioned variables.
The RNN operator will compose its output from step outputs in step scopes.
The RNN operator will compose its output from step outputs in each of the step scopes.
### Memory and Ex-memory
Let's give more details about memory and ex-memory via a simply example:
Let's give more details about memory and ex-memory using a simple example:
$$
h_t = U h_{t-1} + W x_t
$$,
where $h_t$ and $h_{t-1}$ are the memory and ex-memory of step $t$'s respectively.
where $h_t$ and $h_{t-1}$ are the memory and ex-memory (previous memory) of step $t$ respectively.
In the implementation, we can make an ex-memory variable either "refers to" the memory variable of the previous step,
or copy the value of the previous memory value to the current ex-memory variable.
In the implementation, we can make an ex-memory variable either "refer to" the memory variable of the previous step,
or copy the memory value of the previous step to the current ex-memory variable.
### Usage in Python
For more information on Block, please refer to the [design doc](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/block.md).
We can define an RNN's step-net using Block:
We can define an RNN's step-net using a Block:
```python
import paddle as pd
X = some_op() # x is some operator's output, and is a LoDTensor
X = some_op() # x is some operator's output and is a LoDTensor
a = some_op()
# declare parameters
......@@ -68,7 +68,7 @@ with rnn.stepnet():
x = rnn.add_input(X)
# declare a memory (rnn's step)
h = rnn.add_memory(init=a)
# h.pre_state() means previous memory of rnn
# h.pre_state(), the previous memory of rnn
new_state = pd.add_two( pd.matmul(W, x) + pd.matmul(U, h.pre_state()))
# update current memory
h.update(new_state)
......@@ -80,19 +80,19 @@ out = rnn()
Python API functions in above example:
- `rnn.add_input` indicates the parameter is a variable that will be segmented into step-inputs.
- `rnn.add_memory` creates a variable used as the memory.
- `rnn.add_outputs` mark the variables that will be concatenated across steps into the RNN output.
- `rnn.add_input`: indicates that the parameter is a variable that will be segmented into step-inputs.
- `rnn.add_memory`: creates a variable used as the memory.
- `rnn.add_outputs`: marks the variables that will be concatenated across steps into the RNN output.
### Nested RNN and LoDTensor
An RNN whose step-net includes other RNN operators is known as an *nested RNN*.
For example, we could have a 2-level RNN, where the top level corresponds to paragraphs, and the lower level corresponds to sentences.
For example, we could have a 2-level RNN, where the top level corresponds to paragraphs, and the lower level corresponds to sentences. Each step of the higher level RNN also receives an input from the corresponding step of the lower level, and additionally the output from the previous time step at the same level.
The following figure illustrates the feeding of text into the lower level, one sentence each step, and the feeding of step outputs to the top level. The final top level output is about the whole text.
The following figure illustrates feeding in text into the lower level, one sentence at a step, and the feeding in step outputs to the top level. The final top level output is about the whole text.
<p aligh="center">
<p align="center">
<img src="./images/2_level_rnn.png"/>
</p>
......@@ -110,7 +110,7 @@ a = some_op()
# chapter_data is a set of 128-dim word vectors
# the first level of LoD is sentence
# the second level of LoD is chapter
# the second level of LoD is a chapter
chapter_data = pd.Variable(shape=[None, 128], type=pd.lod_tensor, level=2)
def lower_level_rnn(paragraph):
......@@ -138,14 +138,14 @@ with top_level_rnn.stepnet():
pd.matmul(W0, paragraph_data) + pd.matmul(U0, h.pre_state()))
top_level_rnn.add_outputs(h)
# just output the last step
# output the last step
chapter_out = top_level_rnn(output_all_steps=False)
```
in above example, the construction of the `top_level_rnn` calls `lower_level_rnn`. The input is a LoD Tensor. The top level RNN segments input text data into paragraphs, and the lower level RNN segments each paragraph into sentences.
In the above example, the construction of the `top_level_rnn` calls `lower_level_rnn`. The input is an LoD Tensor. The top level RNN segments input text data into paragraphs, and the lower level RNN segments each paragraph into sentences.
By default, the `RNNOp` will concatenate the outputs from all the time steps,
if the `output_all_steps` set to False, it will only output the final time step.
By default, the `RNNOp` will concatenate the outputs from all the time steps.
If the `output_all_steps` is set to False, it will only output the final time step.
<p align="center">
......
doc/design/ops/sequence_decoder.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/optimizer.md 0 → 100644
浏览文件 @ 988a4a66
## Optimizer Design
### The Problem
A PaddlePaddle program, or a block, is a sequence of operators operating variables. A training program needs to do three kinds of works:
1. the forward pass, which computes intermediate results and the cost(s),
1. the backward pass, which derives gradients from intermediate results and costs, and
1. the optimization pass, which update model parameters to optimize the cost(s).
These works rely on three kinds of operators:
1. forward operators,
1. gradient operators, and
1. optimization operators.
It's true that users should be able to create all these operators manually by calling some low-level API, but it would be much more convenient if they could only describe the forward pass and let PaddlePaddle create the backward and optimization operators automatically.
In this design, we propose a high-level API that automatically derives the optimisation pass and operators from the forward pass.
### High-level Python API to describe the training process
1. User write code to describe the network:
```python
images = layer.data("images")
labels = layer.data("labels")
w1 = pd.var("w1")
b1 = pd.var("b1")
hidden = layer.fc(images, w=w1, b=b1)
cost = layer.mse(hidden, labels)
```
The above code snippet will create forward operators in [Block](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/block.md).
2. Users create a certain kind of Optimizer with some argument.
```python
optimizer = AdagradOptimizer(learing_rate=0.001)
```
3. Users use the optimizer to `minimize` a certain `cost` through updating parameters in parameter_list.
```python
opt_op_list = optimizer.minimize(cost, parameter_list=[w1, b1])
```
The above code snippet will create gradient and optimization operators in Block. The return value of `minimize()` is list of optimization operators that will be run by session.
4. Users use Session/Executor to run this opt_op_list as target to do training.
```python
sess.run(target= opt_op_list, ...)
```
#### Optimizer Python interface:
```python
class Optimizer(object):
"""Optimizer Base class.
"""
def __init__(self):
pass
def create_optimization_pass(self, parameters_and_grads):
"""Add optimization operators to update gradients to variables.
Args:
parameters_and_grads: a list of (variable, gradient) pair to update.
Returns:
optmization_op_list: a list of optimization operator that will update parameter using gradient.
"""
return None
def minimize(self, loss, parameter_list):
"""Add operations to minimize `loss` by updating `parameter_list`.
This method combines interface `append_backward_ops()` and
`create_optimization_pass()` into one.
"""
params_grads = self.create_backward_pass(loss, parameter_list)
update_ops = self.create_optimization_pass(params_grads)
return update_ops
```
Users can inherit the Optimizer above to create their own Optimizer with some special logic, such as AdagradOptimizer.
doc/design/parameter_average.md 0 → 100644
浏览文件 @ 988a4a66
# Averaging Parameter in PaddlePaddle
## Why Averaging
In a large scale machine learning setup where the size of the training data is huge, it could take us a large number of iterations over the training data before we can achieve the optimal values of parameters of our model. Looking at the problem setup, it is desirable if we can obtain the optimal values of parameters by going through the data in as few passes as we can.
Polyak and Juditsky (1992) showed that the test performance of simple average of parameters obtained by Stochastic Gradient Descent (SGD) is as good as that of parameter values that are obtained by training the model over and over again, over the training dataset.
Hence, to accelerate the speed of Stochastic Gradient Descent, Averaged Stochastic Gradient Descent (ASGD) was proposed in Polyak and Juditsky (1992). For ASGD, the running average of parameters obtained by SGD, is used as the estimator for <img src="./images/theta_star.gif"/><br/> . The averaging is done as follows:
<img src="./images/asgd.gif" align="center"/><br/>
We propose averaging for any optimizer similar to how ASGD performs it, as mentioned above.
### How to perform Parameter Averaging in PaddlePaddle
Parameter Averaging in PaddlePaddle works in the following way during training :
1. It will take in an instance of a normal optimizer as an input, e.g. RMSPropOptimizer
2. The optimizer itself is responsible for updating the parameters.
3. The ParameterAverageOptimizer maintains a separate copy of the parameters for itself:
1. In concept, the values of this copy are the average of the values of the parameters in the most recent N batches.
2. However, saving all the N instances of the parameters in memory is not feasible.
3. Therefore, an approximation algorithm is used.
Hence, overall we have have two copies of the parameters: one for the optimizer itself, and one for the ParameterAverageOptimizer. The former should be used in back propagation, while the latter should be used during testing and should be saved.
During the testing/ saving the model phase, we perform the following steps:
1. Perform the delayed operations.
2. Save current values of the parameters to a temporary variable.
3. Replace the values of the parameters with the averaged values.
4. Perform testing and/or save the parameters.
5. Restore the values of the parameters once done.
### How to implement Averaging of Parameter in PaddlePaddle
We can add the ParameterAverageOptimizer op to the graph through Python API. Using this approach, we manually add this op to the graph and direct the output of the optimizer op to this op during training.
**Advantages**:
- Allows for greater flexibility to the users of PaddlePaddle. Using this approach, the users can plug different optimizers into ParameterAverageOptimizer by passing in the optimizer to the op.
- Makes it easy for the users to customize and extend the framework.
**Disadvantages**:
- Implementation requires re-writing the averaging methodology in Python.
### Low-Level implementation
In the new design, we propose to create a new operation for averaging parameter updates (ParameterAverageOptimizer). For now, we can add an op that takes in the following as input:
- the optimizer
- the window_size to keep the updates
The ParameterAverageOptimizer op can be like any other operator with its own CPU/GPU implementation either using Eigen or separate CPU and GPU kernels. As the initial implementation, we can implement the kernel using Eigen following the abstraction pattern implemented for [Operators](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/rmsprop_op.h). We also want to support the case when the Trainer/Optimizer runs on the GPU while ParameterAverageOptimizer runs on a CPU.
The idea of building an op for averaging is in sync with the refactored PaddlePaddle philosophy of using operators to represent any computation unit. The way the op will be added to the computation graph will be decided by the [layer functions](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/python_api.md#layer-function) in Python API.
### Python API implementation for ParameterAverageOptimizer
Based on Polyak and Juditsky (1992), we can generalize the averaging of updates to any optimizer. The input to the op would be the following:
- Any optimizer (RMSProp , AdaGrad etc.)
- A window size. The op keeps accumulating updated parameter values over a window of N batches and takes an average. Move the averaged value to a buffer when window is full to avoid loss of precision.
Using the ParameterAverageOptimizer op, any user can add the operation to their computation graphs. However, this will require a lot of lines of code and we should design Python APIs that support averaging. As per the PaddlePaddle [Python API design](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/python_api.md), the layer functions are responsible for creating operators, operator parameters and variables. Since ParameterAverageOptimizer will be an operator, it makes sense to create it in the layer functions.
We will have a wrapper written in Python that will support the functionality and implement the actual core computation in C++ core as we have done for other [Optimizers](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/operators/rmsprop_op.cc)
#### Creation of the ParameterAverageOptimizer operator
There are two ways for creating the ParameterAverageOptimizer op:
1. We create the op immediately while building the computation graph.
2. We add the op in a lazy manner, just before the backward pass, similar to the way the optimization ops are added.
The proposal is to add the op immediately while building the computation graph.
#### High-level API
In PaddlePaddle Python API, users will primarily rely on [layer functions](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/design/python_api.md#layer-function) to create neural network layers. Hence, we also need to provide parameter average functionality in layer functions.
doc/design/program.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/prune.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/python_api.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/reader/README.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/refactor/session.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/refactorization.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/register_grad_op.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/regularization.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/releasing_process.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/scope.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/selected_rows.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/tensor_array.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/test.dot 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/test.dot.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/design/var_desc.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/faq/local/index_cn.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/faq/parameter/index_cn.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/getstarted/basic_usage/index_cn.rst 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/getstarted/basic_usage/index_en.rst 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/getstarted/basic_usage/parameters.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/getstarted/build_and_install/cmake.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/getstarted/build_and_install/paddleci.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/getstarted/build_and_install/pip_install_cn.rst 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/getstarted/build_and_install/pip_install_en.rst 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/getstarted/index_cn.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/getstarted/index_en.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/dev/build_cn.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/dev/build_en.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/dev/contribute_to_paddle_en.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/howto/dev/contribute_to_paddle_en.md 0 → 120000
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/dev/new_op_cn.md
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/dev/write_docs_cn.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/dev/write_docs_en.rst 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/index_cn.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/index_en.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/optimization/cpu_profiling.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/optimization/cpu_profiling_cn.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/optimization/pprof_1.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/optimization/pprof_2.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/usage/cluster/src/trainer.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/usage/cluster/src/trainer_cn.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/usage/cluster/src/word2vec/api_train_v2.py 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/howto/usage/cluster/src/word2vec/prepare.py 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/index_cn.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/index_en.rst
浏览文件 @ 988a4a66
此差异已折叠。
doc/mobile/cross_compiling_for_android_en.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/mobile/cross_compiling_for_ios_cn.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/mobile/cross_compiling_for_raspberry_en.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/mobile/index_cn.rst 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/mobile/index_en.rst 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
doc/tutorials/image_classification/cifar.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/image_classification/lenet.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/image_classification/plot.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/image_classification/src/plot.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/index_cn.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/index_en.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/rec/ml_dataset_cn.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/rec/ml_dataset_en.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/rec/ml_regression_cn.rst 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/rec/ml_regression_en.rst 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/rec/rec_regression_network.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/sentiment_analysis/index_cn.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/sentiment_analysis/index_en.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/sentiment_analysis/lstm.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/sentiment_analysis/src/lstm.png 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/text_generation/index_cn.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/tutorials/text_generation/index_en.md 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
doc/v1_api_tutorials/README.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
go/.gitignore
浏览文件 @ 988a4a66
此差异已折叠。
go/cmd/master/master.go
浏览文件 @ 988a4a66
此差异已折叠。
go/cmd/pserver/pserver.go
浏览文件 @ 988a4a66
此差异已折叠。
go/glide.lock
浏览文件 @ 988a4a66
此差异已折叠。
go/glide.yaml
浏览文件 @ 988a4a66
此差异已折叠。
go/master/c/client.go
浏览文件 @ 988a4a66
此差异已折叠。
go/master/client.go
浏览文件 @ 988a4a66
此差异已折叠。
go/master/client_internal_test.go
浏览文件 @ 988a4a66
此差异已折叠。
go/master/client_test.go
浏览文件 @ 988a4a66
此差异已折叠。
go/master/etcd_client.go
浏览文件 @ 988a4a66
此差异已折叠。
go/master/service.go
浏览文件 @ 988a4a66
此差异已折叠。
go/proto/.gitignore 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/client/c/cclient.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/client/client.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/client/client_test.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/client/etcd_client.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/etcd_client.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/optimizer.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/optimizer_test.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/service.go
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/service_internal_test.go 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
go/pserver/service_test.go
浏览文件 @ 988a4a66
此差异已折叠。
paddle/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/api/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/api/Util.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/Main.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/Matrix.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/config.h.in
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/error.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/error.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/gradient_machine.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/main.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/matrix.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/tests/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/capi/tests/test_Matrix.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/cuda/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/cuda/include/hl_cnn.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/cuda/include/hl_gpu.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/cuda/include/hl_matrix.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/cuda/src/hl_cuda_cnn.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/cuda/src/hl_cuda_matrix.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/attribute.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/attribute.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/backward.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/backward.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/backward_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/block_desc.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/block_desc.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/data_type.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/ddim.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/ddim.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/executor.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/executor.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/feed_fetch_method.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/feed_fetch_type.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/framework.proto
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/grad_op_builder.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/framework/grad_op_builder_test.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/framework/grad_op_desc_maker.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_rank_table.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_rank_table.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_tensor.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_tensor.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_tensor.md
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_tensor_array.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_tensor_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/lod_tensor_test.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/op_desc.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/op_desc.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/op_info.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/op_proto_maker.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/op_registry.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/op_registry.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/operator.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/operator.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/operator_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/program_desc.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/program_desc.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/program_desc_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/prune.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/prune.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/prune_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/scope.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/scope.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/scope_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/selected_rows.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/selected_rows.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/selected_rows_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/shape_inference.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/shape_inference.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/tensor.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/tensor_array.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/framework/tensor_array.h 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/framework/tensor_array_test.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/framework/tensor_impl.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/tensor_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/tensor_util.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/tensor_util_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/type_defs.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/var_desc.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/var_desc.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/var_type.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/var_type_inference.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/var_type_inference_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/framework/variable.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/BlockExpandOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/ConvOp.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/ConvOpTest.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/CosSimOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/CropOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/DepthwiseConvOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/EigenGemm.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/FunctionTest.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/GemmConvOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/GemmConvOpTest.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/Im2Col.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/Im2ColOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/Im2ColOpGpu.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/Im2ColTest.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/MulOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/PadOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/RowConvOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/ScaleSubRegionOp.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/ScaleSubRegionOp.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/ScaleSubRegionOpGpu.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/ScaleSubRegionOpTest.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/function/SwitchOp.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/CRFLayer.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/CostLayer.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/DotProdLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/FactorizationMachineLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/FactorizationMachineLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/L2DistanceLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/L2DistanceLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/Layer.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/Layer.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNAddtoLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNAddtoLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNBase.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNBatchNormLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNBatchNormLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNConcatLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNConcatLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNLRNLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNLRNLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MKLDNNLayer.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MaxPoolWithMaskLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/MaxPoolWithMaskLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/PoolLayer.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/PoolLayer.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/ROIPoolLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/ROIPoolLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/ScaleSubRegionLayer.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/layers/ScaleSubRegionLayer.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/tests/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/tests/MKLDNNTester.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/tests/mkldnn_branch_net.conf 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/tests/proto_files.txt 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/gserver/tests/sequence_lstm.conf 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/tests/sequence_recurrent_group.py 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/tests/test_Expand.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/gserver/tests/test_MaxPoolingWithMaskOutput.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/BaseMatrix.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/CpuSparseMatrix.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/CpuSparseMatrix.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/MKLDNNMatrix.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/MKLDNNMatrix.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/MathFunctions.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/MathFunctions.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/Matrix.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/Matrix.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/RowBuffer.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/SIMDFunctions.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/SparseMatrix.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/SparseMatrix.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/SparseRowMatrix.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/Storage.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/Vector.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/Vector.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/float16.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/tests/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/tests/TensorCheck.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/tests/test_Matrix.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/tests/test_Tensor.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/tests/test_float16.cpp 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/math/tests/test_float16.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/memory/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/memory/README.md
浏览文件 @ 988a4a66
此差异已折叠。
paddle/memory/detail/meta_cache.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/memory/memcpy.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/memory/memcpy.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/memory/memory.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/memory/memory_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/accuracy_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/accuracy_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/accuracy_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/activation_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/activation_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/activation_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adadelta_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adadelta_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adagrad_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adagrad_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adagrad_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adam_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adam_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adam_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adamax_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adamax_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/adamax_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/array_operator.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/array_to_lod_tensor_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/assign_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/auc_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/auc_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/batch_norm_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/batch_norm_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/batch_norm_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/batch_norm_op.md 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/beam_search_decode_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/beam_search_decode_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/beam_search_decode_op_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/beam_search_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/beam_search_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/bilinear_tensor_product_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/bilinear_tensor_product_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/bilinear_tensor_product_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/cast_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/cast_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/cast_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/chunk_eval_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/chunk_eval_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/clip_by_norm_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/clip_by_norm_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/clip_by_norm_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/clip_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/compare_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/compare_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/compare_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/concat_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/cond_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conditional_block_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv2d_op.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/operators/conv_cudnn_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_cudnn_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_shift_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_shift_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_shift_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_transpose_cudnn_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_transpose_cudnn_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_transpose_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_transpose_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/conv_transpose_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/cos_sim_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/cos_sim_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/crf_decoding_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/crf_decoding_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/crop_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/cross_entropy_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/decayed_adagrad_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/decayed_adagrad_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/decayed_adagrad_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/detail/CMakeLists.txt 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/detail/recv_impl.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/detail/safe_ref.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/detail/send_impl.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/detail/send_recv.proto 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/detail/send_recv_impl.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/detail/simple_block_queue.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/dropout_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/dropout_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/dropout_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/elementwise_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/expand_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/expand_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/expand_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/fc_op.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/operators/feed_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/fetch_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/fill_constant_batch_size_like_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/fill_constant_batch_size_like_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/fill_constant_batch_size_like_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/fill_constant_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/fill_zeros_like_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/ftrl_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/ftrl_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/ftrl_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gather.cu.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gather.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gather_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gather_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gather_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gather_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gru_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gru_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gru_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gru_unit_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gru_unit_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/gru_unit_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/hinge_loss_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/hinge_loss_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/hinge_loss_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/huber_loss_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/huber_loss_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/huber_loss_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/images/batch_norm_fork.dot 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/images/batch_norm_fork.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/images/batch_norm_op_kernel.png 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/increment_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/is_empty_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/l1_norm_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/l1_norm_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/l1_norm_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/linear_chain_crf_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/linear_chain_crf_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/linear_chain_crf_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/load_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lod_array_length_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lod_rank_table_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lod_reset_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lod_reset_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lod_reset_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lod_tensor_to_array_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/log_loss_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/log_loss_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/log_loss_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/logical_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/logical_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/logical_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lookup_table_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lookup_table_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lookup_table_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lrn_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lrn_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lrn_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lstm_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lstm_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lstm_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lstm_unit_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lstm_unit_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/lstm_unit_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/margin_rank_loss_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/margin_rank_loss_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/context_project.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/context_project.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/context_project.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/CMakeLists.txt 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/activation_functions.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/avx_functions.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/gru_cpu_kernel.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/gru_gpu_kernel.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/gru_kernel.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/lstm_cpu_kernel.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/lstm_gpu_kernel.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/detail/lstm_kernel.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/gru_compute.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/gru_compute.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/gru_compute.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/im2col.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/im2col.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/im2col.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/lstm_compute.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/lstm_compute.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/lstm_compute.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/math_function_impl.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/math_function_test.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/matmul.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/maxouting.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/maxouting.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/maxouting.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/pooling.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/pooling.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/pooling.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/selected_rows_functor.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/selected_rows_functor.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/selected_rows_functor.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/selected_rows_functor_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/selected_rows_functor_test.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/sequence2batch.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/sequence2batch.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/sequence2batch.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/sequence_pooling.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/sequence_pooling.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/sequence_pooling.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/softmax.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/softmax.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/softmax.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/softmax_impl.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/unpooling.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/unpooling.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/unpooling.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/vol2col.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/vol2col.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/vol2col.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/math/vol2col_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/matmul_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/matmul_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/matmul_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/max_sequence_len_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/maxout_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/maxout_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/maxout_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/mean_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/mean_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/merge_lod_tensor_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/minus_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/momentum_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/momentum_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/momentum_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/mul_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/mul_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/multiplex_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/multiplex_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/name_convention.md
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nccl/CMakeLists.txt 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nccl/nccl_gpu_common.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nccl/nccl_gpu_common.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nccl_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nccl_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nccl_op_test.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nce_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/nce_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/net_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pad_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_cudnn_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_cudnn_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_cudnn_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_with_index_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_with_index_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/pool_with_index_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/positive_negative_pair_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/positive_negative_pair_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/precision_recall_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/precision_recall_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/prelu_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/proximal_adagrad_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/proximal_adagrad_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/proximal_adagrad_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/proximal_gd_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/proximal_gd_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/proximal_gd_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/rank_loss_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/rank_loss_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/rank_loss_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/recurrent_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/recurrent_op.h 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/operators/recv_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/reduce_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/reduce_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/reduce_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/reshape_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/reshape_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/reshape_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/rmsprop_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/rmsprop_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/rmsprop_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/rnn/recurrent_op_utils.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/operators/rnn/recurrent_op_utils.h 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/operators/rnn_memory_helper_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/roi_pool_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/roi_pool_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/roi_pool_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/save_load_op_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/save_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scale_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scale_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scale_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scatter.cu.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scatter.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scatter_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scatter_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scatter_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/scatter_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/send_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/send_recv_op_test.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/seq_expand_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/seq_expand_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/seq_expand_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_concat_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_concat_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_concat_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_conv_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_conv_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_conv_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_pool_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_slice_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_slice_op.cu 0 → 100755
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sequence_slice_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sgd_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sgd_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sgd_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/shrink_rnn_memory_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sign_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sign_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/softmax_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/softmax_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/split_lod_tensor_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/split_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/squared_l2_norm_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/squared_l2_norm_op.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/squared_l2_norm_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sum_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sum_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/sum_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/tensor.save 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/tensor_array_read_write_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/top_k_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/top_k_op.cu
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/top_k_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/transpose_op.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/transpose_op.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/unpool_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/unpool_op.cu.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/unpool_op.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/operators/while_op.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/adam_optimizer.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/adam_optimizer.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/lr_policy.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/optimizer.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/optimizer.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/sgd_optimizer.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/sgd_optimizer.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/optimizer/tensor.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/parameter/Parameter.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/call_once.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/cuda_helper.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/cudnn_helper.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/device_context.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/device_context.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/dynload/cublas.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/dynload/cudnn.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/dynload/cudnn.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/dynload/nccl.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/dynload/nccl.h 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/enforce.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/enforce.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/enforce_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/environment_test.cc 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/platform/gpu_info.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/gpu_info.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/macros.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/nccl_test.cu 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/place.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/transform.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/platform/variant.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pserver/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pserver/LightNetwork.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pserver/ParameterClient2.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pserver/test/SocketTest.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pybind/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pybind/print_operators_doc.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pybind/protobuf.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pybind/pybind.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/pybind/tensor_py.h
浏览文件 @ 988a4a66
此差异已折叠。
paddle/scripts/check_env.sh 0 → 100755
浏览文件 @ 988a4a66
此差异已折叠。
paddle/scripts/cluster_train_v2/fabric/run.sh 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/scripts/deb/postinst 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/scripts/docker/README.md
浏览文件 @ 988a4a66
此差异已折叠。
paddle/scripts/docker/build.sh
浏览文件 @ 988a4a66
此差异已折叠。
paddle/scripts/submit_local.sh.in
浏览文件 @ 988a4a66
此差异已折叠。
paddle/scripts/travis/build_doc.sh
浏览文件 @ 988a4a66
此差异已折叠。
paddle/string/to_string_test.cc
浏览文件 @ 988a4a66
此差异已折叠。
paddle/testing/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/testing/TestUtil.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/testing/paddle_gtest_main.cc 0 → 100644
浏览文件 @ 988a4a66
此差异已折叠。
paddle/trainer/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/trainer/MergeModel.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/trainer/Trainer.cpp
浏览文件 @ 988a4a66
此差异已折叠。
paddle/trainer/tests/CMakeLists.txt
浏览文件 @ 988a4a66
此差异已折叠。
paddle/trainer/tests/chunking.conf 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/trainer/tests/compare_sparse_data 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/trainer/tests/data_bin_part 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/trainer/tests/gen_proto_data.py 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/trainer/tests/mnist.list 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/trainer/tests/mnist_bin_part 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/trainer/tests/test.txt 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
paddle/trainer/tests/test_CompareTwoOpts.cpp 已删除 100644 → 0
浏览文件 @ ddc25873
此差异已折叠。
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或