Skip to content

P
Paddle

PaddlePaddle / Paddle
大约 1 年 前同步成功

通知 2298
Star 20931
Fork 5422
P
Paddle
未验证 提交 dbd25805 编写于 作者: X Xin Pan 提交者： GitHub
浏览文件
操作

Merge pull request #11358 from panyx0718/new_master 

Merge release 0.12.0 and 0.13.0 to master
上级 1ff47b28 fb3bfc3d
展开全部 隐藏空白更改
内联 并排
Showing with 36269 addition and 17692 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 @ dbd25805
......@@ -19,7 +19,7 @@ BasedOnStyle: Google
IndentWidth: 2
TabWidth: 2
ContinuationIndentWidth: 4
AccessModifierOffset: -2 # The private/protected/public has no indent in class
AccessModifierOffset: -1 # The private/protected/public has no indent in class
Standard: Cpp11
AllowAllParametersOfDeclarationOnNextLine: true
BinPackParameters: false
......
.copyright.hook 0 → 100644
浏览文件 @ dbd25805
from __future__ import absolute_import
from __future__ import print_function
from __future__ import unicode_literals
import argparse
import io, re
import sys, os
import subprocess
import platform
COPYRIGHT = '''
Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
'''
LANG_COMMENT_MARK = None
NEW_LINE_MARK = None
COPYRIGHT_HEADER = None
if platform.system() == "Windows":
NEW_LINE_MARK = "\r\n"
else:
NEW_LINE_MARK = '\n'
COPYRIGHT_HEADER = COPYRIGHT.split(NEW_LINE_MARK)[1]
p = re.search('(\d{4})', COPYRIGHT_HEADER).group(0)
process = subprocess.Popen(["date", "+%Y"], stdout=subprocess.PIPE)
date, err = process.communicate()
date = date.decode("utf-8").rstrip("\n")
COPYRIGHT_HEADER = COPYRIGHT_HEADER.replace(p, date)
def generate_copyright(template, lang='C'):
if lang == 'Python':
LANG_COMMENT_MARK = '#'
else:
LANG_COMMENT_MARK = "//"
lines = template.split(NEW_LINE_MARK)
BLANK = " "
ans = LANG_COMMENT_MARK + BLANK + COPYRIGHT_HEADER + NEW_LINE_MARK
for lino, line in enumerate(lines):
if lino == 0 or lino == 1 or lino == len(lines) - 1: continue
if len(line) == 0:
BLANK = ""
else:
BLANK = " "
ans += LANG_COMMENT_MARK + BLANK + line + NEW_LINE_MARK
return ans + "\n"
def lang_type(filename):
if filename.endswith(".py"):
return "Python"
elif filename.endswith(".h"):
return "C"
elif filename.endswith(".c"):
return "C"
elif filename.endswith(".hpp"):
return "C"
elif filename.endswith(".cc"):
return "C"
elif filename.endswith(".cpp"):
return "C"
elif filename.endswith(".cu"):
return "C"
elif filename.endswith(".cuh"):
return "C"
elif filename.endswith(".go"):
return "C"
elif filename.endswith(".proto"):
return "C"
else:
print("Unsupported filetype %s", filename)
exit(0)
PYTHON_ENCODE = re.compile("^[ \t\v]*#.*?coding[:=][ \t]*([-_.a-zA-Z0-9]+)")
def main(argv=None):
parser = argparse.ArgumentParser(
description='Checker for copyright declaration.')
parser.add_argument('filenames', nargs='*', help='Filenames to check')
args = parser.parse_args(argv)
retv = 0
for filename in args.filenames:
fd = io.open(filename, encoding="utf-8")
first_line = fd.readline()
second_line = fd.readline()
if "COPYRIGHT (C)" in first_line.upper(): continue
if first_line.startswith("#!") or PYTHON_ENCODE.match(
second_line) != None or PYTHON_ENCODE.match(first_line) != None:
continue
original_contents = io.open(filename, encoding="utf-8").read()
new_contents = generate_copyright(
COPYRIGHT, lang_type(filename)) + original_contents
print('Auto Insert Copyright Header {}'.format(filename))
retv = 1
with io.open(filename, 'w') as output_file:
output_file.write(new_contents)
return retv
if __name__ == '__main__':
exit(main())
.gitignore
浏览文件 @ dbd25805
paddle/operators/check_t.save
paddle/operators/check_tensor.ls
paddle/operators/tensor.save
python/paddle/v2/fluid/tests/book/image_classification_resnet.inference.model/
python/paddle/v2/fluid/tests/book/image_classification_vgg.inference.model/
python/paddle/v2/fluid/tests/book/label_semantic_roles.inference.model/
*.DS_Store
build/
build_doc/
......@@ -19,13 +25,3 @@ third_party/
# clion workspace.
cmake-build-*
# generated while compiling
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 @ dbd25805
repos:
- repo: https://github.com/Lucas-C/pre-commit-hooks.git
sha: v1.0.1
hooks:
......@@ -25,9 +26,33 @@
entry: bash ./.clang_format.hook -i
language: system
files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|proto)$
- repo: local
hooks:
- id: cpplint-cpp-source
name: cpplint
description: Check C++ code style using cpplint.py.
entry: bash ./tools/codestyle/cpplint_pre_commit.hook
language: system
files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx)$
- repo: local
hooks:
- id: pylint-doc-string
name: pylint
description: Check python docstring style using docstring_checker.
entry: bash ./tools/codestyle/pylint_pre_commit.hook
language: system
files: \.(py)$
- repo: https://github.com/PaddlePaddle/pre-commit-golang
sha: 8337620115c25ff8333f1b1a493bd031049bd7c0
hooks:
- id: go-fmt
types:
- go
- repo: local
hooks:
- id: copyright_checker
name: copyright_checker
entry: python ./.copyright.hook
language: system
files: \.(c|cc|cxx|cpp|cu|h|hpp|hxx|proto|py)$
exclude: (?!.*third_party)^.*$ | (?!.*book)^.*$
.travis.yml
浏览文件 @ dbd25805
......@@ -4,53 +4,38 @@ cache:
- $HOME/.ccache
- $HOME/.cache/pip
- $TRAVIS_BUILD_DIR/build/third_party
- $TRAVIS_BUILD_DIR/build_android/third_party
sudo: required
dist: trusty
services:
- docker
os:
- linux
env:
- JOB=build_doc
- JOB=doc
- JOB=check_style
- JOB=build_android
addons:
apt:
packages:
- gcc-4.8
- g++-4.8
- git
- build-essential
- python
- python-pip
- python2.7-dev
- python-wheel
- libboost-dev
- curl
- swig
- graphviz
- clang-format-3.8
- automake
- libtool
- ccache
ssh_known_hosts: 52.76.173.135
ssh_known_hosts: 13.229.163.131
before_install:
- if [[ "$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
# protobuf version.
- sudo pip install -r $TRAVIS_BUILD_DIR/python/requirements.txt
- sudo pip install wheel sphinx==1.5.6 recommonmark sphinx-rtd-theme==0.1.9 virtualenv pre-commit LinkChecker
# For pylint dockstring checker
- sudo pip install pylint pytest astroid isort
- |
function timeout() { perl -e 'alarm shift; exec @ARGV' "$@"; }
script:
- |
timeout 2580 paddle/scripts/travis/${JOB}.sh # 43min timeout
RESULT=$?; if [ $RESULT -eq 0 ] || [ $RESULT -eq 142 ]; then true ;else exit 1; fi;
# 43min timeout
paddle/scripts/paddle_docker_build.sh ${JOB}
if [ $? -eq 0 ] || [ $? -eq 142 ]; then true; else exit 1; fi;
- |
if [[ "$JOB" != "build_doc" ]]; then exit 0; fi;
if [[ "$JOB" != "doc" ]]; then exit 0; fi;
# For document only
if [[ "$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
curl $DEPLOY_DOCS_SH | bash -s $CONTENT_DEC_PASSWD $TRAVIS_BRANCH $DOCS_DIR $DOCS_DIR/build/doc/
notifications:
email:
on_success: change
......
AUTHORS.md
浏览文件 @ dbd25805
| Github account | name |
|---|---|
| abhinavarora | Abhinav Arora |
| backyes | Yan-Fei Wang |
| baiyfbupt | Yi-Fan Bai |
| beckett1124 | Bin Qi |
| Canpio | Jia-Yi Feng |
| chengxiaohua1105 | Xiao-Hua Cheng |
| cxwangyi, yiwangbaidu, wangkuiyi | Yi Wang |
| cxysteven | Xing-Yi Cheng |
| dzhwinter | Zhi-Hong Dong |
| dragonwarrior | Long Wang |
| dyning | Yuning Du |
| emailweixu | Wei Xu |
| gangliao | Gang Liao |
| gongweibao | Wei-Bao Gong |
......@@ -15,6 +18,9 @@
| hedaoyuan | Dao-Yuan He |
| helinwang | He-Lin Wang |
| jacquesqiao | Long-Fei Qiao |
| jczaja | Jacek Czaja |
| JiayiFeng | Jia-Yi Feng |
| kbinias | Krzysztof Binias |
| kuke | Yi-Bing Liu |
| lcy-seso | Ying Cao |
| lipeng-unisound | Peng Li |
......@@ -23,15 +29,20 @@
| llxxxll | Yong-Feng Liu |
| luotao01 | Tao Luo |
| lzhao4ever | Liang Zhao |
| mozga-intel | Mateusz Ozga |
| NHZlX | Zhao-Long Xing |
| Noplz | Yuan Gao |
| pakchoi | Chuan-Jiang Song |
| panyx0718 | Xin Pan |
| pengli09 | Peng Li |
| pkuyym | Ya-Ming Yang |
| pzelazko-intel | Pawel Zelazko |
| QiJune | Jun Qi |
| qingqing01 | Qing-Qing Dang |
| reyoung | Yang Yu |
| Superjom | Chun-Wei Yan |
| tianbingsz | Tian-Bing Xu |
| tpatejko | Tomasz Patejko |
| typhoonzero | Yi Wu |
| wanghaoshuang | Hao-Shuang Wang |
| wangyang59 | Yang Wang |
......
CMakeLists.txt
浏览文件 @ dbd25805
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
......@@ -16,35 +16,31 @@ 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})
project(paddle CXX C)
message(STATUS "CXX compiler: ${CMAKE_CXX_COMPILER}, version: "
"${CMAKE_CXX_COMPILER_ID} ${CMAKE_CXX_COMPILER_VERSION}")
message(STATUS "C compiler: ${CMAKE_C_COMPILER}, version: "
"${CMAKE_C_COMPILER_ID} ${CMAKE_C_COMPILER_VERSION}")
find_package(Sphinx)
if(NOT CMAKE_CROSSCOMPILING)
find_package(CUDA QUIET)
endif(NOT CMAKE_CROSSCOMPILING)
find_package(Git REQUIRED)
find_package(Threads REQUIRED)
if(NOT ANDROID AND NOT IOS)
find_package(Boost QUIET)
endif()
include(simd)
################################ Configurations #######################################
option(WITH_GPU "Compile PaddlePaddle with NVIDIA GPU" ${CUDA_FOUND})
option(WITH_AMD_GPU "Compile PaddlePaddle with AMD GPU" OFF)
option(WITH_AVX "Compile PaddlePaddle with AVX intrinsics" ${AVX_FOUND})
option(WITH_MKL "Compile PaddlePaddle with MKL support." ${AVX_FOUND})
option(WITH_DSO "Compile PaddlePaddle with dynamic linked CUDA" ON)
option(WITH_TESTING "Compile PaddlePaddle with unit testing" ON)
option(WITH_TESTING "Compile PaddlePaddle with unit testing" OFF)
option(WITH_SWIG_PY "Compile PaddlePaddle with inference api" ON)
option(WITH_STYLE_CHECK "Compile PaddlePaddle with style check" ON)
option(WITH_PYTHON "Compile PaddlePaddle with python interpreter" ON)
option(WITH_DOUBLE "Compile PaddlePaddle with double precision" OFF)
option(WITH_RDMA "Compile PaddlePaddle with RDMA support" OFF)
......@@ -55,12 +51,16 @@ option(WITH_COVERAGE "Compile PaddlePaddle with code coverage" OFF)
option(COVERALLS_UPLOAD "Package code coverage data to coveralls" OFF)
option(ON_TRAVIS "Exclude special unit test on Travis CI" OFF)
option(WITH_C_API "Compile PaddlePaddle with C-API(Prediction)" OFF)
option(WITH_FLUID_ONLY "Compile PaddlePaddle fluid only" OFF)
option(WITH_GOLANG "Compile PaddlePaddle with GOLANG" OFF)
option(GLIDE_INSTALL "Download and install go dependencies " ON)
option(USE_NNPACK "Compile PaddlePaddle with NNPACK library" OFF)
option(WITH_DISTRIBUTE "Compile with grpc distributed support" OFF)
option(USE_EIGEN_FOR_BLAS "Use matrix multiplication in Eigen" OFF)
option(EIGEN_USE_THREADS "Compile with multi-threaded Eigen" OFF)
option(WITH_ARM_FP16 "Use half precision support on armv8.2-a cpu" OFF)
option(WITH_FAST_BUNDLE_TEST "Bundle tests that can be run in a single process together to reduce launch overhead" OFF)
option(WITH_CONTRIB "Compile the third-party contributation" OFF)
# CMAKE_BUILD_TYPE
if(NOT CMAKE_BUILD_TYPE)
......@@ -101,12 +101,19 @@ endif()
set(THIRD_PARTY_PATH "${CMAKE_BINARY_DIR}/third_party" CACHE STRING
"A path setting third party libraries download & build directories.")
set(FLUID_INSTALL_DIR "${CMAKE_BINARY_DIR}/fluid_install_dir" CACHE STRING
"A path setting fluid shared and static libraries")
if (WITH_C_API AND WITH_PYTHON)
message(WARNING "It is suggest not embedded a python interpreter in Paddle "
"when using C-API. It will give an unpredictable behavior when using a "
"different Python interpreter from compiling.")
endif()
if (WITH_C_API)
set(WITH_FLUID_ONLY OFF CACHE STRING "Disable install fluid when compile the C_API" FORCE)
endif()
if(MOBILE_INFERENCE)
set(THIRD_PARTY_BUILD_TYPE MinSizeRel)
else()
......@@ -114,13 +121,14 @@ else()
endif()
set(WITH_MKLML ${WITH_MKL})
if (WITH_MKL AND AVX2_FOUND)
set(WITH_MKLDNN ON)
else()
message(STATUS "Do not have AVX2 intrinsics and disabled MKL-DNN")
set(WITH_MKLDNN OFF)
if (NOT DEFINED WITH_MKLDNN)
if (WITH_MKL AND AVX2_FOUND)
set(WITH_MKLDNN ON)
else()
message(STATUS "Do not have AVX2 intrinsics and disabled MKL-DNN")
set(WITH_MKLDNN OFF)
endif()
endif()
########################################################################################
include(external/mklml) # download mklml package
......@@ -134,45 +142,54 @@ include(external/openblas) # download, build, install openblas
include(external/mkldnn) # download, build, install mkldnn
include(external/swig) # download, build, install swig
include(external/warpctc) # download, build, install warpctc
include(external/boost) # download boost
include(external/any) # download libn::any
include(external/eigen) # download eigen3
include(external/pybind11) # download pybind11
include(external/nccl)
include(external/cares)
include(external/grpc)
include(external/snappy) # download snappy
include(external/snappystream)
include(external/threadpool)
include(cudnn) # set cudnn libraries, must before configure
include(cupti)
include(configure) # add paddle env configuration
include(generic) # simplify cmake module
include(package) # set paddle packages
include(cpplint) # set paddle c++ style
include(ccache) # set ccache for compilation
include(util) # set unittest and link libs
include(rdma) # set rdma libraries
include(flags) # set paddle compile flags
include(version) # set PADDLE_VERSION
include(coveralls) # set code coverage
include(inference_lib) # add paddle fluid inference libraries
include_directories("${PADDLE_SOURCE_DIR}")
include_directories("${PADDLE_SOURCE_DIR}/paddle/cuda/include")
include_directories("${CMAKE_CURRENT_BINARY_DIR}/proto")
include_directories("${CMAKE_CURRENT_BINARY_DIR}/go/pserver/client/c")
include_directories(${Boost_INCLUDE_DIRS})
set(EXTERNAL_LIBS
${GFLAGS_LIBRARIES}
${GLOG_LIBRARIES}
gflags
glog
${CBLAS_LIBRARIES}
${PROTOBUF_LIBRARY}
${ZLIB_LIBRARIES}
protobuf
zlib
${PYTHON_LIBRARIES}
)
if(WITH_GPU)
include(cuda)
include(cuda)
include(tensorrt)
endif(WITH_GPU)
if(WITH_AMD_GPU)
find_package(HIP)
include(hip)
endif(WITH_AMD_GPU)
if(WITH_MKLML)
list(APPEND EXTERNAL_LIBS ${MKLML_IOMP_LIB})
endif()
......@@ -188,7 +205,7 @@ endif(USE_NNPACK)
add_subdirectory(proto)
if(NOT MOBILE_INFERENCE)
if(NOT MOBILE_INFERENCE AND NOT WITH_FLUID_ONLY)
# "add_subdirectory(go)" should be placed after the following loine,
# because it depends on paddle/optimizer.
add_subdirectory(paddle/optimizer)
......@@ -197,15 +214,26 @@ endif()
# "add_subdirectory(paddle)" and "add_subdirectory(python)" should be
# placed after this block, because they depends on it.
if(WITH_GOLANG)
enable_language(Go)
add_subdirectory(go)
endif(WITH_GOLANG)
set(PADDLE_PYTHON_BUILD_DIR "${CMAKE_CURRENT_BINARY_DIR}/python/build")
set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
set(CMAKE_C_FLAGS_RELWITHDEBINFO "-O3 -g -DNDEBUG")
add_subdirectory(paddle)
if(WITH_PYTHON)
add_subdirectory(python)
add_subdirectory(python)
endif()
if(WITH_DOC)
find_package(Sphinx REQUIRED)
find_python_module(recommonmark REQUIRED)
add_subdirectory(doc)
endif()
if (WITH_CONTRIB)
add_subdirectory(paddle/contrib)
endif()
CODE_OF_CONDUCT.md 0 → 100644
浏览文件 @ dbd25805
# Contributor Covenant Code of Conduct
## Our Pledge
In the interest of fostering an open and welcoming environment, we as contributors and maintainers pledge to making participation in our project and our community a harassment-free experience for everyone, regardless of age, body size, disability, ethnicity, gender identity and expression, level of experience, nationality, personal appearance, race, religion, or sexual identity and orientation.
## Our Standards
Examples of behavior that contributes to creating a positive environment include:
* Using welcoming and inclusive language
* Being respectful of differing viewpoints and experiences
* Gracefully accepting constructive criticism
* Focusing on what is best for the community
* Showing empathy towards other community members
Examples of unacceptable behavior by participants include:
* The use of sexualized language or imagery and unwelcome sexual attention or advances
* Trolling, insulting/derogatory comments, and personal or political attacks
* Public or private harassment
* Publishing others' private information, such as a physical or electronic address, without explicit permission
* Other conduct which could reasonably be considered inappropriate in a professional setting
## Our Responsibilities
Project maintainers are responsible for clarifying the standards of acceptable behavior and are expected to take appropriate and fair corrective action in response to any instances of unacceptable behavior.
Project maintainers have the right and responsibility to remove, edit, or reject comments, commits, code, wiki edits, issues, and other contributions that are not aligned to this Code of Conduct, or to ban temporarily or permanently any contributor for other behaviors that they deem inappropriate, threatening, offensive, or harmful.
## Scope
This Code of Conduct applies both within project spaces and in public spaces when an individual is representing the project or its community. Examples of representing a project or community include using an official project e-mail address, posting via an official social media account, or acting as an appointed representative at an online or offline event. Representation of a project may be further defined and clarified by project maintainers.
## Enforcement
Instances of abusive, harassing, or otherwise unacceptable behavior may be reported by contacting the project team at paddle-dev@baidu.com. The project team will review and investigate all complaints, and will respond in a way that it deems appropriate to the circumstances. The project team is obligated to maintain confidentiality with regard to the reporter of an incident. Further details of specific enforcement policies may be posted separately.
Project maintainers who do not follow or enforce the Code of Conduct in good faith may face temporary or permanent repercussions as determined by other members of the project's leadership.
## Attribution
This Code of Conduct is adapted from the [Contributor Covenant][homepage], version 1.4, available at [http://contributor-covenant.org/version/1/4][version]
[homepage]: http://contributor-covenant.org
[version]: http://contributor-covenant.org/version/1/4/
CODE_OF_CONDUCT_cn.md 0 → 100644
浏览文件 @ dbd25805
# 参与者公约
## 我们的保证
为了促进一个开放透明且友好的环境，我们作为贡献者和维护者保证：无论年龄、种族、民族、性别认同和表达（方式）、体型、身体健全与否、经验水平、国籍、个人表现、宗教或性别取向，参与者在我们项目和社区中都免于骚扰。
## 我们的标准
有助于创造正面环境的行为包括但不限于：
* 使用友好和包容性语言
* 尊重不同的观点和经历
* 耐心地接受建设性批评
* 关注对社区最有利的事情
* 友善对待其他社区成员
身为参与者不能接受的行为包括但不限于：
* 使用与性有关的言语或是图像，以及不受欢迎的性骚扰
* 捣乱/煽动/造谣的行为或进行侮辱/贬损的评论，人身攻击及政治攻击
* 公开或私下的骚扰
* 未经许可地发布他人的个人资料，例如住址或是电子地址
* 其他可以被合理地认定为不恰当或者违反职业操守的行为
## 我们的责任
项目维护者有责任为「可接受的行为」标准做出诠释，以及对已发生的不被接受的行为采取恰当且公平的纠正措施。
项目维护者有权利及责任去删除、编辑、拒绝与本行为标准有所违背的评论(comments)、提交(commits)、代码、wiki 编辑、问题(issues)和其他贡献，以及项目维护者可暂时或永久性的禁止任何他们认为有不适当、威胁、冒犯、有害行为的贡献者。
## 使用范围
当一个人代表该项目或是其社区时，本行为标准适用于其项目平台和公共平台。
代表项目或是社区的情况，举例来说包括使用官方项目的电子邮件地址、通过官方的社区媒体账号发布或线上或线下事件中担任指定代表。
该项目的呈现方式可由其项目维护者进行进一步的定义及解释。
## 强制执行
可以通过paddle-dev@baidu.com，来联系项目团队来举报滥用、骚扰或其他不被接受的行为。
任何维护团队认为有必要且适合的所有投诉都将进行审查及调查，并做出相对应的回应。项目小组有对事件回报者有保密的义务。具体执行的方针近一步细节可能会单独公布。
没有切实地遵守或是执行本行为标准的项目维护人员，可能会因项目领导人或是其他成员的决定，暂时或是永久地取消其参与资格。
## 来源
本行为标准改编自[贡献者公约][主页]，版本 1.4
可在此观看https://www.contributor-covenant.org/zh-cn/version/1/4/code-of-conduct.html
[主页]: https://www.contributor-covenant.org
CONTRIBUTING.md
浏览文件 @ dbd25805
# Contribute Code
You are welcome to contribute to project PaddlePaddle. To contribute to PaddlePaddle, you have to agree with the
[PaddlePaddle Contributor License Agreement](https://gist.github.com/wangkuiyi/0c22c7b1bd3bb7eb27d76f85c3a3e329).
We sincerely appreciate your contribution. This document explains our workflow and work style.
## Workflow
......@@ -55,6 +58,8 @@ PaddlePaddle uses this [Git branching model](http://nvie.com/posts/a-successful-
create mode 100644 233
```
NOTE: The `yapf` installed by `pip install pre-commit` and `conda install -c conda-forge pre-commit` is slightly different. Paddle developers use `pip install pre-commit`.
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).
......
Dockerfile
浏览文件 @ dbd25805
# A image for building paddle binaries
# Use cuda devel base image for both cpu and gpu environment
FROM nvidia/cuda:8.0-cudnn5-devel-ubuntu16.04
# When you modify it, please be aware of cudnn-runtime version
# and libcudnn.so.x in paddle/scripts/docker/build.sh
FROM nvidia/cuda:8.0-cudnn7-devel-ubuntu16.04
MAINTAINER PaddlePaddle Authors <paddle-dev@baidu.com>
ARG UBUNTU_MIRROR
......@@ -21,15 +23,16 @@ ENV HOME /root
COPY ./paddle/scripts/docker/root/ /root/
RUN apt-get update && \
apt-get install -y \
git python-pip python-dev openssh-server bison libnccl-dev \
apt-get install -y --allow-downgrades \
git python-pip python-dev openssh-server bison \
libnccl2=2.1.2-1+cuda8.0 libnccl-dev=2.1.2-1+cuda8.0 \
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 \
automake locales clang-format swig cmake \
liblapack-dev liblapacke-dev \
clang-3.8 llvm-3.8 libclang-3.8-dev \
net-tools libtool && \
net-tools libtool ccache && \
apt-get clean -y
# Install Go and glide
......@@ -44,6 +47,17 @@ ENV PATH=${PATH}:${GOROOT}/bin:${GOPATH}/bin
# install glide
RUN curl -s -q https://glide.sh/get | sh
# Install TensorRT
# following TensorRT.tar.gz is not the default official one, we do two miny changes:
# 1. Remove the unnecessary files to make the library small. TensorRT.tar.gz only contains include and lib now,
# and its size is only one-third of the official one.
# 2. Manually add ~IPluginFactory() in IPluginFactory class of NvInfer.h, otherwise, it couldn't work in paddle.
# See https://github.com/PaddlePaddle/Paddle/issues/10129 for details.
RUN wget -qO- http://paddlepaddledeps.bj.bcebos.com/TensorRT-4.0.0.3.Ubuntu-16.04.4.x86_64-gnu.cuda-8.0.cudnn7.0.tar.gz | \
tar -xz -C /usr/local && \
cp -rf /usr/local/TensorRT/include /usr && \
cp -rf /usr/local/TensorRT/lib /usr
# git credential to skip password typing
RUN git config --global credential.helper store
......@@ -52,15 +66,22 @@ RUN localedef -i en_US -f UTF-8 en_US.UTF-8
# FIXME: due to temporary ipykernel dependency issue, specify ipykernel jupyter
# version util jupyter fixes this issue.
RUN pip install --upgrade pip && \
# specify sphinx version as 1.5.6 and remove -U option for [pip install -U
# sphinx-rtd-theme] since -U option will cause sphinx being updated to newest
# version(1.7.1 for now), which causes building documentation failed.
RUN easy_install -U pip && \
pip install -U wheel && \
pip install -U docopt PyYAML sphinx && \
pip install -U sphinx-rtd-theme==0.1.9 recommonmark
pip install -U docopt PyYAML sphinx==1.5.6 && \
pip install sphinx-rtd-theme==0.1.9 recommonmark
RUN pip install pre-commit 'ipython==5.3.0' && \
pip install 'ipykernel==4.6.0' 'jupyter==1.0.0' && \
pip install opencv-python
#For docstring checker
RUN pip install pylint pytest astroid isort
COPY ./python/requirements.txt /root/
RUN pip install -r /root/requirements.txt
......@@ -83,6 +104,3 @@ RUN echo 'root:root' | chpasswd
RUN sed -ri 's/^PermitRootLogin\s+.*/PermitRootLogin yes/' /etc/ssh/sshd_config
RUN sed -ri 's/UsePAM yes/#UsePAM yes/g' /etc/ssh/sshd_config
EXPOSE 22
# development image default do build work
CMD ["bash", "/paddle/paddle/scripts/docker/build.sh"]
Dockerfile.android
浏览文件 @ dbd25805
......@@ -21,23 +21,13 @@ RUN apt-get update && \
wget curl tar unzip gcc g++ locales clang-format-3.8 swig cmake && \
apt-get clean -y
# Install Go and glide
RUN wget -qO- go.tgz https://storage.googleapis.com/golang/go1.8.1.linux-amd64.tar.gz | \
tar -xz -C /usr/local && \
mkdir /root/gopath && \
mkdir /root/gopath/bin && \
mkdir /root/gopath/src
ENV GOROOT=/usr/local/go GOPATH=/root/gopath
# should not be in the same line with GOROOT definition, otherwise docker build could not find GOROOT.
ENV PATH=${PATH}:${GOROOT}/bin:${GOPATH}/bin
# git credential to skip password typing
RUN git config --global credential.helper store
# Fix locales to en_US.UTF-8
RUN localedef -i en_US -f UTF-8 en_US.UTF-8
RUN pip install --upgrade pip && \
RUN pip install --upgrade pip==9.0.3 && \
pip install -U 'protobuf==3.1.0' && \
pip install -U wheel sphinx && \
pip install pre-commit
......@@ -50,5 +40,3 @@ RUN mkdir -p ${ANDROID_TOOLCHAINS_DIR} && \
unzip -q android-ndk-r14b-linux-x86_64.zip && \
mv android-ndk-r14b ${ANDROID_NDK_HOME} && \
rm -rf /opt/android-ndk-tmp
CMD ["bash", "/paddle/paddle/scripts/docker/build_android.sh"]
LICENSE
浏览文件 @ dbd25805
......@@ -188,7 +188,7 @@ Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.
Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
......
README.md
浏览文件 @ dbd25805
......@@ -2,8 +2,8 @@
[![Build Status](https://travis-ci.org/PaddlePaddle/Paddle.svg?branch=develop)](https://travis-ci.org/PaddlePaddle/Paddle)
[![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://doc.paddlepaddle.org/develop/doc/)
[![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://doc.paddlepaddle.org/develop/doc_cn/)
[![Documentation Status](https://img.shields.io/badge/docs-latest-brightgreen.svg?style=flat)](http://www.paddlepaddle.org/docs/develop/documentation/en/getstarted/index_en.html)
[![Documentation Status](https://img.shields.io/badge/中文文档-最新-brightgreen.svg)](http://www.paddlepaddle.org/docs/develop/documentation/zh/getstarted/index_cn.html)
[![Coverage Status](https://coveralls.io/repos/github/PaddlePaddle/Paddle/badge.svg?branch=develop)](https://coveralls.io/github/PaddlePaddle/Paddle?branch=develop)
[![Release](https://img.shields.io/github/release/PaddlePaddle/Paddle.svg)](https://github.com/PaddlePaddle/Paddle/releases)
[![License](https://img.shields.io/badge/license-Apache%202-blue.svg)](LICENSE)
......@@ -36,7 +36,8 @@ Please refer to our [release announcement](https://github.com/PaddlePaddle/Paddl
examples:
- Optimized math operations through SSE/AVX intrinsics, BLAS libraries
(e.g. MKL, ATLAS, cuBLAS) or customized CPU/GPU kernels.
(e.g. MKL, OpenBLAS, cuBLAS) or customized CPU/GPU kernels.
- Optimized CNN networks through MKL-DNN library.
- Highly optimized recurrent networks which can handle **variable-length**
sequence without padding.
- Optimized local and distributed training for models with high dimensional
......@@ -61,32 +62,32 @@ Please refer to our [release announcement](https://github.com/PaddlePaddle/Paddl
## Installation
It is recommended to check out the
[Docker installation guide](http://doc.paddlepaddle.org/develop/doc/getstarted/build_and_install/docker_install_en.html)
[Docker installation guide](http://www.paddlepaddle.org/docs/develop/documentation/fluid/en/build_and_install/docker_install_en.html)
before looking into the
[build from source guide](http://doc.paddlepaddle.org/develop/doc/getstarted/build_and_install/build_from_source_en.html).
[build from source guide](http://www.paddlepaddle.org/docs/develop/documentation/fluid/en/build_and_install/build_from_source_en.html).
## Documentation
We provide [English](http://doc.paddlepaddle.org/develop/doc/) and
[Chinese](http://doc.paddlepaddle.org/doc_cn/) documentation.
We provide [English](http://www.paddlepaddle.org/docs/develop/documentation/en/getstarted/index_en.html) and
[Chinese](http://www.paddlepaddle.org/docs/develop/documentation/zh/getstarted/index_cn.html) documentation.
- [Deep Learning 101](http://book.paddlepaddle.org/index.html)
- [Deep Learning 101](http://www.paddlepaddle.org/docs/develop/book/01.fit_a_line/index.html)
You might want to start from this online interactive book that can run in a Jupyter Notebook.
- [Distributed Training](http://doc.paddlepaddle.org/develop/doc/howto/usage/cluster/cluster_train_en.html)
- [Distributed Training](http://www.paddlepaddle.org/docs/develop/documentation/en/howto/cluster/index_en.html)
You can run distributed training jobs on MPI clusters.
- [Distributed Training on Kubernetes](http://doc.paddlepaddle.org/develop/doc/howto/usage/k8s/k8s_en.html)
- [Distributed Training on Kubernetes](http://www.paddlepaddle.org/docs/develop/documentation/en/howto/cluster/multi_cluster/k8s_en.html)
You can also run distributed training jobs on Kubernetes clusters.
- [Python API](http://doc.paddlepaddle.org/develop/doc/api/index_en.html)
- [Python API](http://www.paddlepaddle.org/docs/develop/api/en/overview.html)
Our new API enables much shorter programs.
- [How to Contribute](http://doc.paddlepaddle.org/develop/doc/howto/dev/contribute_to_paddle_en.html)
- [How to Contribute](http://www.paddlepaddle.org/docs/develop/documentation/fluid/en/dev/contribute_to_paddle_en.html)
We appreciate your contributions!
......
benchmark/IntelOptimizedPaddle.md
浏览文件 @ dbd25805
......@@ -7,11 +7,11 @@ Machine:
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)
PaddlePaddle:
- paddlepaddle/paddle:0.11.0 (for MKLML and MKL-DNN)
- MKL-DNN tag v0.11
- MKLML 2018.0.1.20171007
- paddlepaddle/paddle:latest-openblas (for OpenBLAS)
- paddlepaddle/paddle:0.11.0-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.
......@@ -19,7 +19,10 @@ On each machine, we will test and compare the performance of training on single
## Benchmark Model
### Server
#### Training
Test on batch size 64, 128, 256 on Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Pay attetion that the speed below includes forward, backward and parameter update time. So we can not directly compare the data with the benchmark of caffe `time` [command](https://github.com/PaddlePaddle/Paddle/blob/develop/benchmark/caffe/image/run.sh#L9), which only contain forward and backward. The updating time of parameter would become very heavy when the weight size are large, especially on alexnet.
Input image size - 3 * 224 * 224, Time: images/second
......@@ -53,5 +56,57 @@ Input image size - 3 * 224 * 224, Time: images/second
<img src="figs/googlenet-cpu-train.png" width="500">
- AlexNet
| BatchSize | 64 | 128 | 256 |
|--------------|--------| ------ | -------|
| OpenBLAS | 45.62 | 72.79 | 107.22 |
| MKLML | 66.37 | 105.60 | 144.04 |
| MKL-DNN | 399.00 | 498.94 | 626.53 |
<img src="figs/alexnet-cpu-train.png" width="500">
#### Inference
Test on batch size 1, 2, 4, 8, 16 on Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
- VGG-19
| BatchSize | 1 | 2 | 4 | 8 | 16 |
|-----------|-------|-------|-------|-------|-------|
| OpenBLAS | 1.10 | 1.96 | 3.62 | 3.63 | 2.25 |
| MKLML | 5.58 | 9.80 | 15.15 | 21.21 | 28.67 |
| MKL-DNN | 75.07 | 88.64 | 82.58 | 92.29 | 96.75 |
<img src="figs/vgg-cpu-infer.png" width="500">
- ResNet-50
| BatchSize | 1 | 2 | 4 | 8 | 16 |
|-----------|-------|--------|--------|--------|--------|
| OpenBLAS | 3.31 | 6.72 | 11.59 | 13.17 | 9.27 |
| MKLML | 6.33 | 12.02 | 22.88 | 40.53 | 63.09 |
| MKL-DNN | 107.83| 148.84 | 177.78 | 189.35 | 217.69 |
<img src="figs/resnet-cpu-infer.png" width="500">
- GoogLeNet
| BatchSize | 1 | 2 | 4 | 8 | 16 |
|-----------|--------|--------|--------|--------|--------|
| OpenBLAS | 12.06 | 23.56 | 34.48 | 36.45 | 23.12 |
| MKLML | 22.74 | 41.56 | 81.22 | 133.47 | 210.53 |
| MKL-DNN | 175.10 | 272.92 | 450.70 | 512.00 | 600.94 |
<img src="figs/googlenet-cpu-infer.png" width="500">
- AlexNet
| BatchSize | 1 | 2 | 4 | 8 | 16 |
|-----------|--------|--------|--------|--------|--------|
| OpenBLAS | 3.53 | 6.23 | 15.04 | 26.06 | 31.62 |
| MKLML | 21.32 | 36.55 | 73.06 | 131.15 | 192.77 |
| MKL-DNN | 442.91 | 656.41 | 719.10 | 847.68 | 850.51 |
<img src="figs/alexnet-cpu-infer.png" width="500">
### Laptop
TBD
benchmark/figs/resnet-cpu-train.png
查看替换文件 @ 1ff47b28 浏览文件 @ dbd25805
benchmark/figs/resnet-cpu-train.png

19.8 KB | W: | H:

benchmark/figs/resnet-cpu-train.png

17.6 KB | W: | H:

benchmark/figs/resnet-cpu-train.png
benchmark/figs/resnet-cpu-train.png
benchmark/figs/resnet-cpu-train.png
benchmark/figs/resnet-cpu-train.png
  • 2-up
  • Swipe
  • Onion skin
benchmark/figs/vgg-cpu-train.png
查看替换文件 @ 1ff47b28 浏览文件 @ dbd25805
benchmark/figs/vgg-cpu-train.png

17.9 KB | W: | H:

benchmark/figs/vgg-cpu-train.png

16.7 KB | W: | H:

benchmark/figs/vgg-cpu-train.png
benchmark/figs/vgg-cpu-train.png
benchmark/figs/vgg-cpu-train.png
benchmark/figs/vgg-cpu-train.png
  • 2-up
  • Swipe
  • Onion skin
benchmark/fluid/README.md 0 → 100644
浏览文件 @ dbd25805
# Fluid Benchmark
This directory contains several models configurations and tools that used to run
Fluid benchmarks for local and distributed training.
## Run the Benchmark
To start, run the following command to get the full help message:
```bash
python fluid_benchmark.py --help
```
Currently supported `--model` argument include:
* mnist
* resnet
* you can chose to use different dataset using `--data_set cifar10` or
`--data_set flowers`.
* vgg
* stacked_dynamic_lstm
* machine_translation
* Run the following command to start a benchmark job locally:
```bash
python fluid_benchmark.py --model mnist --device GPU
```
You can choose to use GPU/CPU training. With GPU training, you can specify
`--gpus <gpu_num>` to run multi GPU training.
* Run distributed training with parameter servers:
* start parameter servers:
```bash
PADDLE_TRAINING_ROLE=PSERVER PADDLE_PSERVER_PORT=7164 PADDLE_PSERVER_IPS=127.0.0.1 PADDLE_TRAINERS=1 PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist --device GPU --update_method pserver
```
* start trainers:
```bash
PADDLE_TRAINING_ROLE=TRAINER PADDLE_PSERVER_PORT=7164 PADDLE_PSERVER_IPS=127.0.0.1 PADDLE_TRAINERS=1 PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist --device GPU --update_method pserver
```
* Run distributed training using NCCL2
```bash
PADDLE_PSERVER_PORT=7164 PADDLE_TRAINER_IPS=192.168.0.2,192.168.0.3 PADDLE_CURRENT_IP=127.0.0.1 PADDLE_TRAINER_ID=0 python fluid_benchmark.py --model mnist --device GPU --update_method nccl2
```
## Run Distributed Benchmark on Kubernetes Cluster
We provide a script `kube_gen_job.py` to generate Kubernetes yaml files to submit
distributed benchmark jobs to your cluster. To generate a job yaml, just run:
```bash
python kube_gen_job.py --jobname myjob --pscpu 4 --cpu 8 --gpu 8 --psmemory 20 --memory 40 --pservers 4 --trainers 4 --entry "python fluid_benchmark.py --model mnist --parallel 1 --device GPU --update_method pserver " --disttype pserver
```
Then the yaml files are generated under directory `myjob`, you can run:
```bash
kubectl create -f myjob/
```
The job shall start.
## Notes for Run Fluid Distributed with NCCL2 and RDMA
Before running NCCL2 distributed jobs, please check that whether your node has multiple network
interfaces, try to add the environment variable `export NCCL_SOCKET_IFNAME=eth0` to use your actual
network device.
To run high-performance distributed training, you must prepare your hardware environment to be
able to run RDMA enabled network communication, please check out [this](https://github.com/PaddlePaddle/Paddle/blob/develop/doc/fluid/howto/cluster/nccl2_rdma_training.md)
note for details.
benchmark/fluid/fluid_benchmark.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import argparse
import cProfile
import time
import os
import numpy as np
import paddle.fluid as fluid
import paddle.fluid.core as core
import paddle.fluid.profiler as profiler
import paddle.fluid.transpiler.distribute_transpiler as distribute_transpiler
BENCHMARK_MODELS = [
"machine_translation", "resnet", "vgg", "mnist", "stacked_dynamic_lstm"
]
def parse_args():
parser = argparse.ArgumentParser('Fluid model benchmarks.')
parser.add_argument(
'--model',
type=str,
choices=BENCHMARK_MODELS,
default='resnet',
help='The model to run benchmark with.')
parser.add_argument(
'--batch_size', type=int, default=32, help='The minibatch size.')
parser.add_argument(
'--learning_rate',
type=float,
default=0.001,
help='The minibatch size.')
# TODO(wuyi): add "--use_fake_data" option back.
parser.add_argument(
'--skip_batch_num',
type=int,
default=5,
help='The first num of minibatch num to skip, for better performance test'
)
parser.add_argument(
'--iterations', type=int, default=80, help='The number of minibatches.')
parser.add_argument(
'--pass_num', type=int, default=100, help='The number of passes.')
parser.add_argument(
'--data_format',
type=str,
default='NCHW',
choices=['NCHW', 'NHWC'],
help='The data data_format, now only support NCHW.')
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help='The device type.')
parser.add_argument(
'--gpus',
type=int,
default=1,
help='If gpus > 1, will use ParallelExecutor to run, else use Executor.')
parser.add_argument(
'--data_set',
type=str,
default='flowers',
choices=['cifar10', 'flowers'],
help='Optional dataset for benchmark.')
parser.add_argument(
'--infer_only', action='store_true', help='If set, run forward only.')
parser.add_argument(
'--use_cprof', action='store_true', help='If set, use cProfile.')
parser.add_argument(
'--use_nvprof',
action='store_true',
help='If set, use nvprof for CUDA.')
parser.add_argument(
'--no_test',
action='store_false',
help='If set, test the testset during training.')
parser.add_argument(
'--memory_optimize',
action='store_true',
help='If set, optimize runtime memory before start.')
parser.add_argument(
'--use_fake_data',
action='store_true',
help='If set ommit the actual read data operators.')
parser.add_argument(
'--profile', action='store_true', help='If set, profile a few steps.')
parser.add_argument(
'--update_method',
type=str,
default='local',
choices=['local', 'pserver', 'nccl2'],
help='Choose parameter update method, can be local, pserver, nccl2.')
args = parser.parse_args()
return args
def append_nccl2_prepare(trainer_id):
if trainer_id >= 0:
# append gen_nccl_id at the end of startup program
trainer_id = int(os.getenv("PADDLE_TRAINER_ID"))
port = os.getenv("PADDLE_PSERVER_PORT")
worker_ips = os.getenv("PADDLE_TRAINER_IPS")
worker_endpoints = []
for ip in worker_ips.split(","):
worker_endpoints.append(':'.join([ip, port]))
num_trainers = len(worker_endpoints)
current_endpoint = os.getenv("PADDLE_CURRENT_IP") + ":" + port
worker_endpoints.remove(current_endpoint)
nccl_id_var = fluid.default_startup_program().global_block().create_var(
name="NCCLID",
persistable=True,
type=fluid.core.VarDesc.VarType.RAW)
fluid.default_startup_program().global_block().append_op(
type="gen_nccl_id",
inputs={},
outputs={"NCCLID": nccl_id_var},
attrs={
"endpoint": current_endpoint,
"endpoint_list": worker_endpoints,
"trainer_id": trainer_id
})
return nccl_id_var, num_trainers, trainer_id
else:
raise Exception("must set positive PADDLE_TRAINER_ID env variables for "
"nccl-based dist train.")
def dist_transpile(trainer_id):
if trainer_id < 0:
return None, None
# the port of all pservers, needed by both trainer and pserver
port = os.getenv("PADDLE_PSERVER_PORT", "6174")
# comma separated ips of all pservers, needed by trainer and
# pserver
pserver_ips = os.getenv("PADDLE_PSERVER_IPS", "")
eplist = []
for ip in pserver_ips.split(","):
eplist.append(':'.join([ip, port]))
pserver_endpoints = ",".join(eplist)
# total number of workers/trainers in the job, needed by
# trainer and pserver
trainers = int(os.getenv("PADDLE_TRAINERS"))
# the IP of the local machine, needed by pserver only
current_endpoint = os.getenv("PADDLE_CURRENT_IP", "") + ":" + port
# the role, should be either PSERVER or TRAINER
training_role = os.getenv("PADDLE_TRAINING_ROLE")
t = distribute_transpiler.DistributeTranspiler()
t.transpile(trainer_id, pservers=pserver_endpoints, trainers=trainers)
if training_role == "PSERVER":
pserver_program = t.get_pserver_program(current_endpoint)
pserver_startup_program = t.get_startup_program(current_endpoint,
pserver_program)
return pserver_program, pserver_startup_program
elif training_role == "TRAINER":
train_program = t.get_trainer_program()
return train_program, fluid.default_startup_program()
else:
raise ValueError(
'TRAINING_ROLE environment variable must be either TRAINER or PSERVER'
)
def test(exe, inference_program, test_reader, feeder, batch_acc):
accuracy_evaluator = fluid.metrics.Accuracy()
for batch_id, data in enumerate(test_reader()):
acc = exe.run(inference_program,
feed=feeder.feed(data),
fetch_list=[batch_acc])
accuracy_evaluator.update(value=np.array(acc), weight=len(data))
return accuracy_evaluator.eval()
# TODO(wuyi): replace train, train_parallel, test functions with new trainer
# API once it is ready.
def train(avg_loss, infer_prog, optimizer, train_reader, test_reader, batch_acc,
args, train_prog, startup_prog):
if os.getenv("PADDLE_TRAINING_ROLE") == "PSERVER":
place = core.CPUPlace()
exe = fluid.Executor(place)
exe.run(startup_prog)
exe.run(train_prog)
return
if args.use_fake_data:
raise Exception(
"fake data is not supported in single GPU test for now.")
place = core.CPUPlace() if args.device == 'CPU' else core.CUDAPlace(0)
exe = fluid.Executor(place)
exe.run(startup_prog)
feed_var_list = [
var for var in train_prog.global_block().vars.itervalues()
if var.is_data
]
feeder = fluid.DataFeeder(feed_var_list, place)
iters, num_samples, start_time = 0, 0, time.time()
for pass_id in range(args.pass_num):
train_losses = []
for batch_id, data in enumerate(train_reader()):
if iters == args.skip_batch_num:
start_time = time.time()
num_samples = 0
if iters == args.iterations:
break
loss = exe.run(train_prog,
feed=feeder.feed(data),
fetch_list=[avg_loss])
iters += 1
num_samples += len(data)
train_losses.append(loss)
print("Pass: %d, Iter: %d, Loss: %f\n" %
(pass_id, iters, np.mean(train_losses)))
train_elapsed = time.time() - start_time
examples_per_sec = num_samples / train_elapsed
print('\nTotal examples: %d, total time: %.5f, %.5f examples/sec\n' %
(num_samples, train_elapsed, examples_per_sec))
print("Pass: %d, Loss: %f" % (pass_id, np.mean(train_losses)))
# evaluation
if not args.no_test and batch_acc != None:
pass_test_acc = test(exe, infer_prog, test_reader, feeder,
batch_acc)
print(", Test Accuracy: %f" % pass_test_acc)
print("\n")
# TODO(wuyi): add warmup passes to get better perf data.
exit(0)
# TODO(wuyi): replace train, train_parallel, test functions with new trainer
# API once it is ready.
def train_parallel(avg_loss, infer_prog, optimizer, train_reader, test_reader,
batch_acc, args, train_prog, startup_prog, nccl_id_var,
num_trainers, trainer_id):
feed_var_list = [
var for var in train_prog.global_block().vars.itervalues()
if var.is_data
]
# generate fake:
if args.use_fake_data:
for var in feed_var_list:
v = startup_prog.global_block().clone_variable(var)
var.persistable = True
v.persistable = True
real_shape = list(var.shape)
real_shape[0] = args.batch_size / args.gpus
startup_prog.global_block().append_op(
outputs={"Out": v},
type="fill_constant",
attrs={"shape": real_shape,
"value": 1.0,
"dtype": var.dtype})
place = core.CPUPlace() if args.device == 'CPU' else core.CUDAPlace(0)
if nccl_id_var and trainer_id == 0:
#FIXME(wuyi): wait other trainer to start listening
time.sleep(30)
startup_exe = fluid.Executor(place)
startup_exe.run(startup_prog)
strategy = fluid.ExecutionStrategy()
strategy.num_threads = 1
strategy.allow_op_delay = False
exe = fluid.ParallelExecutor(
True,
avg_loss.name,
exec_strategy=strategy,
num_trainers=num_trainers,
trainer_id=trainer_id)
feeder = fluid.DataFeeder(feed_var_list, place)
for pass_id in range(args.pass_num):
num_samples = 0
iters = 0
start_time = time.time()
for batch_id, data in enumerate(train_reader()):
if args.profile and pass_id == 0 and batch_id == 5:
profiler.start_profiler("All")
elif args.profile and pass_id == 0 and batch_id == 10:
profiler.stop_profiler("total", "/tmp/profile_%d" % trainer_id)
if iters == args.skip_batch_num:
start_time = time.time()
num_samples = 0
if iters == args.iterations:
break
if args.use_fake_data:
loss, = exe.run([avg_loss.name])
else:
loss, = exe.run([avg_loss.name], feed=feeder.feed(data))
if args.update_method == "pserver":
exe.bcast_params()
num_samples += len(data)
iters += 1
if batch_id % 1 == 0:
print("Pass %d, batch %d, loss %s" %
(pass_id, batch_id, np.array(loss)))
train_elapsed = time.time() - start_time
examples_per_sec = num_samples / train_elapsed
print('\nTotal examples: %d, total time: %.5f, %.5f examples/sed\n' %
(num_samples, train_elapsed, examples_per_sec))
if not args.no_test and batch_acc != None:
test_acc = test(startup_exe, infer_prog, test_reader, feeder,
batch_acc)
print("Pass: %d, Test Accuracy: %f\n" % (pass_id, test_acc))
exit(0)
def print_arguments(args):
vars(args)['use_nvprof'] = (vars(args)['use_nvprof'] and
vars(args)['device'] == 'GPU')
print('----------- resnet Configuration Arguments -----------')
for arg, value in sorted(vars(args).iteritems()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def main():
args = parse_args()
print_arguments(args)
# the unique trainer id, starting from 0, needed by trainer
# only
nccl_id_var, num_trainers, trainer_id = (
None, 1, int(os.getenv("PADDLE_TRAINER_ID", "-1")))
if args.use_cprof:
pr = cProfile.Profile()
pr.enable()
model_def = __import__("models.%s" % args.model, fromlist=["models"])
train_args = list(model_def.get_model(args))
train_args.append(args)
# Run optimizer.minimize(avg_loss)
train_args[2].minimize(train_args[0])
if args.memory_optimize:
fluid.memory_optimize(fluid.default_main_program())
if args.update_method == "pserver":
train_prog, startup_prog = dist_transpile(trainer_id)
if not train_prog:
raise Exception(
"Must configure correct environments to run dist train.")
train_args.extend([train_prog, startup_prog])
if args.gpus > 1 and os.getenv("PADDLE_TRAINING_ROLE") == "TRAINER":
train_args.extend([nccl_id_var, num_trainers, trainer_id])
train_parallel(*train_args)
train(*train_args)
exit(0)
# for other update methods, use default programs
train_args.append(fluid.default_main_program())
train_args.append(fluid.default_startup_program())
if args.update_method == "nccl2":
nccl_id_var, num_trainers, trainer_id = append_nccl2_prepare(trainer_id)
if args.gpus == 1:
# NOTE: parallel executor use profiler interanlly
if args.use_nvprof and args.device == 'GPU':
with profiler.cuda_profiler("cuda_profiler.txt", 'csv') as nvprof:
train(*train_args)
else:
train(*train_args)
else:
if args.device == "CPU":
raise Exception("Only support GPU perf with parallel exe")
train_args.extend([nccl_id_var, num_trainers, trainer_id])
train_parallel(*train_args)
if __name__ == "__main__":
main()
benchmark/fluid/kube_gen_job.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import yaml
import copy
import argparse
import random
import os
from kube_templates import pserver, trainer, envs
def parse_args():
parser = argparse.ArgumentParser(description='Generate dist job yamls.')
parser.add_argument(
'--jobname', default="paddlejob", help='unique job name')
parser.add_argument(
'--cpu', default=1, type=int, help='CPU cores per trainer node')
parser.add_argument(
'--pscpu', default=1, type=int, help='CPU cores per pserver node')
parser.add_argument(
'--gpu', default=0, type=int, help='num of GPUs per node')
parser.add_argument(
'--image',
default="bootstrapper:5000/fluid_benchmark:gpu",
help='num of GPUs per node')
parser.add_argument(
'--pservers', default=1, type=int, help='num of pservers')
parser.add_argument(
'--trainers', default=1, type=int, help='num of trainers')
parser.add_argument('--memory', default=1, type=int, help='trainer memory')
parser.add_argument(
'--psmemory', default=1, type=int, help='pserver memory')
parser.add_argument(
'--port', default=30236, type=int, help='num of trainers')
parser.add_argument(
'--entry', default="python train.py", help='command to run')
parser.add_argument(
'--fluid', default=1, type=int, help='whether is fluid job')
parser.add_argument(
'--rdma', action='store_ture', help='whether mount rdma libs')
parser.add_argument(
'--disttype',
default="pserver",
type=str,
choices=['pserver', 'nccl2', 'local'],
help='pserver or nccl2 or local')
args = parser.parse_args()
return args
def gen_job():
ps = pserver
tn = trainer
args = parse_args()
ps_container = ps["spec"]["template"]["spec"]["containers"][0]
tn_container = tn["spec"]["template"]["spec"]["containers"][0]
if args.fluid == 1:
ps_container["command"] = \
["paddle_k8s", "start_fluid"]
tn_container["command"] = \
["paddle_k8s", "start_fluid"]
ps["metadata"]["name"] = args.jobname + "-pserver"
ps["spec"]["template"]["metadata"]["labels"][
"paddle-job-pserver"] = args.jobname
tn["metadata"]["name"] = args.jobname + "-trainer"
tn["spec"]["template"]["metadata"]["labels"]["paddle-job"] = args.jobname
ps_container["image"] = args.image
tn_container["image"] = args.image
ps_container["resources"]["requests"]["cpu"] = str(args.pscpu)
ps_container["resources"]["requests"]["memory"] = str(args.psmemory) + "Gi"
ps_container["resources"]["limits"]["cpu"] = str(args.pscpu)
ps_container["resources"]["limits"]["memory"] = str(args.psmemory) + "Gi"
tn_container["resources"]["requests"]["cpu"] = str(args.cpu)
tn_container["resources"]["requests"]["memory"] = str(args.memory) + "Gi"
tn_container["resources"]["limits"]["cpu"] = str(args.cpu)
tn_container["resources"]["limits"]["memory"] = str(args.memory) + "Gi"
if args.gpu > 0:
tn_container["resources"]["requests"][
"alpha.kubernetes.io/nvidia-gpu"] = str(args.gpu)
tn_container["resources"]["limits"][
"alpha.kubernetes.io/nvidia-gpu"] = str(args.gpu)
ps["spec"]["replicas"] = int(args.pservers)
tn["spec"]["parallelism"] = int(args.trainers)
tn["spec"]["completions"] = int(args.trainers)
ps_container["ports"][0]["name"] = "jobport-" + str(args.port)
ps_container["ports"][0]["containerPort"] = args.port
spreadport = random.randint(40000, 60000)
tn_container["ports"][0]["name"] = "spr-" + str(spreadport)
tn_container["ports"][0]["containerPort"] = spreadport
envs.append({"name": "PADDLE_JOB_NAME", "value": args.jobname})
envs.append({"name": "TRAINERS", "value": str(args.trainers)})
envs.append({"name": "PSERVERS", "value": str(args.pservers)})
envs.append({"name": "ENTRY", "value": args.entry})
envs.append({"name": "PADDLE_INIT_PORT", "value": str(args.port)})
envs.append({"name": "PADDLE_PSERVER_PORT", "value": str(args.port)})
# NOTE: these directories below are cluster specific, please modify
# this settings before you run on your own cluster.
envs.append({
"name": "LD_LIBRARY_PATH",
"value":
"/usr/local/lib:/usr/local/nvidia/lib64:/usr/local/rdma/lib64:/usr/lib64/mlnx_ofed/valgrind"
})
volumes = [{
"name": "nvidia-driver",
"hostPath": {
"path": "/usr/local/nvidia/lib64"
}
}]
volumeMounts = [{
"mountPath": "/usr/local/nvidia/lib64",
"name": "nvidia-driver"
}]
if args.rdma:
volumes.extend([{
"name": "ibetc",
"hostPath": {
"path": "/etc/libibverbs.d"
}
}, {
"name": "iblibs",
"hostPath": {
"path": "/usr/local/rdma"
}
}, {
"name": "valgrind",
"hostPath": {
"path": "/usr/lib64/mlnx_ofed/valgrind"
}
}])
volumeMounts.extend([{
"mountPath": "/etc/libibverbs.d",
"name": "ibetc"
}, {
"mountPath": "/usr/local/rdma",
"name": "iblibs"
}, {
"mountPath": "/usr/lib64/mlnx_ofed/valgrind",
"name": "valgrind"
}])
# append shm for NCCL2
volumes.append({"name": "dshm", "emptyDir": {"medium": "Memory"}})
volumeMounts.append({"mountPath": "/dev/shm", "name": "dshm"})
tn["spec"]["template"]["spec"]["volumes"] = volumes
tn_container["volumeMounts"] = volumeMounts
ps_container["env"] = envs
ps_container["env"].append({"name": "TRAINING_ROLE", "value": "PSERVER"})
tn_container["env"] = envs
if args.disttype == "pserver":
tn_container["env"].append({
"name": "TRAINING_ROLE",
"value": "TRAINER"
})
elif args.disttype == "nccl2" or args.disttype == "local":
# NCCL2 have no training role, set to plain WORKER
tn_container["env"].append({"name": "TRAINING_ROLE", "value": "WORKER"})
os.mkdir(args.jobname)
if args.disttype == "pserver":
with open("%s/pserver.yaml" % args.jobname, "w") as fn:
yaml.dump(ps, fn)
with open("%s/trainer.yaml" % args.jobname, "w") as fn:
yaml.dump(tn, fn)
if __name__ == "__main__":
gen_job()
benchmark/fluid/kube_templates/__init__.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from pserver import pserver
from trainer import trainer
__all__ = ["pserver", "trainer", "envs"]
envs = [
# envs that don't need to change
{
"name": "GLOG_v",
"value": "0"
},
{
"name": "GLOG_logtostderr",
"value": "1"
},
{
"name": "TOPOLOGY",
"value": ""
},
{
"name": "TRAINER_PACKAGE",
"value": "/workspace"
},
{
"name": "PADDLE_INIT_NICS",
"value": "eth2"
},
{
"name": "NAMESPACE",
"valueFrom": {
"fieldRef": {
"fieldPath": "metadata.namespace"
}
}
},
{
"name": "POD_IP",
"valueFrom": {
"fieldRef": {
"fieldPath": "status.podIP"
}
}
},
{
"name": "PADDLE_CURRENT_IP",
"valueFrom": {
"fieldRef": {
"fieldPath": "status.podIP"
}
}
}
]
benchmark/fluid/kube_templates/pserver.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
pserver = {
"apiVersion": "extensions/v1beta1",
"kind": "ReplicaSet",
"metadata": {
"name": "jobname-pserver"
},
"spec": {
"replicas": 1,
"template": {
"metadata": {
"labels": {
"paddle-job-pserver": "jobname"
}
},
"spec": {
"hostNetwork": True,
"imagePullSecrets": [{
"name": "job-registry-secret"
}],
"containers": [{
"name": "pserver",
"image": "",
"imagePullPolicy": "Always",
"ports": [{
"name": "jobport-1",
"containerPort": 1
}],
"env": [],
"command": ["paddle_k8s", "start_pserver"],
"resources": {
"requests": {
"memory": "10Gi",
"cpu": "4"
},
"limits": {
"memory": "10Gi",
"cpu": "4"
}
}
}]
}
}
}
}
benchmark/fluid/kube_templates/trainer.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
trainer = {
"apiVersion": "batch/v1",
"kind": "Job",
"metadata": {
"name": "jobname-pserver"
},
"spec": {
"parallelism": 4,
"completions": 4,
"template": {
"metadata": {
"labels": {
"paddle-job": "jobname"
}
},
"spec": {
"hostNetwork": True,
"imagePullSecrets": [{
"name": "job-registry-secret"
}],
"restartPolicy": "Never",
"containers": [{
"name": "trainer",
"image": "",
"imagePullPolicy": "Always",
# to let container set rlimit
"securityContext": {
"privileged": True
# TODO(wuyi): use below specific cap instead of privileged,
# using privileged will cause all GPU device are visible
# in the container.
# "capabilities": {
# "add": ["SYS_RESOURCE"]
# }
},
"ports": [{
"name": "jobport-1",
"containerPort": 1
}],
"env": [],
"command": ["paddle_k8s", "start_trainer", "v2"],
"resources": {
"requests": {
"memory": "10Gi",
"cpu": "4",
},
"limits": {
"memory": "10Gi",
"cpu": "4",
}
}
}]
}
}
}
}
benchmark/fluid/models/__init__.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
__all__ = [
"machine_translation", "resnet", "vgg", "mnist", "stacked_dynamic_lstm"
]
benchmark/fluid/models/machine_translation.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""seq2seq model for fluid."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import numpy as np
import argparse
import time
import distutils.util
import paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
import paddle.fluid.framework as framework
from paddle.fluid.executor import Executor
def lstm_step(x_t, hidden_t_prev, cell_t_prev, size):
def linear(inputs):
return fluid.layers.fc(input=inputs, size=size, bias_attr=True)
forget_gate = fluid.layers.sigmoid(x=linear([hidden_t_prev, x_t]))
input_gate = fluid.layers.sigmoid(x=linear([hidden_t_prev, x_t]))
output_gate = fluid.layers.sigmoid(x=linear([hidden_t_prev, x_t]))
cell_tilde = fluid.layers.tanh(x=linear([hidden_t_prev, x_t]))
cell_t = fluid.layers.sums(input=[
fluid.layers.elementwise_mul(
x=forget_gate, y=cell_t_prev), fluid.layers.elementwise_mul(
x=input_gate, y=cell_tilde)
])
hidden_t = fluid.layers.elementwise_mul(
x=output_gate, y=fluid.layers.tanh(x=cell_t))
return hidden_t, cell_t
def seq_to_seq_net(embedding_dim, encoder_size, decoder_size, source_dict_dim,
target_dict_dim, is_generating, beam_size, max_length):
"""Construct a seq2seq network."""
def bi_lstm_encoder(input_seq, gate_size):
# Linear transformation part for input gate, output gate, forget gate
# and cell activation vectors need be done outside of dynamic_lstm.
# So the output size is 4 times of gate_size.
input_forward_proj = fluid.layers.fc(input=input_seq,
size=gate_size * 4,
act=None,
bias_attr=False)
forward, _ = fluid.layers.dynamic_lstm(
input=input_forward_proj, size=gate_size * 4, use_peepholes=False)
input_reversed_proj = fluid.layers.fc(input=input_seq,
size=gate_size * 4,
act=None,
bias_attr=False)
reversed, _ = fluid.layers.dynamic_lstm(
input=input_reversed_proj,
size=gate_size * 4,
is_reverse=True,
use_peepholes=False)
return forward, reversed
src_word_idx = fluid.layers.data(
name='source_sequence', shape=[1], dtype='int64', lod_level=1)
src_embedding = fluid.layers.embedding(
input=src_word_idx,
size=[source_dict_dim, embedding_dim],
dtype='float32')
src_forward, src_reversed = bi_lstm_encoder(
input_seq=src_embedding, gate_size=encoder_size)
encoded_vector = fluid.layers.concat(
input=[src_forward, src_reversed], axis=1)
encoded_proj = fluid.layers.fc(input=encoded_vector,
size=decoder_size,
bias_attr=False)
backward_first = fluid.layers.sequence_pool(
input=src_reversed, pool_type='first')
decoder_boot = fluid.layers.fc(input=backward_first,
size=decoder_size,
bias_attr=False,
act='tanh')
def lstm_decoder_with_attention(target_embedding, encoder_vec, encoder_proj,
decoder_boot, decoder_size):
def simple_attention(encoder_vec, encoder_proj, decoder_state):
decoder_state_proj = fluid.layers.fc(input=decoder_state,
size=decoder_size,
bias_attr=False)
decoder_state_expand = fluid.layers.sequence_expand(
x=decoder_state_proj, y=encoder_proj)
concated = fluid.layers.concat(
input=[encoder_proj, decoder_state_expand], axis=1)
attention_weights = fluid.layers.fc(input=concated,
size=1,
act='tanh',
bias_attr=False)
attention_weights = fluid.layers.sequence_softmax(
input=attention_weights)
weigths_reshape = fluid.layers.reshape(
x=attention_weights, shape=[-1])
scaled = fluid.layers.elementwise_mul(
x=encoder_vec, y=weigths_reshape, axis=0)
context = fluid.layers.sequence_pool(input=scaled, pool_type='sum')
return context
rnn = fluid.layers.DynamicRNN()
cell_init = fluid.layers.fill_constant_batch_size_like(
input=decoder_boot,
value=0.0,
shape=[-1, decoder_size],
dtype='float32')
cell_init.stop_gradient = False
with rnn.block():
current_word = rnn.step_input(target_embedding)
encoder_vec = rnn.static_input(encoder_vec)
encoder_proj = rnn.static_input(encoder_proj)
hidden_mem = rnn.memory(init=decoder_boot, need_reorder=True)
cell_mem = rnn.memory(init=cell_init)
context = simple_attention(encoder_vec, encoder_proj, hidden_mem)
decoder_inputs = fluid.layers.concat(
input=[context, current_word], axis=1)
h, c = lstm_step(decoder_inputs, hidden_mem, cell_mem, decoder_size)
rnn.update_memory(hidden_mem, h)
rnn.update_memory(cell_mem, c)
out = fluid.layers.fc(input=h,
size=target_dict_dim,
bias_attr=True,
act='softmax')
rnn.output(out)
return rnn()
if not is_generating:
trg_word_idx = fluid.layers.data(
name='target_sequence', shape=[1], dtype='int64', lod_level=1)
trg_embedding = fluid.layers.embedding(
input=trg_word_idx,
size=[target_dict_dim, embedding_dim],
dtype='float32')
prediction = lstm_decoder_with_attention(trg_embedding, encoded_vector,
encoded_proj, decoder_boot,
decoder_size)
label = fluid.layers.data(
name='label_sequence', shape=[1], dtype='int64', lod_level=1)
cost = fluid.layers.cross_entropy(input=prediction, label=label)
avg_cost = fluid.layers.mean(x=cost)
feeding_list = ["source_sequence", "target_sequence", "label_sequence"]
return avg_cost, feeding_list
def to_lodtensor(data, place):
seq_lens = [len(seq) for seq in data]
cur_len = 0
lod = [cur_len]
for l in seq_lens:
cur_len += l
lod.append(cur_len)
flattened_data = np.concatenate(data, axis=0).astype("int64")
flattened_data = flattened_data.reshape([len(flattened_data), 1])
lod_t = core.LoDTensor()
lod_t.set(flattened_data, place)
lod_t.set_lod([lod])
return lod_t, lod[-1]
def lodtensor_to_ndarray(lod_tensor):
dims = lod_tensor.get_dims()
ndarray = np.zeros(shape=dims).astype('float32')
for i in xrange(np.product(dims)):
ndarray.ravel()[i] = lod_tensor.get_float_element(i)
return ndarray
def get_model(args):
embedding_dim = 512
encoder_size = 512
decoder_size = 512
dict_size = 30000
beam_size = 3
max_length = 250
avg_cost, feeding_list = seq_to_seq_net(
embedding_dim,
encoder_size,
decoder_size,
dict_size,
dict_size,
False,
beam_size=beam_size,
max_length=max_length)
# clone from default main program
inference_program = fluid.default_main_program().clone()
optimizer = fluid.optimizer.Adam(learning_rate=args.learning_rate)
train_batch_generator = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.wmt14.train(dict_size), buf_size=1000),
batch_size=args.batch_size)
test_batch_generator = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.wmt14.test(dict_size), buf_size=1000),
batch_size=args.batch_size)
return avg_cost, inference_program, optimizer, train_batch_generator, \
test_batch_generator, None
benchmark/fluid/models/mnist.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import numpy as np
import argparse
import time
import cProfile
import paddle
import paddle.fluid as fluid
import paddle.fluid.profiler as profiler
SEED = 1
DTYPE = "float32"
# random seed must set before configuring the network.
# fluid.default_startup_program().random_seed = SEED
def cnn_model(data):
conv_pool_1 = fluid.nets.simple_img_conv_pool(
input=data,
filter_size=5,
num_filters=20,
pool_size=2,
pool_stride=2,
act="relu")
conv_pool_2 = fluid.nets.simple_img_conv_pool(
input=conv_pool_1,
filter_size=5,
num_filters=50,
pool_size=2,
pool_stride=2,
act="relu")
# TODO(dzhwinter) : refine the initializer and random seed settting
SIZE = 10
input_shape = conv_pool_2.shape
param_shape = [reduce(lambda a, b: a * b, input_shape[1:], 1)] + [SIZE]
scale = (2.0 / (param_shape[0]**2 * SIZE))**0.5
predict = fluid.layers.fc(
input=conv_pool_2,
size=SIZE,
act="softmax",
param_attr=fluid.param_attr.ParamAttr(
initializer=fluid.initializer.NormalInitializer(
loc=0.0, scale=scale)))
return predict
def get_model(args):
# Input data
images = fluid.layers.data(name='pixel', shape=[1, 28, 28], dtype=DTYPE)
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
# Train program
predict = cnn_model(images)
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
# Evaluator
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(
input=predict, label=label, total=batch_size_tensor)
# inference program
inference_program = fluid.default_main_program().clone()
# Optimization
opt = fluid.optimizer.AdamOptimizer(
learning_rate=0.001, beta1=0.9, beta2=0.999)
# Reader
train_reader = paddle.batch(
paddle.dataset.mnist.train(), batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.dataset.mnist.test(), batch_size=args.batch_size)
return avg_cost, inference_program, opt, train_reader, test_reader, batch_acc
benchmark/fluid/models/resnet.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import functools
import numpy as np
import time
import cProfile, pstats, StringIO
import paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
import paddle.fluid.profiler as profiler
def conv_bn_layer(input, ch_out, filter_size, stride, padding, act='relu'):
conv1 = fluid.layers.conv2d(
input=input,
filter_size=filter_size,
num_filters=ch_out,
stride=stride,
padding=padding,
act=None,
bias_attr=False)
return fluid.layers.batch_norm(input=conv1, act=act)
def shortcut(input, ch_out, stride):
ch_in = input.shape[1] # if args.data_format == 'NCHW' else input.shape[-1]
if ch_in != ch_out:
return conv_bn_layer(input, ch_out, 1, stride, 0, None)
else:
return input
def basicblock(input, ch_out, stride):
short = shortcut(input, ch_out, stride)
conv1 = conv_bn_layer(input, ch_out, 3, stride, 1)
conv2 = conv_bn_layer(conv1, ch_out, 3, 1, 1, act=None)
return fluid.layers.elementwise_add(x=short, y=conv2, act='relu')
def bottleneck(input, ch_out, stride):
short = shortcut(input, ch_out * 4, stride)
conv1 = conv_bn_layer(input, ch_out, 1, stride, 0)
conv2 = conv_bn_layer(conv1, ch_out, 3, 1, 1)
conv3 = conv_bn_layer(conv2, ch_out * 4, 1, 1, 0, act=None)
return fluid.layers.elementwise_add(x=short, y=conv3, act='relu')
def layer_warp(block_func, input, ch_out, count, stride):
res_out = block_func(input, ch_out, stride)
for i in range(1, count):
res_out = block_func(res_out, ch_out, 1)
return res_out
def resnet_imagenet(input, class_dim, depth=50, data_format='NCHW'):
cfg = {
18: ([2, 2, 2, 1], basicblock),
34: ([3, 4, 6, 3], basicblock),
50: ([3, 4, 6, 3], bottleneck),
101: ([3, 4, 23, 3], bottleneck),
152: ([3, 8, 36, 3], bottleneck)
}
stages, block_func = cfg[depth]
conv1 = conv_bn_layer(input, ch_out=64, filter_size=7, stride=2, padding=3)
pool1 = fluid.layers.pool2d(
input=conv1, pool_type='avg', pool_size=3, pool_stride=2)
res1 = layer_warp(block_func, pool1, 64, stages[0], 1)
res2 = layer_warp(block_func, res1, 128, stages[1], 2)
res3 = layer_warp(block_func, res2, 256, stages[2], 2)
res4 = layer_warp(block_func, res3, 512, stages[3], 2)
pool2 = fluid.layers.pool2d(
input=res4,
pool_size=7,
pool_type='avg',
pool_stride=1,
global_pooling=True)
out = fluid.layers.fc(input=pool2, size=class_dim, act='softmax')
return out
def resnet_cifar10(input, class_dim, depth=32, data_format='NCHW'):
assert (depth - 2) % 6 == 0
n = (depth - 2) // 6
conv1 = conv_bn_layer(
input=input, ch_out=16, filter_size=3, stride=1, padding=1)
res1 = layer_warp(basicblock, conv1, 16, n, 1)
res2 = layer_warp(basicblock, res1, 32, n, 2)
res3 = layer_warp(basicblock, res2, 64, n, 2)
pool = fluid.layers.pool2d(
input=res3, pool_size=8, pool_type='avg', pool_stride=1)
out = fluid.layers.fc(input=pool, size=class_dim, act='softmax')
return out
def get_model(args):
model = resnet_cifar10
if args.data_set == "cifar10":
class_dim = 10
if args.data_format == 'NCHW':
dshape = [3, 32, 32]
else:
dshape = [32, 32, 3]
model = resnet_cifar10
else:
class_dim = 102
if args.data_format == 'NCHW':
dshape = [3, 224, 224]
else:
dshape = [224, 224, 3]
model = resnet_imagenet
input = fluid.layers.data(name='data', shape=dshape, dtype='float32')
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
predict = model(input, class_dim)
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(
input=predict, label=label, total=batch_size_tensor)
inference_program = fluid.default_main_program().clone()
with fluid.program_guard(inference_program):
inference_program = fluid.io.get_inference_program(
target_vars=[batch_acc, batch_size_tensor])
optimizer = fluid.optimizer.Momentum(learning_rate=0.01, momentum=0.9)
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.train(),
buf_size=5120),
batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.dataset.cifar.test10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.test(),
batch_size=args.batch_size)
return avg_cost, inference_program, optimizer, train_reader, test_reader, batch_acc
benchmark/fluid/models/stacked_dynamic_lstm.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import cPickle
import os
import random
import time
import numpy
import paddle
import paddle.dataset.imdb as imdb
import paddle.fluid as fluid
import paddle.batch as batch
import paddle.fluid.profiler as profiler
word_dict = imdb.word_dict()
def crop_sentence(reader, crop_size):
unk_value = word_dict['<unk>']
def __impl__():
for item in reader():
if len([x for x in item[0] if x != unk_value]) < crop_size:
yield item
return __impl__
def get_model(args):
lstm_size = 512
emb_dim = 512
crop_size = 1500
data = fluid.layers.data(
name="words", shape=[1], lod_level=1, dtype='int64')
sentence = fluid.layers.embedding(
input=data, size=[len(word_dict), emb_dim])
sentence = fluid.layers.fc(input=sentence, size=lstm_size, act='tanh')
rnn = fluid.layers.DynamicRNN()
with rnn.block():
word = rnn.step_input(sentence)
prev_hidden = rnn.memory(value=0.0, shape=[lstm_size])
prev_cell = rnn.memory(value=0.0, shape=[lstm_size])
def gate_common(
ipt,
hidden,
size, ):
gate0 = fluid.layers.fc(input=ipt, size=size, bias_attr=True)
gate1 = fluid.layers.fc(input=hidden, size=size, bias_attr=False)
gate = fluid.layers.sums(input=[gate0, gate1])
return gate
forget_gate = fluid.layers.sigmoid(
x=gate_common(word, prev_hidden, lstm_size))
input_gate = fluid.layers.sigmoid(
x=gate_common(word, prev_hidden, lstm_size))
output_gate = fluid.layers.sigmoid(
x=gate_common(word, prev_hidden, lstm_size))
cell_gate = fluid.layers.tanh(
x=gate_common(word, prev_hidden, lstm_size))
cell = fluid.layers.sums(input=[
fluid.layers.elementwise_mul(
x=forget_gate, y=prev_cell), fluid.layers.elementwise_mul(
x=input_gate, y=cell_gate)
])
hidden = fluid.layers.elementwise_mul(
x=output_gate, y=fluid.layers.tanh(x=cell))
rnn.update_memory(prev_cell, cell)
rnn.update_memory(prev_hidden, hidden)
rnn.output(hidden)
last = fluid.layers.sequence_pool(rnn(), 'last')
logit = fluid.layers.fc(input=last, size=2, act='softmax')
loss = fluid.layers.cross_entropy(
input=logit,
label=fluid.layers.data(
name='label', shape=[1], dtype='int64'))
loss = fluid.layers.mean(x=loss)
# add acc
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(input=logit, label=fluid.layers.data(name='label', \
shape=[1], dtype='int64'), total=batch_size_tensor)
inference_program = fluid.default_main_program().clone()
with fluid.program_guard(inference_program):
inference_program = fluid.io.get_inference_program(
target_vars=[batch_acc, batch_size_tensor])
adam = fluid.optimizer.Adam()
train_reader = batch(
paddle.reader.shuffle(
crop_sentence(imdb.train(word_dict), crop_size), buf_size=25000),
batch_size=args.batch_size)
test_reader = batch(
paddle.reader.shuffle(
crop_sentence(imdb.test(word_dict), crop_size), buf_size=25000),
batch_size=args.batch_size)
return loss, inference_program, adam, train_reader, test_reader, batch_acc
def to_lodtensor(data, place):
seq_lens = [len(seq) for seq in data]
cur_len = 0
lod = [cur_len]
for l in seq_lens:
cur_len += l
lod.append(cur_len)
flattened_data = numpy.concatenate(data, axis=0).astype("int64")
flattened_data = flattened_data.reshape([len(flattened_data), 1])
res = fluid.LoDTensor()
res.set(flattened_data, place)
res.set_lod([lod])
return res
benchmark/fluid/models/vgg.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""VGG16 benchmark in Fluid"""
from __future__ import print_function
import sys
import time
import numpy as np
import paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
import argparse
import functools
def vgg16_bn_drop(input):
def conv_block(input, num_filter, groups, dropouts):
return fluid.nets.img_conv_group(
input=input,
pool_size=2,
pool_stride=2,
conv_num_filter=[num_filter] * groups,
conv_filter_size=3,
conv_act='relu',
conv_with_batchnorm=True,
conv_batchnorm_drop_rate=dropouts,
pool_type='max')
conv1 = conv_block(input, 64, 2, [0.3, 0])
conv2 = conv_block(conv1, 128, 2, [0.4, 0])
conv3 = conv_block(conv2, 256, 3, [0.4, 0.4, 0])
conv4 = conv_block(conv3, 512, 3, [0.4, 0.4, 0])
conv5 = conv_block(conv4, 512, 3, [0.4, 0.4, 0])
drop = fluid.layers.dropout(x=conv5, dropout_prob=0.5)
fc1 = fluid.layers.fc(input=drop, size=512, act=None)
bn = fluid.layers.batch_norm(input=fc1, act='relu')
drop2 = fluid.layers.dropout(x=bn, dropout_prob=0.5)
fc2 = fluid.layers.fc(input=drop2, size=512, act=None)
return fc2
def get_model(args):
if args.data_set == "cifar10":
classdim = 10
if args.data_format == 'NCHW':
data_shape = [3, 32, 32]
else:
data_shape = [32, 32, 3]
else:
classdim = 102
if args.data_format == 'NCHW':
data_shape = [3, 224, 224]
else:
data_shape = [224, 224, 3]
# Input data
images = fluid.layers.data(name='pixel', shape=data_shape, dtype='float32')
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
# Train program
net = vgg16_bn_drop(images)
predict = fluid.layers.fc(input=net, size=classdim, act='softmax')
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
# Evaluator
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(
input=predict, label=label, total=batch_size_tensor)
# inference program
inference_program = fluid.default_main_program().clone()
with fluid.program_guard(inference_program):
inference_program = fluid.io.get_inference_program(
target_vars=[batch_acc, batch_size_tensor])
# Optimization
optimizer = fluid.optimizer.Adam(learning_rate=args.learning_rate)
# data reader
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.train(),
buf_size=5120),
batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.dataset.cifar.test10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.test(),
batch_size=args.batch_size)
return avg_cost, inference_program, optimizer, train_reader, test_reader, batch_acc
benchmark/fluid/run.sh 0 → 100644
浏览文件 @ dbd25805
#!/bin/bash
# This script benchmarking the PaddlePaddle Fluid on
# single thread single GPU.
#export FLAGS_fraction_of_gpu_memory_to_use=0.0
export CUDNN_PATH=/paddle/cudnn_v5
# disable openmp and mkl parallel
#https://github.com/PaddlePaddle/Paddle/issues/7199
export MKL_NUM_THREADS=1
export OMP_NUM_THREADS=1
ht=`lscpu |grep "per core"|awk -F':' '{print $2}'|xargs`
if [ $ht -eq 1 ]; then # HT is OFF
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,0,0"
fi
if [ -z "$OMP_DYNAMIC" ]; then
export OMP_DYNAMIC="FALSE"
fi
else # HT is ON
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,1,0"
fi
fi
# disable multi-gpu if have more than one
export CUDA_VISIBLE_DEVICES=0
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH=$CUDNN_PATH:$LD_LIBRARY_PATH
# only query the gpu used
nohup stdbuf -oL nvidia-smi \
--id=${CUDA_VISIBLE_DEVICES} \
--query-gpu=timestamp \
--query-compute-apps=pid,process_name,used_memory \
--format=csv \
--filename=mem.log \
-l 1 &
# mnist
# mnist gpu mnist 128
FLAGS_benchmark=true stdbuf -oL python fluid/mnist.py \
--device=GPU \
--batch_size=128 \
--skip_batch_num=5 \
--iterations=500 \
2>&1 | tee -a mnist_gpu_128.log
# vgg16
# gpu cifar10 128
FLAGS_benchmark=true stdbuf -oL python fluid/vgg16.py \
--device=GPU \
--batch_size=128 \
--skip_batch_num=5 \
--iterations=30 \
2>&1 | tee -a vgg16_gpu_128.log
# flowers gpu 128
FLAGS_benchmark=true stdbuf -oL python fluid/vgg16.py \
--device=GPU \
--batch_size=32 \
--data_set=flowers \
--skip_batch_num=5 \
--iterations=30 \
2>&1 | tee -a vgg16_gpu_flowers_32.log
# resnet50
# resnet50 gpu cifar10 128
FLAGS_benchmark=true stdbuf -oL python fluid/resnet50.py \
--device=GPU \
--batch_size=128 \
--data_set=cifar10 \
--model=resnet_cifar10 \
--skip_batch_num=5 \
--iterations=30 \
2>&1 | tee -a resnet50_gpu_128.log
# resnet50 gpu flowers 64
FLAGS_benchmark=true stdbuf -oL python fluid/resnet50.py \
--device=GPU \
--batch_size=64 \
--data_set=flowers \
--model=resnet_imagenet \
--skip_batch_num=5 \
--iterations=30 \
2>&1 | tee -a resnet50_gpu_flowers_64.log
# lstm
# lstm gpu imdb 32 # tensorflow only support batch=32
FLAGS_benchmark=true stdbuf -oL python fluid/stacked_dynamic_lstm.py \
--device=GPU \
--batch_size=32 \
--skip_batch_num=5 \
--iterations=30 \
--hidden_dim=512 \
--emb_dim=512 \
--crop_size=1500 \
2>&1 | tee -a lstm_gpu_32.log
# seq2seq
# seq2seq gpu wmb 128
FLAGS_benchmark=true stdbuf -oL python fluid/machine_translation.py \
--device=GPU \
--batch_size=128 \
--skip_batch_num=5 \
--iterations=30 \
2>&1 | tee -a lstm_gpu_128.log
benchmark/paddle/image/alexnet.py
浏览文件 @ dbd25805
#!/usr/bin/env python
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from paddle.trainer_config_helpers import *
......@@ -6,10 +18,24 @@ height = 227
width = 227
num_class = 1000
batch_size = get_config_arg('batch_size', int, 128)
gp = get_config_arg('layer_num', int, 1)
is_infer = get_config_arg("is_infer", bool, False)
num_samples = get_config_arg('num_samples', int, 2560)
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,
'num_samples': num_samples
}
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,
......@@ -31,7 +57,7 @@ net = img_pool_layer(input=net, pool_size=3, stride=2)
# conv2
net = img_conv_layer(
input=net, filter_size=5, num_filters=256, stride=1, padding=2, groups=1)
input=net, filter_size=5, num_filters=256, stride=1, padding=2, groups=gp)
net = img_cmrnorm_layer(input=net, size=5, scale=0.0001, power=0.75)
net = img_pool_layer(input=net, pool_size=3, stride=2)
......@@ -40,11 +66,11 @@ net = img_conv_layer(
input=net, filter_size=3, num_filters=384, stride=1, padding=1)
# conv4
net = img_conv_layer(
input=net, filter_size=3, num_filters=384, stride=1, padding=1, groups=1)
input=net, filter_size=3, num_filters=384, stride=1, padding=1, groups=gp)
# conv5
net = img_conv_layer(
input=net, filter_size=3, num_filters=256, stride=1, padding=1, groups=1)
input=net, filter_size=3, num_filters=256, stride=1, padding=1, groups=gp)
net = img_pool_layer(input=net, pool_size=3, stride=2)
net = fc_layer(
......@@ -59,6 +85,9 @@ net = fc_layer(
layer_attr=ExtraAttr(drop_rate=0.5))
net = fc_layer(input=net, size=1000, act=SoftmaxActivation())
lab = data_layer('label', num_class)
loss = cross_entropy(input=net, label=lab)
outputs(loss)
if is_infer:
outputs(net)
else:
lab = data_layer('label', num_class)
loss = cross_entropy(input=net, label=lab)
outputs(loss)
benchmark/paddle/image/googlenet.py
浏览文件 @ dbd25805
......@@ -7,13 +7,15 @@ num_class = 1000
batch_size = get_config_arg('batch_size', int, 128)
use_gpu = get_config_arg('use_gpu', bool, True)
is_infer = get_config_arg("is_infer", bool, False)
num_samples = get_config_arg('num_samples', int, 2560)
args = {
'height': height,
'width': width,
'color': True,
'num_class': num_class,
'is_infer': is_infer
'is_infer': is_infer,
'num_samples': num_samples
}
define_py_data_sources2(
"train.list" if not is_infer else None,
......
benchmark/paddle/image/plotlog.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import sys
import argparse
import matplotlib.pyplot as plt
def parse_args():
parser = argparse.ArgumentParser('Parse Log')
parser.add_argument(
'--file_path', '-f', type=str, help='the path of the log file')
parser.add_argument(
'--sample_rate',
'-s',
type=float,
default=1.0,
help='the rate to take samples from log')
parser.add_argument(
'--log_period', '-p', type=int, default=1, help='the period of log')
args = parser.parse_args()
return args
def parse_file(file_name):
loss = []
error = []
with open(file_name) as f:
for i, line in enumerate(f):
line = line.strip()
if not line.startswith('pass'):
continue
line_split = line.split(' ')
if len(line_split) != 5:
continue
loss_str = line_split[2][:-1]
cur_loss = float(loss_str.split('=')[-1])
loss.append(cur_loss)
err_str = line_split[3][:-1]
cur_err = float(err_str.split('=')[-1])
error.append(cur_err)
accuracy = [1.0 - err for err in error]
return loss, accuracy
def sample(metric, sample_rate):
interval = int(1.0 / sample_rate)
if interval > len(metric):
return metric[:1]
num = len(metric) / interval
idx = [interval * i for i in range(num)]
metric_sample = [metric[id] for id in idx]
return metric_sample
def plot_metric(metric,
batch_id,
graph_title,
line_style='b-',
line_label='y',
line_num=1):
plt.figure()
plt.title(graph_title)
if line_num == 1:
plt.plot(batch_id, metric, line_style, label=line_label)
else:
for i in range(line_num):
plt.plot(batch_id, metric[i], line_style[i], label=line_label[i])
plt.xlabel('batch')
plt.ylabel(graph_title)
plt.legend()
plt.savefig(graph_title + '.jpg')
plt.close()
def main():
args = parse_args()
assert args.sample_rate > 0. and args.sample_rate <= 1.0, "The sample rate should in the range (0, 1]."
loss, accuracy = parse_file(args.file_path)
batch = [args.log_period * i for i in range(len(loss))]
batch_sample = sample(batch, args.sample_rate)
loss_sample = sample(loss, args.sample_rate)
accuracy_sample = sample(accuracy, args.sample_rate)
plot_metric(loss_sample, batch_sample, 'loss', line_label='loss')
plot_metric(
accuracy_sample,
batch_sample,
'accuracy',
line_style='g-',
line_label='accuracy')
if __name__ == '__main__':
main()
benchmark/paddle/image/provider.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import io, os
import random
import numpy as np
......@@ -14,6 +28,7 @@ def initHook(settings, height, width, color, num_class, **kwargs):
else:
settings.data_size = settings.height * settings.width
settings.is_infer = kwargs.get('is_infer', False)
settings.num_samples = kwargs.get('num_samples', 2560)
if settings.is_infer:
settings.slots = [dense_vector(settings.data_size)]
else:
......@@ -23,7 +38,7 @@ def initHook(settings, height, width, color, num_class, **kwargs):
@provider(
init_hook=initHook, min_pool_size=-1, cache=CacheType.CACHE_PASS_IN_MEM)
def process(settings, file_list):
for i in xrange(2560 if settings.is_infer else 1024):
for i in xrange(settings.num_samples):
img = np.random.rand(1, settings.data_size).reshape(-1, 1).flatten()
if settings.is_infer:
yield img.astype('float32')
......
benchmark/paddle/image/resnet.py
浏览文件 @ dbd25805
......@@ -7,13 +7,15 @@ 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)
num_samples = get_config_arg('num_samples', int, 2560)
args = {
'height': height,
'width': width,
'color': True,
'num_class': num_class,
'is_infer': is_infer
'is_infer': is_infer,
'num_samples': num_samples
}
define_py_data_sources2(
"train.list" if not is_infer else None,
......
benchmark/paddle/image/run_mkl_infer.sh 0 → 100755
浏览文件 @ dbd25805
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},num_samples=256" \
> /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 vgg 19 $batchsize $use_mkldnn
infer resnet 50 $batchsize $use_mkldnn
infer googlenet v1 $batchsize $use_mkldnn
infer alexnet 2 $batchsize $use_mkldnn
done
done
benchmark/paddle/image/run_mkl_train.sh 0 → 100755
浏览文件 @ dbd25805
set -e
function train() {
unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
topology=$1
layer_num=$2
bs=$3
use_mkldnn=$4
if [ $4 == "True" ]; then
thread=1
log="logs/train-${topology}-${layer_num}-mkldnn-${bs}.log"
elif [ $4 == "False" ]; then
thread=`nproc`
# each trainer_count use only 1 core to avoid conflict
log="logs/train-${topology}-${layer_num}-${thread}mklml-${bs}.log"
else
echo "Wrong input $4, use True or False."
exit 0
fi
args="batch_size=${bs},layer_num=${layer_num}"
config="${topology}.py"
paddle train --job=time \
--config=$config \
--use_mkldnn=$use_mkldnn \
--use_gpu=False \
--trainer_count=$thread \
--log_period=10 \
--test_period=100 \
--config_args=$args \
2>&1 | tee ${log}
avg_time=`tail ${log} -n 1 | awk -F ' ' '{print $8}' | sed 's/avg=//'`
fps=`awk 'BEGIN{printf "%.2f",('$bs' / '$avg_time' * 1000)}'`
echo "FPS: $fps images/sec" 2>&1 | tee -a ${log}
}
if [ ! -f "train.list" ]; then
echo " " > train.list
fi
if [ ! -d "logs" ]; then
mkdir logs
fi
# 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
train alexnet 2 $batchsize $use_mkldnn
done
done
benchmark/paddle/image/run_mkldnn_infer.sh 已删除 100755 → 0
浏览文件 @ 1ff47b28
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_train.sh 已删除 100755 → 0
浏览文件 @ 1ff47b28
set -e
function train() {
unset OMP_NUM_THREADS MKL_NUM_THREADS OMP_DYNAMIC KMP_AFFINITY
topology=$1
layer_num=$2
bs=$3
use_mkldnn=$4
if [ $4 == "True" ]; then
thread=1
log="logs/train-${topology}-${layer_num}-mkldnn-${bs}.log"
elif [ $4 == "False" ]; then
thread=`nproc`
# each trainer_count use only 1 core to avoid conflict
log="logs/train-${topology}-${layer_num}-${thread}mklml-${bs}.log"
else
echo "Wrong input $4, use True or False."
exit 0
fi
args="batch_size=${bs},layer_num=${layer_num}"
config="${topology}.py"
paddle train --job=time \
--config=$config \
--use_mkldnn=$use_mkldnn \
--use_gpu=False \
--trainer_count=$thread \
--log_period=10 \
--test_period=100 \
--config_args=$args \
2>&1 | tee ${log}
}
if [ ! -f "train.list" ]; then
echo " " > train.list
fi
if [ ! -d "logs" ]; then
mkdir logs
fi
# 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/run_openblas_infer.sh 0 → 100755
浏览文件 @ dbd25805
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() {
export OPENBLAS_MAIN_FREE=1
topology=$1
layer_num=$2
bs=$3
trainers=`nproc`
if [ $trainers -gt $bs ]; then
trainers=$bs
fi
log="logs/infer-${topology}-${layer_num}-${trainers}openblas-${bs}.log"
threads=$((`nproc` / trainers))
if [ $threads -eq 0 ]; then
threads=1
fi
export OPENBLAS_NUM_THREADS=$threads
models_in="models/${topology}-${layer_num}/pass-00000/"
if [ ! -d $models_in ]; then
echo "./run_mkl_infer.sh to save the model first"
exit 0
fi
log_period=$((32 / bs))
paddle train --job=test \
--config="${topology}.py" \
--use_mkldnn=False \
--use_gpu=False \
--trainer_count=$trainers \
--log_period=$log_period \
--config_args="batch_size=${bs},layer_num=${layer_num},is_infer=True,num_samples=256" \
--init_model_path=$models_in \
2>&1 | tee ${log}
# calculate the last 5 logs period time of 160(=32*5) 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 $end`
fps=`awk 'BEGIN{printf "%.2f",(160 / ('$end_sec' - '$start_sec'))}'`
echo "Last 160 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
# inference benchmark
for batchsize in 1 2 4 8 16; do
infer vgg 19 $batchsize
infer resnet 50 $batchsize
infer googlenet v1 $batchsize
infer alexnet 2 $batchsize
done
benchmark/paddle/image/run_openblas_train.sh 0 → 100755
浏览文件 @ dbd25805
set -e
function train() {
export OPENBLAS_NUM_THREADS=1
topology=$1
layer_num=$2
bs=$3
thread=`nproc`
# each trainer_count use only 1 core to avoid conflict
log="logs/train-${topology}-${layer_num}-${thread}openblas-${bs}.log"
args="batch_size=${bs},layer_num=${layer_num}"
config="${topology}.py"
paddle train --job=time \
--config=$config \
--use_mkldnn=False \
--use_gpu=False \
--trainer_count=$thread \
--log_period=3 \
--test_period=30 \
--config_args=$args \
2>&1 | tee ${log}
avg_time=`tail ${log} -n 1 | awk -F ' ' '{print $8}' | sed 's/avg=//'`
fps=`awk 'BEGIN{printf "%.2f",('$bs' / '$avg_time' * 1000)}'`
echo "FPS: $fps images/sec" 2>&1 | tee -a ${log}
}
if [ ! -f "train.list" ]; then
echo " " > train.list
fi
if [ ! -d "logs" ]; then
mkdir logs
fi
# training benchmark
for batchsize in 64 128 256; do
train vgg 19 $batchsize
train resnet 50 $batchsize
train googlenet v1 $batchsize
train alexnet 2 $batchsize
done
benchmark/paddle/image/vgg.py
浏览文件 @ dbd25805
......@@ -7,13 +7,15 @@ 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)
num_samples = get_config_arg('num_samples', int, 2560)
args = {
'height': height,
'width': width,
'color': True,
'num_class': num_class,
'is_infer': is_infer
'is_infer': is_infer,
'num_samples': num_samples
}
define_py_data_sources2(
"train.list" if not is_infer else None,
......
benchmark/paddle/rnn/imdb.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import print_function
import six.moves.cPickle as pickle
import gzip
......
benchmark/paddle/rnn/provider.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import io, os
import random
import numpy as np
......
benchmark/tensorflow/image/alexnet.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from six.moves import xrange # pylint: disable=redefined-builtin
from datetime import datetime
import math
......
benchmark/tensorflow/image/alexnet_multi_gpu.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from six.moves import xrange # pylint: disable=redefined-builtin
from datetime import datetime
import math
......
benchmark/tensorflow/image/googlenet.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from six.moves import xrange
from datetime import datetime
import math
......
benchmark/tensorflow/image/googlenet_multi_gpu.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from six.moves import xrange # pylint: disable=redefined-builtin
from datetime import datetime
import math
......
benchmark/tensorflow/image/smallnet_mnist_cifar.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from six.moves import xrange # pylint: disable=redefined-builtin
from datetime import datetime
import math
......
benchmark/tensorflow/machine_translation.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
benchmark/tensorflow/mnist.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import time
import numpy as np
import tensorflow as tf
import paddle.v2 as paddle
DTYPE = tf.float32
def parse_args():
parser = argparse.ArgumentParser("mnist model benchmark.")
parser.add_argument(
'--batch_size', type=int, default=128, help='The minibatch size.')
parser.add_argument(
'--iterations', type=int, default=35, help='The number of minibatches.')
parser.add_argument(
'--pass_num', type=int, default=5, help='The number of passes.')
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help='The device type.')
args = parser.parse_args()
return args
def run_benchmark(args):
def weight_variable(dtype, shape):
initial = tf.truncated_normal(shape, stddev=0.1, dtype=dtype)
return tf.Variable(initial)
def bias_variable(dtype, shape):
initial = tf.constant(0.1, shape=shape, dtype=dtype)
return tf.Variable(initial)
device = '/cpu:0' if args.device == 'CPU' else '/device:GPU:0'
with tf.device(device):
images = tf.placeholder(DTYPE, shape=(None, 28, 28, 1))
labels = tf.placeholder(tf.int64, shape=(None, ))
# conv1, relu, pool1
conv1_weights = weight_variable(DTYPE, [5, 5, 1, 20])
conv1_bias = bias_variable(DTYPE, [20])
conv1 = tf.nn.conv2d(
images, conv1_weights, strides=[1, 1, 1, 1], padding="VALID")
relu1 = tf.nn.relu(tf.nn.bias_add(conv1, conv1_bias))
pool1 = tf.nn.max_pool(
relu1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")
# conv2, relu, pool2
conv2_weights = weight_variable(DTYPE, [5, 5, 20, 50])
conv2_bias = bias_variable(DTYPE, [50])
conv2 = tf.nn.conv2d(
pool1, conv2_weights, strides=[1, 1, 1, 1], padding="VALID")
relu2 = tf.nn.relu(tf.nn.bias_add(conv2, conv2_bias))
pool2 = tf.nn.max_pool(
relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")
# FC
pool_shape = pool2.get_shape().as_list()
hidden_dim = reduce(lambda a, b: a * b, pool_shape[1:], 1)
reshape = tf.reshape(pool2, shape=(tf.shape(pool2)[0], hidden_dim))
fc_weights = weight_variable(DTYPE, [hidden_dim, 10])
fc_bias = bias_variable(DTYPE, [10])
logits = tf.matmul(reshape, fc_weights) + fc_bias
# Get prediction
prediction = tf.nn.softmax(logits)
# Loss
one_hot_labels = tf.one_hot(labels, depth=10)
cost = -tf.reduce_sum(tf.log(prediction) * one_hot_labels, [1])
avg_cost = tf.reduce_mean(cost)
# Get accuracy
correct = tf.equal(tf.argmax(prediction, 1), labels)
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))
# metrics, g_accuracy
with tf.variable_scope("reset_metrics_accuracy_scope") as scope:
g_accuracy = tf.metrics.accuracy(
labels, tf.argmax(
prediction, axis=1))
vars = tf.contrib.framework.get_variables(
scope, collection=tf.GraphKeys.LOCAL_VARIABLES)
g_accuracy_reset_op = tf.variables_initializer(vars)
# Optimizer
opt = tf.train.AdamOptimizer(
learning_rate=0.001, beta1=0.9, beta2=0.999)
train_op = opt.minimize(avg_cost)
# train_op = tf.train.AdamOptimizer(1e-4).minimize(avg_cost)
train_reader = paddle.batch(
paddle.dataset.mnist.train(), batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.dataset.mnist.test(), batch_size=args.batch_size)
def eval_test():
sess.run(g_accuracy_reset_op)
for batch_id, data in enumerate(test_reader()):
images_data = np.array(
map(lambda x: np.transpose(x[0].reshape([1, 28, 28]), axes=[1,2,0]), data)).astype("float32")
labels_data = np.array(map(lambda x: x[1], data)).astype("int64")
loss, acc, g_acc = sess.run(
[avg_cost, accuracy, g_accuracy],
feed_dict={images: images_data,
labels: labels_data})
return g_acc[1]
config = tf.ConfigProto(
intra_op_parallelism_threads=1, inter_op_parallelism_threads=1)
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
init_g = tf.global_variables_initializer()
init_l = tf.local_variables_initializer()
sess.run(init_g)
sess.run(init_l)
for pass_id in range(args.pass_num):
sess.run(g_accuracy_reset_op)
pass_start = time.time()
for batch_id, data in enumerate(train_reader()):
images_data = np.array(
map(lambda x: np.transpose(x[0].reshape([1, 28, 28]), axes=[1,2,0]), data)).astype("float32")
labels_data = np.array(map(lambda x: x[1], data)).astype(
"int64")
start = time.time()
_, loss, acc, g_acc = sess.run(
[train_op, avg_cost, accuracy, g_accuracy],
feed_dict={images: images_data,
labels: labels_data})
end = time.time()
print("pass=%d, batch=%d, loss=%f, error=%f, elapse=%f" %
(pass_id, batch_id, loss, 1 - acc, (end - start) / 1000))
pass_end = time.time()
test_avg_acc = eval_test()
print(
"pass=%d, training_avg_accuracy=%f, test_avg_acc=%f, elapse=%f"
% (pass_id, g_acc[1], test_avg_acc,
(pass_end - pass_start) / 1000))
def print_arguments(args):
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).iteritems()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
if __name__ == '__main__':
args = parse_args()
print_arguments(args)
run_benchmark(args)
benchmark/tensorflow/resnet.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
based on https://github.com/tensorflow/models/blob/master/official/resnet/resnet_model.py
Get help: python resnet.py --help
See performance on flowers: python resnet.py
Train on cifar10: python resnet.py --data=cifar10 --with_test
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import time
import numpy as np
import paddle.v2 as paddle
import tensorflow as tf
DTYPE = tf.float32
def parse_args():
parser = argparse.ArgumentParser('Convolution model benchmark.')
parser.add_argument(
'--model',
type=str,
choices=['resnet'],
default='resnet',
help='The model architecture.')
parser.add_argument(
'--batch_size', type=int, default=32, help='The minibatch size.')
parser.add_argument(
'--use_fake_data',
action='store_true',
help='use real data or fake data')
parser.add_argument(
'--skip_batch_num',
type=int,
default=5,
help='The first num of minibatch num to skip, for better performance test'
)
parser.add_argument(
'--iterations',
type=int,
default=105,
help='The number of minibatches.')
parser.add_argument(
'--pass_num', type=int, default=300, help='The number of passes.')
parser.add_argument(
'--order',
type=str,
default='NHWC',
choices=['NCHW', 'NHWC'],
help='The data order, now only support NCHW.')
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help='The device type.')
parser.add_argument(
'--data',
type=str,
default='flowers102',
choices=['flowers102', 'cifar10'],
help='The kinds of data.')
parser.add_argument(
'--infer_only', action='store_true', help='If set, run forward only.')
parser.add_argument(
'--use_cprof', action='store_true', help='If set, use cProfile.')
parser.add_argument(
'--with_test',
action='store_true',
help='If set, test the testset during training.')
parser.add_argument(
'--use_nvprof',
action='store_true',
help='If set, use nvprof for CUDA.')
args = parser.parse_args()
return args
def print_arguments(args):
vars(args)['use_nvprof'] = (vars(args)['use_nvprof'] and
vars(args)['device'] == 'GPU')
vars(args)['iterations'] = vars(args)['pass_num'] * 1000 if vars(args)[
'with_test'] else vars(args)['iterations']
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).iteritems()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def fixed_padding(inputs, kernel_size, data_format):
"""Pads the input along the spatial dimensions independently of input size.
Args:
inputs: A tensor of size [batch, channels, height_in, width_in] or
[batch, height_in, width_in, channels] depending on data_format.
kernel_size: The kernel to be used in the conv2d or max_pool2d operation.
Should be a positive integer.
data_format: The input format ('channels_last' or 'channels_first').
Returns:
A tensor with the same format as the input with the data either intact
(if kernel_size == 1) or padded (if kernel_size > 1).
"""
pad_total = kernel_size - 1
pad_beg = pad_total // 2
pad_end = pad_total - pad_beg
if data_format == 'channels_first':
padded_inputs = tf.pad(inputs, [[0, 0], [0, 0], [pad_beg, pad_end],
[pad_beg, pad_end]])
else:
padded_inputs = tf.pad(inputs, [[0, 0], [pad_beg, pad_end],
[pad_beg, pad_end], [0, 0]])
return padded_inputs
def conv2d_fixed_padding(inputs, filters, kernel_size, strides, data_format):
"""Strided 2-D convolution with explicit padding."""
# The padding is consistent and is based only on `kernel_size`, not on the
# dimensions of `inputs` (as opposed to using `tf.layers.conv2d` alone).
# This is consistent with PaddlePaddle.
# In addition, the calculation for output size in TensorFlow can refer:
# https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/framework/common_shape_fns.cc
if strides > 1:
inputs = fixed_padding(inputs, kernel_size, data_format)
return tf.layers.conv2d(
inputs=inputs,
filters=filters,
kernel_size=kernel_size,
strides=strides,
padding=('SAME' if strides == 1 else 'VALID'),
use_bias=False,
kernel_initializer=tf.variance_scaling_initializer(),
data_format=data_format)
def conv_bn(inputs,
filters,
kernel_size,
strides,
is_training,
data_format,
act=True):
# def conv2d_fixed_padding(inputs, filters, kernel_size, strides, data_format):
# set fused=True for a significant performance boost. See
# https://www.tensorflow.org/performance/performance_guide#common_fused_ops
inputs = conv2d_fixed_padding(
inputs=inputs,
filters=filters,
kernel_size=kernel_size,
strides=strides,
data_format=data_format)
inputs = tf.layers.batch_normalization(
inputs=inputs,
axis=1 if data_format == 'channels_first' else 3,
momentum=0.9,
epsilon=1e-05,
center=True,
scale=True,
training=is_training,
fused=True)
if act:
inputs = tf.nn.relu(inputs)
return inputs
def basicblock(inputs, filters, is_training, projection_shortcut, strides,
data_format):
shortcut = inputs
if projection_shortcut is not None:
shortcut = projection_shortcut(inputs)
inputs = conv_bn(inputs, filters, 3, strides, is_training, data_format)
inputs = conv_bn(inputs, filters, 3, 1, is_training, data_format, act=False)
inputs = inputs + shortcut
inputs = tf.nn.relu(inputs)
return inputs
def bottleneck(inputs, filters, is_training, projection_shortcut, strides,
data_format):
shortcut = inputs
if projection_shortcut is not None:
shortcut = projection_shortcut(inputs)
inputs = conv_bn(inputs, filters, 1, strides, is_training, data_format)
inputs = conv_bn(inputs, filters, 3, 1, is_training, data_format, act=False)
inputs = conv_bn(
inputs, filters * 4, 1, 1, is_training, data_format, act=False)
inputs = inputs + shortcut
inputs = tf.nn.relu(inputs)
return inputs
def block_layer(inputs, filters, block_fn, blocks, strides, is_training, name,
data_format):
# Bottleneck blocks end with 4x the number of filters as they start with
filters_out = 4 * filters if block_fn is bottleneck else filters
def projection_shortcut(inputs):
return conv2d_fixed_padding(
inputs=inputs,
filters=filters_out,
kernel_size=1,
strides=strides,
data_format=data_format)
# Only the first block per block_layer uses projection_shortcut and strides
inputs = block_fn(inputs, filters, is_training, projection_shortcut,
strides, data_format)
for _ in range(1, blocks):
inputs = block_fn(inputs, filters, is_training, None, 1, data_format)
return tf.identity(inputs, name)
def resnet_imagenet(depth, class_dim, data_format):
"""Returns the ResNet model for a given size and number of output classes."""
def resnet_generator(block_fn,
layers,
num_classes,
data_format='channels_last'):
if data_format is None:
data_format = ('channels_first'
if tf.test.is_built_with_cuda() else 'channels_last')
def model(inputs, is_training):
"""Constructs the ResNet model given the inputs."""
if data_format == 'channels_first':
# Convert the inputs from channels_last (NHWC) to channels_first (NCHW).
# This provides a large performance boost on GPU. See
# https://www.tensorflow.org/performance/performance_guide#data_formats
inputs = tf.transpose(inputs, [0, 3, 1, 2])
inputs = conv_bn(inputs, 64, 7, 2, is_training, data_format)
inputs = tf.identity(inputs, 'initial_conv')
inputs = tf.layers.max_pooling2d(
inputs=inputs,
pool_size=3,
strides=2,
padding='SAME',
data_format=data_format)
inputs = tf.identity(inputs, 'initial_max_pool')
inputs = block_layer(inputs, 64, block_fn, layers[0], 1,
is_training, 'block_layer1', data_format)
inputs = block_layer(inputs, 128, block_fn, layers[1], 2,
is_training, 'block_layer2', data_format)
inputs = block_layer(inputs, 256, block_fn, layers[2], 2,
is_training, 'block_layer3', data_format)
inputs = block_layer(inputs, 512, block_fn, layers[3], 2,
is_training, 'block_layer4', data_format)
inputs = tf.layers.average_pooling2d(
inputs=inputs,
pool_size=7,
strides=1,
padding='VALID',
data_format=data_format)
inputs = tf.identity(inputs, 'final_avg_pool')
inputs = tf.reshape(inputs,
[-1, 512 if block_fn is basicblock else 2048])
inputs = tf.layers.dense(inputs=inputs, units=num_classes)
inputs = tf.identity(inputs, 'final_dense')
return inputs
return model
model_params = {
18: {
'block': basicblock,
'layers': [2, 2, 2, 2]
},
34: {
'block': basicblock,
'layers': [3, 4, 6, 3]
},
50: {
'block': bottleneck,
'layers': [3, 4, 6, 3]
},
101: {
'block': bottleneck,
'layers': [3, 4, 23, 3]
},
152: {
'block': bottleneck,
'layers': [3, 8, 36, 3]
},
200: {
'block': bottleneck,
'layers': [3, 24, 36, 3]
}
}
if depth not in model_params:
raise ValueError('Not a valid depth:', depth)
params = model_params[depth]
return resnet_generator(params['block'], params['layers'], class_dim,
data_format)
def resnet_cifar10(depth, num_classes, data_format):
if depth % 6 != 2:
raise ValueError('depth must be 6n + 2:', depth)
num_blocks = (depth - 2) // 6
if data_format is None:
data_format = ('channels_first'
if tf.test.is_built_with_cuda() else 'channels_last')
def model(inputs, is_training):
inputs = conv_bn(inputs, 16, 3, 1, is_training, data_format)
inputs = tf.identity(inputs, 'initial_conv')
inputs = block_layer(inputs, 16, basicblock, num_blocks, 1, is_training,
'block_layer1', data_format)
inputs = block_layer(inputs, 32, basicblock, num_blocks, 2, is_training,
'block_layer2', data_format)
inputs = block_layer(inputs, 64, basicblock, num_blocks, 2, is_training,
'block_layer3', data_format)
inputs = tf.layers.average_pooling2d(
inputs=inputs,
pool_size=8,
strides=1,
padding='VALID',
data_format=data_format)
inputs = tf.identity(inputs, 'final_avg_pool')
inputs = tf.reshape(inputs, [-1, 64])
inputs = tf.layers.dense(inputs=inputs, units=num_classes)
inputs = tf.identity(inputs, 'final_dense')
return inputs
return model
def run_benchmark(args, data_format='channels_last', device='/cpu:0'):
"""Our model_fn for ResNet to be used with our Estimator."""
class_dim = 1000
dshape = (None, 224, 224, 3)
pdshape = (3, 224, 224)
if args.data == 'flowers102':
class_dim = 102
dshape = (None, 224, 224, 3)
pdshape = (3, 224, 224)
elif args.data == 'cifar10':
class_dim = 10
dshape = (None, 32, 32, 3)
pdshape = (3, 32, 32)
with tf.device(device):
images = tf.placeholder(DTYPE, shape=dshape)
labels = tf.placeholder(tf.int64, shape=(None, ))
is_training = tf.placeholder('bool')
onehot_labels = tf.one_hot(labels, depth=class_dim)
network = resnet_cifar10(
32, class_dim,
data_format) if args.data == 'cifar10' else resnet_imagenet(
50, class_dim, data_format)
logits = network(inputs=images, is_training=is_training)
cross_entropy = tf.losses.softmax_cross_entropy(
logits=logits, onehot_labels=onehot_labels)
avg_cost = tf.reduce_mean(cross_entropy)
correct = tf.equal(tf.argmax(logits, 1), labels)
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))
lr = 0.1 if args.data == 'cifar10' else 0.01
optimizer = tf.train.MomentumOptimizer(learning_rate=lr, momentum=0.9)
# Batch norm requires update_ops to be added as a train_op dependency.
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
with tf.control_dependencies(update_ops):
train_op = optimizer.minimize(avg_cost)
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10()
if args.data == 'cifar10' else paddle.dataset.flowers.train(),
buf_size=5120),
batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.dataset.cifar.test10()
if args.data == 'cifar10' else paddle.dataset.flowers.test(),
batch_size=100)
def test():
test_accs = []
for batch_id, data in enumerate(test_reader()):
test_images = np.array(
map(lambda x: np.transpose(x[0].reshape(pdshape),
axes=[1, 2, 0]), data)).astype("float32")
test_labels = np.array(map(lambda x: x[1], data)).astype('int64')
test_accs.append(
accuracy.eval(feed_dict={
images: test_images,
labels: test_labels,
is_training: False
}))
print("Pass = %d, Train performance = %f imgs/s, Test accuracy = %f\n" %
(pass_id, num_samples / train_elapsed, np.mean(test_accs)))
config = tf.ConfigProto(
intra_op_parallelism_threads=1, inter_op_parallelism_threads=1)
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
init_g = tf.global_variables_initializer()
init_l = tf.local_variables_initializer()
sess.run(init_g)
sess.run(init_l)
if args.use_fake_data:
data = train_reader().next()
images_data = np.array(
map(lambda x: np.transpose(x[0].reshape(pdshape),
axes=[1, 2, 0]), data)).astype("float32")
labels_data = np.array(map(lambda x: x[1], data)).astype('int64')
iters, num_samples, start_time = 0, 0, 0.0
for pass_id in range(args.pass_num):
if iters == args.iterations:
break
train_accs = []
train_losses = []
for batch_id, data in enumerate(train_reader()):
if iters == args.skip_batch_num:
start_time = time.time()
num_samples = 0
if iters == args.iterations:
break
if not args.use_fake_data:
images_data = np.array(
map(lambda x: np.transpose(x[0].reshape(pdshape),
axes=[1, 2, 0]), data)).astype("float32")
labels_data = np.array(map(lambda x: x[1], data)).astype(
'int64')
_, loss, acc = sess.run([train_op, avg_cost, accuracy],
feed_dict={
images: images_data,
labels: labels_data,
is_training: True
})
iters += 1
train_accs.append(acc)
train_losses.append(loss)
num_samples += len(data)
print("Pass=%d, Iter=%d, Loss=%f, Accuray=%f\n" %
(pass_id, iters, loss, acc))
train_elapsed = time.time() - start_time
print("Pass=%d, Loss=%f, Accuray=%f\n" %
(pass_id, np.mean(train_losses), np.mean(train_accs)))
# evaluation
if args.with_test:
test()
if not args.with_test:
duration = time.time() - start_time
examples_per_sec = num_samples / duration
sec_per_batch = duration / (iters - args.skip_batch_num)
print('Total examples: %d, total time: %.5f' %
(num_samples, duration))
print('%.5f examples/sec, %.5f sec/batch' %
(examples_per_sec, sec_per_batch))
if __name__ == '__main__':
args = parse_args()
print_arguments(args)
if tf.test.is_built_with_cuda():
device = '/device:GPU:0'
if args.order == 'NHWC':
data_format = 'channels_last'
else:
data_format = 'channels_first'
else:
device = '/cpu:0'
if args.order == 'NHWC':
data_format = 'channels_last'
else:
raise ValueError('Only support NHWC order in CPU mode')
run_benchmark(args, data_format, device)
benchmark/tensorflow/rnn/reader.py
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os.path
import io
import numpy as np
......
benchmark/tensorflow/stacked_dynamic_lstm.py 0 → 100644
浏览文件 @ dbd25805
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import numpy as np
import argparse
import time
import tensorflow as tf
import paddle.v2 as paddle
def parse_args():
parser = argparse.ArgumentParser("LSTM model benchmark.")
parser.add_argument(
'--batch_size',
type=int,
default=32,
help='The sequence number of a batch data. (default: %(default)d)')
parser.add_argument(
'--stacked_num',
type=int,
default=5,
help='Number of lstm layers to stack. (default: %(default)d)')
parser.add_argument(
'--embedding_dim',
type=int,
default=512,
help='Dimension of embedding table. (default: %(default)d)')
parser.add_argument(
'--hidden_dim',
type=int,
default=512,
help='Hidden size of lstm unit. (default: %(default)d)')
parser.add_argument(
'--pass_num',
type=int,
default=10,
help='Epoch number to train. (default: %(default)d)')
parser.add_argument(
'--learning_rate',
type=float,
default=0.0002,
help='Learning rate used to train. (default: %(default)f)')
parser.add_argument(
'--infer_only', action='store_true', help='If set, run forward only.')
args = parser.parse_args()
return args
def print_arguments(args):
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).iteritems()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def dynamic_lstm_model(dict_size,
embedding_dim,
hidden_dim,
stacked_num,
class_num=2,
is_train=True):
word_idx = tf.placeholder(tf.int64, shape=[None, None])
sequence_length = tf.placeholder(tf.int64, shape=[None, ])
embedding_weights = tf.get_variable('word_embeddings',
[dict_size, embedding_dim])
embedding = tf.nn.embedding_lookup(embedding_weights, word_idx)
lstm_cell = tf.nn.rnn_cell.LSTMCell(
num_units=hidden_dim, use_peepholes=False)
stacked_cell = tf.nn.rnn_cell.MultiRNNCell([lstm_cell] * stacked_num)
# final_state [LSTMTuple(c, h), LSTMTuple(c, h) ...] total stacked_num LSTMTuples
_, final_state = tf.nn.dynamic_rnn(
cell=stacked_cell,
inputs=embedding,
dtype=tf.float32,
sequence_length=sequence_length)
w = tf.Variable(
tf.truncated_normal([hidden_dim, class_num]), dtype=tf.float32)
bias = tf.Variable(
tf.constant(
value=0.0, shape=[class_num], dtype=tf.float32))
prediction = tf.matmul(final_state[-1][1], w) + bias
if not is_train:
return (word_idx, sequence_length), tf.nn.softmax(prediction)
label = tf.placeholder(tf.int64, shape=[None, ])
loss = tf.nn.softmax_cross_entropy_with_logits(
labels=tf.one_hot(label, 2), logits=prediction)
avg_loss = tf.reduce_mean(loss)
correct_count = tf.equal(tf.argmax(prediction, 1), label)
acc = tf.reduce_mean(tf.cast(correct_count, tf.float32))
with tf.variable_scope("reset_metrics_accuracy_scope") as scope:
g_acc = tf.metrics.accuracy(label, tf.argmax(prediction, axis=1))
vars = tf.contrib.framework.get_variables(
scope, collection=tf.GraphKeys.LOCAL_VARIABLES)
reset_op = tf.variables_initializer(vars)
return (word_idx, sequence_length, label), avg_loss, acc, g_acc, reset_op
def padding_data(data, padding_size, value):
data = data + [value] * padding_size
return data[:padding_size]
def train(args):
word_dict = paddle.dataset.imdb.word_dict()
dict_size = len(word_dict)
feeding_list, avg_loss, acc, g_acc, reset_op = dynamic_lstm_model(
dict_size, args.embedding_dim, args.hidden_dim, args.stacked_num)
adam_optimizer = tf.train.AdamOptimizer(learning_rate=args.learning_rate)
train_op = adam_optimizer.minimize(avg_loss)
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.imdb.train(word_dict), buf_size=25000),
batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.imdb.test(word_dict), buf_size=25000),
batch_size=args.batch_size)
def do_validation(sess):
sess.run(reset_op)
for batch_id, data in enumerate(test_reader()):
word_idx = map(lambda x: x[0], data)
sequence_length = np.array(
[len(seq) for seq in word_idx]).astype('int64')
maxlen = np.max(sequence_length)
word_idx = [padding_data(seq, maxlen, 0) for seq in word_idx]
word_idx = np.array(word_idx).astype('int64')
label = np.array(map(lambda x: x[1], data)).astype('int64')
_, loss, fetch_acc, fetch_g_acc = sess.run(
[train_op, avg_loss, acc, g_acc],
feed_dict={
feeding_list[0]: word_idx,
feeding_list[1]: sequence_length,
feeding_list[2]: label
})
return fetch_g_acc[1]
config = tf.ConfigProto(
intra_op_parallelism_threads=1, inter_op_parallelism_threads=1)
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
init_g = tf.global_variables_initializer()
init_l = tf.local_variables_initializer()
sess.run(init_l)
sess.run(init_g)
for pass_id in xrange(args.pass_num):
# clear accuracy local variable
sess.run(reset_op)
pass_start_time = time.time()
words_seen = 0
for batch_id, data in enumerate(train_reader()):
word_idx = map(lambda x: x[0], data)
sequence_length = np.array(
[len(seq) for seq in word_idx]).astype('int64')
words_seen += np.sum(sequence_length)
maxlen = np.max(sequence_length)
word_idx = [padding_data(seq, maxlen, 0) for seq in word_idx]
word_idx = np.array(word_idx).astype('int64')
label = np.array(map(lambda x: x[1], data)).astype('int64')
_, loss, fetch_acc, fetch_g_acc = sess.run(
[train_op, avg_loss, acc, g_acc],
feed_dict={
feeding_list[0]: word_idx,
feeding_list[1]: sequence_length,
feeding_list[2]: label
})
print("pass_id=%d, batch_id=%d, loss: %f, acc: %f, avg_acc: %f"
% (pass_id, batch_id, loss, fetch_acc, fetch_g_acc[1]))
pass_end_time = time.time()
time_consumed = pass_end_time - pass_start_time
words_per_sec = words_seen / time_consumed
test_acc = do_validation(sess)
print("pass_id=%d, test_acc: %f, words/s: %f, sec/pass: %f" %
(pass_id, test_acc, words_per_sec, time_consumed))
if __name__ == '__main__':
args = parse_args()
print_arguments(args)
if args.infer_only:
pass
else:
train(args)
benchmark/tensorflow/vgg.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/cblas.cmake
浏览文件 @ dbd25805
......@@ -3,7 +3,7 @@
# 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 MKLML, ATLAS, OPENBLAS, REFERENCE
# CBLAS_PROVIDER # one of MKLML, 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.
......@@ -17,7 +17,7 @@ if(WITH_MKLML AND MKLML_INC_DIR AND MKLML_LIB)
set(CBLAS_INC_DIR ${MKLML_INC_DIR})
set(CBLAS_LIBRARIES ${MKLML_LIB})
add_definitions(-DPADDLE_USE_MKLML)
add_definitions(-DPADDLE_WITH_MKLML)
add_definitions(-DLAPACK_FOUND)
message(STATUS "Found cblas and lapack in MKLML "
......@@ -25,42 +25,6 @@ if(WITH_MKLML AND MKLML_INC_DIR AND MKLML_LIB)
return()
endif()
## Then find atlas.
set(ATLAS_ROOT $ENV{ATLAS_ROOT} CACHE PATH "Folder contains Atlas")
set(ATLAS_INCLUDE_SEARCH_PATHS
${ATLAS_ROOT}/include
/usr/include
/usr/include/atlas)
set(ATLAS_LIB_SEARCH_PATHS
${ATLAS_ROOT}/lib
/usr/lib
/usr/lib/blas/atlas
/usr/lib/atlas
/usr/lib/atlas-base # special for ubuntu 14.04.
)
find_path(ATLAS_INC_DIR NAMES cblas.h
PATHS ${ATLAS_INCLUDE_SEARCH_PATHS})
find_path(ATLAS_CLAPACK_INC_DIR NAMES clapack.h
PATHS ${ATLAS_INCLUDE_SEARCH_PATHS})
find_library(ATLAS_CBLAS_LIB NAMES cblas libcblas.so.3
PATHS ${ATLAS_LIB_SEARCH_PATHS})
find_library(ATLAS_CLAPACK_LIB NAMES lapack_atlas liblapack_atlas.so.3
PATHS ${ATLAS_LIB_SEARCH_PATHS})
if(ATLAS_CLAPACK_INC_DIR AND ATLAS_INC_DIR AND ATLAS_CBLAS_LIB AND ATLAS_CLAPACK_LIB)
set(CBLAS_FOUND ON)
set(CBLAS_PROVIDER ATLAS)
set(CBLAS_INC_DIR ${ATLAS_INC_DIR} ${ATLAS_CLAPACK_INC_DIR})
set(CBLAS_LIBRARIES ${ATLAS_CLAPACK_LIB} ${ATLAS_CBLAS_LIB})
add_definitions(-DPADDLE_USE_ATLAS)
add_definitions(-DLAPACK_FOUND)
message(STATUS "Found ATLAS (include: ${ATLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
message(STATUS "Found lapack in ATLAS (include: ${ATLAS_CLAPACK_INC_DIR})")
return()
endif()
## Then find openblas.
set(OPENBLAS_ROOT $ENV{OPENBLAS_ROOT} CACHE PATH "Folder contains Openblas")
set(OPENBLAS_INCLUDE_SEARCH_PATHS
......@@ -98,29 +62,33 @@ endif()
## Then find the reference-cblas. www.netlib.org/blas/
set(REFERENCE_CBLAS_ROOT $ENV{REFERENCE_CBLAS_ROOT} CACHE PATH
"Folder contains reference-cblas")
set(REFERENCE_CBLAS_INCLUDE_SEARCH_PATHS
${REFERENCE_CBLAS_ROOT}/include
/usr/include
/usr/include/cblas
)
set(REFERENCE_CBLAS_LIB_SEARCH_PATHS
${REFERENCE_CBLAS_ROOT}/lib
/usr/lib
/usr/lib/blas/reference/
/usr/lib/reference/
)
if(NOT CMAKE_CROSSCOMPILING)
set(REFERENCE_CBLAS_INCLUDE_SEARCH_PATHS
${REFERENCE_CBLAS_ROOT}/include
/usr/include
/usr/include/cblas
)
set(REFERENCE_CBLAS_LIB_SEARCH_PATHS
${REFERENCE_CBLAS_ROOT}/lib
/usr/lib
/usr/lib/blas/reference/
/usr/lib/reference/
)
else()
# Disable the finding of reference cblas under host's system path
set(REFERENCE_CBLAS_INCLUDE_SEARCH_PATHS ${REFERENCE_CBLAS_ROOT}/include)
set(REFERENCE_CBLAS_LIB_SEARCH_PATHS ${REFERENCE_CBLAS_ROOT}/lib)
endif()
find_path(REFERENCE_CBLAS_INCLUDE_DIR NAMES cblas.h PATHS
${REFERENCE_CBLAS_INCLUDE_SEARCH_PATHS})
find_library(REFERENCE_CBLAS_LIBRARY NAMES cblas PATHS
${REFERENCE_CBLAS_LIB_SEARCH_PATHS})
if (REFERENCE_CBLAS_INCLUDE_DIR AND REFERENCE_CBLAS_LIBRARY)
if(REFERENCE_CBLAS_INCLUDE_DIR AND REFERENCE_CBLAS_LIBRARY)
set(CBLAS_FOUND ON)
set(CBLAS_PROVIDER REFERENCE)
set(CBLAS_INC_DIR ${REFERENCE_CBLAS_INCLUDE_DIR})
......
cmake/configure.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/cpplint.cmake 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
cmake/cross_compiling/android.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/cross_compiling/host.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/cross_compiling/ios.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/cuda.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/cupti.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/boost.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/cares.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/eigen.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/gflags.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/glog.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/grpc.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/gtest.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/mkldnn.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/mklml.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/nccl.cmake 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
cmake/external/openblas.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/protobuf.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/pybind11.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/python.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/snappy.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/snappystream.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/swig.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/threadpool.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/warpctc.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/external/zlib.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/generic.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/hip.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/inference_lib.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
cmake/make_resource.py
浏览文件 @ dbd25805
此差异已折叠。
cmake/system.cmake
浏览文件 @ dbd25805
此差异已折叠。
cmake/tensorrt.cmake 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
doc/api/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/data_provider/dataprovider_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/data_provider/dataprovider_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/data_provider/src/mnist_config.py 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/data_provider/src/mnist_train.txt 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/data_provider/src/train.list 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/predict/src/predict_sample.py 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/predict/swig_py_paddle_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v1/predict/swig_py_paddle_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/config/layer.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/data/data_reader.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/data/dataset.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/data_feeder.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/evaluator.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/executor.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/initializer.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/layers.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/nets.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/optimizer.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/param_attr.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/profiler.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/api/v2/fluid/regularizer.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/api.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/auto_gradient_check.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/block.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/build_system/README.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/evaluator.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/executor.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/file_manager/README.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/file_manager/pfs/pfsclient.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/file_manager/src/filemanager.graffle 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/file_manager/src/filemanager.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/float16.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/functions_operators_layers.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/gan_api.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/images/replica.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/images/two_phase_commit.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/mkldnn/README.MD 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/model_format.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/ops/rnn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/ops/sequence_decoder.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/optimizer.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/parameter_average.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/program.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/python_api.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/refactor/parameter_server.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/refactor/session.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/refactor/src/local_architecture.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/refactorization.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/regularization.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/releasing_process.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/scope.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/speech/README.MD 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/design/var_desc.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/build_and_install/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/local/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/local/src/reduce_min_pool_size.py 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/local/src/word2vec_config.py 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/local/src/word2vec_dataprovider.py 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/model/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/faq/parameter/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/fluid/CMakeLists.txt 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/CMakeLists.txt 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/clip.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/data/data_reader.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/data/dataset.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/data_feeder.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/evaluator.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/executor.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/gen_doc.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/gen_doc.sh 0 → 100755
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/initializer.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/io.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/layers.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/metrics.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/nets.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/optimizer.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/param_attr.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/profiler.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/api/regularizer.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/build_from_source_cn.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/build_from_source_en.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/docker_install_cn.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/docker_install_en.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/index_cn.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/index_en.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/paddleci.png 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/pip_install_cn.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/build_and_install/pip_install_en.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/algorithm/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/algorithm/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/algorithm/parameter_average.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/README.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/block.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/cpp_data_feeding.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/executor.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/functions_operators_layers.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/images/multiple_reader.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/images/readers.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/lod_tensor.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/parallel_executor.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/program.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/scope.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/tensor.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concepts/var_desc.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/channel.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/concurrent_programming.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/csp.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/go_op.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/images/channel_recv.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/images/channel_send.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/images/select_op_workflow.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/parallel_do.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/concurrent/select_op.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/data_type/float16.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/data_type/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/data_type/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/README.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/async_update.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/distributed_architecture.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/mpi_enabled_design.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/multi_cpu.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/parameter_server.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/async_pserver.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/async_pserver.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/async_update.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/async_update.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/local_architecture.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/lookup_table.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/lookup_table_training.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/mpi_module.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/multi-threads.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/remote_executor.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/remote_executor.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/sparse_update.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dist_train/src/sparse_update.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dynamic_rnn/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dynamic_rnn/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dynamic_rnn/rnn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dynamic_rnn/rnn_design.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/dynamic_rnn/rnn_design_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/execution/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/execution/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/execution/switch.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/interface/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/interface/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/memory/README.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/memory/images/control_flow_graph.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/memory/images/dataflow_equations.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/memory/images/deep_learning.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/memory/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/memory/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/memory/memory_optimization.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/backward.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/batch_norm_op.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/evaluator.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/net_op_design.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/optimizer.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/python_api.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/modules/regularization.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/api.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/fluid-compiler.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/fluid-compiler.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/fluid.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/fluid_compiler.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/motivation/refactorization.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/multi_devices/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/multi_devices/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/multi_devices/kernel_hint_design.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/multi_devices/kernel_selection.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/multi_devices/operator_kernel_type.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/network/deep_speech_2.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/network/images/beam_search.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/network/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/network/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/network/sequence_decoder.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/onnx/images/project_structure.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/onnx/onnx_convertor.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/others/auto_gradient_check.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/design/others/gan_api.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/api_doc_std_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/api_doc_std_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/ci_build_whl.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/contribute_to_paddle_cn.md 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/contribute_to_paddle_en.md 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/name_convention.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/new_op_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/new_op_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/new_op_kernel.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/op_markdown_format.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/releasing_process_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/releasing_process_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/src/fc.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/support_new_device.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/use_eigen_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/use_eigen_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/write_docs_cn.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/dev/write_docs_en.rst 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/faq/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/faq/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/getstarted/concepts/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/getstarted/concepts/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/getstarted/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/getstarted/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/getstarted/quickstart_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/getstarted/quickstart_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/cluster/fluid_cluster_train_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/cluster/fluid_cluster_train_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/cluster/nccl2_rdma_training.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/inference/inference_support_in_fluid.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/benchmark/README.md 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/benchmark/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/benchmark/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/benchmark/vgg16/README.md 0 → 120000
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/cpu_profiling_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/cpu_profiling_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/timeline.jpeg 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/timeline.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/optimization/tracing.jpeg 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/performance/error_clip.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/performance/images/profiler.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/performance/profiler.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/third_party/images/multigpu_allreduce.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/third_party/mkldnn_fluid.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/howto/third_party/paddle_nccl.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/1.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/2.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/2_level_rnn.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/2_level_rnn.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/3.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/4.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/LoDTensor.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/asgd.gif 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/batch_norm_fork.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/batch_norm_fork.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/batch_norm_op_kernel.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/beam_search.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/ci_build_whl.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/compile_run_time.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/compiler.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/compiler.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/control_flow_graph.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/dataflow_equations.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/dcgan.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/deep_learning.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/dist-graph.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/dist-graph.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/distributed_architecture.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/distributed_architecture.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/ds2_network.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/executor.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/feed_forward.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/feed_forward_regularized.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/fluid-compiler.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/fluid-compiler.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/fluid_examples.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/fluid_module_1.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/fluid_module_2.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/graph_construction_example.bash 0 → 100755
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/graph_construction_example.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/graph_construction_example_all.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/l1_regularization.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/l2_regularization.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/layer.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/local-graph.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/local-graph.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/local_architecture.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/local_architecture.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/lookup_table.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/lookup_table_training.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/loss_equation.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/multi-threads.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/multi-threads@3x.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/multigpu_allreduce.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/multigpu_allreduce.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/multigpu_before_convert.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/multigpu_before_convert.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/multiple_reader.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/op.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/op_op_with_kern_class_diagram.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/op_with_kernel.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/operator1.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/operator2.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/paddle-compile.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/paddle-compile.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/place.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/pprof_1.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/pprof_2.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/print_fluid_program.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/profiler.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/program_desc1.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/program_desc2.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/raw_input.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/readers.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/remote_executor.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/remote_executor.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/rnn.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/rnn.jpg 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/rnn.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/rnn_2level_data.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/rnn_2level_data.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/scope_variable_tensor.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/single-thread@3x.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/sorted_input.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/sparse_update.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/sparse_update.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/test.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/test.dot.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/theta_star.gif 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/timeline.jpeg 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/tracing.jpeg 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/transpiler.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/images/user_interface.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/fluid/read_source.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/getstarted/build_and_install/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/getstarted/build_and_install/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/getstarted/concepts/src/train.py 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/getstarted/concepts/use_concepts_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/getstarted/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/getstarted/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/deep_model/rnn/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/deep_model/rnn/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/build_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/build_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/contribute_to_paddle_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/new_layer_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/new_layer_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/new_op_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/new_op_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/use_eigen_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/use_eigen_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/write_docs_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/dev/write_docs_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/optimization/cpu_profiling.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/optimization/cpu_profiling_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/optimization/gpu_profiling_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/optimization/gpu_profiling_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/cluster/cluster_train_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/cluster/cluster_train_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/cmd_parameter/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/cmd_parameter/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/k8s_aws_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/k8s_basis_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/k8s_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/k8s_distributed_cn.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/k8s_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/src/Dockerfile 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/src/k8s-paddle-arch.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/howto/usage/k8s/src/k8s_train/Dockerfile 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/index_en.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/mobile/CMakeLists.txt 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/mobile/cross_compiling_for_ios_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/mobile/index_cn.rst
浏览文件 @ dbd25805
此差异已折叠。
doc/mobile/index_en.rst
浏览文件 @ dbd25805
此差异已折叠。
doc/templates/conf.py.cn.in
浏览文件 @ dbd25805
此差异已折叠。
doc/templates/conf.py.en.in
浏览文件 @ dbd25805
此差异已折叠。
doc/templates/layout.html
浏览文件 @ dbd25805
此差异已折叠。
doc/v1_api_tutorials/README.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/v1_api_tutorials/gan/gan.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/v1_api_tutorials/gan/index_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/v1_api_tutorials/gan/mnist_sample.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/v1_api_tutorials/gan/uniform_sample.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/v1_api_tutorials/quick_start/index_cn.rst 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/v1_api_tutorials/quick_start/index_en.md 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc/v2/CMakeLists.txt 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/CMakeLists.txt 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/config/layer.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/data.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/data/data_reader.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/data/dataset.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/data/image.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/api/overview.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/build_from_source_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/build_from_source_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/docker_install_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/docker_install_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/pip_install_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/build_and_install/pip_install_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/design/interface/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/design/interface/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/design/mkl/mkl_packed.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/design/mkl/mkldnn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/contribute_to_paddle_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/new_layer_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/new_layer_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/src/doc_en.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/write_docs_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/dev/write_docs_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/build_and_install/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/build_and_install/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/cluster/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/local/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/local/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/local/src/reduce_min_pool_size.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/local/src/word2vec_config.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/local/src/word2vec_dataprovider.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/model/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/model/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/parameter/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/faq/parameter/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/concepts/src/infer.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/concepts/src/train.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/concepts/use_concepts_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/concepts/use_concepts_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/quickstart_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/getstarted/quickstart_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/compile_paddle_lib_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/compile_paddle_lib_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/images/csr.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/images/sequence_data.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/images/workflow_of_CAPI.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/organization_of_the_inputs_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/organization_of_the_inputs_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/workflow_of_capi_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/capi/workflow_of_capi_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/cmd_argument_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/cmd_argument_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/fabric_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/fabric_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/k8s_aws_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/k8s_aws_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/k8s_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/k8s_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/openmpi_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/multi_cluster/openmpi_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/preparations_cn.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/preparations_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/src/Dockerfile 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/src/trainer.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/src/trainer_cn.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/src/word2vec/api_train_v2.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cluster/src/word2vec/prepare.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cmd_parameter/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/cmd_parameter/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/optimization/gpu_profiling_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/optimization/gpu_profiling_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/rnn/hierarchical_layer_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/rnn/hierarchical_layer_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/rnn/hrnn_rnn_api_compare_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/rnn/hrnn_rnn_api_compare_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/rnn/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/rnn/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/howto/rnn/recurrent_group_en.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/FullyConnected.jpg 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/add_security_group.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/bi_lstm.jpg 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/checkpointing.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/create_efs.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/csr.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/data_dispatch.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/dataset.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/dataset.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/doc_en.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/efs_mount.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/encoder-decoder-attention-model.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/engine.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/file_storage.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/file_storage.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/glossary_rnn.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/glossary_rnn_with_memory.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/gradients.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/init_lock.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/init_lock.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/k8s-paddle-arch.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/layers.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/managed_policy.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/matrix.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/nvvp1.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/nvvp2.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/nvvp3.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/nvvp4.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/overview.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-cloud-in-data-center.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-etcd.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-etcd.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-model-sharding.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-model-sharding.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-ps-0.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-ps-1.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-ps.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-task-queues.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-task-queues.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-task-states.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/paddle-task-states.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/ps_cn.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/ps_en.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/pserver_and_trainer.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/pserver_init.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/pserver_init.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/route53_create_recordset.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/route53_create_zone.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/sequence_data.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/simple_full_hierarchical_recurrent.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/simple_full_recurrent.dot 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/submit-job.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/submit-job.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/trainer.graffle 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/trainer.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/trainer_cn.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/worker_security_group.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/images/workflow_of_CAPI.png 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/index_cn.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc/v2/index_en.rst 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
doc_theme/static/css/override.css 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc_theme/static/images/PP_w.png 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc_theme/static/js/paddle_doc_init.js 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc_theme/templates/breadcrumbs.html 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc_theme/templates/layout.html 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
doc_theme/templates/search.html 已删除 100644 → 0
浏览文件 @ 1ff47b28
此差异已折叠。
go/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/cmd/master/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/cmd/master/master.go
浏览文件 @ dbd25805
此差异已折叠。
go/cmd/pserver/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/cmd/pserver/pserver.go
浏览文件 @ dbd25805
此差异已折叠。
go/connection/conn.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/master/c/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/master/c/client.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/client.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/client_internal_test.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/client_test.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/etcd_client.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/inmem_store.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/service.go
浏览文件 @ dbd25805
此差异已折叠。
go/master/service_internal_test.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/client/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/client/c/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/client/c/cclient.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/client/client.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/client/client_test.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/client/etcd_client.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/etcd_client.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/optimizer.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/optimizer_test.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/service.go
浏览文件 @ dbd25805
此差异已折叠。
go/pserver/service_test.go
浏览文件 @ dbd25805
此差异已折叠。
go/utils/networkhelper/helper.go
浏览文件 @ dbd25805
此差异已折叠。
paddle/.gitignore
浏览文件 @ dbd25805
此差异已折叠。
paddle/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Arguments.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/ConfigParser.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Evaluator.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/GradientMachine.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Internal.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Matrix.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/PaddleAPI.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/PaddleAPIPrivate.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Parameter.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/ParameterOptimizer.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/ParameterUpdater.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/SequenceGenerator.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Trainer.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Util.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/Vector.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/test/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
paddle/api/test/testTrainConfig.py
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/Arguments.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/Main.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/Matrix.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/Vector.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/arguments.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/capi.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/capi_private.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/error.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/error.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/examples/model_inference/dense/mnist_v2.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/gradient_machine.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/gradient_machine.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/main.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/matrix.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/tests/test_Matrix.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/tests/test_Vector.cpp
浏览文件 @ dbd25805
此差异已折叠。
paddle/capi/vector.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/CMakeLists.txt 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/float16/.gitignore 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/float16/README.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/float16/float16_benchmark.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/float16/float16_inference_demo.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/float16/float16_transpiler.py 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/float16/run_float16_demo.sh 0 → 100755
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/inference/CMakeLists.txt 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/inference/README.md 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/inference/paddle_inference_api.cc 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/inference/paddle_inference_api.h 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/inference/paddle_inference_api_impl.cc 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/inference/paddle_inference_api_impl.h 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/contrib/inference/test_paddle_inference_api.cc 0 → 100644
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/CMakeLists.txt
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_aggregate.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_base.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_batch_norm.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_cnn.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_cpu_gru.cuh
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_cpu_lstm.cuh
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_cuda.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_cuda.ph
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_cuda_cudnn.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_functions.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_gpu.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_gpu_gru.cuh
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_gpu_lstm.cuh
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_gru_ops.cuh
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_lstm.h
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_lstm_ops.cuh
浏览文件 @ dbd25805
此差异已折叠。
paddle/cuda/include/hl_matrix.h
浏览文件 @ dbd25805
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