Created by: Noplz

• fix bug of 'shuf' and 'gshuf' in demo/recommendation/preprocess.sh

• Remove those extra trailing empty lines.

• Rephrase the first paragraph

• remove random flag

• modify format of cluster_train_cn.md

• add link of demo/recommendation

• fix dead links and refine tutorials/index

• modify markdown format of cluster_train_en.md and add link of demo/recommendation

• remove gpu memory alloc

• modify details

• add refer link of Kubernetes in cluster_train_cn.md

• add refer link of Kubernetes in cluster_train_en.md

• Clean BatchNorm Code.

• add Layer::createFunction

• fix dead links on quick_start

• travis for check broken links

• Added support for cudnn v6 and cuda 8.0

• Matrix API refactor, when passing parameters, convert shared_ptr (MatrixPtr) to reference or raw matrix (Matrix & or Matrix *) contextProjectionForward contextProjectionBackward contextProjectionBackwardData contextProjectionBackwardWeight classificationError The mul functions would be updated later.

• Refactor MUL functions, pass object reference instead of shared_ptr.

• set -e for docs.sh

• update docker_install_en.rst

• Add priorbox layer gpu unit test.

• Add excluded_chunk_types to ChunkEvaluator

The chunks of types in excluded_chunk_types will not be counted in ChunkEvaluator. This is useful for tasks such as SRL, in which chunks of type V (verb) will not be taken into account in evaluation.

• move TEST from test_matrixCompare.cpp to cross_map_normal_op_test.cpp

• Refine Code

• fix style issues

• Add const in ParameterUpdater init

• follow comments

• Fix bug in config_parse.py when batch_norm layer is used in RecurrentLayerGroup.

• change float to real in two test

Change float in test_ConvTrans and test_ConvUnify to real.

• Start doing mnist_train_api

• Remove not used params in GradientMachine::start

• Add GradientMachine::start/finish to API

• Remove unused cost parameter in ParameterUpdater

• Adding namespace in timing macros

Sometime those macros are used under different namespaces. We need to use namespace ::paddle to make it compile correctly.

Change-Id: I57a6d6ec8cd0d680b584aab62d72a35c226a24a4

• close log info in BN.

• fix array style problem

• refine docs.sh

• Change type float to real.

• Support user specified label input in tests

• refine docs.sh

• change type to bool.

• 按照评审人的意见进行了修改

• Update index_cn.md

• modify

• modify load_missing_parameter_strategy

• modify the url reference

correct the original to "../text_generation/index_cn.md"

• Wait for reading data.

• A tiny fix in PyDataProvider2

• hidden decorator kwargs in DataProvider.init

• also add unit test for this.

• Try to read data in mnist

• Use numpy in DenseScanner.

• Try expose ParamUpdater::update

• A simplest train file for mnist added.

• A tiny fix in PyDataProvider2

• hidden decorator kwargs in DataProvider.init

• also add unit test for this.

• change std::vector to const reference

• Add const to GradientMachine::eval

• A tiny fix in PyDataProvider2

• hidden decorator kwargs in DataProvider.init

• also add unit test for this.

• Also add const to makeEvaluator

• Change member variables from public to protected

• Expose Evaluator API

• Complete a very simple mnist demo.

• Fix typo in PyDataProvider2.py

• Fix data provider bug in srl demo

Once encoutering a single sample whose size is larger than batch size, the provider will yield empty batch and terminate the current pass unexpectedly if can_over_batch_size=False.

• Remove utils/CommandLineParser.h

• Change float to real in NormLayer.h

• Add AverageOptimizer, Add save parameter

• Update index_en.md

删除了关于train.sh文件路径的说明

• Update index_cn.md

按照review建议进行了修改

• Add gitignore

• fix conflicts not found by git diff

• Fix gflags 2.1 issue

• Always use copy method for numpy.

• Make this demo support GPU

• Test GPU

• Test on GPU

• Remove bazel from dockerfile and docs

• Add comments.

• Remove all bazel tests in third party

• Modified VM OS from Debian to CoreOS in doc.

• add translation for cmd_parameter

• Bug fix for BatchNorm when testing more than models in test mode.

• add config_parser in trainer_config_helpers to seperate trainer config

• add config_parser_utils

• Remove unnecessary import in api_train.py

• enable clang-format

• modify documents

• modify documents

• follow comments: remove flags control

• move TestMain.cpp into utils and fix some tests for check

• fix some tests which only need paddle::initMain in main

• Add c++11 build python binding package

• Add common.h and remove DisableCopy and Typedefs

• follow some comments with issue 973

• follow some comments whit pr 854 after committed

• Update arguments_cn.md

• Update detail_introduction_cn.md

• Update detail_introduction_en.md

• add prefetch and onPassEnd to PaddleApi.h

• fix bug:

• gradient_clipping_threshold should be allowed to set with parameter-grain

• add cluster train for quick_start

• Rename math.py to layer_math.py

• Fix #903 (closed)

• Add externel google's dependencies

• Add other extern dependencies

• add remote updater in api and swig

• Fix merge errors.

• Update index_cn.rst

• Update external libs

• Refine cmake file names

• Remove paddle internals

• Refine CMakeLists

• Update external dependencies cmake

• Fix glog check type unmatch

• Add external_project_dependencies for targets

• Fix external denpendencies build order

• do not init localUpdater

• Fix warpctc header directory

• Clean Travis ci and fix bug

• Check md5 when download word embedding.

Fix #1008 (closed)

• udw nullptr as localUpdater

• add new_layer_cn doc

• fix format problem in new_layer_cn.rst

• move TestMain.cpp to testing

• fix error when WITH_TESTING=OFF

• remove gtest_main_library

• fix link order error

• add python version evaluate function

• change LOG to log

• modify follow comment

• add first paddle function example for ContextProjectionForward operator, by going through Daoyuan's excellent paddle function design.

• add ContextProjectionBackward, ContextProjectionBackwardData, ContextProjectionBackwardWeightw

• add Context Projection functions to Layer's forward and backward, resolve merge conflicts

• combine data/weight to ContextProjectionBackward for clean code

• update interface of context projection functions, Tensor -> Matrix/Vector

• clean up unused code.

• Also check baidu.dict md5

• refine print information

• Add jupyter notebook support in Docker images.

• add translated chinese docs into catalog

• add paddle_test_util static lib to simplify unit test.

• add TestUtil.h and TestUtil.cpp, moving from gserver/tests/ to testing/

• Change run_all => entrypoint.

• fix dead links, reduce image size

• revise new_layer_cn and fix a few problems in new_layer_en

• change format

• a missing character in line 32

• Add extern python interp

• Add system configure

• Update external dependencies

• clean travis ci

• Clean cmake

• Fix merge conflict bug and glog

• Add python in travis ci for paddle version command

• Add wheel for paddle version command

• remove double spaces, remove chinese character "，"

• Update travis ci pip

• Move Execepts into arch/osx dir

• Revise common to Common

• Clean travis ci

• Add find system's swig