* Add Preface

* Add demo code

* Save file

* Refine code

* seems can work

* use elementwise strategy

* Use ElementwiseComputeEx

* Add comments

* extract functions from operator

* Refine code

* Follow comment

* code refine

* follow comments

* follow comments

* polish dist se resnext ut

* update

* update

* update

* avoid cpu initializer differ

* change to use executor for now

* update by comment

* remove lr decay use para exe, should fix para exe bug later

* update by comment

* fix the order of sum

* add doc

* check whether need to copy

* follow comments

* fix the result of unit test

* fix resnext

* compare the result of PE and Exe

* compare the result of reduce and allreduce

enforce test_parallel_executor_transformer and test_image_classification_resnet to run in serial (#12545)

* api clean

* update API.spec

* "cherry picked cpp tests"

* "cherry picked"

* "cherry picked tests"

* "merge develop branch"

* "add gradient register"

* "make some enhance"

* "better format"

* "fix typo"

* "fix reuse"

* "fix get expected kernel"

* "change the mkldnn code"

* "fix mkldnn"

* "fix mkldnn failed test"

* "add comment"

* reopen test_parallel_executor_transformer

* save the CPU runing time

* "remove inplace in single op"

* "fix ci"

* "add transpiler case"

* fix conflict

* "fix reshape"

* "delete reshape inplace attr"

* "follo the comments"

* "rerun ci"

better test

fix fluid_benchmark.py

* add flatten op

* enable dist se resnext

* small batch size

* refine regularization for selected_rows

* clean lookup_table

* refine rpc_server_test

* temporally disable rpc_server_test

* fix rpc_server_test

* add unit test

* update test_pe_mnist threshold

* clean code

* Hiden some LoDTensor ralatted ops' Python wrapper.

* hide variable API

* edit API.spec