The detail of 2 bugs:

tools/manylinux1/Dockerfile cannot build Cuda9 Cuda10 image on
CI. Fixed it and tested manually. Will test on CI after this PR is
merged.

tools/manylinux1/build_scripts/install_nccl2.sh doesn't have
Cuda10 version. Added it

test=develop

* delay infershape test=develop

* fall back subblock to paddle test=develop

* fix edge cases test=develop

* remove output duplicates test=develop

* handle reshape2_grad infershape test=develop

* for debug

* test=develop, memory optimize for dygraph using shared_ptr

* test=develop, fix travis ci showed error

* test=develop, fix bug for recurrent usage of varbase

* test=develop, init varbase when it need to be Add

* test=develop, fix problem of recurrent gradient

* test=develop, add gradient test for recurrent varbase usage

fix flower-dataset unclose error

* use python api to reduce python time cost, test=develop

* fix travis ci, test=develop

* fix Py_None error,test=develop

* test=develop, refine ocr attention model

* test=develop, refine code, remove cpu only part

test=develop, refine code, remove cpu only part

* refine doc for prelu

test=develop

test=develop

test=develop

* update and polish hash op doc

test=develop

* update api spec

test=develop

* expand support tensor attribute; test=develop

* fix bug ; test=develop

* fix uni test bug; test=develop

* fix copy bug; test=develop

* refine expand_times default value; test=develop

* for debug

* test=develop, memory optimize for dygraph using shared_ptr

* test=develop, fix travis ci showed error

* test=develop, fix bug for recurrent usage of varbase

* test=develop, init varbase when it need to be Add

* support some input tensor remain on cpu; test=develop

* fix input = none; test=develop

* fix unfound bug; test=develop

* fix proto None case; test=develop

* fix bug; test=develop

* fix proto null bug; test=develop

* remove conv check; test=develop

* fix test bug; test=develop

* move fill constant; test=develop

* no change in proto; test=develop

* fix bug; test=develop

* change attr detph name; test=develop

* remove remain cpu; test=develop

* fix bug; test=develop

* merge develop; test=develop

* fix one_hot bug; test=develop

* fix bug; test=develop

* fix bug; test=develop

* fix bug; test=develop

* fix python api bug; test=develop

Refactor the code to be OpenMP friendly

test=develop

* revise conv layer cudnn algo choose test=develop

* update for code style test=develop

* update for code style test=develop

test=develop

test=develop

* add support for cuda9 on windows test=develop
* use different git address for cuda9 compatible on windows 

test=develop

* Enable elementwise_div operator test=develop

* Fix update date test=develop

The cdn.bcebos.com can be unstable randomly for unknown reason, restore it to bj.bcebos.com.

Note the append_batch_size variable is doing prepend. We should
change the name, but due to backward compatibility, I suggest to
change at v2.0. Not now.

test=develop

* add no_grad decorator to dygraph, test=develop

* add unittest,test=develop

 use numpy assignment to restore parameters at build-once

* fix the bug of mobilenet-ssd INT8 inference without overloading GetHash
test=develop

* remove the out_grad->format() in TransposeMKLDNNGradOpKernel
test=develop

* fix the dll not found issue on windows

* add example to get_startup_program()
* fix example to get_startup_program()

Split the unittest test_dist_mmist into multiple unittests (test_dist_mnist, test_dist_mnist_nccl and test_dist_mnist_lars) to avoid timeout (#17707)

* test=develop modify some initializer api

* test=develop modify API.spec

* test=develop modify API.spec

* test=develop modify API.spec

* test=develop modify API.spec

* fix prepare context redundant code problem, optimize executor by caching create_varaiables
test=develop

* cache sub_scope, program, var when use_program_cache=True is set

* make fetch_list runable with variables, add more unittest for use_program_cache

test=develop