Add Chinese translation for k8s_aws

Unify REGISTER_OP and REGISTER_OPERATOR

Add cutoff parameter to word_dict for IMDB

fix pool and conv3d unittest when paddle is not build with gpu

Split byref op

Remove redundent pserver block

Correctly implement destructor of ParallelExecutor

Fix FetchTensor on CPU

Add 4 docs to fluid/dev

Clean MultiDevicesGraphBuilder

Remove `dense_vector` from v2 api documents

Use mutex to stablize ncclCtxMap

Add neural transformer leanring rate decay function

Fix bugs in local_scopes

* Add topk Python API.

* Add unit test.

* Remove the repeated API.

Disabling Channel Test to fix CI

Add parallel executor tests

fuse batch norm for conv operator 

- Added Epsilon (preventing softmax output from being too small) for softmax MKLDNN op

Fix conv_mkldnn_op.cc which is causing CI failure