* fix dist train reduce mode

* fix previous fix

* fix dist transpile with memopt

* update api.spec

* polish dist transpile api

* update apispec

* update apispec

* refine dist train RPC deps

* clean up

* clean up

* fix ut

* remove input for fetch_barrier

* follow comments

* fix auc layer and add check for auc op

* use input to check if states are inited

* optimize code

* dist transpiler add control dependency var between send and recv

* fix async deps

* follow comments and refine

* fix deps connect for rpc ops

Add dependency to send recv

* Disable in_place in batch_norm API.

Optimize selected rows for dist lookup table with rwlock 

* remove patchelf in windowls

* "follow comment"

* Add three modes for prelu_op.

* dist transpile can pass startup program by argument

* update API.spec

Add flatten op interface and enhance APIs about detection to support variable-length image. (#12422)

* add flatten api&enhance detection api

* unify shape_op data type

* update API.spec

add export_for_deployment flag to save_inference_model

* hide misc APIs

* update

* fix transformer test

* update API.spec

* fix dist train with rmsprop

* add rmsprop transpiler test

* update by comment

* fix the numpy version has to be lower than 1.15

* Add TODO comments