The English of the previous API is bad.

* Adding the distributed implementation for machine translation

* re-running CI

* Updated the code style

neural machine translation model support beam search with while op

* limit variable type to lod tensor in memory optimization transpiler

* refine policy

* support while operator

* fix random seed and training data order

* refine get_cfgs method to support multi while operators

* refine codes

2. Fix evaluator using 'reduce_sum' op instead of 'mean' op

* add memory optimization transpiler demo

* add memory benchmark compile option

* add gflags instead of macro

* refine code

And write an unittest for it

* "fix decode bug"

* "follow commnet"

* "fix error"

* "fix hook bug"

* fix based comment

* fix copyright

* fix based on comment

Add distribution implement of image classification

* registery is only used by layers
* Rename it to layer_function_generator

Users can use `a+b`, `a*10`.