* Fix parallel.do with batch norm

* Change log level

* CopyShare AllPlaces

* disable nccl test

* bring back parallel_do test

* init polynomial_decay

* test polynomial_decay

* complete polynomial_decay

* fix conditional block op

* init scalar-switch-case-op

* switch op can compile

* complete forward switch_op

* add GetMatchCaseIndex

* add switch_grad_op

* init switch Python API

* add test_switch

* support set block list in python

* fix scope problem

* complete test

* optimize test

* optimize test

* rm backward part

* clear grad op

* polynomial_decay use switch op

* revert conditional_block_op and reshape_op

* add piecewise_decay and test

* fix piecewise_decay

* try to use condition op for switch

* can work

* clean old code

* revert

* rm switch_op.cc

* optimize code

* add attr is_scalar_condition for condition_block_op

* fix comment

* fix comment

* add export

* refine get cfgs method in memory optimization transpiler

* clean code

Impl scalar switch case op with condition op

* set up python code

* fix bug

* add cc file

* fix cmake

* add inference test for label semantic role

* fix

* address comments

* address comments

* address comments

* address comments

* add use_cuda

* First basic implementation

* Add infer example for image_classification

* Address review comments: round 1

Add a new type `Reader` for `VarDesc`, which can holds more than one
LoDTensor.

init debuger.

* init

* add delete operator

* debug

* add wait

* clean code

* fix bug

* fix bug

* refine code

* remove unused code

fetch_var for getting the values of a variable with given name
get_var for getting the Variable with given name

* initial commit

* add get_parameters method

* add get_parameters method

* small fix

* address comments

* address comments

* address comments

* fix