* "init benchmark ops"

* "untrack outputs"

* "delete some usused code"

* "benchmark"

* "fix ci"

* "fix op test"

* "fix uint16 missing"

* "fix ci"

* "follow comments"

* "fix ci"

* "follow comments"

* "conficts. merge develop branch"

* repick

* "merge develop branch"

* add normalize switch to box_coder_op

* add default prior box var

* update according to the review

* Auto generate python api for polygon_box_transform op.

* Add unitest.

* fix conflict

* code clean

* Add python wrapper for gather op.

* Add unitest for 'rank==1' and fix comments.

* Fix comments.

* Add shape op to get the shape of variable.

* Rename get_shape to shape.

* Add checker for output and fix comments.

2. replace split_method with slice_var_up in func init_splited_variables

* Make bilinear_interp_op support attrs from input.

* Fix python api.

2. Merge the develop repo

* WIP move weight decay

* weight decay ok

* wip

* clean up transpiler

* add details folder

* update

* fix split var test

* follow comments

2. remvoe comment line

2. add align_var_to_block argument to func transpile
3. remove concat and spilt if align_var_to_block is False
4. unittests for simple_dist_transpiler

* Develop a fake dequantized op for fixed-point quantization training framework.

* Add the missing file.

* 1. implement StopAll in ListenAndServOp
2. make pserver receive the SIGINT and SIGTERM from outside
3. add unittests for listen_and_serv_op in python

* 1. add blocking queue set to record queue
2. aware all blocking queue when exit and exit gracefully

* 1. Remove comment lines from blocking_queue.h
2. Implement SignalHandler and move all global vars and funcs into it

* 1. Make code follows the style check
2. Move the SignalHandler out of the unnamed namespace

* 1. Make yapf happy

* 1. Call Stop() in destructor to release the resource allocated by ListendAndServOp
2. Change exit status to EXIT_SUCCESS after handling the signal from outside
3. Remove the mis-usage of REMOVE_ITEM in unittests

* 1. use DISABLE_COPY_AND_ASSIGN
2. use program once macro only

* Add quad transform.

* Fix some syntax error.

* Fix CUDA kernel launch configure.

* Generalize geometry channels.

* Rename QuadTransform to PolygonRestore.

* Rename op.

* Rename op and fix computation.

* Modify CMakeLists.txt for box_restore op.

* Refine code:
1. rename op
2. uncomment unitest on GPU