* total_piece_num

* fw

* fix for review

* code style

* fix bug

* fix bug for data loader

* kStart in FW for data loader

* fix bug

* bp

* bp && fw

* unexpected_run

* boxing state

* fix bug

* ProcessEord & Boxing fixed

* ProcessEord in Actor and fix boxing

* remove useless override

* total_reading_cnt_

* use mut

* redefine fw

* redefine bp

* remove useless variable

* total_piece_num

* fw

* fix for review

* code style

* fix bug

* fix bug for data loader

* kStart in FW for data loader

* fix bug

* bp

* bp && fw

* unexpected_run

* boxing state

* fix bug

* ProcessEord & Boxing fixed

* ProcessEord in Actor and fix boxing

* remove useless override

* total_reading_cnt_

* use mut

* basic model update

* spell mistake

* fix bug

* remove a log line

* add model_update_kernel_test

* fix some problem in model_update

* function problem

* judge isfinite and change function

* fix batch problem

* remove model update test

* basic model update

* spell mistake

* fix bug

* remove a log line

* add model_update_kernel_test

* fix some problem in model_update

* function problem

* judge isfinite and change function

* fix batch problem

* add filler

* update filler

* fix issues

* rename filler

* add fill in kernel_util.cu

* update fill in gpu version

* update

* rm useless function declaration in kernel_util.h

* fix issues

* use unique_ptr to release memory of cudaMallocHost

* fix issues

* use kernel_util memcpy

* update Fill in kernel_util.cu

fix wrong design of softmax_op

add namespace {} in kernel_test oneflow::test::

relu_kernel test

add #include<random> in util.h

Modified tests with kernel_test_common

define kernel test common

boxing op need midlle blob