Merge pull request !4785 from TronZhang/fix_pass_release_bug

Merge pull request !4769 from wangchangkai/master

Merge pull request !4788 from zyli2020/gpu_bn_grad_perf_optimize

Merge pull request !4557 from caifubi/profiling

Merge pull request !4744 from lichen/support_bert

Merge pull request !4796 from xiefangqi/fix_randomdata_ut

Merge pull request !4797 from lixian/master

Merge pull request !4786 from songhonglei413/roi

Merge pull request !4664 from hangq/master

Merge pull request !4782 from islam_amin/bbox_class

Merge pull request !4724 from mengchunyang/meng

Merge pull request !4768 from chenjianping/lite_dev2

Merge pull request !4757 from liuchao/clean_code

Merge pull request !4765 from pengyongrong/modifyMemoryLeacky

Merge pull request !4764 from zyli2020/gpu_bn_op_perf_optimize

Merge pull request !4766 from chenzupeng/master-lite

Merge pull request !4640 from lyvette/parser

Merge pull request !4689 from zhangbuxue/add_check_for_user_define_bprop_in_Pynative_mode

Merge pull request !4735 from zhouyaqiang0/master

Merge pull request !4763 from gongdaguo/add_lite_models

Merge pull request !4752 from chenzhongming/lite

Merge pull request !4770 from wandongdong/master

Merge pull request !4732 from liuxiao93/fix-SoftmaxCrossEntropyWithLogits-attr-reduction

Merge pull request !4718 from fangzehua/avg0819

Merge pull request !4774 from peixu_ren/custom_pp_ops

Merge pull request !4728 from hewei/fix_tensor_to_string

Merge pull request !4734 from byweng/add_test

Merge pull request !4773 from peixu_ren/custom_bijector

fix pylint

changed doc and code for SmoothL1Loss to be same a dchip. fixed grad kernel

fix ci

Merge pull request !4476 from JesseKLee/cache_prep_branch