Merge pull request !277 from guozhijian/fix_read_with_exact_option

Merge pull request !274 from ZPaC/gpu-backend-supports-multiple-streams

Merge pull request !301 from dengwentao/ci_st

Merge pull request !214 from Maoweiyong/modify-gpu-op-register

Merge pull request !293 from zjun/Modify_all_tbe_op

2.GPU commnication stream and compute stream overlap.

Merge pull request !295 from gongchen/fix_free_cell_obj

Merge pull request !298 from gongchen/fix_allreduce_control_depend

Merge pull request !289 from YuJianfeng/find_op

Merge pull request !287 from vlne-v1/I1EFUD-the-output-is-not-tuple

Merge pull request !299 from wangnan39/modify_annotation_of_depthwise_conv

Merge pull request !302 from kisnwang/fix-workspace-malloc

Merge pull request !297 from yanghaoran/master

Merge pull request !291 from caifubi/dev-getnext-mem-reuse-off

Merge pull request !260 from lianliguang/refactor-padding-strategy

Merge pull request !232 from Xiaoda/model-memory-cost-in-auto-parallel

Merge pull request !288 from kisnwang/fix-nopnode-output-bug

Merge pull request !226 from wangnan39/expend_conv_stride_to_2d

Merge pull request !275 from yoonlee666/master

Merge pull request !281 from SanjayChan/origin/quant_liantiao

Model the memory cost in auto-parallel. It is calculated by the output of operators, plus the parameters. Additionally, modify the graph-operations in auto_parallel to include memory_cost.

Merge pull request !283 from jjfeing/master

Merge pull request !273 from liyong126/mindrecord_subset_sampler_python