Merge pull request !4005 from fangzehua/unsortedsegsum

Merge pull request !4004 from fangzehua/squaresumall

Merge pull request !3789 from riemann_penn/eager_mode_sparse

Merge pull request !3977 from nhussain/multi_dim_concat_2

Merge pull request !4000 from zhangbuxue/improve_bool_for_tensor

Merge pull request !3966 from lichen/add_dropout_distributed_op

Merge pull request !3916 from xychow/remove-loss-scale-range-check

Merge pull request !3940 from zhangbuxue/support_interface_all_and_any_of_tensor

Merge pull request !3915 from Simson/push-to-opensource

Merge pull request !3648 from 张毅辉/zyh_profiler_test

Merge pull request !3715 from BowenK/new_pattern

Merge pull request !3756 from XunDeng/pp_poc_v3

Merge pull request !3370 from xychow/fix-context-dup

Merge pull request !3846 from liuxiao93/Add-ops-SeluSquaredDifference

Merge pull request !3880 from wangqiuliang/fix-gradients-issue-in-pynative

Merge pull request !3820 from zhangbuxue/support_dtype_and_shape_as_attr_in_graph_mode

Merge pull request !3473 from vlne-v1/change_base_class_of_parameter

if funcgraph or metafuncgraph call it recursively, then anf_node should be used as tracking_id to discriminate the first occurcance and the
recursive occurance.
add anf_node to PrimitiveAbstractClosure hash() to reduce cost of GetEvaluatorFor().

ignore the tracking_id to make cse work.

Merge pull request !3744 from Xiaoda/14-add-new-graph-operations-in-auto-parallel

Merge pull request !3837 from liucunwei/add_test_for_mindrecord3

Merge pull request !3553 from tony_liu2/staging

Merge pull request !3537 from riemann_penn/add_sparse_operator

Merge pull request !3727 from xiefangqi/fuzz_issue_pad_01

Merge pull request !3645 from chenzhongming/master

Merge pull request !3671 from ms_yan/vgg_optimize