* remove return in op_test teardown, test=develop
* add check_grad for fp64 and other type
* update white list

* test the equality between running-time and compiling time

* add white_list for compile_vs_runtime check test=develop

* add check of new-registered op on compile&runtime-lodEquality test=develop

* test=develop

* test=develop

* update white-list test=develop

* merge with new code test=develop

* fix code style test=develop

* test=devleop

* test=develop

* test=develop

* test=develop

* add more ops to white_list test=develop

* test=develop

* test=develop

* test=develop

* test=develop

* test=develop

* fix code_style: add comments test=develop

* test=develop

test=develop

Update iou_similarity op to support non-normalized bbox

test=develop

fix flip comments, test=develop
closes:#21799

* add unittests, test=develop

* set dtype of compare op to bool, test=develop

* add selected rows supported in framework

* copy stop_gradient property in clone, test=develop

* fix pruning issue, test=develop

* fix kLookupTablePath issue, test=develop

* sync program, test=develop

* copy subblocks, test=develop

* add unittests, test=develop

test=develop

* test=develop, speed dense calc & communication

test=develop

* Remove self-set accuracy parameters of op tests: max_relative_error
test=develop

* fix error
test=develop

* Remove self-set accuracy parameters of op tests: max_relative_erro
test=develop

* fix error
test=develop

* add Linear layer
test=develop

* update unittest for coverage
test=develop

* Remove self-set accuracy parameters of op tests:atol
test=develop

* keep smaller parameters
test=develop

* fix error
test=develop

* fix post training quantization bug of memory constrained, support the input be different, test=develop

* add some debug flags to auto growth allocator, test=develop

* add comments about auto growth, test=develop

1. Make while_op accept GPU conditional data
2. Add more complex test cases for while_loop API

test=develop

* Remove self-set accuracy parameters of op tests: max_relative_error
test=develop

* fix errors
test=develop

*use dtype to determine whether check_grade is needed, and delete useless class

The fixed bugs:

1. The condition sub-graph is not pruned
2. When backward graph is extremely simple, the whole backward ops are pruned.