* update

* fix bug

* fix ci-windows-openblas

* fix test_partial_sum_op

* fix codestyle

* fix divide zero bug for softmax_with_cross_entropy

* change the single test way

* can run but slow. the most important is that I do not know why it slow

* remove some useless commet

* change the copyright to correct

* remove some useless change

* if repeat_times == 1, we will not use BroadcastKernel

* Optimize the performance of logsumexp

* Support zero-dim tensor

* add autogen code support for affine_grid op

* update op_compat.yaml for affine_grid

* update op_compat.yaml for affine_grid

* fix AffineGridGradInferMeta

* fix CI error

* update AffineGridInferMeta

* delete [-Wno-error=terminate], test=develop

* remove GPUps[-Wterminate],test=develop

* remove some -Wno-, test=develop

* modify ~MatmulDescriptor

* mess

* optimize setup.py

* add ninja

* mv WITH_ASCEND_CL

* mv WITH_ASCEND

* rollback

* remove WITH_ASCEND

* remove WITH_ASCEND

* [bug fix] fix pow composite

* [bug fix] for ci

* [PHI CAPI] support complex dtype kernel

* update

* add seed control

* fix bug

* add autogen code support for matrix_nms.

* update

* add strategy force_sequential_run

* fix

* fix

* fix

* fix

* fix

* mv WITH_ASCEND_CL

* mv WITH_ASCEND

* rollback

* remove matrix_multiply unitest

* remove run_program

* remove FLAGS_USE_STANDALONE_EXECUTOR

* [AMP OP&Test] Fix the logic of calling infer_dtype func in op test

* add fp16

* Isolate DenseTensor::set_type from header file

* Fix selected_rows

* register bf16 for communication ops

* fix bfloat16 type finding compile error in c_allreduce_max_op