* add boardcast_sub

* add boardcast_sub

* refine structure for cuda and rocm

* update

* update

* update

* update

* add pten reduce kernel

* add reduce_sum kernel

* update attribute args and order

* make out dtype undefined

* fix empty input error

* merge develop branch

* rename sum as reduce function

* rename sum as reduce function

* fix reducekernelImpl args error

* add reduce cuda kernel

* modify dims type to const &

* remove unsed log

* fix reduce_all out eigen function error

* remove unused codes

* add the missing sum api define and testcase

* merge develop branch

* fix sum test axis value error

* replace pten mean kernel with reduce_mean

* revcover meam cuda to original implement

* [NPU] reorganization for device API abstraction

* [NPU] delete old files

* [NPU] fix npu_collective_helper

* [NPU] fix collective_helper

* [NPU] fix ut

* [NPU] mod memory allocation and hccl_helper

* [NPU] fix place_type

* [NPU] split enfoce.h

* move acl* call into npu_info

* merge conflict

* fix merge

* merge conflict

* merge conflict

* [XPU] Reorganize xpu device codes in platform, test=develop

* fix xpu_header.h, test=develop

* Modify reduce_op.op.h for xpu2 with kernel primitive api

* add TypeAdapter method for npu_op_runner

* add int64 supporting for elementwise_mul and reduce_sum

* add int64 supporting and UT for expand_v2, scale and reduce_max

* fix bug

* [NPU] fix argsort op, test=develop

* remove debug files, test=develop

* fix typo, test=develop

* address review comments, test=develop

* Update the implement of reduceAnyKernel according to kernel primitive api
* Fix a bug in ReadData, ReadDataBc and ReadDataReduce when NX != 1

[XPU AMP] 1. xpu support gradient acc 2. xpu support create tensor in dygraph 3. xpu support update weight params in amp (#36439)

* - REorder disabling caching

* - compilation fix

* - another compilation fix

* - another compilation fix

* - compilation fix

* - Fix

* - yet another compilation fix

* - suppresingly another compilation fix

* - lint

* - fix after review

* - fix

[oneDNN] disable caching oneDNN primitives in  matmul v2, Reduce grad and elementwise_add grad, expand_v2 (#35132)

* - grad caching disabled of matmul_v1

- compilation fix

- compilation fix

* - reduction removed

* - Matmul v2 disabled caching

* Draft of further changes

* - workaround for reducegrad

* - fixes to UT

* - fix to compilation

* - another fix

* - fix

fix a bug in nlp: text_matching/sentence_transformers when last dim is 1 and reduce mid dim (#34941)

* add reduce_mean_op_npu and test

* remove skip.If

* update

* [NPU] add reduce_prod

* [NPU] delete check_dygraph=False

* [NPU] delete skipIf

* add attrs support or check

* [NPU] delete extra codes for test_reduce_max_op_npu

* [NPU] add attr out_dtype

* first test version

* add test exec;

* add data transfer; test=develop

* add new exec head;

* add memcpy; test=develop

* add python fetch

* add new test

* add graph node; test=develop

* remove useless new executor test; test=develop

* remove gperf dependency; test=develop

* fix compile bugs; test=develop

* remove useless code; test=develop

* remove useless code; test=develop

* add uni test; test=develop

* polish code; test=develop

* polish code; test=develop

* add interpreter cmakefile; test=develop

* remove useless code; test=develop

* support Kunlun2

* support KL2

* support KL2

* [NPU] add reduce_max

* [NPU] delete skipIf

* [NPU] add atrrs support or check

* [NPU] add attr out_dtype

* [NPU] delete debug codes

* added expand_v2 bf16/fp32 kernel

* minor change

* CI fix

* added missing test file

* added formatting

* reduced binary size

* CI fix

* add reduce_sum_op by add self-kernel

* set all ReduceKernel MPType for accuracy

* add float16 test script which input is integer number

* solve reduce sum float16 check_grad problem

* solve conflict and change test script for CI

* change kernel register for CI

* remove all useless template

* modify to complex template types for fill_constant op

* modify to complex template types for py_layer, strided_slice and reduce_sum_op.part

* refine ~npuOpRunner

* implement destructor and forbid copy

* use reference to avoid copy

* use const reference

* relax adam precision

* fix top_k

* fix gather op and add logsumexp op on kunlun

* update xpu depence

* update tests and fix elementwise_add