* init unit test commit, contains register thinking

* support inplace

* get inplaced x.grad

* Try support inplace and hook at the same time

* Support inplace, need debug

* Support inplace successfully

* Inplace use Tensor&, consistent with Tensor*

* fix MapPlainOutputs bug

* fix double grad inplace error

* rename flash_attn_raw to flash_attn_unpadded

* fix static api

* fix static return

* remove contexts in tensor_utils

* update from_blob

* update from_blob

* update from_blob

* fix bug

* fix bug

* add assign composite backward op

* fix log msg

* code style

* fix comp rule

* replace assign with by_pass

* [CINN]Enhance CacheKey hash logic by considering input dtypes (#50557)

* [CINN]Enhance CacheKey hash logic by considering input dtypes

* add unittest

* fix typo

* fix typo

* fix map.at

* fix find

* fix test

* fix cinn cache key structure realize

* using ordered map for attributes

* add test by review advice

---------
Co-authored-by: Njiangcheng <thisjiang@qq.com>

* [prim] enable dygraph_to_static to support custom_vjp

* Pr 50885 (#7)

* [CINN]Enhance CacheKey hash logic by considering input dtypes (#50557)

* [CINN]Enhance CacheKey hash logic by considering input dtypes

* add unittest

* fix typo

* fix typo

* fix map.at

* fix find

* fix test

* fix cinn cache key structure realize

* using ordered map for attributes

* add test by review advice

---------
Co-authored-by: Njiangcheng <thisjiang@qq.com>

* [prim] enable dygraph_to_static to support custom_vjp

* fix code in a dy2static-friendly way.

* [dystatic] add hooker for prim

---------
Co-authored-by: NAurelius84 <zhangliujie@baidu.com>
Co-authored-by: Njiangcheng <thisjiang@qq.com>
Co-authored-by: Ncxxly <chenxx_id@163.com>

* [prim] enable dygraph_to_static to support custom_vjp

* fix cast prim and vjp dtype mapping error bug

* Cxx prim custom vjp (#8)

* [CINN]Enhance CacheKey hash logic by considering input dtypes (#50557)

---------
Co-authored-by: Njiangcheng <thisjiang@qq.com>

* [prim] enable dygraph_to_static to support custom_vjp

* Pr 50885 (#7)

* [CINN]Enhance CacheKey hash logic by considering input dtypes (#50557)

* [CINN]Enhance CacheKey hash logic by considering input dtypes

---------
Co-authored-by: Njiangcheng <thisjiang@qq.com>

* [prim] enable dygraph_to_static to support custom_vjp

* fix code in a dy2static-friendly way.

* [dystatic] add hooker for prim

---------
Co-authored-by: NAurelius84 <zhangliujie@baidu.com>
Co-authored-by: Njiangcheng <thisjiang@qq.com>
Co-authored-by: Ncxxly <chenxx_id@163.com>

* [prim] enable dygraph_to_static to support custom_vjp

* fix cast prim and vjp dtype mapping error bug

* [dy2static-ci] fix dy2static ci errors.

---------
Co-authored-by: NAurelius84 <zhangliujie@baidu.com>
Co-authored-by: Njiangcheng <thisjiang@qq.com>
Co-authored-by: Ncxxly <chenxx_id@163.com>

* [Prim] enable whitelist and blacklist for custom_vjp

* support softmax grad

* remove additional code

* add test back

---------
Co-authored-by: NAurelius84 <zhangliujie@baidu.com>
Co-authored-by: Njiangcheng <thisjiang@qq.com>
Co-authored-by: Ncxxly <chenxx_id@163.com>
Co-authored-by: Nxiongkun <807377414@qq.com>

* test_get_kernel

* add invoke signature

* change reduce_max

* change frobenius_norm

* reset reduce_max according to composite and change reduce_all

* fix the bug when Scalar(*)

* fix 'scalar when support_tensor'

* change code according to review

* change 'keep_signature' to 'manual_signature' and add some erro info

* Delete hardswish_raw op

* fix ut

* modify_yaml

* delete default param

* add output for matmul_double_grad

* add split_with_num composite rule

* add split_with_num composite rule

* add split composite rule

* update

* update test

* update test

* delete split_with_num_grad

* add all_gather and fix conflicts

* fix code format

* fix ut

* fix broadcast ut

* refresh

* compat

* register

* testop

* fix

* fix

* fox

* cast

* cast

* fix

* type

* fix

* out

* cast

* fix

* fix

* fix

* broad

* broad

* broad

* fix

* fix

* fix

* fix

* fix

* broad

* broad

* numel

* fix

* fix

* fix

* fix

* cinn

* fix

* fix

* fix

* fix

* add from_blob

* fix test

* fix test

* fix codestyle

* add gpu test

* fix test

* update

* add comment

* fix comment

* update comment

* fix CI bug

* add thread_local

* update

* fix bug

* fix bug

* fix bug

* fix bug

* fix bug

* fix bug

* fix cmake

* fix CI-Py3 make

* update

* use api_reg

* fix include

* update

* update

* update

* fix bug

* fix bug

* fix bug

* fix bug

* mkldnn->onednn

* fused softplus op + kernel

* remove extra attributes

* add missing handler

* change var name

* Add function node in phi_kernel for MKLDNN

* fix the bug in 'BuildInferVarKernelContext'

* add infer_varkernel_utils.cc

* fix the bug:the first two parametes of 'BuildInferVarKernelContext' can't be template variable

* change the code according to first review

* change the code according to first review

* change the mode of paddle_build.sh

* change 'infer_var_kernel_fn_' to 'get_kerneltype_forvar_fn_'

* add the error information

* fix NotFound infomation warning

* fix NotFound infomation warning

* fix NotFound infomation warning

* add prim erf grad

* add yaml config for prim erf grad

* add math.h

* add cmath

* add math  defines

* use define math

* use define math

* define M_2_SQRTPI

* M_2_SQRTPI math

* try math.h

* fix typro

* remove pow in erf grad

* use new optest

* add fp16 fp32 test

* remove fp16 test

* add softplus double grad

* use constant method

* register custom kernel for all type of custom device

* fix bug

* fix GetKernelInputArgDef

* fix amp bug

* fix TransToPhiPlace

* adapt interpreter_util

* add abs composite backward op

* add missing changes during merge

* modify according to new rules

* local UT OK

* fix typo

* codestyle

* register composite operator

* add fp16 test for abs

* replace experimenta::tensor

* * add comm context for device context

* add broadcast phi operator kernel and api

* add broadcast support dtype, update ut

* fix broadcast bfloat16 type

* fix ut

* update test_collective_broadcast_api timeout to 300

* [cinn] add elementwise_pow backward

* [cinn] update unnitest

* [cinn] update by comments

* [cinn] for ci

* [cinn] for ci

* [cinn] for ci

* [cinn] for ci

* [cinn] for ci

* implement floor_grad by primitive logic

* implement floor_grad by primitive logic

* Merge branch 'develop' into floor_grad

* move DeviceContextPool to phi

* add EmplaceExternalContextFunc

* update namespace

* update cmake

* fix bugs and create context_pool_impl.h

* replace platform::is_xxx_place

* fix bugs

* update generator

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix enforce usage

* Revert "fix enforce usage"

This reverts commit 5f521f08a69713cee506e64a00ec6d9fba709e27.

* fix bugs

* rm XPUDeviceContext and CustomDeviceContext

* fix bugs

* fix fix context init bug

* fix bugs after merge

* fix bugs

* fix name

* fix mutable_data

* update and fix bugs

* fix bugs

* update

* fix bugs

* fix name

* fix bugs

* merge

* fix bugs

* create context_pool in phi/backends

* create context_pool in phi/backends

* fix bugs

* fix xpu bugs

* fix rocm bugs

* fix bugs

* fix bugs

* fix bugs

* fix xpu bugs

* update

* update

* fix bugs

* fix bugs

* comp gather_nd_grad

* fix

* test no cinn

* fix

* fix cinn

* polish codes according #50813

* [getCurrentCUDAStream] Add C++ API getCurrentCUDAStream

* change get->Get

* wrap with macro

* use Get instead of get

* [cinn] concat_grad

* [cinn] concat_grad

* [cinn] concat_grad build success

* [Add PGLBOX] fix unnitest

* [Add PGLBOX] fix unnitest

* [Add PGLBOX] fix codestyle

* [cinn] update by comments

* [cinn] update by comment

* [cinn] add axis check

* flash attn

* seed

* almost

* softmax

* fix workspace

* add unitest; linux only

* fix setup

* fix datatype include

* fix setup typo

* fix def scope

* new error api

* use paddle fork

* fix attr bug; complete ut

* update flash hash

* fix rng reset

* fix offset

* fix comments

* Add comments for #50886

* [Tensor Operants & Prim-Relevant] Tensor supports logical operants

* add prim dynamic unit test

* add prim static unit test

* tmp gather vjp

* support gather

* remove useless code

* fix compiling error

* fix ut

* add eager test

* add eager test

* add seed

* small change

* fix cpu error

* fix transpose op compat

* remove tensor index case

* fix prim_cinn

* small commit

* add cumsum prim backward

* small commit

* skip aixs=None test case

* fix op generante eror

* fix static test error

* remove unused code

* fix static test error

* small commit

* skip cpu float16 test case

* skip eager cpu cumsum float16 test case

* add eager and static UT

* fix ut

* add composite backward rule

* fix error

* fix type error and format error

* add try cpu+float16 test

* fix test bugs

* remove test for cpu+float16 and make y[0] be the grad arg

* add cinn test

* fix UT

* fix the wrong dim of v in test cases

* change y[0] to y[1] for grad in UT

* reshape flatten out

* Disable cinn single test

* use scatter_nd_add

* modify the reshape part of topk_grad

* delete useless build file

* to make the syntax right

* modify bug

* try use of put_along_axis

* remove cinn test

* reformat todo

* add silu composite rule

* fix code style.

* add cinn test

* fix composite grad maker code gen

* add prim in cumsum op test

* remove old test

* fix typro

* pass the static test

* fix typro

* modify optest and delete old test files

* remove normal test_top_k_op test

* fix typro

* pass axis=None test case

* buffer comment

* for debug

* add silu fp16 unit test.

* add static guard

* remove forward prim test

* remove same name axis

* modify the test_top_v2_op.py to pass all local tests

* delete the useless testcase

* fix mistake

* add more testcases to test dtype16 and dtype32

---------
Co-authored-by: NJiabinYang <360788950@qq.com>
Co-authored-by: NGGBond8488 <857631483@qq.com>
Co-authored-by: Nzxcd <228587199@qq.com>
Co-authored-by: NCharles-hit <wanghao107@baidu.com>