* update assign_pos

* update attention_lstm

* update barrier

* update batch_fc

* update beam_search

* update beam_search_decode

* update bilateral_slice

* fix bug

* Handle Structure kernel for InterpreterCore::RunOperator

* fix bug

* fix rocm compile

* fix rocm compile

* Revert "fix rocm compile"

* test

* revert test and update cmake

---------
Co-authored-by: Nchenruibiao <chenruibiao@baidu.com>

* fix kp compile

* test

* Revert "test"

This reverts commit 3a1cbfaa0f23e6e06d3dcd8d0b0c28aa63a98e70.

* update copyright

* update cmake

* update cmake

* update cmake

* update cmake

* update

* update

* update

* update

* update

* fix test

* unify_kernel

* fix compile bugs

* modify macro name

* perfect code according comment

* fix compile bugs

* fix compile bugs

* fix ci bugs

* fix ci bug

* fix ci bugs

* fix ci bugs

* modify code according comment

* rm conv_fusion_op

* generate some static graph ops

* fix the bug of pow

* add REGISTER_ACTIVATION_OP in operators.cmake

* modify the file operators.cmake

* change feed_op to phi kernel

* fix ci bugs

* fix build bugs

* fix ci bugs

* fix compile bugs

* fix ci bugs

* perfect code

* perfect comment code

* fix install bugs

* modify code according comment

* remove visitor in feed_op

* modify according comment

* perfect code according comment

* add infershape

* fix py3 bugs

* fix getexpected kernel type

* fix getexpected kernel type

* fix ci bugs

* add registry for custom device

* fix py3 bugs

* fix floating point error

* fix py3 test bugs

* add new tensor

* fix windows compile bugs

* fix ci bugs

* fix ci bugs

* fix ci bugs

* perfect according comment

* fix ci compile bugs

* add raw tensor

* fix ci bugs

* modify code by comment

* delete String

* remove fc mkldnn hardcode

* remove useless enum of kFCMKLDNN

* fix macro error

* update operators.cmake

* Add infer prune function

* Update phi.cmake

* Update operators.cmake

* add fusion op

* simplify conv_mkldnn op registration

* remove custom type value in conv grad op

* solve transpose2, follow #22402

* fix CI cmake

* update REGISTER_OP_KERNEL of transpose2

* gaussian random

* mkldnn to onednn renaming

* fix merge conflicts

* remove fluid code

* onednn renaming

* gelu fwd

* sort activations

* gelu gradient

* remove unused macros

* merge conflicts

* fix merge conflicts

* remove extra contraint from gelu op

* move mkldnn activation kernel

* fix compile bugs

* fix compile bugs

* deal with conflict

* fix compile bugs

* fix windows compile bugs

* mkldnn unittest fix

* change mutable to alloc

* fix unittest bugs

* modify code according comment

* cmake-lint

* update

* [NPU] add take_along_axis and take_along_axis_grad ops

* [NPU] add take_along_axis and take_along_axis_grad ops

* fix ut because cpu kernel can not be fallbacked

* [NPU] add multinomial op

* fix place

* deal with cann version

* fix for old operator

* change another way

* fix assign kernel bug

* fix xpu kernel select error

* add cudn pinned place

* fix copy error

* fix infrt error

[Phi]Move Relu/Cos/Sin/Tan/Acos/Asin/Atan/Sinh/Cosh/Asinh/Acosh/Atanh kernels in Activation to Phi (#40175)

* move activation op

* adjust code format

* fix compile bugs

* fix ci bugs

* code format adjust

* code format adjust2

* activate ci status

* modify according to comment

* Added cuBlasLtHandle_t to device context.

* Added fused_gemm_epilogue op.

1. Added fused_gemm_epilogue op to leverage cuBlastLt Epilogue.
2. Support fusion Act(X*Y + bias), X'dims >=2 and Y'dims shoule be 2.
2. Act currently only be supported ReLU. (Will add GeLU in the future).

* Added UT to fused_gemm_epilogue op.

* Added LinearAct Pattern

1. Added LinearAct into graph_pattern_detector.* to define (2.)'s
pattern.
2. LinearAct is used to detect act(element_add(matmul_v2(x, w), bias)).
3. act currently only support ReLU (Will support GeLU in the future).

* Added FuseGemmEpiloguePass

1, Added FuseGemmEpiloguePass to handle nn.Linear + Act{ReLU}
fusion (GeLU will be supported in the future).
2. Only support matmul_v2 from nn.Linear.

* Added pybind to BuildStrageter.fuse_gemm_epilogue_.

* Added UT for fuse_gemm_epilogue_pass.

* GeLU support and EpilogueSingleton

1. Added GeLU support to fused_gemm_epilogue op.
2. Added EpilogueSingleton to cache auxiliary pointer.
3. Added related UTs.

* Rename cublaslt_epilogue_opto gemm_epilogue_op.*.

* Added both train and infer pattern to LinearAct.

1. Added support of fwd graph with grap_ops linking to LinearAct.
2. Added related changes to fuse_gemm_epilogue_pass for above
modification.

* Changed CUDA requirement from 11.4 to 11.6 for fuse_gemm_epilogue_pass.

* Added identity activation support to gemm_epilogue_op.

* Added Linear Fusion (matmul_v2 + ele_add)

1. Added matmul_v2 + ele_add pattern to LinearActPattern.
2. Added matmul_v2 + ele_add support to fuse_gemm_epilogue_pass.

* Rename gemm_epilogue_op.* to fused_gemm_epilogue_op.*

* Add fused_gemm_epilogue_grad op.

1. Added fused_gemm_epilogue_grad to support backward epilogue fusion.

* Add UTs to fused_gemm_epilogue_grad_op.

* Change attribute name in fused_gemm_epilogue_grad_op for clearing.

* Allow DX and DBias be dispensable to fused_gemm_epilogue_grad op.

* Added ElementwiseAdd+Matmul+Act graph pattern detection.

* Fuse backward of Linear( Act(x))

1. Added backward fusion pass to Linear( Act(x)).
2. Added backward fusion pass to Linear(x).

* Added UTs to backward fusion of Linear(Act(x)).

* Complete document of arguments to fused_gemm_epilogue_op.

* Made arguments of some functions pass by reference.

* Modify code with review comments.

1. Made arguments of some function pass by reference.
2. Removed redundant code.
3. Followed Google code style to change code.

* Made 'const' code style be consistent

* Fixed random seed of python UTs.

* Set Compiling constrains to cuBlasLt

1. Require CUDA 11.6+
2. Remove fuse_gemm_epilogue related tests when CUDA < 11.6.

* Code Reivew from Paddle

1. Changed arguments name is_first_gemm to without_x_gradient for
clearing.
2. Applied PADDLE_THROW in fused_gemm_epilogue_op.

* Remove EpilogueSingleton

1. Applied ReserveSpace to replace Epilogue for passing auxiliary
pointers between FWD and BWD.

* Fix a logical error and enhance UTs.

1. Added act op count checking in UTs.
2. Fix issue to fuse backward or ReLU(Linear(X)).
3. TODO: solve GELU fusion issues.

* Fix Linear and GeLU fusion issues.

1. Modified graph_detech_pattern to fit with both linear wiht gelu or
relu.
2. Modified data range in Uts to allow negative values.

* Removed fused_gemm_epilogue_op.h.

* Rename namespace pten to phi.

* Rename name of arguments in fused_gemm_epilogue_op

1. bias -> Bias.
2. out -> Out.
3. reserve_space -> ReserveSpace.

* Change EpiloguePassActivationCache as local variable.

1. Removed singleton in EpiloguePassActivationCache.
2. Made EpiloguePassActivationCache as an argument to each pass
functions.

* [KP] Unify .cu and .xpu files with .kps files

* fix CI bug in GPU and modify the list

* fix conflict

* modify the date

* [KP] Add elementwise add xpu, test=develop

* modify the File Permissions

* modify the copyright time

* modify code style

* modify code style

* Add XPU compiler for paddle, test=develop

* clean code

* clean useless code

* clean useless code

* clean useless code

* test

* add include path

* use clang compiler

* xpu2.cmake

* XPU2 compiler passed

* update

* update after pten

* combination the WITH_XPU and WITH_XPU2

* update the fuse operation in WITH_XPU and WITH_XPU2

* update

* update

* update

* fix the merge error

* update

* update the code

* update the code

* add run_kp_kernel flag

* update

* update

* fix prepared type_ bug

* clean and update the code

* reset the kernel_primitives

* update

* clean the code

* delete useless comment

* fix the bug in WITH_XPU

* update

* update

* modify the abi

* delete some useless code

* Parameter automation in xpu compilation

* Parameter automation in xpu compilation

* delete kps in cmake

* delete useless comment

* clean the code

* clean the code

* update cmake file to remove fluid kernel

* add pten declaration.h to where pybind.h used

* fix sync_bn and tensorrt_engine

* refine detection_library

* fix interpreter_core

* support eager legacy

* fit eager legacy for pten

* fall back to cpu if not found kernel

* fix compile problem

* fix compile problem

* refine fallback logic

* fit operator.run()

* fix xpu compile

* fit for new_exec

* add REGISTER_OP_WITHOUT_GRADIENT

* un-cache pt_kernel_context

* fix compile

* fix cudnn

* fix compiling with on_infer

* fix mkldnn

* fix isfinite_v2

* fix xpu problem

* fix op_device

* refine fallback for xpu

* fix xpu compile

* merge develop

* refine code format

* fix compile

* fix compile

* add data_transfer

* fix PreparePtenData

* fix cpu context

* merge develop

* fix compile

* fix error device context

* fix xpu

* fix dev_ctx

* add lstsq gpu kernel

* update

* add docs_en

* modify ut

* fix bugs

* modify example in docs_en

* remove lstsq_op.cu from ROCM cmake

* modify docs_en

* modify docs_en

* modify docs_en

* remove unneccessary TensorCopy

LGTM

* add eigvalsh with is_test

* add eigvalsh op

* fix backward bug

* forward and backward, float and complex, unittest

* remove eigvalsh_helper.h

* remove changes of cusolver.h

* fix unittest

* fix unittest bug

* update code following eigh

* fix test

* update lapack

* pull develop

* update funcor

* fix unittest bug

* fix details

* add tensor_method_func

* fix notes

* add some ops to train ssd on kunlun

* add some ops to train ssd on kunlun

* add some ops to train ssd on kunlun

* update cast op unittest

* update cast op unittest

* update cast op unittest

* update xpu cmake

* update cast unittest

这个PR只包含fused_feedforward前向的代码。

相关kernel实现：fused_dropout_act_bias, fused_residual_dropout_bias, fused_layernorm_residual_dropout_bias

fused_feedforward是一个融合算子，该算子对transformer模型的feed forward层的算子进行融合和封装，使得前端只呈现一个接口，通过融合减少部分访存和kernel launch的时间，以此提升性能。

功能：本PR的目标是提高attention模块的计算性能。
为了减少框架层对op的调度开销，本PR通过在C++层手动实现attention模块，对外提供attention 大op；
为了减少防存开销，本PR采取了两种优化方法：
（1）在q,k,v计算时通过共享输入X，将该处的gemm，transpose和bias add从三次调用减少为一次；
（2）使用kernel融合优化技术，在不同cuda kernel之间通过寄存器传输数据；

* Add QR decomposition op

* Change codes to adapt to new svd_helper

* Update linalg.py

Restore the deleted comma

* Restore the deleted line

* Update linalg.py

* Update linalg.py

* Improve the qr code by reviews

* Update QR based on CI results

* Update qr doc, test=document_fix

* Change unsafe and ill-formed codes

Add sparse_attention OPs, python api will be added in next pr

* init matrix_rank op, add matrix_rank CPU code and test

* add GPU kernel, remove svd_eigen.h

* add CPU kernel when tol is tensor

* add cpu and gpu code when tol is tensor

* fix CI-ROCM error

* add matrix_rank API describe, fix PR-CI-Py3 error

* fix PR-CI-Windows error, add matrix_rank API test

* delete useless comments

* fix review

* add my code in svd_helper.h

* update doc commets

* remove spaces

* Add SVD Op and it's GPU and CPU kernel

* Remove CUDAPlace in test_svd_op, make the test available in CPU package

* modfity the file

* fix windows bug/ fix ROCM / fix test timeout

* for pass the CIs

* improve error report

* for code review

* some modification to test_svd_op

* change python code style

* expose the svd interface for document