cherry-pick PR #42777

* [Dy2Stat]Modify all jit.save path into tempfile

* [Dy2Stat]Modify all jit.save path into tempfile

* [MLU] add mlu new profiler (#41138)

* [MLU] add mlu new profiler

* fix format

* [MLU] support add callback to stream (#41831)

* [MLU] add gather mlu kernel (#41969)

* [MLU] add mlu activation kernels (#41751)

    add class NameScopeHelper for adding namescope info
    添加更多 种类优化器状态的映射
    为 IpuStrategy 添加 compilation_progress_logger option 用于输出 编译进度
    部分代码清理和杂项优化

* Reduce time variation for cuda_managed_memory_test (#42458)

* Disable standalone executor for test_tensordot (#42476)

* fix wrong place

* skip bf16 test if not supported (#42503)

* Fix Release 2.3 Bug

* Fix format

* [Dy2Stat]Fix losting pre/post hook from outermost layer while jit.save

* fix kwargs

* fix unittest

* Extend python einsum interface to make einsum_v2 support multi-operands and switch it to default.

* add opt_einsum dependence

* add yaml and support eager model

* fix by code review

* fix pad3d infer shape

* fix pad3d

* fix pad default value

* fix order

* add unit test

* fix unittest for ci coverage

* add ndhwc check

[cherry-pick 2.3] Add fused_multi_transformer op to optimize transformer generation performance (#42311)

* Add fused_multi_transformer op to optimize transformer generation performance (#41814)

* fix fused_multi_transformer compile failed in cuda arch < sm53 (#42315)

* fix ci timeout

* fix bug of building InferMetaContext (#42211)

* add unitest

* full api fix

* when out is None, go old dygraph mode

* by static check

* first version: support 2-inputs forwards. TODO: 1. backward  2. BroadCast  3. MultiVariable

* time out -> 120

* implement autotune python API

* fix doc

* fix windows error

* fix doc and enable auto-tuning when config is None

* fix windows error

* fix doc

[Eager] Remove retain_grad_flag in accumulation_nade, add is_new_grad args in operator (#42240) (#42290)

cherry-pick #42140

* [Eager] Support div scalar in eager mode

* Updated and remove debug logs

* Remove list, use 'or' directly

* Remove useless statement

* fix kenrel name apperance (#42071)

* Reduce performance influence by record event in python (#42040)

* optimize performance

* fix

* improve coverage

* fix

* fix

* [Eager] Fix CastPyArg2scalar for max value of int64 (#42098)

* [Eager] Fix CastPyArg2Scalar in Long case

* Add more test cases for paddle.clip

* Use PyLong_AsLongLong

* Fix merge conflicts

* add _grad_name and _grad_value for eager tensor

* fix paddle_enforce

* fix paddle_enforce 2

* fix grad_name

* _grad_value return lodtensor rather than tensor

* fix

* sharding_for_eager_tensor (#41415)

* fix_sharding_copy_right (#41849)

add mixed-precission support for ipu

cherry-pick from #41733

* modify xpu.cmake,*test=kunlun (#41832)

* modify xpu.cmake,*test=kunlun

* modify xpu.cmake,*test=kunlun

* modify xpu.cmake,*test=kunlun

* modify xpu.cmake,*test=kunlun

* support bilstm,*test=kunlun

* [cherry-pick]support multi_layer of bilstm,*test=kunlun

* fix_nccl_barrier (#41970)

* be compatible with the old version of alltoall (#42007)
Co-authored-by: NBaibaifan <39549453+Baibaifan@users.noreply.github.com>

* fix utest

* fix time

* double accessor and show_scale

* double accessor and show_scale

* rename

* fix bug in pslib config

* add unittest

* [PHI] Support some c++ api in paddle namespace (#41778)

* support some c++ api in paddle namespace

* change c++ api namespace in custom op

* [Phi] Support setting size of vector<Tensor> for out in yaml (#41576)

* support setting vector out size in yaml

* support setting size of vector<tensor> for out in yaml

* add data transform config for shape and size (#41909)

* fix api_gen bug