* add minimum grad composite rules

* add public python api

* fix format

* fix format

* update testcase

* fix testcase

* fix format

* fix cmakelist.txt

* fix format

* fix param problem

* fix op and composite rule

* fix bf16 cpu support problem

* fix bf16 cpu issue

* fix axis error log

* add axis for maximum

* revert commit

* remove .orig

* fix generic problem

* revert max op

* fix axis error

* fix maximum axis

* fix test_check_output

* fix cinn

* fix minimum maximum axis check

* Reorganize the forward codes of flash-attention.

* Fix forward.

* Remove some noused codes.

* Simplify codes and fix backward.

* Change all LOG(INFO) to VLOG and fix the backward.

* add scale for AF2 flash_attn, much thanks to xreki and shaojie for debug these codes

* decrease the effect of debug print on performance

* Unify the initialize of flashattn arguments.

* Rewirte the reshape of temp_mask and temp_bias.

* API support use_flash_attn.

* Fix compiling error on CI.

* Try to crop the flash-attention lib.

* Correct the condition of whether can use flash-attn.

* Remove the softmax_out argument.

* Remove is_causal.

* Polish codes.

* Fix qkv_transpose_out's shape and scaling of Q * K.

* Update commit of flash-attention.

---------
Co-authored-by: NLiu Yiqun <liuyiqun01@baidu.com>

* Add large dim test of log_softmax

* fix

* delete bf16 of cross entropy

* delete bf16 of cross entropy

* add meshgrid,expand_as, prod and grad bf16 kernel

* fix bf16 for optest

* modify code style

* fix amp test

* [CINN] Fix TestGelu unittest of CINN

* pass if_enable_cinn

* suport device_guard for npu

* fix comment

* fix typo

* Add i1 and i1e op

* resolve merge conflicts

* add bf16 for mode

* remove random seed 666

* try to fix op_type error

* test for me

* try to fix op_type

* fix redundancy code

* add fp,bf for lastdim

* fix some error

* simplify code

* fix shape error

* optype error

* fix skipif bf16

* polish

* polish

* add rules

* modify no kernel yaml parse

* success op generate

* success test_silu_double

* modify bug

* modify static error

* modify silu_grad input

* modify kernel signature

* modify kernel signature

* code style

* code style

* review

* delete opinfo modify

* modify gradOpMaker

* modify gradOpMaker

* modify genarated-j2

* add approve rules

* modify aytograd_functional_static_test

* trans fused batch norm Compute function

* trans batch norm register info to phi

* trans fused batch norm grad Compute

* trans batch norm grad register info

* add sig file

* update sig file

* Update fused_bn_activation_kernel.cu

* Update fused_bn_activation_grad_kernel.cu

* fix

* Rename fused_bn_activation_kernel_grad.cu to fused_bn_activation_kernel.cu

* fix

* fix

* fix CudnnDataType error

* fix

* fix include

* update

* add #if

* add fused bn act to cmakelist.txt

* update  cmakelist

* fix #ifdef error

* add timeout set

* add env set

* fix

* fix

* Update fused_bn_activation_sig.cc

* remove _in_eager_mode_

* remove _in_eager_mode_

* [phi]mv StftKernel to phi

* [phi] fix KernelSignature

* [phi]fix arr error

* [phi] Disable check_dygraph

* [phi]fix include

* [phi] rewrite mutable_data, add output register

* [phi] fix  Alloc

* [phi] fix Alloc again

* [phi] fix mutable_data

* [phi] fix onesided_out Resize

* correct 1th to 1st

* correct 1th to 1st

* fix typo

* fix typos

* add check ops for prim

* fix pow and concat composite registration

* modify log

* add note and remove useless code

* remove useless code

* modify program to check

* remove useless note

* relocate regularizer.py

* fix bug

* fix bug

* fix bug

* relocate the import

* replace _regularization_coeff with coeff

* remove the L1DecayRegularizer and L2DecayRegularizer

* update

* Update backward.h

* Update composite_backward_api.h

* Update tensor_utils.cc

* Update backward.cc

* update

* stype

* update

* add ctest

* code stype

* added base code for i0 and i0e

* added grad base code for i0 and i0e

* added i0 and i0e python code

* added ops and backward yaml config

* added i0 and i0e cpu kernel, but not test.

* added i0 and i0e code and unitest files

* added test files

* added i0/i0e gpu implementation code

* updated code style

* updated code style

* fixed unitests code

* updated i0 with eigen3

* fixed bug and added more test cases

* refactor: fixed static graph bug

* refactor: removed i0 and i0e from op_compat

* refactor: updated code style

* refactor: updated op_compat.yaml

* refactor: updated op_compat.yaml

* refactor: fixed op name mapping and optimize unittest case

* refactor: manually implement i0 / i0e

* refactor: added grad kernel for i0 / i0e,didn't finish

* Update math.py

* refactor: added equation to doc in English and added comments for computing i0 / i0e gradient

* refactor: removed eigen implementation

* refactor: finished i0 / i0e cpu and gpu op

* refactor: updated code style

* fix: find  a bug but not fix

* fix: incorrect unittest cases

* update: updated code style and remove my file

* update: updated unittest case

* fix: fixed sign error

* fix: fixed mistakes when merging

* refactor: updated code style

* refactor: remove unused code

* refactor: updated code style

* [Dy2St]Fix x grad names when high order gradient

* Polish error msg

* Add inputs var to backward in dy2st

* Fix error

* Get grad names for backward API

* Fix save load

* Polish code

* Add ut

* [prim] fix not support optional grad bugs in higher order autodiff

* [prim] remove duplicate fill_any_like caused by infershape_for_composite

* fix _strip_grad_suffix_ bugs in higher-order autodiff

* [prim] create output for test_static_prim.cc

---------
Co-authored-by: N0x45f <wangzhen45@baidu.com>

* move DataLoader to paddle.io. test=develop

* Fix div error when dtype is int64 in static mode

* Fix out dtype

* add index_put api

* fix value broadcast in backward and add test case in static

* add timeout=120s for index_put

* add op_compat for index_put

* add inplace index_put test

* add test case when index tensor in indices is int32 when indices.size less than x.dims

* add index_put api backward in cpu place

* add backward test case

* refactor code to delete some duplicated code

* replace reshape with resize for decrease extra memcpy

* add datatype flag in backward yaml

* fix bug in documentation

* Update python/paddle/tensor/manipulation.py

---------
Co-authored-by: NLigoml <39876205+Ligoml@users.noreply.github.com>