* inplace tol

* code style

Update Save/Load Interface to 2.0

* Work for fa-2 padded fwd. Code to be cleaned.

* Work for fa2 unpadded fwd.

* Work for padded-bwd, dk get small diff on np.random.seed(0)

* Anyway I pass paddle's utest, except return softmax without dropout.

* Clean code.

* Modify interface.

* Clean code and add some check.

* Easy compile for dev.

* Fix ci.

* Fix ci-build.

* Add std c++17 option again.

* Limit max job when compiling fa2.

* Remove const_cast

* Add fwd params, to be cleaned.

* Clean code.

* Add bwd params.

* Clean code.

* Add enforce.

* Use v2.0.4

* Pass RNG state to fa2 capi

* Fix review.

* Add assert

* Skip compile for sm less than 80.

* replace embedding

* replace sparse_embedding

* fix some bugs

* del embedding

* repalce layers.embedding

* fix type error

* fix bug: feed_with_place should consider variable existence

* fix

* fix build scope

* change method to set feed var name

* remove feed_with_place to placeholder

* fix

* rename to data

* fix

* fix

* refacot aot

* update

* fix bugs

* remove some test

* fix bug

* fix bug

* fix bug

* fix bug

* update

* [clang-tidy] enable modernize-use-emplace

* Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into modernize_use_emplace

* add decorator for new_ir_test

* fix bug and only test in ci-coverage

* fix bug and only test in ci-coverage

* fix

* fix bugs

* fix

* fix

* WIP: start writing combined indexing get

* list/tuple/Variable

* getitem 80%

* add setitem

* add some unittest for setitem

* lazy import

* fix some setitem error

* fix advance indexing with decreasing axes; fix strided_slice input name

* combine int-tensor getitem is ok (without boolean support & broadcast); add getitem unittest for static

* add broadcast & parse bool tensor for __getitem

* [change getitem] _getitem_impl_ to _getitem_static, not deleting the former one

* refine new getitem; fix ut in variable/var_base

* add __getitem__ ut in dygraph

* re-dispatch getitem for Py/CPP; fix strided_slice decrease axes error in dygraph

* fix ut; support tensor in slice

* [change setitem] _setitem_impl_ to _setitem_static, not deleting the former one

* remove some UT (for some, temporarily)

* add IndexError to solve timeout problem in static-mode

* 1.temply forbideen all-False bool-indexput; 2.setitem_static will return new variable

* xpu uses old stratege

* rename dy2st setitem ut to avoid same-name problem

* dy2st for new combined index

* ut case for combine-index with dy2st

* open ut with all-false-bool setitem

* remove useless doc and _getitem_impl_

* change static res

* fix static xpu

* fix fp 16

* bf16 rtol

* fixed input

* code style

* add interface

* add code

* add code

* add code

* add code

* fix bug

* fix bug

* add var prefix

* add code

* add code

* add code

* fix compile bug

* fix bug

* refine code

* refine code

* refine code

* refine code

* fix bug

* add code

* add code

* fix bug

* add code

* add code

* refine code

* refine code

* fix bug

* add code

* fix bug in phi__kernel_utils

* refine code

* fix bug

* open flag

* refine code

* fix bug

* fix bug

* refine code

* fix bug

[Inference] Replace groupNorm when data types are bf16 and fp16, and data format is NHWC implementation. (#55399)

* finish

* cpergroup odd

* fix bf16

* single channel

* code style

* jingdu duiqi

* add head_file

* add bf16 head file

* bf16 2

* bf16

* bf16 head

* bf16 compile

* py test

* bf16 compile

* bf16 compile

* unset py test

* nhwc

* test

* mean var

* bf16 success

* su

* ctest success

* use is_same_as

* is_same

* use is_same

* rtol

* gpu_stream

* del sigmod

* fix bfloat16 type

* use cuda_bf16_hpp

* use_cuda_arch

* bfloat162float2

* del inplace_tol

* del max_releative_tol

* temp store

* jingdu duiqi

* temp store

* plugin

* jingdu duiqi

* duiqi

* include cuda.h

* del half

* half single

* ci

* add const

* ci

* cudamemset

* del printf

* fp16 test

* add half compute

* del br16 ci

* del ci

* ci approve

* del fluid include

* fix security bug

* Update autoparallel DistributedDataLoader

* add places for engine.dataloder()

* add ir core

* add test

* modify name

* merge

* add test for __eq__

* shield  test for __eq__

* --amend

* Update new_ir_compiler.cc

* done replaced in d2s models, except lac/transformer/sentiment

* replace cinn_test

* fix d2s transfomer model

* fix cinn model ut

* new ir support print op

* fix gpu bug

* fix bug

* update

* remove layout to string

* remove usless header

* polish code

* fix bug

* posolis code

修复Paddle With CINN 执行 python单测下op_mapper 时报gflags符号undefined问题。

* add ir output check in OpTest

* add ir grad check in op test

* fix legacy name converter bug

* add more unittest

* fix

* fix warprnn op bug

* add whit list

* polish code

* polish code

* fix unique and close op bug

* fix bug

* update

* fix new ir unique is close bug

* remove useless code

* use new stringattr api

---------
Co-authored-by: Nkangguangli <kangguangli@hotmail.com>
Co-authored-by: Nzhangbo9674 <zhangbo54@baidu.com>

[CodeStyle] replace `assert np.allclose` with `np.testing.assert_allclose` and `assert np.array_equal` with `np.testing.assert_array_equal` (#55385)

* update TestAPI arguments to enable param_attr and bias_attr in test_layer_norm_op

* add bf16 condition in test_layer_norm_op

* add fast_math condition

* support construct network using ir in Python

* fix py3 bugs

* change source return type to OpResult

Fix imperative out scale unittest