[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

* [Paddle-TRT] add flip op

* Paddle-TRT support float64  in/out type, support fill_any_like_op in int64

[Paddle-TRT] Convert 0D tensor to 1D tensor, increase the shape tensor's number count when collecting shape (#55503)

* make 0-D tensor to 1-D tensor to support Grounding-SAM and add shape check

* recover identity_op_clean_pass.cc

* onednn: remove fc+eltwiseadd fusion pass
* onednn: remove post-sum fusion in fc kernel
* onednn: tests: make unfused add run into f32

* fix cudnn 8.7+ bug on cudnnConvolutionBiasActivationForward

* save_optimized_model_pass support gpu

Issue #55016

* Fix TRT multihead matmul UT failure

* support_conv2d

* remove comment

* check code style

* add former Test

* check code style

* add unittest

* fix log

* change unittest

---------
Co-authored-by: zhoutianzi666 <17801055074@163.com>

* [Paddle-TRT] add assign op

* copy dense_tensor.h to inference lib

* update

* update

* rewrite identity_op_clean_pass

* fix

* adjust identity_op_clean_pass order in gpu passes

* fix ut

* squash pick the poc code
* fix build after rebase
* fix int8 conv and fc uts
* Fix and clean-up Get_SRC_Scale_Memory
* fix floating point fc uts
* fix test_analyzer_int8_googlenet
* test_analyzer_int8_mobilenetv1
* fix int8 mobilenet v2 and v3
* fix build error after rebase
* [oneDNN] rename library version
* fix conv bias datatype
* try to fix import error
* fix rebase error
* [oneDNN] pack library into python wheel
* add MKLDNN_SHARED_LIB_3 to env_dict
* fix test_analyzer_bert
* fix fill_constant op kernel
* fix ernie and matmul op ut
* fix softplus ut
* fix conv+relu6 fusion ut
* fix hardswish fusion
* fix quant+transpose fusion ut
* fixsgd ut
* fix int8 matmul with flatten
* fix fc+scale fusion
* fix conv/matmul+gelu fusion uts
* fix rebase error
* Revert "fix conv/matmul+gelu fusion uts"
This reverts commit 47eb5e49972bd8f7271a233def9bfb3e98ce78e1.
* upgrade to onednn v3.1
* remove older version onednn
* use densetensor::data() for achieving mean and var in layernorm impl
* comments for atol of integer tests
* fix clang-format
* Revert "remove older version onednn"
This reverts commit 783e57ddfd4401254596eae7d47adb9b03590c09.
* improve binary handle
* fix expand kernel
* Revert "use densetensor::data() for achieving mean and var in layernorm impl"
* always use forward_inference for conv
* remove activation scales
* rollback changes to mkldnn.cmake
* address comments
* port changes to dequantize kernel
* fix merge error
* fix fused_elementwise_kernel
* upgrade onednn version to v3.1.1
* fix some approval error
* fix error msg format
* remove old onednn libs
* try to fix symbolic link issue
* fix cinn test case segfault
* do not explicit link test with onednn
* remove unnecessary changes
* integrate CINN with onednn v3
* link with mkldnn project
* fix cinn build file

---------
Co-authored-by: NTomasz Socha <tomasz.socha@intel.com>
Co-authored-by: NChen, Xinyu1 <xinyu1.chen@intel.com>
Co-authored-by: Ntianshuo78520a <707759223@qq.com>

* [clang-tidy] enable readability-container-size-empty check

* fix test_custom_kernel Failed

* add clang-tid-10 in dockerfile

* add clang-tidy in dockerfile

* fix bug

* fix WITH_SHARED_IR option type

* rename WITH_INFERENCE_NVTX to WITH_NVTX and fix compile bug

* update

* commit

* del inplace op in memory_optimize_pass.cc

* check code style

* commit

* commit

* commit

* commit

* final commit

* use hook to collect shape and shape value

* fix clang-format

* 'fix_clang-format'

* fix remaining errors

* format

* empty commit, re-trigger all ci

* empty commit, re-trigger all ci

---------
Co-authored-by: NSigureMo <sigure.qaq@gmail.com>

* add einsum layer

* set_value update

* support ValueTensor's rank != Input'rank & update topk

* update range to avoid coredump

* fix addShape error

* Dims definition differ between 7.2 and 8.0+

* Update test_trt_convert_top_k_v2.py

* update top_k

* Update test_trt_convert_top_k_v2.py