* [C++ API GetAllocator] Add C++ GetAllocator interface

* move api to accurate directory

* add cumsum prim backward

* skip aixs=None test case

* fix op generante eror

* fix static test error

* remove unused code

* fix static test error

* skip cpu float16 test case

* skip eager cpu cumsum float16 test case

* add cinn test

* reshape flatten out

* Disable cinn single test

* remove cinn test

* reformat todo

* add prim in cumsum op test

* remove old test

* fix typro

* fix typro

* fix typro

* pass axis=None test case

* remove forward prim test

* remove same name axis

* add unittest for nn.DropOut2D

* add fp16

* add fp16 in docs of temporal_shift_op.cc

* Update test_dropout_op.py

* forbid tensorrt_engine op's output is a persistable var

* fix concat bug

* recommit for ci

* support transpose and reshape

* support reshpe, transpose, cast vjp

* merge develop

* recover unused file

* remove prim base

* support problem

* remove additional status settting

* remove additional status settting

* fix ut

* fix ut

* fix ut

* fix no grad branch

* add more test

* disable fp16 in cpu

* fix test

* add inferface of get registered phi kernels

* change KernelType to KernelKey

* add test

* refactor code

* [XPU] add fp16 support for shape op.

* [XPU] add fp16 support for lookup_table_v2 op.

* update approval list: add qingshu's id.

* remove utils

* remove utils

* remove utils

* remove utils

* Update get_data_from_tensor.h

* Update rnn_functor.h

* Update rnn_grad_kernel.cu.cc

* Update rnn_kernel.cu.cc

* Update rnn_kernel.cc

* Update rnn_grad_kernel.cu.cc

* Update rnn_functor.h

* Update rnn_kernel.cu.cc

* Update rnn_kernel.cc

* remove utils

* Update rnn_functor.h

* remove utils

* remove utils

* remove utils

* remove utils

* remove utils

* Update rnn_functor.h

* Update unsqueeze_op.h

* Update utils.h

* roll back

* Update tensor_utils.h

* Update tensor_utils.h

* Update tensor_utils.h

* Update tensor_utils.h

* Update tensor_utils.h

* use TensorToVector

* use TensorToVector

* use TensorToVector

* use TensorToVector

* use TensorToVector

* Update rnn_kernel.cc

* Update rnn_grad_kernel.cc

* Update rnn_functor.h

* Update rnn_grad_kernel.cu.cc

* Update rnn_kernel.cu.cc

* Update rnn_functor.h

* Update rnn_grad_kernel.cu.cc

* Update rnn_kernel.cu.cc

* Update rnn_functor.h

* Update rnn_grad_kernel.cu.cc

* Update rnn_kernel.cu.cc

* add TensorToVector

* roll back

* Update tensor_utils.h

* Update rnn_functor.h

* Update rnn_grad_kernel.cu.cc

* Update tensor_utils.h

* Update rnn_kernel.cu.cc

* Update rnn_grad_kernel.cc

* Update rnn_kernel.cc

* Update rnn_grad_kernel.cu.cc

* Update rnn_kernel.cu.cc

* Update rnn_grad_kernel.cc

* Update rnn_kernel.cc

* TensorCopySync to phi::Copy

* fix codestyle

* rnn_kernel.cc: add ;

* replace all GetDataFromTensor with phi::GetVectorFromTensor

* delete include of util.h

* add sm version check

* use GetGPUComputeCapability

* [Tensor Operants & Prim] Tensor pow API uses elementwise_pow

* unittest change to fill_constant+elementwise_pow

* [Error Msg] Polish error message when GPU kernel not found

* Only test in GPU environment

* conflict

* add UpdateSliceAttrs

* [kunlun] support reduce_scatter

* uncomment unittest

* update xccl to 1.0.10

* add TypeUniquer and IrContext

* refine include code

* add Type, TypeBase

* add built-in type

* add bulit-in Float32Type

* refine ut

* refine code

* refine code

* delete type_base

* rename ImplType to StorageType

* rename ImplType to StorageType

* add macros util for register type

* add macros util for register type

* refine name

* refine name

* change storage manager

* add multi_thread for ir_ctx

* rwlock_2_spinlock, add REGISTER_TYPE_2_IRCONTEXT

* DECLARE_TYPE_UTILITY_FUNCTOR

* refine ircontext singleton

* del destructor for ParametricStorageManager

* refine code

* Add necessary logs for debugging

* refine ir_context instance

* refine type get interface

* refine code by comment

* [XPU] bind op scatter_nd_add

* [XPU] add more data type for op: clip, transpose2 & assign_value

* register bfloat16 datatype for squared l2 norm

* register bfloat16 datatype for softmax with upper triangular mask

* register bfloat16 for tril triu cuda kernel

* [mv fleet] mv fleet to distributed

* [mv fleet] for ci

* [mv fleet] for ci

* [mv fleet] solve ci of version

* implement of matmul using cublasLt instead of cublas

* Update matmul_kernel_impl_via_blasLt.h

---------
Co-authored-by: Nzhangbopd <1299246947@qq.com>
Co-authored-by: NBo Zhang <105368690+zhangbopd@users.noreply.github.com>
Co-authored-by: NLiu Yiqun <liuyiqun01@baidu.com>

* Enable matmul + bias fusion in fused_gat_attention.

* Add a variable to control whether using fused matmul + bias.

* rename elementwise_heaviside to heaviside

* delete __init__.py

* fix bug

* allow fall back to fp16 when int8

* refine code

* refine code

* refine code

* ConvertToFusedOp

* change static to inline
Co-authored-by: NTomasz Socha <tomasz.socha@intel.com>

---------
Co-authored-by: NTomasz Socha <tomasz.socha@intel.com>

* change amp with to_prim

* fix prim amp

* fix rules

* fix liear

* add amp test

* add test

* disable this test on cpu

* disable this test on cpu

---------
Co-authored-by: Ncyber-pioneer <chenzhuo@tju.edu.cn>