* fix py::array_t calling bug

* fix typo, implemention->implementation, test=document_fix

* tmp gather vjp

* support gather

* remove useless code

* fix compiling error

* fix ut

* add eager test

* add eager test

* add seed

* fix cpu error

* fix transpose op compat

* remove tensor index case

* fix prim_cinn

* fix ut

* support bw invoke fw

* fix scale in static_backward.yaml

* fix the bug in tensorrt/convert

* move 'scale','sign' into ops.yaml

* add scale_grad of scale in op_compat.yaml

* change generated_static_op in CMakeLists.txt

* fix flatten op map

* remove prim op all list

* add op map info of full_like

* polish code

* support prim test in OpTest

* fix cmake

* fix op test

* fix test_input_spec

* disable cinn in reduce_sum unit test

* add bfloat16 dtype for sum

* polish code

* add clear jit program function

* convert grad out from tensor to numpy

* remove unnecessary code

* add only_prim flag

* fix flag

* fix op test

* fix optest comp inplace error

* fix op test

* fix op test with guard

* add initialization of check_comp flag

* fix comp inplace error in op test

* rename check_comp with check_prim and add bfloat16 dtype convert

* rename comp_op_type to prim_op_type

* rename comp to prim

* remove useless code

* skip ci check for only prim

* add no_grad_vars and grad_outputs in prim test

* fix var_dict

* fix op test for only_prim

* fix dy2static bugs

* polish some code

* testcase init commit

* skip sign conflict

* fix year

* decouple_memory

* perfect memory utils

* fix ci bugs

* fix inference bugs

* fix custom test bugs

* fix converage bugs

* modify code according comment

* modify namespace

* deal with compile bugs

* process unit test matched test_p*

* fix ci bug

* fix ci bugs

* fix codestyle

* remove all tests about pe and restore some irrelated tests

* delete test_parallel_executor_test_while_train.py

* assert len(self.places) == 1

* remove with_data_parallel in unittest

* remove multi-card related checks

* remove with_data_parallel in test_multiprocess_dataloader*

* fix error hint

* fix typo

* Flatten input_spec when check neg shape

* Fix cycle import

* move sequence_padding to phi

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix bugs

* fix buga

* fix bugs

* revert and update phi::XPUContext

* add static custom back_list

* rm comments

* fix log

* fix comment

* fix composite mean op map

* fix composite check output

* init layer_norm

* init layer_norm

* map output from composite rule to origin op

* add dropout op map

* add input map check

* polish log

* modify rules

* success test_forward

* modify test without cinn

* modify cinn test

* modify cinn test

* except fp64

* except fp64

* delete flatten

* delete unused change

* review

* pass cpu test

* code style

* delete flatten fp16 error

* modify flatten test

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

* Optimize the ernie inference performance on xpu

* fix enable runtime cache logic

* when op's input shape has changed, should create a new runtime context

* fix

* set flag when input shape has changed

* Check -1 shape for input_spec and program when prim or cinn enabled

* Polish neg shape check

* Polish code

* Fix UT

* Fix UT in static

* polish tensor operants implementation

* change year, 2021->2023

* autogen tensor.h and tensor_api.cc

* polish CMakeLists logic

* cancel tensor.h auto-gen

* clean useless parameter

* delete tensor_api.cc

* [RM FLUID] trainer_pass&heter_trainer_pass
* [RM FLUID] rm distributed_strategy

* autogen phi tensor operants

* autogen operants base

* autogen operants_manager

* polish code style

* add cmake output

* fix cmake depency error

* fix cmake file path error

* decouple prim/api/api.yaml in phi CMakeLists

* polish CMakeLists of phi/api/lib

* [dy2static] support fallback for whole graph. (stage 1)

* bug fix

* bug fix and add a new unittest

* fix code by code review

* fix coverage

* [dy2static] ParameterRecorder Part 2

* Parameter Recorder - 2 bug fix

* bugfix: fix the dygraph go into _jst.ld() errors.

* fix ci error.

* fix ci error

* update include directory

* fully support C++ extension, pass unittest

* fix include directory

* support both extension and operator in one file

* polish testcase

* add jit unittest

* update third_party.cmake, pass CI test

* fix cmake

* fix setup

* fix inference, fix unittest precision

* fix unittest precision

* fix inference_lib cmake

* try fix setup, try fix inference_lib

* try fix inference_lib pybind

* fix mix_op_extension, fix inference_lib

* fix mix_op_extension, fix inference_lib

* change cmake

* change cmake

* add compile flags

* add Python.h headerfile

* add test_custom_plugin_creater cmake

* comment compile flag

* pass all CI

* pass all CI

* comment compile flag

* try solve test_custom_plugin_creater link error

* try solve test_custom_plugin_creater link error

* polish codes

* remove windows compile flag

* remove python_include_path

* update pybind11, 2.4.3->2.6.0

* update pybind11, 2.6.0->2.10.0

* update pybind11, 2.10.0->2.6.0b1

* update pybind11, 2.6.0b1->2.6.0, start fix unittest

* fix pybind11 2.6.0 VarBase print error

* fix pybind11 2.6.0 VarBase print error

* handle PADDLE_ON_INFERENCE

* modify according to reviewer

* fix cmake

* cmake decouple pybind_util when not ON_INFER

* cmake decouple pybind_util when not ON_INFER

* remove copy of inference_lib.cmake

* change pybind.cc headerfile fluid->phi

* polish tensor operants implementation

* change year, 2021->2023

* move save_op to fluid

* fix namespace

* move_load_kernel

* fix kernel_register

* move serialization to fluid

* fix test

* fix bugs

* check win

* fix random error

* fix manage