* support fp16 for maxout op

* format code

* change api

* add test for static float16

* format code

* formatting code

* atol alignment

* experiment—1

* experiment-2

* experiment-3

* format code

* trans fused_feedward Compute function to phi

* add register info

* remove maxfunctor

* move fused feedward to phi

* remove sig file

* remove fliud include

* add include

* add include

* add sig file

* add output register info

* fix sig file

* Update fused_feedforward_sig.cc

* fix grad kernel

* update output register info

* fix

* open fused_feedforward static build

* add optional and fix code style

* fix output info for fused attention

* add optional param

* merge

* support OD level and skip dynamic loss scaling for bf16

* [Fix CppExtension Unittest] Change CUDAExtension to CppExtension if necessary

* Temporarily test cpp_extension under GPU

* Split mixed_extension unittest

* add jacobian and hessian in paddle.autograd

* disable unitest 'func_multi_input' for bug in high-order gradient of multiply

* add dimension checks

* add support for 0-D tensor

* change return type from Jacobian to Hessian in hessian function

* refine Jacobian _flatten function for single xs

* refine support for 0-D tensor

* 1. add 'func_multi_input' unitest for multiply_grad_kernel bug fixed
already.
2. support non-inplace math operation via magical method overwriting.

* add unitest for math operation and raise error when 0-D tensor is indexed

* add ndim check on ys and xs according to is_batched, and add one unitest

* refine docstring of jacobian and hessian

* move paddle.incubate.autograd.Jacobian/Hessian to paddle.incubate.autograd.functional.Jacobian/Hessian

* remove single_input unitest case because numerical differentiation is wrong

* remove 3 unitest for numerical result(reference result) is wrong

* 1. rename autodiff.py to autograd.py
2. increase TIMEOUT to 100

* cancel modification for functional Jacobian/Hessian

* 1. use tuple as return type instead of list
2. refine docstring

* add more unitest case to improve coverage

* remove 2 unitest of Hessian for numerical result is wrong

* remove 1 unitest of Hessian for numerical result is wrong

* remove 1 unitest of Hessian for numerical result is wrong

* change unit test to shape check

* correct doc and replace incubate API to stable API in _grad

* modify concat_grad add sum comp rule

* modify opcompat

* updata Adamw.py

out.backward()  -> loss.backward()

* Update adamw.py

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* [XPU] remove scale_loss in parallel.py

* [XPU] throw Unimplemented when using Reducer

* update

* fix bug

* update cmake3.16 to 3.18

* test

* Update Dockerfile.ubuntu

* add pad op

* add_some_code

* modify some code

* add some code

* add some code

* modify some code

* add some code

* modify some code

* Update composite_backward_api.h

* modify some code

* add some code

* add some code

* add some code

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* rm paddle_install_dir

* update test=document_fix

* update test=document_fix

* update

* add test

* update

* update

* update

* update

* update

* update

* update

* update

* update

* add todo commont

* add scatter_nd_add comp

* add scatter_nd_add prim

* fix

* fix

* add public_python_api in TestScatterNdAddSimpleOp setup function

* fix composite_backward_api.h

* fix composite_backward

* add test cases

* fix composite_backward_api.h, unittest

* Fix fused_attention_op and fused_feedforward_op bugs in xpu

* Fix d_x alloc errors for fused_feedforward_grad_kernel

* delete prim flag for matmul_2_grad

* delete prim flag for matmul_2_grad

* recover multiply prune

* add rules

* add rules

* add rules

* modify approve rules

* fix conflict

* recover block

* add test cases, test=allcase

* fix test cases, test=allcase

* fix test cases, test=allcase

* assert_allclose, test=allcase

* 1e-5 to 1e-4, test=allcase

* change rtol from 1e-4 to 1e-3, test=allcase

* modify numel in lbfgs and add a new test case. test=develop

* change param 'lr' to 'learning_rate' in lbfgs and its test

* add opt LBFGS and change test

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* test,test=develop

* optimize embedding deterministic mode

* fix compile error

* change FLAGS_cudnn_deterministic to int64

* fix 700 error

* add ut

* fix ut

* fix ut

* fix win32 ci

* fix flags with PHI_DEFINE_EXPORTED_int64

* add leaky relu composite rule

* add public python api

* unset default negative slope

* fix unittest case

[Bug Fixs] fix bugs when using cast<int64_t, int32_t> in xpu/cross_entropy kernels, *test=kunlun (#53325)