* add interpretercore

* refine backward program id

* add code

* refine program

* refine code

* create forward/backward_program by prog2graph2prog method

* test, do not care

* refine code

* refine code

* refine code

* test, do not care

* add interpretorcore

* add scope

* refine scope create method

* add jit for new_exe

* solve conflict

* delete unused code

* polish code

* polish code

* refine scope in inplace

* refine for datatransfer

* refine _rebuild_from_desc

* refine control eager deletion attr

* refine used_for_jit

* refine jit for infer

* op size0 use ori program

* polish code

* refine jit

* refine run_program_op ut

* refine inplace

* refine control

* refine graph helper

* refine control

* refine inplace

* refine buffer_share_inplace_pass

* polish code

* polish code

* refine usage for compilerProgram

* refine control

* test

* test core cache

* refine code

* refine io.py

* increase test_seq2seq timeout

* refine convert program

* refine interpretercore_cache release

* delete buildinplace

* refine partial_program && io

* refine code for io

* test

* test

* test

* unify the size and size() by jst

* fix bugs

* bug fix.

* fix bugs

* change all_close -> np.testing.assert_allclose

* while support for python container.
It is convenient to convert more dynamic graph codes into static graphs.

* cond support python container

* 1. make select_input output shape = input[1]
2. add warning in while_loop risky assign

* fix 2 problem in GPT export:
1. a bug in while_op no_need_copy_var, which causes gpu memory leakage
2. a bug in undefined_var where the stop_gradient should be False.

* change name by code review

* format

* while support for python container.
It is convenient to convert more dynamic graph codes into static graphs.

* cond support python container

* optimize python api to speed up eager exec

* optimize python api to speed up eager exec

* optimize python api to speed up eager exec

* Make sure that the output of whilep must exist in the input

* insert assign in block(0)

* add unittest.

* fix_shape

* [Eager]Support Lazy initialization for nn.Lazyer

* fix

* add control flag and input shapes for new dygraph

* fix file mode

* improve code coverage

* fix a bug in statstic

* fix according to review

* optimize performance

* fix

* 1. add prepare_gradient_aggregation in PartialProgramLayer

* 1. draft

* fix ci problem

* Fix abnormal format of message when raise KeyError in Dy2St

* Format code

* Format code

* Add UT

* Rename method

* add name attribute to code-block, test=document_fix

* remove redundant labels, test=document_fix

* remove redundant labels (from upstream), test=document_fix

* more COPY-FROM (try multiple code example), test=document_fix

* empty commit, try to trigger PR-CI-build

* fix some `Examples:` format issues

* fix some ci errors

* [CustomDevice] add custom ccl 2/2

* update

* update

* update launch

* release the convert_len()

* fix bugs

* fix bugs

* fix start error.

* when shape==0, we should also call paddle.shape

* just a test

* remove split assign test

* remove other useless code related to split assign

* [Eager] call final_state_slice under eager mode

* rm useless comments

* use eager final state instead intermidiate state

* update fill_constant yaml

* update fill_constant yaml

* modify wrapped_infermeta_gen logic to fix special case

* fix slice in manipulation

* use fill_constant_

* modify slice infermeta

* rm final_state_conv2d

* use final_state_slice

* use final_state_slice only

* polish slice, use final state

* add paddle_throw for SplitInferMeta

* rm fill_constant_ temply

* recover array_equal, not allclose

* recover original code

* Fix to_static error when call to_static to convert a method of a common class

* raise typerror when class no inherits from layer

* Fix @to_static

* jit.save support peropty serilization

* extract set property function

* fix property test file name

* fix typing error

* fix typing error

* fix test coverage

* fix behavior of device_id=None in Tensor.cuda

* fix CI

* Fix jit.save error when save params combined

* Change dict_value to list

* fix the outputs of net is x,x

* add unittest for duplicate output

* fix

* fix _infer_program use the original program not the amp program.

* get _***program_id back and avoid duplicate cache
ing

* fix

* Replace with dygraph op calling method.

* fix the outputs of net is x,x

* add unittest for duplicate output

* fix

* change NameVisitor to FunctionScopeAnalysis

* polish the logic of undefined var in while_loop. create vars after body execution

* replace old NameVisitor in while and fix all CI

* Togather with CreateVariableTransformer

* add create_variable_transformer

* fix bugs

* merge

* fix some error, TODO: ForNodePreTransform ahead

* merge for unite PR

* fix conflict with base_transformer PR

* fix ci errors, fix [for i in range()] error

* fix according to code review

* Polish visit function in transformer

* Call generic_visit first in visit_While/For

* Remove comments

* Polish utils.py, move some transformer to base_transformer

* add support for control flow block analysis

* move FunctionNameLivenessAnalysis into utils

* pass test_ifelse.py

* remove duplicate data_layer_not_check

* pass the test_ifelse.py

* fix unittest error .

* fix all ci error in first version

* temporay disable CreateVariableTransformer

* fix ci errors

* fix function name liveness analysis bugs

* modifty def cond

* fix

* fix ci error - v2

* fix by code review

* change return_name_ids -> return_name

* change fluid.mean to paddle.mean

* reverse some old code examples

* refine tensor.dtype for bloat16

* refine test

* revert

* refine bfloat16 print

* Add BaseTransformer for dy2st error message

* Fix return_transformer error

* Polish dy2st error info in runtime

* Fix UT error

* Polish runtime error code

* [Dy2Stat]Polish break/continue statement transformer logic

* [Dy2Stat]Enhance nonlocal machanism while returning single var

* [Dy2Stat]Enhance nonlocal machanism while returning single var

* save multi program into one param and seperate model

* export class property