* make params_grads order same bewteen dynamic and static mode

* revert inplace clip

* use sorted attribute to control

* tiny fix

* fix find loss_grad_op

* remove the optimizer base and learning rate base

* fix bug

* fix bug

* open

* update

* Update autoparallel DistributedDataLoader

* add places for engine.dataloder()

* simple reaplce

* for debug

* fix bugs

* fix some bugs

* del fill_constant_batch_size_like

* add shard tensor api

* add DistAttr api

* add unittest for coverage

* fix process mesh sample code

* fix checking error

* Support nvprof for auto parallel

* Fix CI errors

* Fix CI errors

* remove allreduce before c_allgather

* update reshard insert_fill_constant_op func

* insert_fill_constant_op add shape arg

* add skip_gc_vars for 1f1b schedule mode

* add pp_degree and pp_stage

* base rule

* add sharidng merge

* add sharidng axis merge

* define unified data class for inferencing dist_attr

* test wrap DistTensorSpec in dygraph mode

* matmul main logic done

* define unified data class for inferencing dist_attr

---------
Co-authored-by: NYichen Zhang <zhangyichen03@baidu.com>

* auto parallel support pipeline scheduler with standalone executor

* rm check_fetch

* update cmakelist and flags env

* rm set micro batch id

* rm import

* update utils func

* raise error when merge tensor for return_numpy is False

* fix _pipeline_opt

* fix unittest

* bump ruff to 0.0.271 and update config

* exclude third_party

* bump ruff to 0.0.272

* refine config

* eager call all2all to avoid p2p hang in lazy init

* update

* [AutoParallel] add 1F1B

* rm amp

* [AutoParallel] update while control_flow with pipeline

* update process group instantiate

* fix micro_bsz for reshard

* update api for micro batch size

* add strategy for dp optimization

* [Auto Parallel] Reorganize the fold structure

* [Auto Parallel] Fix some import errors

* global view process_group

* fix import

* fix attr

* fix tunner init comm

* [CustomDevice] fix auto_paralell

* update

* update

* update

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* fixed cyclic reference that caused patial import

* fixed bad change

* fix bad import

* fix bad import

* fix bad import

* fix ut failed caused by change in_dynamic_mode

* fix ut failed caused by change in_dynamic_mode

* fixed usage of in_dynamic_mode() or in_dygraph_mode()

* revert python3 to python in .pre-commit-config.yaml

* fix merge conflicts

* rm cmake npu

* Update generic.cmake

* Update generic.cmake

* rm npu

* rm use_npu

* rm npuid

* rm use_npu

* rm npuid

* delete npupinned

* roll back sth.

* roll back sth.

* delete npupinned

* roll back sth.

* roll back sth.

* rm npu

* rollback something

* rollback npu identity

* rollback npu identity

* move DataLoader to paddle.io. test=develop

* support 0-D output and 0-D as indice in __getitem__

* fix tests

* fix inference and UT

* add unittest for setitem

* fix xpu test

* fix xpu 0-d

* add flash randomness control

* fix VLOG undefied

* add random control for fused dropout add

* add __init__

* add o2 tune

* add unittest

* fix error

* set unittest timeout

1. modify set_value op, use Scalars to represent attr `values`, instead of a bunch of attributs of various types; (#52408)

2. add program converter and set_value op as an example, which provides the functionality to convert `paddle::framework::ProgramDesc` between old and new formats(the differences are mainly some operators with incompatible updates in the definition);
3. program version and operator version map now are always saved when serializing `paddle::framework::ProgramDesc` to identify the version;
3. provide an option `legacy_format=false` in  serialization of `paddle::framework::ProgramDesc`, it decided whether to convert ProgramDesc back to a legacy format, which is compatible for paddle 2.4.2 or earlier versions to load and execute;
4. deserialization of `paddle::framework::ProgramDesc` is now automatically detecting whether the bytes it receives is in legacy format(contains any of the operators that has been incompatibly updated and have any attribute of type `Scalar`) and convert it to new format. But if you want a faithful deserialization without the automatic conversion, you can use protobuf's deserialization instead. Though it is not recommended, it can be used for the purpose of testing.

* [Auto Parallel] Speedup the completion process

* [Auto Parallel] Skip the property of dist_context when deepcopying

* [Auto Parallel] Remove the unnecessary print

* [Auto Parallel] Move some changes from 2.4 branch to develop

* Update engine.py

* [Auto Parallel] Fix a bug