Refactor pipeline_parallel code

[ForFishes]

PR Category

Distributed Strategy

PR Types

Others

Description

Split the large pipeline_parallel.py file (3995 lines) into multiple modules for better maintainability:

• pipeline_impl/init.py: Export all public interfaces
• pipeline_impl/utils.py: Utility functions and classes
• pipeline_impl/hook_system.py: Hook system
• pipeline_impl/p2p_handle.py: P2P communication
• pipeline_impl/no_pipeline.py: NoPipelineParallel class
• pipeline_impl/base_pipeline.py: PipelineParallel base class
• pipeline_impl/interleave_pipeline.py: PipelineParallelWithInterleave class
• pipeline_impl/fthenb_pipeline.py: PipelineParallelWithInterleaveFthenB class
• pipeline_impl/balanced_memory.py: VPPFhenBInBalancedMemory class

The original pipeline_parallel.py is now a simple redirect file that re-exports all public interfaces from pipeline_impl sub-package, ensuring backward compatibility.

是否引起精度变化

否

[paddle-bot]

你的PR提交成功，感谢你对开源项目的贡献!
请关注后续CI自动化测试结果，详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

root seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}