separate mtp head & loss for pp balance

[Wennie396]

PR types

New features

PR changes

Others

Description

把mtp的lmhead和loss拆分到上一个pp stage，减少最后一个pp stage的时间，使pp更均衡

[paddle-bot]

Thanks for your contribution!

[codecov-commenter]

Codecov Report

❌ Patch coverage is 57.14286% with 3 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@733171e). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
paddleformers/trainer/trainer.py	0.00%	2 Missing ⚠️
paddleformers/transformers/gpt_provider.py	80.00%	1 Missing ⚠️

❌ Your patch status has failed because the patch coverage (57.14%) is below the target coverage (75.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #4327   +/-   ##
==========================================
  Coverage           ?   39.13%           
==========================================
  Files              ?      474           
  Lines              ?    89340           
  Branches           ?        0           
==========================================
  Hits               ?    34967           
  Misses             ?    54373           
  Partials           ?        0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Wennie396]

/re-run all-failed