Add constraint for sharding overlap

[iosmers]

PR types

Function optimization

PR changes

Others

Description

为sharding stage1 overlap优化添加一些约束项

1. allgather优化只针对stage1 v2
2. broadcast overlap优化只针对stage1 v1
3. logging_step必须大于1
4. 使用Use_Casual_mask环境变量避免DToH的sync

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 12 lines in your changes are missing coverage. Please review.

Project coverage is 54.25%. Comparing base (7aaa788) to head (9214ba6).
Report is 2 commits behind head on develop.

Files	Patch %	Lines
paddlenlp/trainer/training_args.py	0.00%	12 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #8494      +/-   ##
===========================================
- Coverage    54.25%   54.25%   -0.01%     
===========================================
  Files          617      617              
  Lines        96368    96380      +12     
===========================================
- Hits         52288    52287       -1     
- Misses       44080    44093      +13     

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