[AutoParallel] add parameter enable_stage1_tensor_fusion_blanced_save_load and enable_stage1_tensor_fusion

[AndSonder]

PR types

Others

PR changes

Others

Description

给静半添加 enable_stage1_tensor_fusion_blanced_save_load 和 enable_stage1_tensor_fusion 参数

相关 PR：

• [AutoParallel] optimize sharding stage1 tensor fusion save&load strategy Paddle#70309

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 3 lines in your changes missing coverage. Please review.

Project coverage is 52.77%. Comparing base (7197b79) to head (e203dff).
Report is 5 commits behind head on develop.

Files with missing lines	Patch %	Lines
paddlenlp/trainer/training_args.py	0.00%	3 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #9714      +/-   ##
===========================================
+ Coverage    52.12%   52.77%   +0.65%     
===========================================
  Files          724      718       -6     
  Lines       117142   112377    -4765     
===========================================
- Hits         61056    59306    -1750     
+ Misses       56086    53071    -3015     

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