[LLM]add ktotrainer

[lugimzzz]

PR types

New features

PR changes

Others

Description

新增支持KTO Trainer，开源模型适配在下一个PR

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 10.69767% with 384 lines in your changes missing coverage. Please review.

Project coverage is 52.97%. Comparing base (140ea48) to head (280fae9).
Report is 3 commits behind head on develop.

Files with missing lines	Patch %	Lines
paddlenlp/trl/kto_trainer.py	9.96%	262 Missing ⚠️
paddlenlp/trl/kto_criterion.py	12.71%	103 Missing ⚠️
paddlenlp/transformers/tensor_parallel_utils.py	0.00%	15 Missing ⚠️
paddlenlp/trl/dpo_criterion.py	0.00%	4 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #9393      +/-   ##
===========================================
- Coverage    53.00%   52.97%   -0.04%     
===========================================
  Files          674      676       +2     
  Lines       107427   107838     +411     
===========================================
+ Hits         56945    57123     +178     
- Misses       50482    50715     +233     

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

[wawltor]

LGTM