[Refactor] SFTTrainer SFTConfig

[ZHUI]

PR types

Others

PR changes

Others

Description

Others

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 79.87421% with 32 lines in your changes missing coverage. Please review.

Project coverage is 53.07%. Comparing base (3ef14dc) to head (6356a56).
Report is 2 commits behind head on develop.

Files with missing lines	Patch %	Lines
paddlenlp/trl/sft_config.py	45.00%	22 Missing ⚠️
paddlenlp/trl/utils.py	0.00%	9 Missing ⚠️
paddlenlp/datasets/dataset.py	66.66%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #9318      +/-   ##
===========================================
- Coverage    53.09%   53.07%   -0.02%     
===========================================
  Files          699      699              
  Lines       110471   110348     -123     
===========================================
- Hits         58651    58569      -82     
+ Misses       51820    51779      -41     

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

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ gongel
❌ ZHUI
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[wawltor]

预训练的config没有考虑到

[ZHUI]

你说预训练的这部分也重构进入到主库吗？

[wawltor]

这里放到paddlenlp library我想到的一个问题是，如果用户想新增相关的参数就必须要改动library，因此最近developllm工具必须适配最新的library，看起来会导致一些兼容问题。

[ZHUI]

可以引导 用户使用

1. 正式版lib + 正式版 llm
2. dev lib + dev llm

不要混用。

实际是上，大部分更新的参数，用户并不会使用到。