[LLM] fix bug when loss is None in llama modeling.py

[cqulilujia]

PR types

Bug Fixs

PR changes

Models

Description

修复llama模型在训练过程中由异常数据导致的loss为None时引发的异常， #8342 中考虑到了这一问题，但loss设置为了float而非tensor形式。
另可参考： #8348 #8299

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 54.29%. Comparing base (87e4c4f) to head (e1d30a0).

Files	Patch %	Lines
paddlenlp/transformers/llama/modeling.py	0.00%	4 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #8459      +/-   ##
===========================================
- Coverage    54.29%   54.29%   -0.01%     
===========================================
  Files          617      617              
  Lines        96339    96341       +2     
===========================================
- Hits         52312    52308       -4     
- Misses       44027    44033       +6     

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

[cqulilujia]

@ZHUI 麻烦帮忙review

[ZHUI]

LGTM