[LLM] support sparse attention for LLAMA

[GuoxiaWang]

PR types

New features

PR changes

Models

Description

[LLM] support sparse attention for LLAMA

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 14.89362% with 40 lines in your changes missing coverage. Please review.

Project coverage is 54.17%. Comparing base (a8d446c) to head (61f2314).
Report is 239 commits behind head on develop.

Files with missing lines	Patch %	Lines
paddlenlp/transformers/llama/modeling_pp.py	10.25%	35 Missing ⚠️
paddlenlp/transformers/llama/fusion_ops.py	0.00%	4 Missing ⚠️
paddlenlp/transformers/llama/modeling.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #8592      +/-   ##
===========================================
- Coverage    54.18%   54.17%   -0.02%     
===========================================
  Files          625      625              
  Lines        98942    98972      +30     
===========================================
+ Hits         53612    53615       +3     
- Misses       45330    45357      +27     

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

[lugimzzz]

lgtm