[XPU] qwen2 supports flash_attn on XPU

[will-jl944]

PR types

New features

PR changes

Models

Description

qwen2 supports flash_attn on XPU

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 53.09%. Comparing base (2c1387f) to head (1446e8a).
Report is 217 commits behind head on develop.

Files with missing lines	Patch %	Lines
paddlenlp/transformers/qwen2/modeling.py	50.00%	3 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #9549      +/-   ##
===========================================
+ Coverage    52.85%   53.09%   +0.24%     
===========================================
  Files          700      701       +1     
  Lines       110418   111257     +839     
===========================================
+ Hits         58359    59070     +711     
- Misses       52059    52187     +128     

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

[wawltor]

LGTM