chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Ascend attention backend(PA&MLA) (#7722)

Browse Source 
Co-authored-by: Maksim <makcum888e@mail.ru>
Co-authored-by: VDV1985 <vladdv85@mail.ru>
This commit is contained in:
ronnie_zheng
2025-07-03 09:23:19 -07:00
committed by GitHub
co-authored by Maksim VDV1985
parent b58226510f
commit 1e0e549766
17 changed files with 842 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/models/deepseek_v2.py
+3 -1
View File
@@ -956,7 +956,9 @@ class DeepseekV2AttentionMLA(nn.Module):
else:
return AttnForwardMethod.MLA
if self.attention_backend == "flashinfer":
if self.attention_backend == "ascend":
return AttnForwardMethod.MLA
elif self.attention_backend == "flashinfer":
# Flashinfer MLA: Do not absorb when enabling ragged prefill
if (
not self.flashinfer_mla_disable_ragged