biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[XPU] Quick fix for TritonMLA to remove cuda hardcode (#39088)

Browse Source 
Signed-off-by: Chendi Xue <chendi.xue@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
Chendi.Xue
2026-04-07 11:17:58 -05:00
committed by GitHub
parent 7310555482
commit 92b9afeecd
2 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm/model_executor/layers/fused_moe/unquantized_fused_moe_method.py
View File

@@ -222,7 +222,7 @@ class UnquantizedFusedMoEMethod(FusedMoEMethodBase, CustomOp):
self.cpu_fused_moe = cpu_fused_moe.CPUFusedMOE(layer)
else:
self.cpu_fused_moe = cpu_fused_moe.CPUFusedMOE(layer)
elif current_platform.is_xpu():
elif self.unquantized_backend == UnquantizedMoeBackend.XPU:
w13 = layer.w13_weight
w2 = layer.w2_weight