A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 4, 10:00 UTC
[Bugfix] Disable FlashInfer CUTLASS MoE on SM110 (Jetson Thor AGX) (#40808)
Signed-off-by: Stefano Castagnetta <scastagnetta@nvidia.com>
Stefano Castagnetta committed
May 1, 2026 at 05:08 UTC
b542bdf7fb3611879b30917e03bef62963496e83
1 file changed
+1
-1
vllm/model_executor/layers/fused_moe/flashinfer_cutlass_moe.py
+1
-1
@@ -129,7 +129,7 @@ class FlashInferExperts(mk.FusedMoEExpertsModular):
129
and (
130
p.is_device_capability(90)
131
or p.is_device_capability_family(100)
132
- or p.is_device_capability_family(110)
132
+ # SM110 excluded: flashinfer-ai/flashinfer#3134
133
or p.is_device_capability_family(120)
134
)
135
and has_flashinfer_cutlass_fused_moe()