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 3, 04:15 UTC
Fix FlashInfer TRTLLM NvFP4 monolithic MoE routing (#43223)
Signed-off-by: zhangxin81 <115389973+zhangxin81@users.noreply.github.com>
zhangxin81 committed
May 21, 2026 at 16:17 UTC
edafea35550fab0b185b885711ec048dfd2e1a4d
1 file changed
+1
-1
vllm/model_executor/layers/fused_moe/config.py
+1
-1
@@ -159,7 +159,7 @@ def get_routing_method_type(
159
160
if scoring_func == "softmax":
161
if renormalize:
162
- return RoutingMethodType.Renormalize
162
+ return RoutingMethodType.RenormalizeNaive
163
else:
164
return RoutingMethodType.Default
165