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 12, 02:15 UTC
[Bugfix] Fix flashinfer ar+norm kernel not available issue (#29960)
Signed-off-by: elvischenv <219235043+elvischenv@users.noreply.github.com>
elvischenv committed
Dec 4, 2025 at 02:50 UTC
afe9eb408ee1191cd57a68d46b6ce2860b1b41e1
1 file changed
+2
-1
vllm/compilation/fix_functionalization.py
+2
-1
@@ -104,7 +104,8 @@ class FixFunctionalizationPass(VllmInductorPass):
104
mutated_args = {1: "result"}
105
self.defunctionalize(graph, node, mutated_args)
106
elif (
107
- at_target
107
+ hasattr(torch.ops.vllm, "flashinfer_trtllm_fused_allreduce_norm")
108
+ and at_target
109
== torch.ops.vllm.flashinfer_trtllm_fused_allreduce_norm.default
110
):
111
mutated_args = {