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 8, 03:45 UTC
[Bugfix] fix activation in cpu_fused_moe_torch call (#34696)
Signed-off-by: Marek Michalowski <marek.michalowski@arm.com>
Marek Michalowski committed
Feb 18, 2026 at 07:39 UTC
e89a91d9275cd8ac086fe04476b41675a9ebbd5c
1 file changed
+1
-1
vllm/model_executor/layers/fused_moe/cpu_fused_moe.py
+1
-1
@@ -402,7 +402,7 @@ class CPUFusedMOE:
402
input,
403
topk_weights,
404
topk_ids,
405
- activation,
405
+ activation.value,
406
global_num_experts,
407
skip_weighted,
408
)