From 70f8b967242633f4cf38a456127f102664d1da3e Mon Sep 17 00:00:00 2001 From: Mengqing Cao Date: Thu, 15 May 2025 14:16:31 +0800 Subject: [PATCH] [Bugfix] Fix FusedMoEPrepareAndFinalize for cuda-disalike backends (#18178) Signed-off-by: Mengqing Cao --- vllm/model_executor/layers/fused_moe/layer.py | 1 + 1 file changed, 1 insertion(+) diff --git a/vllm/model_executor/layers/fused_moe/layer.py b/vllm/model_executor/layers/fused_moe/layer.py index d083e0040c0e2..0b3c02d1ba28f 100644 --- a/vllm/model_executor/layers/fused_moe/layer.py +++ b/vllm/model_executor/layers/fused_moe/layer.py @@ -43,6 +43,7 @@ if current_platform.is_cuda_alike(): from .pplx_prepare_finalize import PplxPrepareAndFinalize else: fused_experts = None # type: ignore + FusedMoEPrepareAndFinalize = None # type: ignore if is_rocm_aiter_moe_enabled(): from vllm.model_executor.layers.fused_moe.rocm_aiter_fused_moe import ( # noqa: E501 rocm_aiter_biased_group_topk as grouped_topk)