From 5e973209aaf5fa15459555eaa42bcd20ea63aa0d Mon Sep 17 00:00:00 2001 From: Zijing Liu Date: Thu, 13 Nov 2025 03:30:04 -0800 Subject: [PATCH] [BugFix] Fix type error when assign a trition kernel tensor to a torch.nn.Parameter (#28603) Signed-off-by: Zijing Liu --- vllm/model_executor/layers/quantization/mxfp4.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/vllm/model_executor/layers/quantization/mxfp4.py b/vllm/model_executor/layers/quantization/mxfp4.py index 0f69a18a1f3fd..5552c1ae5edf8 100644 --- a/vllm/model_executor/layers/quantization/mxfp4.py +++ b/vllm/model_executor/layers/quantization/mxfp4.py @@ -755,8 +755,8 @@ class Mxfp4MoEMethod(FusedMoEMethodBase): self.w13_weight = w13_weight self.w2_weight = w2_weight - layer.w13_weight = w13_weight - layer.w2_weight = w2_weight + layer.w13_weight = Parameter(w13_weight.data, requires_grad=False) + layer.w2_weight = Parameter(w2_weight.data, requires_grad=False) else: raise ValueError(f"Unsupported backend: {self.mxfp4_backend}")