From 572b25b03ea5893305d4b25a25150ef129717605 Mon Sep 17 00:00:00 2001 From: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com> Date: Wed, 8 Jul 2026 09:05:03 -0400 Subject: [PATCH] [Bug] Fix Batched DeepGEMM (#47884) Signed-off-by: Robert Shaw Co-authored-by: Robert Shaw Co-authored-by: Claude --- .../layers/fused_moe/experts/batched_deep_gemm_moe.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/layers/fused_moe/experts/batched_deep_gemm_moe.py b/vllm/model_executor/layers/fused_moe/experts/batched_deep_gemm_moe.py index 275f80c68f..90972fc7f0 100644 --- a/vllm/model_executor/layers/fused_moe/experts/batched_deep_gemm_moe.py +++ b/vllm/model_executor/layers/fused_moe/experts/batched_deep_gemm_moe.py @@ -210,7 +210,7 @@ def persistent_masked_m_silu_mul_quant( DeepGemmQuantScaleFMT.UE8M0, ] - device_capability = current_platform.get_device_capability(device_id=y.device.index) + device_capability = current_platform.get_device_capability() assert device_capability is not None cuda_arch = device_capability.to_int()