From 81e217fe6b5a3030aa5c4d859a2125b81979bee4 Mon Sep 17 00:00:00 2001
From: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Date: Mon, 9 Feb 2026 21:29:39 -0800
Subject: [PATCH] [Bugfix] Fix DP Attention Padding in Dummy Run (#34187)

Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Benjamin Chislett <bchislett@nvidia.com>
---
 vllm/v1/worker/gpu_model_runner.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/vllm/v1/worker/gpu_model_runner.py b/vllm/v1/worker/gpu_model_runner.py
index a7c2a8800..0e2e381f2 100644
--- a/vllm/v1/worker/gpu_model_runner.py
+++ b/vllm/v1/worker/gpu_model_runner.py
@@ -4787,6 +4787,7 @@ class GPUModelRunner(
             pad_attn = cudagraph_runtime_mode == CUDAGraphMode.FULL
             attn_metadata, _ = self._build_attention_metadata(
                 num_tokens=num_tokens_unpadded,
+                num_tokens_padded=num_tokens_padded if pad_attn else None,
                 num_reqs=num_reqs_padded,
                 max_query_len=max_query_len,
                 ubatch_slices=ubatch_slices_padded if pad_attn else ubatch_slices,