From cdaed2e160b7e55c9aafa03c16686b826e7f891b Mon Sep 17 00:00:00 2001
From: Chenfei Zhang <chenfeiz@nvidia.com>
Date: Fri, 12 Jun 2026 13:55:09 -0700
Subject: [PATCH] fix(perf): PR #14851 already removed the bad
 `is_sliding_window`/`mMaxSeqLenKv` logic on `o

Signed-off-by: Chenfei Zhang <chenfeiz@nvidia.com>
---
 cpp/tensorrt_llm/kernels/trtllmGenKernels/fmha/fmhaKernels.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/cpp/tensorrt_llm/kernels/trtllmGenKernels/fmha/fmhaKernels.h b/cpp/tensorrt_llm/kernels/trtllmGenKernels/fmha/fmhaKernels.h
index 423f3de064cf..3429c8819654 100644
--- a/cpp/tensorrt_llm/kernels/trtllmGenKernels/fmha/fmhaKernels.h
+++ b/cpp/tensorrt_llm/kernels/trtllmGenKernels/fmha/fmhaKernels.h
@@ -306,7 +306,7 @@ class TllmGenFmhaKernel
     inline static std::vector<int> const kDefaultWarmupPrefillBatchSizeCandidates
         = {1, 2, 3, 4, 6, 8, 12, 16, 24, 32, 48, 64, 128, 256};
     inline static std::vector<int> const kDefaultWarmupSeqLenQkvCandidates
-        = {1, 128, 512, 1024, 2048, 4096, 8192, 16384, 32768};
+        = {1, 128, 256, 512, 1024, 2048, 3072, 4096, 8192, 16384, 32768};
 
     static std::vector<int> makeWarmupCandidateSizes(std::vector<int> const& defaultCandidateSizes, int maxSize)
     {