From c51df43005726a09c6eb7348e8c1b00501c70a8e Mon Sep 17 00:00:00 2001 From: Wei Zhao <51183510+wzhao18@users.noreply.github.com> Date: Sun, 3 May 2026 12:19:59 -0400 Subject: [PATCH] Disable flashinfer autotune temporarily due to correctness issues (#41524) Signed-off-by: wzhao18 --- vllm/config/vllm.py | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/vllm/config/vllm.py b/vllm/config/vllm.py index 0146ee4c14..88e6660e21 100644 --- a/vllm/config/vllm.py +++ b/vllm/config/vllm.py @@ -209,7 +209,9 @@ OPTIMIZATION_LEVEL_01 = { "use_inductor_graph_partition": False, }, "kernel_config": { - "enable_flashinfer_autotune": True, + # Disabled for now due to correctness issues: + # https://github.com/flashinfer-ai/flashinfer/issues/3197 + "enable_flashinfer_autotune": False, }, } OPTIMIZATION_LEVEL_02 = { @@ -229,7 +231,9 @@ OPTIMIZATION_LEVEL_02 = { "use_inductor_graph_partition": False, }, "kernel_config": { - "enable_flashinfer_autotune": True, + # Disabled for now due to correctness issues: + # https://github.com/flashinfer-ai/flashinfer/issues/3197 + "enable_flashinfer_autotune": False, }, } OPTIMIZATION_LEVEL_03 = {