From f9b954ddb1896292bfcfb8dcd6786894e29d7f68 Mon Sep 17 00:00:00 2001 From: Baizhou Zhang Date: Wed, 5 Aug 2026 18:52:57 -0700 Subject: [PATCH] [CI] Temporarily disable prefill cuda graph for qwen3.5 nightly test (#33772) --- test/registered/8-gpu-models/test_qwen35.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/test/registered/8-gpu-models/test_qwen35.py b/test/registered/8-gpu-models/test_qwen35.py index bf7fb2d01..f8e48e757 100644 --- a/test/registered/8-gpu-models/test_qwen35.py +++ b/test/registered/8-gpu-models/test_qwen35.py @@ -30,7 +30,7 @@ class TestQwen35(unittest.TestCase): "--tool-call-parser=qwen3_coder", "--mem-fraction-static=0.8", ] - dp_args = ["--dp=8", "--enable-dp-attention"] + dp_args = ["--dp=8", "--enable-dp-attention", "--disable-prefill-cuda-graph"] mtp_args = [ "--speculative-algorithm=EAGLE", "--speculative-num-steps=3",