From 0b6591ae0a3e9f3094ee3a127bd0698baaca63c1 Mon Sep 17 00:00:00 2001 From: tzhouam Date: Sat, 3 Oct 2026 22:06:27 +0800 Subject: [PATCH] test: fit Qwen2.5 online expansion on its declared two GPUs Signed-off-by: tzhouam --- tests/e2e/online_serving/test_qwen2_5_omni_expansion.py | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/tests/e2e/online_serving/test_qwen2_5_omni_expansion.py b/tests/e2e/online_serving/test_qwen2_5_omni_expansion.py index 389b0b6ba..de31172b6 100644 --- a/tests/e2e/online_serving/test_qwen2_5_omni_expansion.py +++ b/tests/e2e/online_serving/test_qwen2_5_omni_expansion.py @@ -21,7 +21,14 @@ models = ["Qwen/Qwen2.5-Omni-7B"] # Single CI deploy YAML; rocm/xpu deltas are picked automatically via the # platforms: section in vllm_omni/deploy/ci/qwen2_5_omni.yaml. -stage_configs = [modify_stage_config(get_deploy_config_path("ci/qwen2_5_omni.yaml"))] +stage_configs = [ + modify_stage_config( + get_deploy_config_path("ci/qwen2_5_omni.yaml"), + # This suite requests two cards. The general CI overlay puts Code2Wav + # on a third card; share the Talker's card within the existing budgets. + updates={"stages": {2: {"devices": "1"}}}, + ) +] # Create parameter combinations for model and stage config test_params = [