[CI] Stabilize memory-sensitive compile and structured output tests (#49749)

Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Co-authored-by: OpenAI Codex <noreply@openai.com>
This commit is contained in:
Jiangyun Zhu
2026-07-24 16:45:09 -04:00
committed by GitHub
co-authored by OpenAI Codex
parent 972848f276
commit 9e6746b3c7
2 changed files with 4 additions and 0 deletions
@@ -64,6 +64,8 @@ class TestSetting:
"bfloat16",
"--max-model-len",
"2048",
"--gpu-memory-utilization",
"0.98",
],
pp_size=1,
tp_size=1,
@@ -212,6 +212,7 @@ class CarDescription(BaseModel):
PARAMS_MODELS_BACKENDS_TOKENIZER_MODE,
)
def test_structured_output(
request: pytest.FixtureRequest,
backend: str,
tokenizer_mode: str,
model_name: str,
@@ -242,6 +243,7 @@ def test_structured_output(
speculative_config=speculative_config,
**platform_args,
)
request.addfinalizer(llm.llm_engine.engine_core.shutdown)
#
# Test 1: Generate JSON output based on a provided schema