Increase timeout for Diffusion Model Test and Diffusion Images API LoRA E2E in Buildkite configurations to 30 minutes. Also, introduce a new timeout constant for diffusion initialization in relevant test files to enhance stability during testing. Update OmniServer initialization to utilize the new timeout settings.

This commit is contained in:
Taichang Zhou committed 2026-03-17 23:43:46 +08:00
1 parent 004c60a031
commit 8a31e7865d
6 files changed
+24 -7

No files matched your search

@@ -24,6 +24,7 @@ os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "1"
# This test is specific to Z-Image LoRA behavior. Keep it focused on a single
# model to reduce runtime and avoid extra downloads.
models = ["Tongyi-MAI/Z-Image-Turbo"]
DIFFUSION_INIT_TIMEOUT_S = 600
@pytest.mark.parametrize("model_name", models)
@@ -76,7 +77,11 @@ def test_diffusion_model(model_name: str, tmp_path: Path):
)
return str(adapter_dir)
m = Omni(model=model_name)
m = Omni(
model=model_name,
stage_init_timeout=DIFFUSION_INIT_TIMEOUT_S,
init_timeout=DIFFUSION_INIT_TIMEOUT_S,
)
try:
# high resolution may cause OOM on L4
height = 256
@@ -28,6 +28,7 @@ from tests.utils import hardware_test
os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
MODEL = "Tongyi-MAI/Z-Image-Turbo"
DIFFUSION_INIT_TIMEOUT_S = 600
PROMPT = "a photo of a cat sitting on a laptop keyboard"
@@ -37,7 +38,17 @@ SEED = 42
@pytest.fixture(scope="module")
def omni_server():
with OmniServer(MODEL, ["--num-gpus", "1"]) as server:
with OmniServer(
MODEL,
[
"--num-gpus",
"1",
"--stage-init-timeout",
str(DIFFUSION_INIT_TIMEOUT_S),
"--init-timeout",
str(DIFFUSION_INIT_TIMEOUT_S),
],
) as server:
yield server