A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 14, 05:30 UTC
[Flaky CI] Increase timeout tolerance for test_mp_crash_detection+test_default_mm_lora_chat_completions (#23028)
Signed-off-by: mgoin <mgoin64@gmail.com>
Michael Goin committed
Aug 16, 2025 at 14:33 UTC
3253ae765ef4dc0604a6f3ed3a1dcd61fdda6bda
2 files changed
+4
-3
tests/entrypoints/openai/test_default_mm_loras.py
+2
-1
@@ -48,7 +48,8 @@ def multimodal_server(): # noqa: F811
48
f"{{\"audio\": \"{AUDIO_LORA_PATH}\"}}",
49
]
50
51
- with RemoteOpenAIServer(MULTIMODAL_MODEL_NAME, args) as remote_server:
51
+ with RemoteOpenAIServer(MULTIMODAL_MODEL_NAME, args,
52
+ max_wait_seconds=480) as remote_server:
53
yield remote_server
54
55
tests/mq_llm_engine/test_error_handling.py
+2
-2
@@ -255,8 +255,8 @@ async def test_mp_crash_detection(monkeypatch: pytest.MonkeyPatch):
255
pass
256
end = time.perf_counter()
257
258
- assert end - start < 60, (
259
- "Expected vLLM to gracefully shutdown in <60s "
258
+ assert end - start < 100, (
259
+ "Expected vLLM to gracefully shutdown in <100s "
260
"if there is an error in the startup.")
261
262