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 12, 11:00 UTC
[CI/Build] Skip ray tests on ROCm (#29556)
Signed-off-by: Ryan Rock <ryan.rock@amd.com>
Ryan Rock committed
Nov 27, 2025 at 09:01 UTC
bab438ff3e7bd93f861e66a60c6cbefe42af0d1a
1 file changed
+5
tests/v1/distributed/test_async_llm_dp.py
+5
@@ -12,6 +12,7 @@ from vllm import SamplingParams
12
from vllm.config import VllmConfig
13
from vllm.engine.arg_utils import AsyncEngineArgs
14
from vllm.inputs import PromptType
15
+from vllm.platforms import current_platform
16
from vllm.sampling_params import RequestOutputKind
17
from vllm.v1.engine.async_llm import AsyncLLM
18
from vllm.v1.engine.core_client import DPAsyncMPClient
@@ -84,6 +85,10 @@ async def test_load(
85
if async_scheduling and data_parallel_backend == "ray":
86
# TODO(NickLucche) Re-enable when async scheduling is supported
87
pytest.skip("Async scheduling is not supported with ray")
88
+ elif data_parallel_backend == "ray" and current_platform.is_rocm():
89
+ pytest.skip(
90
+ "Ray as the distributed executor backend is not supported with ROCm."
91
+ )
92
stats_loggers = {}
93
94
@dataclass