xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-13 11:05:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Test] Add non-MoE DP test coverage (#28235)

Browse Source 
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Matthew Bonanni 2025-11-06 15:59:57 -05:00 committed by GitHub
parent da855b42d2
commit ca90f50304
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 19 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
tests/v1/distributed/test_async_llm_dp.py
View File

@ -20,13 +20,6 @@ from vllm.v1.metrics.stats import IterationStats, MultiModalCacheStats, Schedule
DP_SIZE = int(os.getenv("DP_SIZE", 2)) DP_SIZE = int(os.getenv("DP_SIZE", 2))
engine_args = AsyncEngineArgs(
model="ibm-research/PowerMoE-3b",
enforce_eager=True,
tensor_parallel_size=int(os.getenv("TP_SIZE", 1)),
data_parallel_size=DP_SIZE,
)
async def generate( async def generate(
engine: AsyncLLM, engine: AsyncLLM,
@ -65,6 +58,13 @@ async def generate(
return count, request_id return count, request_id
@pytest.mark.parametrize(
"model",
[
"ibm-research/PowerMoE-3b",
"hmellor/tiny-random-LlamaForCausalLM",
],
)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"output_kind", "output_kind",
[ [
@ -76,7 +76,10 @@ async def generate(
@pytest.mark.parametrize("async_scheduling", [True, False]) @pytest.mark.parametrize("async_scheduling", [True, False])
@pytest.mark.asyncio @pytest.mark.asyncio
async def test_load( async def test_load(
output_kind: RequestOutputKind, data_parallel_backend: str, async_scheduling: bool model: str,
output_kind: RequestOutputKind,
data_parallel_backend: str,
async_scheduling: bool,
): ):
if async_scheduling and data_parallel_backend == "ray": if async_scheduling and data_parallel_backend == "ray":
# TODO(NickLucche) Re-enable when async scheduling is supported # TODO(NickLucche) Re-enable when async scheduling is supported
@ -107,8 +110,14 @@ async def test_load(
with ExitStack() as after: with ExitStack() as after:
prompt = "This is a test of data parallel" prompt = "This is a test of data parallel"
engine_args.data_parallel_backend = data_parallel_backend engine_args = AsyncEngineArgs(
engine_args.async_scheduling = async_scheduling model=model,
enforce_eager=True,
tensor_parallel_size=int(os.getenv("TP_SIZE", 1)),
data_parallel_size=DP_SIZE,
data_parallel_backend=data_parallel_backend,
async_scheduling=async_scheduling,
)
engine = AsyncLLM.from_engine_args( engine = AsyncLLM.from_engine_args(
engine_args, stat_loggers=[SimpleStatsLogger] engine_args, stat_loggers=[SimpleStatsLogger]
) )