biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Core] Fix engine-use-ray broken (#4105)

Browse Source
This commit is contained in:
SangBin Cho
2024-04-16 14:24:53 +09:00
committed by GitHub
parent 37e84a403d
commit 4e7ee664e2
2 changed files with 16 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
tests/async_engine/test_api_server.py
View File

@@ -25,21 +25,30 @@ def _query_server_long(prompt: str) -> dict:
@pytest.fixture
def api_server(tokenizer_pool_size: int):
def api_server(tokenizer_pool_size: int, engine_use_ray: bool,
worker_use_ray: bool):
script_path = Path(__file__).parent.joinpath(
"api_server_async_engine.py").absolute()
uvicorn_process = subprocess.Popen([
commands = [
sys.executable, "-u",
str(script_path), "--model", "facebook/opt-125m", "--host",
"127.0.0.1", "--tokenizer-pool-size",
str(tokenizer_pool_size)
])
]
if engine_use_ray:
commands.append("--engine-use-ray")
if worker_use_ray:
commands.append("--worker-use-ray")
uvicorn_process = subprocess.Popen(commands)
yield
uvicorn_process.terminate()
@pytest.mark.parametrize("tokenizer_pool_size", [0, 2])
def test_api_server(api_server, tokenizer_pool_size: int):
@pytest.mark.parametrize("worker_use_ray", [False, True])
@pytest.mark.parametrize("engine_use_ray", [False, True])
def test_api_server(api_server, tokenizer_pool_size: int, worker_use_ray: bool,
engine_use_ray: bool):
"""
Run the API server and test it.