xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-18 04:45:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

[gpt-oss] Enable unit test for response API harmony integration (#23533)

Browse Source 
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
This commit is contained in:
Chen Zhang 2025-08-26 18:23:26 -07:00 committed by GitHub
parent 2c2b140ae8
commit eb1995167e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 28 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
tests/entrypoints/openai/test_response_api_with_harmony.py
View File

@ -11,18 +11,25 @@ from openai import BadRequestError, NotFoundError, OpenAI
from ...utils import RemoteOpenAIServer from ...utils import RemoteOpenAIServer
pytest.skip(allow_module_level=True, reason="gpt-oss can't run on CI yet.")
MODEL_NAME = "openai/gpt-oss-20b" MODEL_NAME = "openai/gpt-oss-20b"
DTYPE = "bfloat16"
@pytest.fixture(scope="module") @pytest.fixture(scope="module")
def server(): def monkeypatch_module():
from _pytest.monkeypatch import MonkeyPatch
mpatch = MonkeyPatch()
yield mpatch
mpatch.undo()
@pytest.fixture(scope="module")
def server(monkeypatch_module: pytest.MonkeyPatch):
args = ["--enforce-eager", "--tool-server", "demo"] args = ["--enforce-eager", "--tool-server", "demo"]
with RemoteOpenAIServer(MODEL_NAME, args) as remote_server: with monkeypatch_module.context() as m:
yield remote_server m.setenv("VLLM_ENABLE_RESPONSES_API_STORE", "1")
with RemoteOpenAIServer(MODEL_NAME, args) as remote_server:
yield remote_server
@pytest_asyncio.fixture @pytest_asyncio.fixture
@ -269,10 +276,11 @@ async def test_stateful_multi_turn(client: OpenAI, model_name: str):
@pytest.mark.asyncio @pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME]) @pytest.mark.parametrize("model_name", [MODEL_NAME])
async def test_streaming(client: OpenAI, model_name: str): async def test_streaming(client: OpenAI, model_name: str):
# TODO: Add back when web search and code interpreter are available in CI
prompts = [ prompts = [
"tell me a story about a cat in 20 words", "tell me a story about a cat in 20 words",
"What is 13 * 24? Use python to calculate the result.", # "What is 13 * 24? Use python to calculate the result.",
"When did Jensen found NVIDIA? Search it and answer the year only.", # "When did Jensen found NVIDIA? Search it and answer the year only.",
] ]
for prompt in prompts: for prompt in prompts:
@ -281,15 +289,15 @@ async def test_streaming(client: OpenAI, model_name: str):
input=prompt, input=prompt,
reasoning={"effort": "low"}, reasoning={"effort": "low"},
tools=[ tools=[
{ # {
"type": "web_search_preview" # "type": "web_search_preview"
}, # },
{ # {
"type": "code_interpreter", # "type": "code_interpreter",
"container": { # "container": {
"type": "auto" # "type": "auto"
} # }
}, # },
], ],
stream=True, stream=True,
) )
@ -317,6 +325,7 @@ async def test_streaming(client: OpenAI, model_name: str):
@pytest.mark.asyncio @pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME]) @pytest.mark.parametrize("model_name", [MODEL_NAME])
@pytest.mark.skip(reason="Web search tool is not available in CI yet.")
async def test_web_search(client: OpenAI, model_name: str): async def test_web_search(client: OpenAI, model_name: str):
response = await client.responses.create( response = await client.responses.create(
model=model_name, model=model_name,
@ -331,6 +340,7 @@ async def test_web_search(client: OpenAI, model_name: str):
@pytest.mark.asyncio @pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME]) @pytest.mark.parametrize("model_name", [MODEL_NAME])
@pytest.mark.skip(reason="Code interpreter tool is not available in CI yet.")
async def test_code_interpreter(client: OpenAI, model_name: str): async def test_code_interpreter(client: OpenAI, model_name: str):
response = await client.responses.create( response = await client.responses.create(
model=model_name, model=model_name,
@ -436,6 +446,7 @@ async def test_function_calling(client: OpenAI, model_name: str):
@pytest.mark.asyncio @pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME]) @pytest.mark.parametrize("model_name", [MODEL_NAME])
@pytest.mark.flaky(reruns=5)
async def test_function_calling_multi_turn(client: OpenAI, model_name: str): async def test_function_calling_multi_turn(client: OpenAI, model_name: str):
tools = [ tools = [
{ {