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

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

105
tests/entrypoints/pooling/openai/test_rerank.py
View File

@@ -25,15 +25,18 @@ def server():
def test_rerank_texts(server: RemoteOpenAIServer, model_name: str):
query = "What is the capital of France?"
documents = [
"The capital of Brazil is Brasilia.", "The capital of France is Paris."
"The capital of Brazil is Brasilia.",
"The capital of France is Paris.",
]
rerank_response = requests.post(server.url_for("rerank"),
json={
"model": model_name,
"query": query,
"documents": documents,
})
rerank_response = requests.post(
server.url_for("rerank"),
json={
"model": model_name,
"query": query,
"documents": documents,
},
)
rerank_response.raise_for_status()
rerank = RerankResponse.model_validate(rerank_response.json())
@@ -49,16 +52,14 @@ def test_top_n(server: RemoteOpenAIServer, model_name: str):
query = "What is the capital of France?"
documents = [
"The capital of Brazil is Brasilia.",
"The capital of France is Paris.", "Cross-encoder models are neat"
"The capital of France is Paris.",
"Cross-encoder models are neat",
]
rerank_response = requests.post(server.url_for("rerank"),
json={
"model": model_name,
"query": query,
"documents": documents,
"top_n": 2
})
rerank_response = requests.post(
server.url_for("rerank"),
json={"model": model_name, "query": query, "documents": documents, "top_n": 2},
)
rerank_response.raise_for_status()
rerank = RerankResponse.model_validate(rerank_response.json())
@@ -71,28 +72,26 @@ def test_top_n(server: RemoteOpenAIServer, model_name: str):
@pytest.mark.parametrize("model_name", [MODEL_NAME])
def test_rerank_max_model_len(server: RemoteOpenAIServer, model_name: str):
query = "What is the capital of France?" * 100
documents = [
"The capital of Brazil is Brasilia.", "The capital of France is Paris."
"The capital of Brazil is Brasilia.",
"The capital of France is Paris.",
]
rerank_response = requests.post(server.url_for("rerank"),
json={
"model": model_name,
"query": query,
"documents": documents
})
rerank_response = requests.post(
server.url_for("rerank"),
json={"model": model_name, "query": query, "documents": documents},
)
assert rerank_response.status_code == 400
# Assert just a small fragments of the response
assert "Please reduce the length of the input." in \
rerank_response.text
assert "Please reduce the length of the input." in rerank_response.text
def test_invocations(server: RemoteOpenAIServer):
query = "What is the capital of France?"
documents = [
"The capital of Brazil is Brasilia.", "The capital of France is Paris."
"The capital of Brazil is Brasilia.",
"The capital of France is Paris.",
]
request_args = {
@@ -101,23 +100,25 @@ def test_invocations(server: RemoteOpenAIServer):
"documents": documents,
}
rerank_response = requests.post(server.url_for("rerank"),
json=request_args)
rerank_response = requests.post(server.url_for("rerank"), json=request_args)
rerank_response.raise_for_status()
invocation_response = requests.post(server.url_for("invocations"),
json=request_args)
invocation_response = requests.post(
server.url_for("invocations"), json=request_args
)
invocation_response.raise_for_status()
rerank_output = rerank_response.json()
invocation_output = invocation_response.json()
assert rerank_output.keys() == invocation_output.keys()
for rerank_result, invocations_result in zip(rerank_output["results"],
invocation_output["results"]):
for rerank_result, invocations_result in zip(
rerank_output["results"], invocation_output["results"]
):
assert rerank_result.keys() == invocations_result.keys()
assert rerank_result["relevance_score"] == pytest.approx(
invocations_result["relevance_score"], rel=0.05)
invocations_result["relevance_score"], rel=0.05
)
# TODO: reset this tolerance to 0.01 once we find
# an alternative to flash_attn with bfloat16
@@ -125,34 +126,36 @@ def test_invocations(server: RemoteOpenAIServer):
@pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME])
async def test_activation(server: RemoteOpenAIServer, model_name: str):
async def get_outputs(activation):
query = "What is the capital of France?"
documents = [
"The capital of Brazil is Brasilia.",
"The capital of France is Paris."
"The capital of France is Paris.",
]
response = requests.post(server.url_for("rerank"),
json={
"model": model_name,
"query": query,
"documents": documents,
"activation": activation
})
response = requests.post(
server.url_for("rerank"),
json={
"model": model_name,
"query": query,
"documents": documents,
"activation": activation,
},
)
outputs = response.json()
return torch.tensor([x['relevance_score'] for x in outputs["results"]])
return torch.tensor([x["relevance_score"] for x in outputs["results"]])
default = await get_outputs(activation=None)
w_activation = await get_outputs(activation=True)
wo_activation = await get_outputs(activation=False)
assert torch.allclose(default, w_activation,
atol=1e-2), "Default should use activation."
assert not torch.allclose(
w_activation, wo_activation,
atol=1e-2), "wo_activation should not use activation."
assert torch.allclose(
F.sigmoid(wo_activation), w_activation, atol=1e-2
), "w_activation should be close to activation(wo_activation)."
assert torch.allclose(default, w_activation, atol=1e-2), (
"Default should use activation."
)
assert not torch.allclose(w_activation, wo_activation, atol=1e-2), (
"wo_activation should not use activation."
)
assert torch.allclose(F.sigmoid(wo_activation), w_activation, atol=1e-2), (
"w_activation should be close to activation(wo_activation)."
)