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

83
tests/entrypoints/openai/test_completion_with_prompt_embeds.py
View File

@@ -9,6 +9,7 @@ import openai # use the official client for correctness check
import pytest
import pytest_asyncio
import torch
# downloading lora to test lora requests
from openai import BadRequestError
from transformers import AutoConfig
@@ -23,8 +24,9 @@ CONFIG = AutoConfig.from_pretrained(MODEL_NAME)
@pytest.fixture(scope="module", params=["use-lora"])
def default_server_args(request: pytest.FixtureRequest,
opt125_lora_files: str) -> list[str]:
def default_server_args(
request: pytest.FixtureRequest, opt125_lora_files: str
) -> list[str]:
args = [
# use half precision for speed and memory savings in CI environment
"--dtype",
@@ -42,18 +44,20 @@ def default_server_args(request: pytest.FixtureRequest,
lora_module_1 = {
"name": LORA_SERVING_MODEL_NAME,
"path": opt125_lora_files,
"base_model_name": MODEL_NAME
"base_model_name": MODEL_NAME,
}
args.extend([
"--enable-lora",
"--lora-module",
json.dumps(lora_module_1),
"--max-lora-rank",
"64",
"--max-cpu-loras",
"2",
])
args.extend(
[
"--enable-lora",
"--lora-module",
json.dumps(lora_module_1),
"--max-lora-rank",
"64",
"--max-cpu-loras",
"2",
]
)
return args
@@ -67,7 +71,7 @@ EXAMPLE_PROMPTS = [
def _encode_embeds(embeds: torch.Tensor):
buffer = io.BytesIO()
torch.save(embeds, buffer)
return base64.b64encode(buffer.getvalue()).decode('utf-8')
return base64.b64encode(buffer.getvalue()).decode("utf-8")
@pytest.fixture(scope="module")
@@ -79,8 +83,7 @@ def example_prompt_embeds(hf_runner):
return [_encode_embeds(item) for item in example_embeddings]
@pytest.fixture(scope="module",
params=["", "--disable-frontend-multiprocessing"])
@pytest.fixture(scope="module", params=["", "--disable-frontend-multiprocessing"])
def server_with_prompt_embeds(default_server_args, request):
if request.param:
default_server_args.append(request.param)
@@ -110,7 +113,8 @@ async def test_completions_with_prompt_embeds(
prompt="", # Add empty prompt as required parameter
max_tokens=5,
temperature=0.0,
extra_body={"prompt_embeds": encoded_embeds})
extra_body={"prompt_embeds": encoded_embeds},
)
assert len(completion.choices[0].text) >= 1
assert completion.choices[0].prompt_logprobs is None
@@ -120,7 +124,8 @@ async def test_completions_with_prompt_embeds(
prompt="", # Add empty prompt as required parameter
max_tokens=5,
temperature=0.0,
extra_body={"prompt_embeds": [encoded_embeds, encoded_embeds2]})
extra_body={"prompt_embeds": [encoded_embeds, encoded_embeds2]},
)
assert len(completion.choices) == 2
assert len(completion.choices[0].text) >= 1
assert len(completion.choices[1].text) >= 1
@@ -131,7 +136,8 @@ async def test_completions_with_prompt_embeds(
prompt="", # Add empty prompt as required parameter
max_tokens=5,
temperature=0.0,
extra_body={"prompt_embeds": encoded_embeds})
extra_body={"prompt_embeds": encoded_embeds},
)
single_output = single_completion.choices[0].text
stream = await client_with_prompt_embeds.completions.create(
@@ -140,7 +146,8 @@ async def test_completions_with_prompt_embeds(
max_tokens=5,
temperature=0.0,
stream=True,
extra_body={"prompt_embeds": encoded_embeds})
extra_body={"prompt_embeds": encoded_embeds},
)
chunks = []
finish_reason_count = 0
async for chunk in stream:
@@ -159,12 +166,12 @@ async def test_completions_with_prompt_embeds(
max_tokens=5,
temperature=0.0,
stream=True,
extra_body={"prompt_embeds": [encoded_embeds, encoded_embeds2]})
extra_body={"prompt_embeds": [encoded_embeds, encoded_embeds2]},
)
chunks_stream_embeds: list[list[str]] = [[], []]
finish_reason_count = 0
async for chunk in stream:
chunks_stream_embeds[chunk.choices[0].index].append(
chunk.choices[0].text)
chunks_stream_embeds[chunk.choices[0].index].append(chunk.choices[0].text)
if chunk.choices[0].finish_reason is not None:
finish_reason_count += 1
assert finish_reason_count == 2
@@ -179,7 +186,8 @@ async def test_completions_with_prompt_embeds(
prompt="This is a prompt",
max_tokens=5,
temperature=0.0,
extra_body={"prompt_embeds": encoded_embeds})
extra_body={"prompt_embeds": encoded_embeds},
)
assert len(completion.choices) == 2
completion_text_only = await client_with_prompt_embeds.completions.create(
model=model_name,
@@ -192,18 +200,18 @@ async def test_completions_with_prompt_embeds(
prompt="",
max_tokens=5,
temperature=0.0,
extra_body={"prompt_embeds": encoded_embeds})
extra_body={"prompt_embeds": encoded_embeds},
)
# Embeddings responses should be handled first
assert completion_mixed.choices[0].text == completion_embeds_only.choices[
0].text
assert completion_mixed.choices[1].text == completion_text_only.choices[
0].text
assert completion_mixed.choices[0].text == completion_embeds_only.choices[0].text
assert completion_mixed.choices[1].text == completion_text_only.choices[0].text
@pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME, LORA_SERVING_MODEL_NAME])
async def test_completions_errors_with_prompt_embeds(
client_with_prompt_embeds: openai.AsyncOpenAI, model_name: str):
client_with_prompt_embeds: openai.AsyncOpenAI, model_name: str
):
# Test error case: invalid prompt_embeds
with pytest.raises(BadRequestError):
await client_with_prompt_embeds.completions.create(
@@ -211,7 +219,8 @@ async def test_completions_errors_with_prompt_embeds(
model=model_name,
max_tokens=5,
temperature=0.0,
extra_body={"prompt_embeds": "invalid_base64"})
extra_body={"prompt_embeds": "invalid_base64"},
)
@pytest.mark.asyncio
@@ -233,7 +242,8 @@ async def test_completions_with_logprobs_and_prompt_embeds(
temperature=0.0,
echo=False,
logprobs=logprobs_arg,
extra_body={"prompt_embeds": encoded_embeds})
extra_body={"prompt_embeds": encoded_embeds},
)
logprobs = completion.choices[0].logprobs
assert logprobs is not None
@@ -252,7 +262,8 @@ async def test_completions_with_logprobs_and_prompt_embeds(
temperature=0.0,
echo=False,
logprobs=logprobs_arg,
extra_body={"prompt_embeds": [encoded_embeds, encoded_embeds2]})
extra_body={"prompt_embeds": [encoded_embeds, encoded_embeds2]},
)
assert len(completion.choices) == 2
for choice in completion.choices:
@@ -262,8 +273,7 @@ async def test_completions_with_logprobs_and_prompt_embeds(
assert len(logprobs.token_logprobs) == 5
assert len(logprobs.top_logprobs) == 5
for top_logprobs in logprobs.top_logprobs[1:]:
assert max(logprobs_arg,
1) <= len(top_logprobs) <= logprobs_arg + 1
assert max(logprobs_arg, 1) <= len(top_logprobs) <= logprobs_arg + 1
assert len(logprobs.tokens) == 5
@@ -280,8 +290,5 @@ async def test_prompt_logprobs_raises_error(
prompt="",
max_tokens=5,
temperature=0.0,
extra_body={
"prompt_embeds": encoded_embeds,
"prompt_logprobs": True
},
extra_body={"prompt_embeds": encoded_embeds, "prompt_logprobs": True},
)