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

117
tests/entrypoints/openai/test_transcription_validation.py
View File

@@ -17,8 +17,12 @@ from ...utils import RemoteOpenAIServer
MODEL_NAME = "openai/whisper-large-v3-turbo"
SERVER_ARGS = ["--enforce-eager"]
MISTRAL_FORMAT_ARGS = [
"--tokenizer_mode", "mistral", "--config_format", "mistral",
"--load_format", "mistral"
"--tokenizer_mode",
"mistral",
"--config_format",
"mistral",
"--load_format",
"mistral",
]
@@ -36,8 +40,8 @@ async def client(server):
@pytest.mark.asyncio
@pytest.mark.parametrize(
"model_name",
["openai/whisper-large-v3-turbo", "mistralai/Voxtral-Mini-3B-2507"])
"model_name", ["openai/whisper-large-v3-turbo", "mistralai/Voxtral-Mini-3B-2507"]
)
async def test_basic_audio(mary_had_lamb, model_name):
server_args = ["--enforce-eager"]
@@ -52,10 +56,11 @@ async def test_basic_audio(mary_had_lamb, model_name):
file=mary_had_lamb,
language="en",
response_format="text",
temperature=0.0)
temperature=0.0,
)
out = json.loads(transcription)
out_text = out['text']
out_usage = out['usage']
out_text = out["text"]
out_usage = out["usage"]
assert "Mary had a little lamb," in out_text
assert out_usage["seconds"] == 16, out_usage["seconds"]
@@ -74,8 +79,9 @@ async def test_basic_audio_gemma(foscolo):
file=foscolo,
language="it",
response_format="text",
temperature=0.0)
out = json.loads(transcription)['text']
temperature=0.0,
)
out = json.loads(transcription)["text"]
assert "da cui vergine nacque Venere" in out
@@ -85,24 +91,21 @@ async def test_non_asr_model(winning_call):
model_name = "JackFram/llama-68m"
with RemoteOpenAIServer(model_name, SERVER_ARGS) as remote_server:
client = remote_server.get_async_client()
res = await client.audio.transcriptions.create(model=model_name,
file=winning_call,
language="en",
temperature=0.0)
res = await client.audio.transcriptions.create(
model=model_name, file=winning_call, language="en", temperature=0.0
)
err = res.error
assert err["code"] == 400 and not res.text
assert err[
"message"] == "The model does not support Transcriptions API"
assert err["message"] == "The model does not support Transcriptions API"
@pytest.mark.asyncio
async def test_bad_requests(mary_had_lamb, client):
# invalid language
with pytest.raises(openai.BadRequestError):
await client.audio.transcriptions.create(model=MODEL_NAME,
file=mary_had_lamb,
language="hh",
temperature=0.0)
await client.audio.transcriptions.create(
model=MODEL_NAME, file=mary_had_lamb, language="hh", temperature=0.0
)
@pytest.mark.asyncio
@@ -114,17 +117,18 @@ async def test_long_audio_request(mary_had_lamb, client):
repeated_audio = np.tile(audio, 10)
# Repeated audio to buffer
buffer = io.BytesIO()
sf.write(buffer, repeated_audio, sr, format='WAV')
sf.write(buffer, repeated_audio, sr, format="WAV")
buffer.seek(0)
transcription = await client.audio.transcriptions.create(
model=MODEL_NAME,
file=buffer,
language="en",
response_format="text",
temperature=0.0)
temperature=0.0,
)
out = json.loads(transcription)
out_text = out['text']
out_usage = out['usage']
out_text = out["text"]
out_usage = out["usage"]
counts = out_text.count("Mary had a little lamb")
assert counts == 10, counts
assert out_usage["seconds"] == 161, out_usage["seconds"]
@@ -135,10 +139,8 @@ async def test_completion_endpoints(client):
# text to text model
res = await client.chat.completions.create(
model=MODEL_NAME,
messages=[{
"role": "system",
"content": "You are a helpful assistant."
}])
messages=[{"role": "system", "content": "You are a helpful assistant."}],
)
err = res.error
assert err["code"] == 400
assert err["message"] == "The model does not support Chat Completions API"
@@ -157,16 +159,19 @@ async def test_streaming_response(winning_call, client):
file=winning_call,
response_format="json",
language="en",
temperature=0.0)
res = await client.audio.transcriptions.create(model=MODEL_NAME,
file=winning_call,
language="en",
temperature=0.0,
stream=True,
timeout=30)
temperature=0.0,
)
res = await client.audio.transcriptions.create(
model=MODEL_NAME,
file=winning_call,
language="en",
temperature=0.0,
stream=True,
timeout=30,
)
# Reconstruct from chunks and validate
async for chunk in res:
text = chunk.choices[0]['delta']['content']
text = chunk.choices[0]["delta"]["content"]
transcription += text
assert transcription == res_no_stream.text
@@ -180,9 +185,9 @@ async def test_stream_options(winning_call, client):
language="en",
temperature=0.0,
stream=True,
extra_body=dict(stream_include_usage=True,
stream_continuous_usage_stats=True),
timeout=30)
extra_body=dict(stream_include_usage=True, stream_continuous_usage_stats=True),
timeout=30,
)
final = False
continuous = True
async for chunk in res:
@@ -190,7 +195,7 @@ async def test_stream_options(winning_call, client):
# final usage sent
final = True
else:
continuous = continuous and hasattr(chunk, 'usage')
continuous = continuous and hasattr(chunk, "usage")
assert final and continuous
@@ -198,27 +203,31 @@ async def test_stream_options(winning_call, client):
async def test_sampling_params(mary_had_lamb, client):
"""
Compare sampling with params and greedy sampling to assert results
are different when extreme sampling parameters values are picked.
are different when extreme sampling parameters values are picked.
"""
transcription = await client.audio.transcriptions.create(
model=MODEL_NAME,
file=mary_had_lamb,
language="en",
temperature=0.8,
extra_body=dict(seed=42,
repetition_penalty=1.9,
top_k=12,
top_p=0.4,
min_p=0.5,
frequency_penalty=1.8,
presence_penalty=2.0))
extra_body=dict(
seed=42,
repetition_penalty=1.9,
top_k=12,
top_p=0.4,
min_p=0.5,
frequency_penalty=1.8,
presence_penalty=2.0,
),
)
greedy_transcription = await client.audio.transcriptions.create(
model=MODEL_NAME,
file=mary_had_lamb,
language="en",
temperature=0.0,
extra_body=dict(seed=42))
extra_body=dict(seed=42),
)
assert greedy_transcription.text != transcription.text
@@ -226,15 +235,16 @@ async def test_sampling_params(mary_had_lamb, client):
@pytest.mark.asyncio
async def test_audio_prompt(mary_had_lamb, client):
prompt = "This is a speech, recorded in a phonograph."
#Prompts should not omit the part of original prompt while transcribing.
# Prompts should not omit the part of original prompt while transcribing.
prefix = "The first words I spoke in the original phonograph"
transcription = await client.audio.transcriptions.create(
model=MODEL_NAME,
file=mary_had_lamb,
language="en",
response_format="text",
temperature=0.0)
out = json.loads(transcription)['text']
temperature=0.0,
)
out = json.loads(transcription)["text"]
assert prefix in out
transcription_wprompt = await client.audio.transcriptions.create(
model=MODEL_NAME,
@@ -242,6 +252,7 @@ async def test_audio_prompt(mary_had_lamb, client):
language="en",
response_format="text",
prompt=prompt,
temperature=0.0)
out_prompt = json.loads(transcription_wprompt)['text']
temperature=0.0,
)
out_prompt = json.loads(transcription_wprompt)["text"]
assert prefix in out_prompt