[V1] Enable multi-input by default (#15799)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-04-12 16:52:39 +08:00
committed by GitHub
parent f069f3ea74
commit d9fc8cd9da
21 changed files with 214 additions and 105 deletions

View File

@@ -12,7 +12,9 @@ from ...utils import RemoteOpenAIServer
MODEL_NAME = "fixie-ai/ultravox-v0_5-llama-3_2-1b"
TEST_AUDIO_URLS = [
AudioAsset("winning_call").url,
AudioAsset("mary_had_lamb").url,
]
MAXIMUM_AUDIOS = 2
@pytest.fixture(scope="module")
@@ -24,6 +26,8 @@ def server():
"5",
"--enforce-eager",
"--trust-remote-code",
"--limit-mm-per-prompt",
f"audio={MAXIMUM_AUDIOS}",
]
with RemoteOpenAIServer(MODEL_NAME, args) as remote_server:
@@ -46,7 +50,7 @@ def base64_encoded_audio() -> dict[str, str]:
@pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME])
@pytest.mark.parametrize("audio_url", TEST_AUDIO_URLS)
@pytest.mark.parametrize("audio_url", [TEST_AUDIO_URLS[0]])
async def test_single_chat_session_audio(client: openai.AsyncOpenAI,
model_name: str, audio_url: str):
messages = [{
@@ -100,7 +104,7 @@ async def test_single_chat_session_audio(client: openai.AsyncOpenAI,
@pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME])
@pytest.mark.parametrize("audio_url", TEST_AUDIO_URLS)
@pytest.mark.parametrize("audio_url", [TEST_AUDIO_URLS[0]])
async def test_single_chat_session_audio_base64encoded(
client: openai.AsyncOpenAI, model_name: str, audio_url: str,
base64_encoded_audio: dict[str, str]):
@@ -158,7 +162,7 @@ async def test_single_chat_session_audio_base64encoded(
@pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME])
@pytest.mark.parametrize("audio_url", TEST_AUDIO_URLS)
@pytest.mark.parametrize("audio_url", [TEST_AUDIO_URLS[0]])
async def test_single_chat_session_input_audio(
client: openai.AsyncOpenAI, model_name: str, audio_url: str,
base64_encoded_audio: dict[str, str]):
@@ -330,28 +334,21 @@ async def test_chat_streaming_input_audio(client: openai.AsyncOpenAI,
@pytest.mark.asyncio
@pytest.mark.parametrize("model_name", [MODEL_NAME])
@pytest.mark.parametrize("audio_url", TEST_AUDIO_URLS)
@pytest.mark.parametrize(
"audio_urls", [TEST_AUDIO_URLS, TEST_AUDIO_URLS + [TEST_AUDIO_URLS[0]]])
async def test_multi_audio_input(client: openai.AsyncOpenAI, model_name: str,
audio_url: str,
base64_encoded_audio: dict[str, str]):
audio_urls: list[str]):
messages = [{
"role":
"user",
"content": [
{
*({
"type": "audio_url",
"audio_url": {
"url": audio_url
}
},
{
"type": "input_audio",
"input_audio": {
"data": base64_encoded_audio[audio_url],
"format": "wav"
}
},
} for audio_url in audio_urls),
{
"type": "text",
"text": "What's happening in this audio?"
@@ -359,20 +356,30 @@ async def test_multi_audio_input(client: openai.AsyncOpenAI, model_name: str,
],
}]
with pytest.raises(openai.BadRequestError): # test multi-audio input
await client.chat.completions.create(
if len(audio_urls) > MAXIMUM_AUDIOS:
with pytest.raises(openai.BadRequestError): # test multi-audio input
await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
temperature=0.0,
)
# the server should still work afterwards
completion = await client.completions.create(
model=model_name,
prompt=[0, 0, 0, 0, 0],
max_tokens=5,
temperature=0.0,
)
completion = completion.choices[0].text
assert completion is not None and len(completion) >= 0
else:
chat_completion = await client.chat.completions.create(
model=model_name,
messages=messages,
max_completion_tokens=10,
temperature=0.0,
)
# the server should still work afterwards
completion = await client.completions.create(
model=model_name,
prompt=[0, 0, 0, 0, 0],
max_tokens=5,
temperature=0.0,
)
completion = completion.choices[0].text
assert completion is not None and len(completion) >= 0
message = chat_completion.choices[0].message
assert message.content is not None and len(message.content) >= 0