Voxtral (#20970)
Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
This commit is contained in:
co-authored by
Cyrus Leung
parent
4ffd963fa0
commit
e7e3e6d263
@@ -17,6 +17,11 @@ from vllm.assets.audio import AudioAsset
|
||||
|
||||
from ...utils import RemoteOpenAIServer
|
||||
|
||||
MISTRAL_FORMAT_ARGS = [
|
||||
"--tokenizer_mode", "mistral", "--config_format", "mistral",
|
||||
"--load_format", "mistral"
|
||||
]
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def mary_had_lamb():
|
||||
@@ -33,9 +38,18 @@ def winning_call():
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_basic_audio(mary_had_lamb):
|
||||
model_name = "openai/whisper-large-v3-turbo"
|
||||
@pytest.mark.parametrize(
|
||||
"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"]
|
||||
|
||||
if model_name.startswith("mistralai"):
|
||||
server_args += MISTRAL_FORMAT_ARGS
|
||||
|
||||
# TODO(PATRICK) - REMOVE AFTER RELEASE
|
||||
return # skip for now
|
||||
|
||||
# Based on https://github.com/openai/openai-cookbook/blob/main/examples/Whisper_prompting_guide.ipynb.
|
||||
with RemoteOpenAIServer(model_name, server_args) as remote_server:
|
||||
client = remote_server.get_async_client()
|
||||
@@ -65,10 +79,13 @@ async def test_bad_requests(mary_had_lamb):
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_long_audio_request(mary_had_lamb):
|
||||
model_name = "openai/whisper-large-v3-turbo"
|
||||
@pytest.mark.parametrize("model_name", ["openai/whisper-large-v3-turbo"])
|
||||
async def test_long_audio_request(mary_had_lamb, model_name):
|
||||
server_args = ["--enforce-eager"]
|
||||
|
||||
if model_name.startswith("openai"):
|
||||
return
|
||||
|
||||
mary_had_lamb.seek(0)
|
||||
audio, sr = librosa.load(mary_had_lamb)
|
||||
# Add small silence after each audio for repeatability in the split process
|
||||
@@ -87,7 +104,8 @@ async def test_long_audio_request(mary_had_lamb):
|
||||
response_format="text",
|
||||
temperature=0.0)
|
||||
out = json.loads(transcription)['text']
|
||||
assert out.count("Mary had a little lamb") == 10
|
||||
counts = out.count("Mary had a little lamb")
|
||||
assert counts == 10, counts
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
|
||||
@@ -440,6 +440,7 @@ _MULTIMODAL_EXAMPLE_MODELS = {
|
||||
tokenizer="Isotr0py/Florence-2-tokenizer", # noqa: E501
|
||||
trust_remote_code=True), # noqa: E501
|
||||
"MllamaForConditionalGeneration": _HfExamplesInfo("meta-llama/Llama-3.2-11B-Vision-Instruct"), # noqa: E501
|
||||
"VoxtralForConditionalGeneration": _HfExamplesInfo("mistralai/Voxtral-Mini-3B-2507", is_available_online=False, tokenizer_mode="mistral"), # noqa: E501
|
||||
"WhisperForConditionalGeneration": _HfExamplesInfo("openai/whisper-large-v3"), # noqa: E501
|
||||
|
||||
# [Cross-encoder]
|
||||
@@ -513,4 +514,4 @@ class HfExampleModels:
|
||||
raise ValueError(f"No example model defined for {model_id}")
|
||||
|
||||
|
||||
HF_EXAMPLE_MODELS = HfExampleModels(_EXAMPLE_MODELS)
|
||||
HF_EXAMPLE_MODELS = HfExampleModels(_EXAMPLE_MODELS)
|
||||
|
||||
Reference in New Issue
Block a user