154 lines
5.3 KiB
Python
154 lines
5.3 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
|
|
|
|
"""
|
|
E2E Online tests for Qwen2.5-Omni model with video input and audio output.
|
|
"""
|
|
|
|
import os
|
|
|
|
import pytest
|
|
|
|
from tests.helpers.mark import hardware_test
|
|
from tests.helpers.media import generate_synthetic_audio, generate_synthetic_image, generate_synthetic_video
|
|
from tests.helpers.runtime import OmniServerParams, dummy_messages_from_mix_data
|
|
from tests.helpers.stage_config import get_deploy_config_path, modify_stage_config
|
|
|
|
os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
|
|
|
|
|
|
models = ["Qwen/Qwen2.5-Omni-7B"]
|
|
|
|
# Single CI deploy YAML; rocm/xpu deltas are picked automatically via the
|
|
# platforms: section in vllm_omni/deploy/ci/qwen2_5_omni.yaml.
|
|
stage_configs = [
|
|
modify_stage_config(
|
|
get_deploy_config_path("ci/qwen2_5_omni.yaml"),
|
|
# This suite requests two cards. Share the Talker's card with
|
|
# Code2Wav and leave room for its whole-sequence DiT workspace.
|
|
updates={"stages": {1: {"gpu_memory_utilization": 0.3}, 2: {"devices": "1"}}},
|
|
)
|
|
]
|
|
|
|
# Create parameter combinations for model and stage config
|
|
test_params = [
|
|
OmniServerParams(model=model, stage_config_path=stage_config) for model in models for stage_config in stage_configs
|
|
]
|
|
|
|
|
|
def get_system_prompt():
|
|
return {
|
|
"role": "system",
|
|
"content": [
|
|
{
|
|
"type": "text",
|
|
"text": (
|
|
"You are Qwen, a virtual human developed by the Qwen Team, "
|
|
"Alibaba Group, capable of perceiving auditory and visual inputs, "
|
|
"as well as generating text and speech."
|
|
),
|
|
}
|
|
],
|
|
}
|
|
|
|
|
|
def get_prompt(prompt_type="text_only"):
|
|
prompts = {
|
|
"text_only": "What is the capital of China? Answer in 20 words.",
|
|
"mix": "What is recited in the audio? What is in this image? Describe the video briefly.",
|
|
}
|
|
return prompts.get(prompt_type, prompts["text_only"])
|
|
|
|
|
|
def get_max_batch_size(size_type="few"):
|
|
batch_sizes = {"few": 5, "medium": 100, "large": 256}
|
|
return batch_sizes.get(size_type, 5)
|
|
|
|
|
|
@pytest.mark.slow
|
|
@pytest.mark.omni
|
|
@hardware_test(res={"cuda": "L4", "rocm": "MI325"}, num_cards=2)
|
|
@pytest.mark.parametrize("omni_server", test_params, indirect=True)
|
|
def test_mix_to_text_audio_001(omni_server, online_client) -> None:
|
|
"""
|
|
Test multi-modal input processing and text/audio output generation via OpenAI API.
|
|
Deploy Setting: default yaml
|
|
Input Modal: text + audio + video + image
|
|
Output Modal: text + audio
|
|
Input Setting: stream=True
|
|
Datasets: single request
|
|
"""
|
|
|
|
video_data_url = f"data:video/mp4;base64,{generate_synthetic_video(224, 224, 300)['base64']}"
|
|
image_data_url = f"data:image/jpeg;base64,{generate_synthetic_image(224, 224)['base64']}"
|
|
audio_data_url = f"data:audio/wav;base64,{generate_synthetic_audio(5, 1, sample_rate=16000)['base64']}"
|
|
messages = dummy_messages_from_mix_data(
|
|
system_prompt=get_system_prompt(),
|
|
video_data_url=video_data_url,
|
|
image_data_url=image_data_url,
|
|
audio_data_url=audio_data_url,
|
|
content_text=get_prompt("mix"),
|
|
)
|
|
|
|
request_config = {
|
|
"model": omni_server.model,
|
|
"messages": messages,
|
|
"stream": True,
|
|
"key_words": {
|
|
"video": ["sphere", "globe", "circle", "round", "ball"],
|
|
},
|
|
}
|
|
|
|
# Test single completion
|
|
online_client.send_omni_request(request_config)
|
|
|
|
|
|
@pytest.mark.slow
|
|
@pytest.mark.omni
|
|
@hardware_test(res={"cuda": "L4", "rocm": "MI325"}, num_cards=2)
|
|
@pytest.mark.parametrize("omni_server", test_params, indirect=True)
|
|
@pytest.mark.parametrize("stream", [False, True])
|
|
def test_audio_in_video_boundaries(omni_server, online_client, stream: bool) -> None:
|
|
"""Interleaved video/audio input must not treat audio boundaries as image placeholders."""
|
|
# More than two seconds exercises multiple interleaved video/audio chunks.
|
|
video_data_url = f"data:video/mp4;base64,{generate_synthetic_video(224, 224, 128, embed_audio=True)['base64']}"
|
|
messages = dummy_messages_from_mix_data(
|
|
system_prompt=get_system_prompt(),
|
|
video_data_url=video_data_url,
|
|
content_text="Describe what you see and hear in this video in one sentence.",
|
|
)
|
|
responses = online_client.send_omni_request(
|
|
{
|
|
"model": omni_server.model,
|
|
"messages": messages,
|
|
"stream": stream,
|
|
"modalities": ["text"],
|
|
"use_audio_in_video": True,
|
|
}
|
|
)
|
|
assert responses[0].text_content
|
|
|
|
|
|
@pytest.mark.slow
|
|
@pytest.mark.omni
|
|
@pytest.mark.parametrize("omni_server", test_params, indirect=True)
|
|
def test_text_to_text_001(omni_server, online_client) -> None:
|
|
"""
|
|
Test text input processing and text/audio output generation via OpenAI API.
|
|
Deploy Setting: default yaml
|
|
Input Modal: text
|
|
Output Modal: text
|
|
Datasets: few requests
|
|
"""
|
|
messages = dummy_messages_from_mix_data(system_prompt=get_system_prompt(), content_text=get_prompt())
|
|
|
|
request_config = {
|
|
"model": omni_server.model,
|
|
"messages": messages,
|
|
"stream": False,
|
|
"modalities": ["text"],
|
|
"key_words": {"text": ["beijing"]},
|
|
}
|
|
|
|
online_client.send_omni_request(request_config, request_num=get_max_batch_size())
|