[Feature] Support online inference

[Gaohan123]

Purpose

This PR supports online serving for vllm-omni, which resolves #38 and part of #33 . It has been firstly verified on Qwen2.5-omni.

Test Plan

Here I both tests offline and online inference. First please follow README.md in root repo to finish installation.

Online inference

vllm serve Qwen/Qwen2.5-Omni-7B --omni --port 8091

Follow README.md in examples/online_serving

Offline inference

Follow README.md in examples/offline_inference/qwen2_5_omni

Test Result

Online inference

For request from python:

python openai_chat_completion_client_for_multimodal_generation.py
Chat completion output from text: Well, it usually has input modules for data, processing units like neural networks or algorithms, output for generated audio, and scalability through parallel computing or distributed systems.If you want to know more about any part of this, feel free to ask.
Audio saved to audio_0.wav

For request from curl:

bash run_curl_multimodal_generation.sh
Output of request: "Well, it usually has input modules for data, processing units like neural networks or algorithms, output for generated audio, and scalability through parallel computing or distributed systems.If you want to know more about any part of this, feel free to ask."

Offline inference

For single request:

Request ID: 0, Text saved to outputs/00000.txt
Request ID: 0, Saved audio to outputs/output_0.wav

For multiple requests:

Request ID: 0, Text saved to outputs/00000.txt
Request ID: 1, Text saved to outputs/00001.txt
Request ID: 2, Text saved to outputs/00002.txt
...
Request ID: 0, Saved audio to outputs/output_0.wav
Request ID: 1, Saved audio to outputs/output_1.wav
Request ID: 2, Saved audio to outputs/output_2.wav
...

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/hsliuustc0106/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

[tzhouam]

looks good to me