[BugFix] Fix the issue of thinker requests being preempted, causing shape mismatch.

[amy-why-3459]

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

Fix the issue of thinker requests being preempted, causing shape mismatch

Test Plan

pytest -sv tests/e2e/online_serving/test_qwen3_omni_expansion.py -m "full_model" --run-level "full_model"

vllm bench serve \
    --omni \
  --dataset-name random \
  --port 28889 \
  --max-concurrency 32 \
  --model /home/models/Qwen3-Omni-30B-A3B-Instruct \
  --endpoint /v1/chat/completions \
  --backend openai-chat-omni \
  --num-warmups 2 \
  --num-prompts 128 \
  --random-input-len 2500 \
  --ignore-eos \
  --percentile-metrics ttft,tpot,itl,e2el,audio_ttfp,audio_rtf \
  --random-output-len 900 \
  --extra_body '{"modalities": ["text", "audio"]}'

Test Result

=============================== warnings summary ===============================
<frozen importlib._bootstrap>:488
  <frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyPacked has no __module__ attribute

<frozen importlib._bootstrap>:488
  <frozen importlib._bootstrap>:488: DeprecationWarning: builtin type SwigPyObject has no __module__ attribute

../../../usr/local/lib/python3.12/dist-packages/torch/jit/_script.py:365: 14 warnings
  /usr/local/lib/python3.12/dist-packages/torch/jit/_script.py:365: DeprecationWarning: `torch.jit.script_method` is deprecated. Please switch to `torch.compile` or `torch.export`.
    warnings.warn(

tests/e2e/online_serving/test_qwen3_omni_expansion.py::test_text_to_audio_001[default]
  /usr/local/lib/python3.12/dist-packages/pydub/utils.py:14: DeprecationWarning: 'audioop' is deprecated and slated for removal in Python 3.13
    import audioop

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
--- Running Summary
=========== 36 passed, 3 skipped, 17 warnings in 1433.71s (0:23:53) ============

============ Serving Benchmark Result ============
Successful requests:                     128
Failed requests:                         0
Maximum request concurrency:             32
Benchmark duration (s):                  783.97
Request throughput (req/s):              0.16
Peak concurrent requests:                37.00
----------------End-to-end Latency----------------
Mean E2EL (ms):                          176080.43
Median E2EL (ms):                        174200.13
P99 E2EL (ms):                           299217.79
================== Text Result ===================
Total input tokens:                      321792
Total generated tokens:                  75756
Output token throughput (tok/s):         96.63
Peak output token throughput (tok/s):    928.00
Peak concurrent requests:                37.00
Total Token throughput (tok/s):          507.09
---------------Time to First Token----------------
Mean TTFT (ms):                          1527.28
Median TTFT (ms):                        289.07
P99 TTFT (ms):                           5655.24
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms):                          58.93
Median TPOT (ms):                        36.80
P99 TPOT (ms):                           327.84
---------------Inter-token Latency----------------
Mean ITL (ms):                           25.41
Median ITL (ms):                         21.15
P99 ITL (ms):                            104.51
================== Audio Result ==================
Total audio duration generated(s):       23292.16
Total audio frames generated:            559011675
Audio throughput(audio duration/s):      29.71
---------------Time to First Packet---------------
Mean AUDIO_TTFP (ms):                    6244.77
Median AUDIO_TTFP (ms):                  3812.04
P99 AUDIO_TTFP (ms):                     14192.98
-----------------Real Time Factor-----------------
Mean AUDIO_RTF:                          1.10
Median AUDIO_RTF:                        1.01
P99 AUDIO_RTF:                           1.57
==================================================

---

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. Please provide the test scripts & test commands. Please state the reasons if your codes don't require additional test scripts. For test file guidelines, please check the test style doc
• The test results. Please paste the results comparison before and after, or the e2e results.
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model. Please run mkdocs serve to sync the documentation editions to ./docs.
• (Optional) Release notes update. If your change is user-facing, please update the release notes draft.

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

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a07f07fdd0

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[hsliuustc0106]

I also met this problem when I try to send request using vllm bench --omni