一、运行环境： 1、kylin v10 sp3 x86,海光k100_ai dcu, 2、docker Version: 20.10.24 3、gpustack: v2.1.0 4、K100_AI 驱动：rock-6.3.16-V1.1.0c.run 5、DTK:DTK-25.04.2-Kylinv10SP3-x86_64.tar.gz 二、问题 vllm运行Qwen3-TTS-12Hz-1.7B-CustomVoice时，报如下错误，请问是什么原因，能否提供解决方法或思路。 2026-03-30 23:25:36.942439+08:00 - gpustack.worker.backends.base - INFO - Preparing model files... 2026-03-30 23:25:36.947244+08:00 - gpustack.worker.backends.base - INFO - Model files are ready. 2026-03-30 23:25:36.947339+08:00 - gpustack.worker.serve\_manager - INFO - Provisioning model instance qwen3-tts-12hz-1.7b-customvoice-gHmMH 2026-03-30 23:25:36.947381+08:00 - gpustack.worker.backends.vllm - INFO - Starting vLLM model instance: qwen3-tts-12hz-1.7b-customvoice-gHmMH 2026-03-30 23:25:36.947701+08:00 - gpustack.worker.backends.vllm - INFO - Creating vLLM container workload: qwen3-tts-12hz-1.7b-customvoice-gHmMH 2026-03-30 23:25:36.947766+08:00 - gpustack.worker.backends.vllm - INFO - With image: harbor.sourcefind.cn:5443/dcu/admin/base/vllm:0.9.2-ubuntu22.04-dtk25.04.2-1226-das1.7-py3.10-20251226, command: [], arguments: [vllm serve /var/lib/gpustack/models/Qwen3-TTS-12Hz-1.7B-CustomVoice --omni --host 172.16.156.19 --port 40008 --served-model-name qwen3-tts-12hz-1.7b-customvoice], ports: [40008], envs(inconsistent input items mean unchangeable): OMP\_NUM\_THREADS=1 RUNAI\_STREAMER\_LOG\_LEVEL=INFO RUNAI\_STREAMER\_LOG\_TO\_STDERR=1 RUNAI\_STREAMER\_MEMORY\_LIMIT=0 SAFETENSORS\_FAST\_GPU=1 2026-03-30 23:25:36.973836+08:00 - 3821 - gpustack\_runtime.deployer.docker - INFO - Mirrored deployment enabled, using self Container dfa7f6698f39 for options mirroring 2026-03-30 23:25:37.402462+08:00 - gpustack.worker.backends.vllm - INFO - Created vLLM container workload: qwen3-tts-12hz-1.7b-customvoice-gHmMH 2026-03-30 23:25:37.402731+08:00 - gpustack.worker.serve\_manager - INFO - Finished provisioning model instance qwen3-tts-12hz-1.7b-customvoice-gHmMH \_ \_\_\_\_ \_\_ \_\_ \_\_\_ | | / / / / / / |/ / | | / / / / / / /|\_/ / | |/ / /\_\_\_/ /\_\_\_/ / / / |\_\_\_/\_\_\_\_\_/\_\_\_\_\_/\_/ /\_/ DTK version 25.04.2 vllm version 0.9.2 [32m8[0m DCU were detected in the container. [31mNOTE:[0m Make sure the user is in the 39 group. For more information, please go to https://developer.sourcefind.cn/ INFO 03-30 23:25:41 [\_\_init\_\_.py:245] Automatically detected platform rocm. usage: vllm [-h] [-v] {chat,complete,serve,bench,collect-env,run-batch} ... vllm: error: unrecognized arguments: --omni