[Bug]: qwen3-omni realtime audio return random voice and noise

[ram16g]

Your current environment

The output of python collect_env.py

WARNING 01-29 11:48:49 [mooncake_connector.py:18] Mooncake not available, MooncakeOmniConnector will not work
WARNING 01-29 11:48:49 [yuanrong_connector.py:14] Datasystem not available, YuanrongConnector will not work
WARNING 01-29 11:48:49 [envs.py:76] Flash Attention library "flash_attn" not found, using pytorch attention implementation
==============================
        System Info
==============================
OS                           : Ubuntu 24.04.3 LTS (x86_64)
GCC version                  : (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
Clang version                : Could not collect
CMake version                : version 3.28.3
Libc version                 : glibc-2.39

==============================
       PyTorch Info
==============================
PyTorch version              : 2.9.1+cu128
Is debug build               : False
CUDA used to build PyTorch   : 12.8
ROCM used to build PyTorch   : N/A

==============================
      Python Environment
==============================
Python version               : 3.12.12 | packaged by Anaconda, Inc. | (main, Oct 21 2025, 20:16:04) [GCC 11.2.0] (64-bit runtime)
Python platform              : Linux-6.14.0-37-generic-x86_64-with-glibc2.39

==============================
       CUDA / GPU Info
==============================
Is CUDA available            : True
CUDA runtime version         : 13.0.88
CUDA_MODULE_LOADING set to   :
GPU models and configuration :
GPU 0: NVIDIA RTX PRO 6000 Blackwell Server Edition
GPU 1: NVIDIA RTX PRO 6000 Blackwell Server Edition

Nvidia driver version        : 580.95.05
cuDNN version                : Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.9.10.1
/usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.10.1
/usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.10.1
/usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.10.1
/usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.10.1
/usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.10.1
/usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.10.1
/usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.10.1
HIP runtime version          : N/A
MIOpen runtime version       : N/A
Is XNNPACK available         : True

==============================
          CPU Info
==============================
架构：                                   x86_64
CPU 运行模式：                           32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
字节序：                                 Little Endian
CPU:                                     208
在线 CPU 列表：                          0-207
厂商 ID：                                GenuineIntel
型号名称：                               Intel(R) Xeon(R) Platinum 8473C
CPU 系列：                               6
型号：                                   143
每个核的线程数：                         2
每个座的核数：                           52
座：                                     2
步进：                                   8
CPU(s) scaling MHz:                      23%
CPU 最大 MHz：                           3800.0000
CPU 最小 MHz：                           800.0000
BogoMIPS：                               4200.00
标记：                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cat_l2 cdp_l3 intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect user_shstk avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req hfi vnmi avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr ibt amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
虚拟化：                                 VT-x
L1d 缓存：                               4.9 MiB (104 instances)
L1i 缓存：                               3.3 MiB (104 instances)
L2 缓存：                                208 MiB (104 instances)
L3 缓存：                                210 MiB (2 instances)
NUMA 节点：                              2
NUMA 节点0 CPU：                         0-51,104-155
NUMA 节点1 CPU：                         52-103,156-207
Vulnerability Gather data sampling:      Not affected
Vulnerability Ghostwrite:                Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Mitigation; IBPB before exit to userspace

==============================
Versions of relevant libraries
==============================
[pip3] flashinfer-python==0.5.3
[pip3] numpy==2.2.6
[pip3] nvidia-cublas-cu12==12.8.4.1
[pip3] nvidia-cuda-cupti-cu12==12.8.90
[pip3] nvidia-cuda-nvrtc-cu12==12.8.93
[pip3] nvidia-cuda-runtime-cu12==12.8.90
[pip3] nvidia-cudnn-cu12==9.10.2.21
[pip3] nvidia-cudnn-frontend==1.17.0
[pip3] nvidia-cufft-cu12==11.3.3.83
[pip3] nvidia-cufile-cu12==1.13.1.3
[pip3] nvidia-curand-cu12==10.3.9.90
[pip3] nvidia-cusolver-cu12==11.7.3.90
[pip3] nvidia-cusparse-cu12==12.5.8.93
[pip3] nvidia-cusparselt-cu12==0.7.1
[pip3] nvidia-cutlass-dsl==4.3.4
[pip3] nvidia-ml-py==13.590.44
[pip3] nvidia-nccl-cu12==2.27.5
[pip3] nvidia-nvjitlink-cu12==12.8.93
[pip3] nvidia-nvshmem-cu12==3.3.20
[pip3] nvidia-nvtx-cu12==12.8.90
[pip3] onnxruntime==1.23.2
[pip3] pyzmq==27.1.0
[pip3] torch==2.9.1
[pip3] torchaudio==2.9.1
[pip3] torchsde==0.2.6
[pip3] torchvision==0.24.1
[pip3] transformers==4.57.3
[pip3] triton==3.5.1
[conda] flashinfer-python                    0.5.3            pypi_0           pypi
[conda] numpy                                2.2.6            pypi_0           pypi
[conda] nvidia-cublas-cu12                   12.8.4.1         pypi_0           pypi
[conda] nvidia-cuda-cupti-cu12               12.8.90          pypi_0           pypi
[conda] nvidia-cuda-nvrtc-cu12               12.8.93          pypi_0           pypi
[conda] nvidia-cuda-runtime-cu12             12.8.90          pypi_0           pypi
[conda] nvidia-cudnn-cu12                    9.10.2.21        pypi_0           pypi
[conda] nvidia-cudnn-frontend                1.17.0           pypi_0           pypi
[conda] nvidia-cufft-cu12                    11.3.3.83        pypi_0           pypi
[conda] nvidia-cufile-cu12                   1.13.1.3         pypi_0           pypi
[conda] nvidia-curand-cu12                   10.3.9.90        pypi_0           pypi
[conda] nvidia-cusolver-cu12                 11.7.3.90        pypi_0           pypi
[conda] nvidia-cusparse-cu12                 12.5.8.93        pypi_0           pypi
[conda] nvidia-cusparselt-cu12               0.7.1            pypi_0           pypi
[conda] nvidia-cutlass-dsl                   4.3.4            pypi_0           pypi
[conda] nvidia-ml-py                         13.590.44        pypi_0           pypi
[conda] nvidia-nccl-cu12                     2.27.5           pypi_0           pypi
[conda] nvidia-nvjitlink-cu12                12.8.93          pypi_0           pypi
[conda] nvidia-nvshmem-cu12                  3.3.20           pypi_0           pypi
[conda] nvidia-nvtx-cu12                     12.8.90          pypi_0           pypi
[conda] pyzmq                                27.1.0           pypi_0           pypi
[conda] torch                                2.9.1            pypi_0           pypi
[conda] torchaudio                           2.9.1            pypi_0           pypi
[conda] torchsde                             0.2.6            pypi_0           pypi
[conda] torchvision                          0.24.1           pypi_0           pypi
[conda] transformers                         4.57.3           pypi_0           pypi
[conda] triton                               3.5.1            pypi_0           pypi

==============================
         vLLM Info
==============================
ROCM Version                 : Could not collect
vLLM Version                 : 0.14.0
vLLM-Omni Version            : 0.14.0rc1 (git sha: 741f7e2)
vLLM Build Flags:
  CUDA Archs: Not Set; ROCm: Disabled
GPU Topology:
        GPU0    GPU1    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NODE    52-103,156-207  1               N/A
GPU1    NODE     X      52-103,156-207  1               N/A

Legend:

  X    = Self
  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
  PIX  = Connection traversing at most a single PCIe bridge
  NV#  = Connection traversing a bonded set of # NVLinks

==============================
     Environment Variables
==============================
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
LD_LIBRARY_PATH=/home/kemove/miniconda3/envs/vllm-omni/lib/python3.12/site-packages/cv2/../../lib64:
TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_kemove

Your code version

The commit id or version of vllm

The commit id or version of vllm-omni

🐛 Describe the bug

访问 v1/chat/completions 这个端点，我提供了一段录音

call_recording_2026-01-29T02-28-24.wav

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[david6666666]

serve command and request command

[ram16g]

vllm serve $MODEL_PATH
--omni
--port 8002
--stage-configs-path ~/call/vllm-omni/vllm_omni/model_executor/stage_configs/qwen3_omni_moe_single_gpu_fp8kv_async_chunk.yaml
--served-model-name qwen3-omni
--attention-config '{"backend": "FLASHINFER"}'

yaml 文件内容：

async_chunk: true
stage_args:

• stage_id: 0
  stage_type: llm
  runtime:
  devices: "0"
  max_batch_size: 2
  engine_args:
  model_stage: thinker
  model_arch: Qwen3OmniMoeForConditionalGeneration
  worker_cls: vllm_omni.worker.gpu_ar_worker.GPUARWorker
  scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
  gpu_memory_utilization: 0.70
  enforce_eager: false
  attention_backend: FLASHINFER
  trust_remote_code: true
  engine_output_type: latent
  distributed_executor_backend: "mp"
  enable_prefix_caching: false
  max_num_batched_tokens: 32768
  hf_config_name: thinker_config
  tensor_parallel_size: 1
  kv_cache_dtype: fp8
  custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker_async_chunk
  final_output: true
  final_output_type: text
  is_comprehension: true
  default_sampling_params:
  temperature: 0.4
  top_p: 0.9
  top_k: 1
  max_tokens: 2048
  seed: 42
  detokenize: True
  repetition_penalty: 1.05

• stage_id: 1
  stage_type: llm
  runtime:
  devices: "0"
  max_batch_size: 2
  engine_args:
  model_stage: talker
  model_arch: Qwen3OmniMoeForConditionalGeneration
  worker_cls: vllm_omni.worker.gpu_ar_worker.GPUARWorker
  scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
  gpu_memory_utilization: 0.12
  enforce_eager: false
  attention_backend: FLASHINFER

请求：

curl -X POST http://192.168.103.121:8002/v1/chat/completions
-H "Content-Type: application/json"
-d '{
"model": "qwen3-omni",
"messages": [
{
"role": "system",
"content": "你是一个语音助手。用户会用语音和你对话，你只需要简短回复一句话即可。"
},
{
"role": "user",
"content": [
{
"type": "text",
"text": "[语音消息]"
},
{
"type": "input_audio",
"input_audio": {
"data": "UklGRiQAAABXQVZFZm10...(Base64编码的WAV音频数据)",
"format": "wav"
}
}
]
}
],
"temperature": 0.7,
"max_tokens": 2048,
"stream": true,
"modalities": ["audio"],
"audio": {
"format": "wav"
}
}'

[david6666666]

cc @amy-why-3459

[amy-why-3459]

Can you provide the commit_id you used?

[amy-why-3459]

If you are using audio input, perhaps you can try using the commit_id after this PR: #1020. In this PR, I fixed the precision issue for multimodal input.

[ram16g]

The latest code fixed the issue, but a startup error occurred.

log.txt

● Update(~/ai_call/vllm-omni/vllm_omni/model_executor/models/qwen3_omni/qwen3_omni_moe_code_predictor_mtp.py)
⎿  Added 3 lines, removed 1 line
513 # Input for this layer: [last_talker_hidden, prev_layer_embed]
514
515 # Forward through code_predictor model
516 - position_ids = self.position_ids_buffer[layer_idx][: (layer_idx + 2) * batch_size] # [1, seq_len]
516 + # Compute position_ids dynamically to avoid torch.compile specializing batch_size
517 + seq_len = layer_idx + 2
518 + position_ids = torch.arange(seq_len, device=current_input.device, dtype=torch.int64).repeat(batch_size)
519 begin_pos += layer_idx + 2
520 outputs = self.model(
521 inputs_embeds=current_input,

After these changes, it starts successfully.

[ZeldaHuang]

The latest code fixed the issue, but a startup error occurred.

log.txt

● Update(~/ai_call/vllm-omni/vllm_omni/model_executor/models/qwen3_omni/qwen3_omni_moe_code_predictor_mtp.py) ⎿  Added 3 lines, removed 1 line 513 # Input for this layer: [last_talker_hidden, prev_layer_embed] 514 515 # Forward through code_predictor model 516 - position_ids = self.position_ids_buffer[layer_idx][: (layer_idx + 2) * batch_size] # [1, seq_len] 516 + # Compute position_ids dynamically to avoid torch.compile specializing batch_size 517 + seq_len = layer_idx + 2 518 + position_ids = torch.arange(seq_len, device=current_input.device, dtype=torch.int64).repeat(batch_size) 519 begin_pos += layer_idx + 2 520 outputs = self.model( 521 inputs_embeds=current_input,

After these changes, it starts successfully.

@ram16g Thanks for the fix , I think it's a bug when max_batch_size=2 after #1005 merged, can you submit a PR to fix it?

[ram16g]

#1102