[Bug]: 4xH800 Qwen/Qwen3-Next-80B-A3B-Instruct MTP, benchmark failed mixed_qkv_spec.view shape '[5, -1, 2048]' is invalid for input of size 104448

[david6666666]

Your current environment

The output of python collect_env.py

INFO 09-12 08:15:44 [__init__.py:216] Automatically detected platform cuda.
Collecting environment information...
uv is set
==============================
        System Info
==============================
OS                           : Ubuntu 24.04.1 LTS (x86_64)
GCC version                  : (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0
Clang version                : Could not collect
CMake version                : version 3.31.4
Libc version                 : glibc-2.39

==============================
       PyTorch Info
==============================
PyTorch version              : 2.8.0+cu128
Is debug build               : False
CUDA used to build PyTorch   : 12.8
ROCM used to build PyTorch   : N/A

==============================
      Python Environment
==============================
Python version               : 3.12.3 (main, Nov  6 2024, 18:32:19) [GCC 13.2.0] (64-bit runtime)
Python platform              : Linux-5.15.0-1053-nvidia-x86_64-with-glibc2.39

==============================
       CUDA / GPU Info
==============================
Is CUDA available            : True
CUDA runtime version         : 12.8.61
CUDA_MODULE_LOADING set to   : LAZY
GPU models and configuration : 
GPU 0: NVIDIA H800
GPU 1: NVIDIA H800
GPU 2: NVIDIA H800
GPU 3: NVIDIA H800
GPU 4: NVIDIA H800
GPU 5: NVIDIA H800
GPU 6: NVIDIA H800
GPU 7: NVIDIA H800

Nvidia driver version        : 535.161.08
cuDNN version                : Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.7.0
/usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.7.0
HIP runtime version          : N/A
MIOpen runtime version       : N/A
Is XNNPACK available         : True

==============================
          CPU Info
==============================
Architecture:                       x86_64
CPU op-mode(s):                     32-bit, 64-bit
Address sizes:                      52 bits physical, 57 bits virtual
Byte Order:                         Little Endian
CPU(s):                             224
On-line CPU(s) list:                0-223
Vendor ID:                          GenuineIntel
BIOS Vendor ID:                     Intel(R) Corporation
Model name:                         Intel(R) Xeon(R) Platinum 8480C
BIOS Model name:                    Intel(R) Xeon(R) Platinum 8480C  CPU @ 2.0GHz
BIOS CPU family:                    179
CPU family:                         6
Model:                              143
Thread(s) per core:                 2
Core(s) per socket:                 56
Socket(s):                          2
Stepping:                           8
CPU(s) scaling MHz:                 100%
CPU max MHz:                        3800.0000
CPU min MHz:                        800.0000
BogoMIPS:                           4000.00
Flags:                              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 invpcid_single intel_ppin cdp_l2 ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi 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 avx_vnni avx512_bf16 wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku ospke waitpkg avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq la57 rdpid bus_lock_detect cldemote movdiri movdir64b enqcmd fsrm md_clear serialize tsxldtrk pconfig arch_lbr amx_bf16 avx512_fp16 amx_tile amx_int8 flush_l1d arch_capabilities
Virtualization:                     VT-x
L1d cache:                          5.3 MiB (112 instances)
L1i cache:                          3.5 MiB (112 instances)
L2 cache:                           224 MiB (112 instances)
L3 cache:                           210 MiB (2 instances)
NUMA node(s):                       2
NUMA node0 CPU(s):                  0-55,112-167
NUMA node1 CPU(s):                  56-111,168-223
Vulnerability Gather data sampling: 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 Retbleed:             Not affected
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass:    Mitigation; Speculative Store Bypass disabled via prctl and seccomp
Vulnerability Spectre v1:           Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:           Mitigation; Enhanced IBRS, IBPB conditional, RSB filling, PBRSB-eIBRS SW sequence
Vulnerability Srbds:                Not affected
Vulnerability Tsx async abort:      Not affected

==============================
Versions of relevant libraries
==============================
[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-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-nccl-cu12==2.27.3
[pip3] nvidia-nvjitlink-cu12==12.8.93
[pip3] nvidia-nvtx-cu12==12.8.90
[pip3] pyzmq==27.1.0
[pip3] torch==2.8.0
[pip3] torchaudio==2.8.0
[pip3] torchvision==0.23.0
[pip3] transformers==4.56.1
[pip3] triton==3.4.0
[conda] Could not collect

==============================
         vLLM Info
==============================
ROCM Version                 : Could not collect
vLLM Version                 : 0.10.2rc3.dev1+g12a8414d8 (git sha: 12a8414d8)
vLLM Build Flags:
  CUDA Archs: 7.5 8.0 8.6 9.0 10.0 12.0+PTX; ROCm: Disabled
GPU Topology:
        GPU0    GPU1    GPU2    GPU3    GPU4    GPU5    GPU6    GPU7    NIC0    NIC1    NIC2    NIC3    NIC4    NIC5    NIC6    NIC7    NIC8    NIC9    NIC10    NIC11   CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NV8     NV8     NV8     NV8     NV8     NV8     NV8     PXB     NODE    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS      SYS     0-55,112-167    0               N/A
GPU1    NV8      X      NV8     NV8     NV8     NV8     NV8     NV8     NODE    NODE    NODE    PXB     NODE    NODE    SYS     SYS     SYS     SYS     SYS      SYS     0-55,112-167    0               N/A
GPU2    NV8     NV8      X      NV8     NV8     NV8     NV8     NV8     NODE    NODE    NODE    NODE    PXB     NODE    SYS     SYS     SYS     SYS     SYS      SYS     0-55,112-167    0               N/A
GPU3    NV8     NV8     NV8      X      NV8     NV8     NV8     NV8     NODE    NODE    NODE    NODE    NODE    PXB     SYS     SYS     SYS     SYS     SYS      SYS     0-55,112-167    0               N/A
GPU4    NV8     NV8     NV8     NV8      X      NV8     NV8     NV8     SYS     SYS     SYS     SYS     SYS     SYS     PXB     NODE    NODE    NODE    NODE     NODE    56-111,168-223  1               N/A
GPU5    NV8     NV8     NV8     NV8     NV8      X      NV8     NV8     SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    PXB     NODE     NODE    56-111,168-223  1               N/A
GPU6    NV8     NV8     NV8     NV8     NV8     NV8      X      NV8     SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    PXB      NODE    56-111,168-223  1               N/A
GPU7    NV8     NV8     NV8     NV8     NV8     NV8     NV8      X      SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    NODE     PXB     56-111,168-223  1               N/A
NIC0    PXB     NODE    NODE    NODE    SYS     SYS     SYS     SYS      X      NODE    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS      SYS
NIC1    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     NODE     X      PIX     NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS      SYS
NIC2    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     NODE    PIX      X      NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS      SYS
NIC3    NODE    PXB     NODE    NODE    SYS     SYS     SYS     SYS     NODE    NODE    NODE     X      NODE    NODE    SYS     SYS     SYS     SYS     SYS      SYS
NIC4    NODE    NODE    PXB     NODE    SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE     X      NODE    SYS     SYS     SYS     SYS     SYS      SYS
NIC5    NODE    NODE    NODE    PXB     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    NODE     X      SYS     SYS     SYS     SYS     SYS      SYS
NIC6    SYS     SYS     SYS     SYS     PXB     NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS      X      NODE    NODE    NODE    NODE     NODE
NIC7    SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     NODE     X      PIX     NODE    NODE     NODE
NIC8    SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     NODE    PIX      X      NODE    NODE     NODE
NIC9    SYS     SYS     SYS     SYS     NODE    PXB     NODE    NODE    SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE     X      NODE     NODE
NIC10   SYS     SYS     SYS     SYS     NODE    NODE    PXB     NODE    SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE     
X       NODE
NIC11   SYS     SYS     SYS     SYS     NODE    NODE    NODE    PXB     SYS     SYS     SYS     SYS     SYS     SYS     NODE    NODE    NODE    NODE    NODE      X

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

NIC Legend:

  NIC0: mlx5_0
  NIC1: mlx5_1
  NIC2: mlx5_2
  NIC3: mlx5_3
  NIC4: mlx5_4
  NIC5: mlx5_5
  NIC6: mlx5_6
  NIC7: mlx5_7
  NIC8: mlx5_8
  NIC9: mlx5_9
  NIC10: mlx5_10
  NIC11: mlx5_11

==============================
     Environment Variables
==============================
NVIDIA_VISIBLE_DEVICES=all
CUBLAS_VERSION=12.8.3.14
NVIDIA_REQUIRE_CUDA=cuda>=9.0
CUDA_CACHE_DISABLE=1
TORCH_CUDA_ARCH_LIST=7.5 8.0 8.6 9.0 10.0 12.0+PTX
NCCL_VERSION=2.25.1
NCCL_NVLS_ENABLE=0
NVIDIA_DRIVER_CAPABILITIES=compute,utility,video
TORCH_NCCL_USE_COMM_NONBLOCKING=0
NVIDIA_PRODUCT_NAME=PyTorch
CUDA_VERSION=12.8.0.038
PYTORCH_VERSION=2.6.0a0+ecf3bae
PYTORCH_BUILD_NUMBER=0
CUDNN_FRONTEND_VERSION=1.9.0
CUDNN_VERSION=9.7.0.66
PYTORCH_HOME=/opt/pytorch/pytorch
LD_LIBRARY_PATH=/usr/local/lib/python3.12/dist-packages/torch/lib:/usr/local/lib/python3.12/dist-packages/torch_tensorrt/lib:/usr/local/cuda/compat/lib:/usr/local/nvidia/lib:/usr/local/nvidia/lib64
NVIDIA_BUILD_ID=134983853
CUDA_DRIVER_VERSION=570.86.10
PYTORCH_BUILD_VERSION=2.6.0a0+ecf3bae
CUDA_HOME=/usr/local/cuda
CUDA_HOME=/usr/local/cuda
CUDA_MODULE_LOADING=LAZY
NVIDIA_REQUIRE_JETPACK_HOST_MOUNTS=
NVIDIA_PYTORCH_VERSION=25.01
TORCH_ALLOW_TF32_CUBLAS_OVERRIDE=1
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_root
VLLM_WORKER_MULTIPROC_METHOD=spawn

According recipes Guide Use MTP: [[Qwen3-Next]](https://docs.vllm.ai/projects/recipes/en/latest/Qwen/Qwen3-Next.html)

🐛 Describe the bug

uv venv
source .venv/bin/activate
uv pip install vllm --extra-index-url https://wheels.vllm.ai/nightly

vllm serve Qwen/Qwen3-Next-80B-A3B-Instruct  \
--tokenizer-mode auto  --gpu-memory-utilization 0.8 \
--speculative-config '{"method": "qwen3_next_mtp", "num_speculative_tokens": 2}' \
--tensor-parallel-size 4 --no-enable-chunked-prefill 2>&1 | tee ./test_mtp.log

vllm bench serve \
  --backend vllm \
  --model Qwen/Qwen3-Next-80B-A3B-Instruct \
  --endpoint /v1/completions \
  --dataset-name random \
  --random-input 2048 \
  --random-output 1024 \
  --max-concurrency 10 \
  --num-prompt 100

Error Log:

auto Use moe config
193 ^[[1;36m(Worker_TP0 pid=35966)^[[0;0m WARNING 09-12 06:46:35 [fused_moe.py:727] Using default MoE config. Performance might be sub-optimal! Config f     ile not found at ['/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/vllm/model_executor/layers/fused_moe/configs/E=512,N=128,device_name=NVIDIA_H800.json']

313 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/v1/executor/multiproc_executor.py", line 649, in worker_busy_loop^M
 314 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     output = func(*args, **kwargs)^M
 315 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^^M
 316 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/utils/_contextlib.py", line 120, in decorate_context^M
 317 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return func(*args, **kwargs)^M
 318 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^M
 319 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/v1/worker/gpu_worker.py", line 436, in execute_model^M
 320 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     output = self.model_runner.execute_model(scheduler_output     ,^M
 321 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^     ^^M
 322 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/utils/_contextlib.py", line 120, in decorate_context^M
 323 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return func(*args, **kwargs)^M
 324 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^M
 325 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/v1/worker/gpu_model_runner.py", line 2064, in execute_model^M
 326 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     model_output = self.model(^M
 327 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]                    ^^^^^^^^^^^^M
 328 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/compilation/cuda_graph.py", line 119, in __call__^M
 329 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self.runnable(*args, **kwargs)^M
 330 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 331 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl^M
 332 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)^M
 333 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 334 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1784, in _call_impl^M
 335 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)^M
 336 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 337 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/model_executor/models/qwen3_next.py", line 1165, in forward^M
 338 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     hidden_states = self.model(input_ids, positions, intermed     iate_tensors,^M
 339 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^     ^^^^^^^^^^^^^^M
 340 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/compilation/decorators.py", line 312, in __call__^M
 341 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     model_output = self.forward(*args, **kwargs)^M
 342 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 343 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/model_executor/models/qwen3_next.py", line 926, in forward^M
 344 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     def forward(^M
 345 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/_dynamo/eval_frame.py", line 375, in __call__^M
 346 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return super().__call__(*args, **kwargs)^M
 347 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 348 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl^M
^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)^M
 350 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 351 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1784, in _call_impl^M
 352 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)^M
 353 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 354 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/_dynamo/eval_frame.py", line 929, in _fn^M
 355 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return fn(*args, **kwargs)^M
 356 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^M
 357 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/fx/graph_module.py", line 848, in call_wrapped^M
 358 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)^M       
 359 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M       
 360 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/fx/graph_module.py", line 424, in __call__^M
 361 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     raise e^M
 362 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/fx/graph_module.py", line 411, in __call__^M
 363 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # 
     type: ignore[misc]^M
 364 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M  
 365 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl^M
 366 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)^M
 367 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 368 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1784, in _call_impl^M
 369 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)^M
 370 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 371 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "<eval_with_key>.98", line 374, in forward^M
 372 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     submod_9 = self.submod_9(getitem_16, s72, getitem_17);  g     etitem_16 = submod_9 = None^M
 373 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M  
 374 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/fx/graph_module.py", line 848, in call_wrapped^M
 375 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)^M       
 376 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M       
 377 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/fx/graph_module.py", line 424, in __call__^M
 378 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     raise e^M
 379 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/fx/graph_module.py", line 411, in __call__^M
 380 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # 
     type: ignore[misc]^M
 381 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M  
 382 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl^M
 383 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)^M
 384 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 385 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/nn/modules/module.py", line 1784, in _call_impl^M
 386 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)^M
 387 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "<eval_with_key>.10", line 5, in forward^M
 389 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     gdn_attention = torch.ops.vllm.gdn_attention(x_55, self_a     ttention_output_4, 'model.layers.4.linear_attn');  x_55 = self_attention_output_4 = gdn_attention = None^M
 390 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^M
 391 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/torch/_ops.py", line 1243, in __call__^M
 392 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     return self._op(*args, **kwargs)^M
 393 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^M
 394 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/model_executor/models/qwen3_next.py", line 1225, in gdn_attention^M
 395 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     self._forward(hidden_states=hidden_states, output=output)     ^M
 396 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site     -packages/vllm/model_executor/models/qwen3_next.py", line 460, in _forward^M
 397 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]     mixed_qkv_spec = mixed_qkv_spec.view(^M
 398 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654]                      ^^^^^^^^^^^^^^^^^^^^^M
 399 ^[[1;36m(Worker_TP1 pid=35967)^[[0;0m ERROR 09-12 06:47:40 [multiproc_executor.py:654] RuntimeError: shape '[5, -1, 2048]' is invalid for input of size 104448^M

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]

@sighingnow @jeejeelee

[jeejeelee]

Try reproducing now

[jeejeelee]

I cannot reproduce the error you mentioned, and I can get normal benchmark results. I'm using vllm commit 7920de0a2af7c663efae29856f1bac061a553b58, and test on H20*4

[chaunceyjiang]

Using the same command as above, I encounter a new error. On 4 * NVIDIA H100 80GB HBM3

(APIServer pid=62285) INFO:     127.0.0.1:39784 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39756 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39804 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39812 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39822 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39796 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:33648 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39834 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39796 - "POST /v1/completions HTTP/1.1" 200 OK
(APIServer pid=62285) INFO:     127.0.0.1:39844 - "POST /v1/completions HTTP/1.1" 200 OK
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654] WorkerProc hit an exception.
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654] Traceback (most recent call last):
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/v1/executor/multiproc_executor.py", line 649, in worker_busy_loop
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     output = func(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return func(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/v1/worker/gpu_worker.py", line 436, in execute_model
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     output = self.model_runner.execute_model(scheduler_output,
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return func(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/v1/worker/gpu_model_runner.py", line 2064, in execute_model
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     model_output = self.model(
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]                    ^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/compilation/cuda_graph.py", line 119, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self.runnable(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 1165, in forward
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     hidden_states = self.model(input_ids, positions, intermediate_tensors,
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/compilation/decorators.py", line 312, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     model_output = self.forward(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 926, in forward
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     def forward(
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py", line 375, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return super().__call__(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py", line 929, in _fn
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return fn(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 848, in call_wrapped
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 424, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     raise e
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 411, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # type: ignore[misc]
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "<eval_with_key>.98", line 374, in forward
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     submod_9 = self.submod_9(getitem_16, s72, getitem_17);  getitem_16 = submod_9 = None
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 848, in call_wrapped
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 424, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     raise e
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 411, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # type: ignore[misc]
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "<eval_with_key>.10", line 5, in forward
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     gdn_attention = torch.ops.vllm.gdn_attention(x_55, self_attention_output_4, 'model.layers.4.linear_attn');  x_55 = self_attention_output_4 = gdn_attention = None
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/_ops.py", line 1243, in __call__
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     return self._op(*args, **kwargs)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 1225, in gdn_attention
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     self._forward(hidden_states=hidden_states, output=output)
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 603, in _forward
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     core_attn_out[:, spec_token_masks] = core_attn_out_spec
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654]     ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654] torch.AcceleratorError: CUDA error: an illegal memory access was encountered
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654] CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
(Worker_TP1 pid=63193) ERROR 09-12 09:27:01 [multiproc_executor.py:654] For debugging consider passing CUDA_LAUNCH_BLOCKING=1

[david6666666]

I cannot reproduce the error you mentioned, and I can get normal benchmark results. I'm using vllm commit 7920de0a2af7c663efae29856f1bac061a553b58, and test on H20*4

still ocuur issue , similar as @chaunceyjiang just encountered error. @jeejeelee

(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654] Traceback (most recent call last):
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/v1/executor/multiproc_executor.py", line 649, in worker_busy_loop
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     output = func(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return func(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/v1/worker/gpu_worker.py", line 436, in execute_model
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     output = self.model_runner.execute_model(scheduler_output,
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return func(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/v1/worker/gpu_model_runner.py", line 2064, in execute_model
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     model_output = self.model(
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                    ^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/compilation/cuda_graph.py", line 119, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self.runnable(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/model_executor/models/qwen3_next.py", line 1169, in forward
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     hidden_states = self.model(input_ids, positions, intermediate_tensors,       
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^       
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/compilation/decorators.py", line 312, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     model_output = self.forward(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/model_executor/models/qwen3_next.py", line 930, in forward
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     def forward(
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py", line 375, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return super().__call__(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py", line 929, in _fn
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return fn(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 848, in call_wrapped
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 424, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     raise e
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 411, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # type: ignore[misc]
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 424, in __call__
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     raise e
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 411, in __call__
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # type: ignore[misc]  
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "<eval_with_key>.10", line 5, in forward
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     gdn_attention = torch.ops.vllm.gdn_attention(x_55, self_attention_output_4, 'model.layers.4.linear_attn');  x_55 = self_attention_output_4 = gdn_attention = None
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/_ops.py", line 1243, in __call__
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._op(*args, **kwargs)
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/model_executor/models/qwen3_next.py", line 1229, in gdn_attention
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     self._forward(hidden_states=hidden_states, output=output)
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/model_executor/models/qwen3_next.py", line 453, in _forward
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     mixed_qkv_spec = mixed_qkv[spec_token_masks]
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                      ~~~~~~~~~^^^^^^^^^^^^^^^^^^
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654] torch.AcceleratorError: CUDA error: an illegal memory access was encountered     
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654] CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654] For debugging consider passing CUDA_LAUNCH_BLOCKING=1
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654] Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654]
(Worker_TP0 pid=48097) ERROR 09-12 09:40:37 [multiproc_executor.py:654] 
ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "<eval_with_key>.98", line 374, in forward
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     submod_9 = self.submod_9(getitem_16, s72, getitem_17);  getitem_16 = submod_9 = None
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 848, in call_wrapped
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 424, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     raise e
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/fx/graph_module.py", line 411, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # type: ignore[misc]  
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "<eval_with_key>.10", line 5, in forward
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     gdn_attention = torch.ops.vllm.gdn_attention(x_55, self_attention_output_4, 'model.layers.4.linear_attn');  x_55 = self_attention_output_4 = gdn_attention = None
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/.venv/lib/python3.12/site-packages/torch/_ops.py", line 1243, in __call__
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     return self._op(*args, **kwargs)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/model_executor/models/qwen3_next.py", line 1229, in gdn_attention
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     self._forward(hidden_states=hidden_states, output=output)
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]   File "/root/.cache/cwq/qwen3-next/vllm/vllm/model_executor/models/qwen3_next.py", line 453, in _forward
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]     mixed_qkv_spec = mixed_qkv[spec_token_masks]
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]                      ~~~~~~~~~^^^^^^^^^^^^^^^^^^
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654] torch.AcceleratorError: CUDA error: an illegal memory access was encountered     
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654] CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654] For debugging consider passing CUDA_LAUNCH_BLOCKING=1
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654] Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
(Worker_TP3 pid=48100) ERROR 09-12 09:40:37 [multiproc_executor.py:654]

[jeejeelee]

Have you tried the latest main branch?

[david6666666]

Yes still occur issue

[chaunceyjiang]

Yes still occur issue

Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654] Traceback (most recent call last):
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/v1/executor/multiproc_executor.py", line 649, in worker_busy_loop
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     output = func(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return func(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/v1/worker/gpu_worker.py", line 436, in execute_model
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     output = self.model_runner.execute_model(scheduler_output,
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return func(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/v1/worker/gpu_model_runner.py", line 2064, in execute_model
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     model_output = self.model(
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]                    ^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/compilation/cuda_graph.py", line 119, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self.runnable(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 1165, in forward
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     hidden_states = self.model(input_ids, positions, intermediate_tensors,
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/compilation/decorators.py", line 312, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     model_output = self.forward(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 926, in forward
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     def forward(
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py", line 375, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return super().__call__(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/_dynamo/eval_frame.py", line 929, in _fn
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return fn(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 848, in call_wrapped
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 424, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     raise e
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 411, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # type: ignore[misc]
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "<eval_with_key>.98", line 451, in forward
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     submod_35 = self.submod_35(getitem_67, s72, getitem_68);  getitem_67 = submod_35 = None
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 848, in call_wrapped
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self._wrapped_call(self, *args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 424, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     raise e
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/fx/graph_module.py", line 411, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return super(self.cls, obj).__call__(*args, **kwargs)  # type: ignore[misc]
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self._call_impl(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return forward_call(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "<eval_with_key>.36", line 5, in forward
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     gdn_attention = torch.ops.vllm.gdn_attention(x_222, self_attention_output_17, 'model.layers.17.linear_attn');  x_222 = self_attention_output_17 = gdn_attention = None
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/opt/conda/envs/vllm/lib/python3.12/site-packages/torch/_ops.py", line 1243, in __call__
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     return self._op(*args, **kwargs)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]            ^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 1225, in gdn_attention
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     self._forward(hidden_states=hidden_states, output=output)
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]   File "/home/jovyan/vllm/vllm/model_executor/models/qwen3_next.py", line 460, in _forward
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]     mixed_qkv_spec = mixed_qkv_spec.view(
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654]                      ^^^^^^^^^^^^^^^^^^^^
(Worker_TP1 pid=102986) ERROR 09-12 10:48:18 [multiproc_executor.py:654] RuntimeError: shape '[9, -1, 2048]' is invalid for input of size 227328

[JaheimLee]

I encountered another error using 4x 3090 and fp8 quantization model(DevQuasar/Qwen.Qwen3-Next-80B-A3B-Instruct-FP8-Dynamic). I found this pr but both qwen3_next.py and qwen3_next_mtp.py have packed_modules_mapping. So maybe it's related to this issue?

(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600] WorkerProc failed to start.
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600] Traceback (most recent call last):
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/v1/executor/multiproc_executor.py", line 574, in worker_main
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     worker = WorkerProc(*args, **kwargs)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/v1/executor/multiproc_executor.py", line 440, in __init__
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     self.worker.load_model()
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     ~~~~~~~~~~~~~~~~~~~~~~^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/v1/worker/gpu_worker.py", line 213, in load_model
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     self.model_runner.load_model(eep_scale_up=eep_scale_up)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/v1/worker/gpu_model_runner.py", line 2387, in load_model
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     self.model = model_loader.load_model(
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]                  ~~~~~~~~~~~~~~~~~~~~~~~^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]         vllm_config=self.vllm_config, model_config=self.model_config)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/model_executor/model_loader/base_loader.py", line 51, in load_model
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     process_weights_after_loading(model, model_config, target_device)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/model_executor/model_loader/utils.py", line 113, in process_weights_after_loading
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     quant_method.process_weights_after_loading(module)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors.py", line 716, in process_weights_after_loading
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     layer.scheme.process_weights_after_loading(layer)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a16_fp8.py", line 59, in process_weights_after_loading
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     prepare_fp8_layer_for_marlin(layer)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/model_executor/layers/quantization/utils/marlin_utils_fp8.py", line 107, in prepare_fp8_layer_for_marlin
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     marlin_qweight = ops.gptq_marlin_repack(b_q_weight=qweight,
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]                                             perm=perm,
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]                                             size_k=part_size_k,
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]                                             size_n=part_size_n,
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]                                             num_bits=8)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/vllm/_custom_ops.py", line 955, in gptq_marlin_repack
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     return torch.ops._C.gptq_marlin_repack(b_q_weight, perm, size_k, size_n,
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]            ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]                                            num_bits)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]                                            ^^^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]   File "/data/lijinghui/uv_projects/.venv/lib/python3.13/site-packages/torch/_ops.py", line 1243, in __call__
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]     return self._op(*args, **kwargs)
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600]            ~~~~~~~~^^^^^^^^^^^^^^^^^
(Worker_TP3_EP3 pid=2136744) ERROR 09-15 09:14:32 [multiproc_executor.py:600] RuntimeError: size_n = 3088 is not divisible by tile_n_size = 64

[sighingnow]

@david6666666 @chaunceyjiang The MTP + guided decoding issue has been fixed by: #24957

Could you please try to merge this PR and take a try again?