Your current environment

============================== System Info

OS : Ubuntu 22.04.5 LTS (x86_64) GCC version : (Ubuntu 11.4.0-1ubuntu1~22.04.3) 11.4.0 Clang version : Could not collect CMake version : Could not collect Libc version : glibc-2.35

============================== PyTorch Info

PyTorch version : 2.11.0+cu130 Is debug build : False CUDA used to build PyTorch : 13.0 ROCM used to build PyTorch : N/A XPU used to build PyTorch : N/A

============================== Python Environment

Python version : 3.12.13 (main, May 3 2026, 03:30:16) [GCC 11.4.0] (64-bit runtime) Python platform : Linux-6.19.9-arch1-1-x86_64-with-glibc2.35

============================== 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 GeForce RTX 3060 GPU 1: NVIDIA GeForce RTX 3060 GPU 2: NVIDIA GeForce RTX 3060

Nvidia driver version : 595.58.03 cuDNN version : Could not collect 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: 43 bits physical, 48 bits virtual Byte Order: Little Endian CPU(s): 64 On-line CPU(s) list: 0-63 Vendor ID: AuthenticAMD Model name: AMD EPYC 7551P 32-Core Processor CPU family: 23 Model: 1 Thread(s) per core: 2 Core(s) per socket: 32 Socket(s): 1 Stepping: 2 Frequency boost: enabled CPU max MHz: 2000.0000 CPU min MHz: 1200.0000 BogoMIPS: 3992.40 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid amd_dcm aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb hw_pstate ssbd ibpb vmmcall fsgsbase bmi1 avx2 smep bmi2 rdseed adx smap clflushopt sha_ni xsaveopt xsavec xgetbv1 clzero xsaveerptr arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif overflow_recov succor smca sev sev_es Virtualization: AMD-V L1d cache: 1 MiB (32 instances) L1i cache: 2 MiB (32 instances) L2 cache: 16 MiB (32 instances) L3 cache: 64 MiB (8 instances) NUMA node(s): 4 NUMA node0 CPU(s): 0-7,32-39 NUMA node1 CPU(s): 8-15,40-47 NUMA node2 CPU(s): 16-23,48-55 NUMA node3 CPU(s): 24-31,56-63 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 Old microcode: Not affected Vulnerability Reg file data sampling: Not affected Vulnerability Retbleed: Mitigation; untrained return thunk; SMT vulnerable Vulnerability Spec rstack overflow: Mitigation; Safe RET 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; Retpolines; IBPB conditional; STIBP disabled; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected 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==0.6.8.post1 [pip3] numpy==2.2.6 [pip3] nvidia-cublas==13.1.0.3 [pip3] nvidia-cuda-cupti==13.0.85 [pip3] nvidia-cuda-nvrtc==13.0.88 [pip3] nvidia-cuda-runtime==13.0.96 [pip3] nvidia-cudnn-cu13==9.19.0.56 [pip3] nvidia-cudnn-frontend==1.18.0 [pip3] nvidia-cufft==12.0.0.61 [pip3] nvidia-cufile==1.15.1.6 [pip3] nvidia-curand==10.4.0.35 [pip3] nvidia-cusolver==12.0.4.66 [pip3] nvidia-cusparse==12.6.3.3 [pip3] nvidia-cusparselt-cu13==0.8.0 [pip3] nvidia-cutlass-dsl==4.4.2 [pip3] nvidia-cutlass-dsl-libs-base==4.4.2 [pip3] nvidia-ml-py==13.595.45 [pip3] nvidia-nccl-cu13==2.28.9 [pip3] nvidia-nvjitlink==13.0.88 [pip3] nvidia-nvshmem-cu13==3.4.5 [pip3] nvidia-nvtx==13.0.85 [pip3] pyzmq==27.1.0 [pip3] torch==2.11.0+cu130 [pip3] torch_c_dlpack_ext==0.1.5 [pip3] torchaudio==2.11.0+cu130 [pip3] torchvision==0.26.0+cu130 [pip3] transformers==5.7.0 [pip3] triton==3.6.0 [conda] Could not collect

============================== vLLM Info

ROCM Version : Could not collect vLLM Version : 0.20.1 vLLM Build Flags: CUDA Archs: 7.5 8.0 8.6 8.9 9.0 10.0 12.0+PTX; ROCm: Disabled; XPU: Disabled GPU Topology: GPU0 GPU1 GPU2 CPU Affinity NUMA Affinity GPU NUMA ID GPU0 X PHB SYS 8-15,40-47 1 N/A GPU1 PHB X SYS 8-15,40-47 1 N/A GPU2 SYS SYS X 16-23,48-55 2 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

VLLM_USE_SIMPLE_KV_OFFLOAD=1 VLLM_PP_LAYER_PARTITION=24,23,17 PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True NVIDIA_REQUIRE_CUDA=cuda>=13.0 brand=unknown,driver>=535,driver<536 brand=grid,driver>=535,driver<536 brand=tesla,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=vapps,driver>=535,driver<536 brand=vpc,driver>=535,driver<536 brand=vcs,driver>=535,driver<536 brand=vws,driver>=535,driver<536 brand=cloudgaming,driver>=535,driver<536 brand=unknown,driver>=550,driver<551 brand=grid,driver>=550,driver<551 brand=tesla,driver>=550,driver<551 brand=nvidia,driver>=550,driver<551 brand=quadro,driver>=550,driver<551 brand=quadrortx,driver>=550,driver<551 brand=nvidiartx,driver>=550,driver<551 brand=vapps,driver>=550,driver<551 brand=vpc,driver>=550,driver<551 brand=vcs,driver>=550,driver<551 brand=vws,driver>=550,driver<551 brand=cloudgaming,driver>=550,driver<551 brand=unknown,driver>=565,driver<566 brand=grid,driver>=565,driver<566 brand=tesla,driver>=565,driver<566 brand=nvidia,driver>=565,driver<566 brand=quadro,driver>=565,driver<566 brand=quadrortx,driver>=565,driver<566brand=nvidiartx,driver>=565,driver<566 brand=vapps,driver>=565,driver<566 brand=vpc,driver>=565,driver<566 brand=vcs,driver>=565,driver<566 brand=vws,driver>=565,driver<566 brand=cloudgaming,driver>=565,driver<566 brand=unknown,driver>=570,driver<571 brand=grid,driver>=570,driver<571 brand=tesla,driver>=570,driver<571 brand=nvidia,driver>=570,driver<571 brand=quadro,driver>=570,driver<571 brand=quadrortx,driver>=570,driver<571 brand=nvidiartx,driver>=570,driver<571 brand=vapps,driver>=570,driver<571 brand=vpc,driver>=570,driver<571 brand=vcs,driver>=570,driver<571 brand=vws,driver>=570,driver<571 brand=cloudgaming,driver>=570,driver<571 brand=unknown,driver>=575,driver<576 brand=grid,driver>=575,driver<576 brand=tesla,driver>=575,driver<576 brand=nvidia,driver>=575,driver<576 brand=quadro,driver>=575,driver<576 brand=quadrortx,driver>=575,driver<576 brand=nvidiartx,driver>=575,driver<576 brand=vapps,driver>=575,driver<576 brand=vpc,driver>=575,driver<576 brand=vcs,driver>=575,driver<576 brand=vws,driver>=575,driver<576 brand=cloudgaming,driver>=575,driver<576 CUDA_VERSION=13.0.2 LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64 NVIDIA_VISIBLE_DEVICES=all NVIDIA_DRIVER_CAPABILITIES=compute,utility VLLM_ENABLE_CUDA_COMPATIBILITY=0 TORCH_CUDA_ARCH_LIST=7.5 8.0 8.6 8.9 9.0 10.0 12.0+PTX VLLM_USAGE_SOURCE=production-docker-image PYTORCH_NVML_BASED_CUDA_CHECK=1 TORCHINDUCTOR_COMPILE_THREADS=1 TORCHINDUCTOR_CACHE_DIR=/tmp/torchinductor_root

🐛 Describe the bug

When using pipeline parallelism (PP > 1) with KV offloading and async scheduling enabled (the default since v0.14.0), the last PP rank crashes with an AssertionError on any request beyond a few thousand tokens.

The error:

(Worker_PP2 pid=617) AssertionError
  File ".../vllm/v1/worker/gpu_model_runner.py", line 4374, in _pp_receive_prev_sampled_token_ids_to_input_batch
    assert not pp.is_last_rank

Full traceback:

(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962] WorkerProc hit an exception.
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962] Traceback (most recent call last):
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   File ".../vllm/v1/executor/multiproc_executor.py", line 944, in worker_busy_loop
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]     output = func(*args, **kwargs)
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   File ".../vllm/v1/worker/worker_base.py", line 332, in execute_model
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]     return self.worker.execute_model(scheduler_output)
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   File ".../vllm/v1/worker/gpu_worker.py", line 803, in execute_model
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]     output = self.model_runner.execute_model(
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   File ".../vllm/v1/worker/gpu_model_runner.py", line 4034, in execute_model
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]     model_output = self._model_forward(
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   [... model forward succeeds ...]
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   File ".../vllm/v1/worker/gpu_worker.py", line 740, in sample_tokens
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]     return self.model_runner.sample_tokens(grammar_output)
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   File ".../vllm/v1/worker/gpu_model_runner.py", line 4131, in sample_tokens
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]     self._pp_receive_prev_sampled_token_ids_to_input_batch()
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]   File ".../vllm/v1/worker/gpu_model_runner.py", line 4374, in _pp_receive_prev_sampled_token_ids_to_input_batch
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962]     assert not pp.is_last_rank
(Worker_PP2 pid=617) ERROR [multiproc_executor.py:962] AssertionError

Root cause

In sample_tokens(), the early-return path (when execute_model_state is None) calls _pp_receive_prev_sampled_token_ids_to_input_batch() without checking is_last_rank:

# gpu_model_runner.py, line ~4131
if self.use_async_scheduling and get_pp_group().world_size > 1:
    self._pp_receive_prev_sampled_token_ids_to_input_batch()

The function asserts not pp.is_last_rank, intended only for non-last ranks:

def _pp_receive_prev_sampled_token_ids_to_input_batch(self) -> None:
    """Receive sampled token ids broadcast from last PP stage"""
    pp = get_pp_group()
    assert not pp.is_last_rank  # <-- last rank hits this
    ...

With KV offloading, the connector can return a no-op output on the last PP rank, setting execute_model_state = None. This pushes the last rank into the early-return path where it hits the assertion. Without KV offloading, the last rank always has model execution results and takes the normal sampling path (which correctly checks is_last_rank at line ~4137).

To reproduce

vllm serve groxaxo/Qwen3.6-27B-GPTQ-Pro-4bit \
  --quantization gptq_marlin \
  --pipeline-parallel-size 3 \
  --kv-offloading-backend native \
  --kv-offloading-size 16 \
  --enable-prefix-caching \
  --mamba-cache-mode align \
  --max-model-len 172480 \
  --gpu-memory-utilization 0.95 \
  --language-model-only \
  --enforce-eager \
  --kv-cache-dtype fp8

Then send any request exceeding ~5K tokens:

import openai

client = openai.OpenAI(api_key="test", base_url="http://localhost:8000/v1")

# This works (short request):
response = client.chat.completions.create(
    model="groxaxo/Qwen3.6-27B-GPTQ-Pro-4bit",
    messages=[{"role": "user", "content": "What is 2+2?"}],
    max_tokens=100,
)

# This crashes (long request triggers KV offloading on last PP rank):
response = client.chat.completions.create(
    model="groxaxo/Qwen3.6-27B-GPTQ-Pro-4bit",
    messages=[{"role": "user", "content": "x " * 6000 + "\nWhat is 2+2?"}],
    max_tokens=100,
)

Short requests (under ~2K tokens) succeed because the connector does not produce a no-op. Longer requests trigger the offloading path on the last PP rank and crash.

Workaround

--no-async-scheduling

Proposed fix

Add an is_last_rank guard at the call site:

if self.use_async_scheduling and get_pp_group().world_size > 1 and not get_pp_group().is_last_rank:
    self._pp_receive_prev_sampled_token_ids_to_input_batch()

Regression context

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.