[Bug]: Exception: Invalid prefix encountered

[Cui-Peng-624]

Your current environment

(vllm) root@autodl-container-f0a8449f5d-af7706f2:/home/cuipeng/Project01# python collect_env.py
INFO 04-30 13:47:51 [__init__.py:239] Automatically detected platform cuda.
Collecting environment information...
PyTorch version: 2.6.0+cu124
Is debug build: False
CUDA used to build PyTorch: 12.4
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.4 LTS (x86_64)
GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
Clang version: Could not collect
CMake version: version 3.22.1
Libc version: glibc-2.35

Python version: 3.12.9 | packaged by Anaconda, Inc. | (main, Feb  6 2025, 18:56:27) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.15.0-94-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: 12.4.131
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: 
GPU 0: NVIDIA vGPU-48GB
GPU 1: NVIDIA vGPU-48GB
GPU 2: NVIDIA vGPU-48GB
GPU 3: NVIDIA vGPU-48GB
GPU 4: NVIDIA vGPU-48GB
GPU 5: NVIDIA vGPU-48GB
GPU 6: NVIDIA vGPU-48GB
GPU 7: NVIDIA vGPU-48GB

Nvidia driver version: 570.124.04
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.9.1.0
/usr/lib/x86_64-linux-gnu/libcudnn_adv.so.9.1.0
/usr/lib/x86_64-linux-gnu/libcudnn_cnn.so.9.1.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_precompiled.so.9.1.0
/usr/lib/x86_64-linux-gnu/libcudnn_engines_runtime_compiled.so.9.1.0
/usr/lib/x86_64-linux-gnu/libcudnn_graph.so.9.1.0
/usr/lib/x86_64-linux-gnu/libcudnn_heuristic.so.9.1.0
/usr/lib/x86_64-linux-gnu/libcudnn_ops.so.9.1.0
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
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
Model name:                         Intel(R) Xeon(R) Platinum 8481C
CPU family:                         6
Model:                              143
Thread(s) per core:                 2
Core(s) per socket:                 56
Socket(s):                          2
Stepping:                           8
Frequency boost:                    enabled
CPU max MHz:                        2001.0000
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 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 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==2.2.5
[pip3] nvidia-cublas-cu12==12.4.5.8
[pip3] nvidia-cuda-cupti-cu12==12.4.127
[pip3] nvidia-cuda-nvrtc-cu12==12.4.127
[pip3] nvidia-cuda-runtime-cu12==12.4.127
[pip3] nvidia-cudnn-cu12==9.1.0.70
[pip3] nvidia-cufft-cu12==11.2.1.3
[pip3] nvidia-curand-cu12==10.3.5.147
[pip3] nvidia-cusolver-cu12==11.6.1.9
[pip3] nvidia-cusparse-cu12==12.3.1.170
[pip3] nvidia-cusparselt-cu12==0.6.2
[pip3] nvidia-nccl-cu12==2.21.5
[pip3] nvidia-nvjitlink-cu12==12.4.127
[pip3] nvidia-nvtx-cu12==12.4.127
[pip3] pyzmq==26.4.0
[pip3] torch==2.6.0
[pip3] torchaudio==2.6.0
[pip3] torchvision==0.21.0
[pip3] transformers==4.51.3
[pip3] triton==3.2.0
[conda] numpy                     2.2.5                    pypi_0    pypi
[conda] nvidia-cublas-cu12        12.4.5.8                 pypi_0    pypi
[conda] nvidia-cuda-cupti-cu12    12.4.127                 pypi_0    pypi
[conda] nvidia-cuda-nvrtc-cu12    12.4.127                 pypi_0    pypi
[conda] nvidia-cuda-runtime-cu12  12.4.127                 pypi_0    pypi
[conda] nvidia-cudnn-cu12         9.1.0.70                 pypi_0    pypi
[conda] nvidia-cufft-cu12         11.2.1.3                 pypi_0    pypi
[conda] nvidia-curand-cu12        10.3.5.147               pypi_0    pypi
[conda] nvidia-cusolver-cu12      11.6.1.9                 pypi_0    pypi
[conda] nvidia-cusparse-cu12      12.3.1.170               pypi_0    pypi
[conda] nvidia-cusparselt-cu12    0.6.2                    pypi_0    pypi
[conda] nvidia-nccl-cu12          2.21.5                   pypi_0    pypi
[conda] nvidia-nvjitlink-cu12     12.4.127                 pypi_0    pypi
[conda] nvidia-nvtx-cu12          12.4.127                 pypi_0    pypi
[conda] pyzmq                     26.4.0                   pypi_0    pypi
[conda] torch                     2.6.0                    pypi_0    pypi
[conda] torchaudio                2.6.0                    pypi_0    pypi
[conda] torchvision               0.21.0                   pypi_0    pypi
[conda] transformers              4.51.3                   pypi_0    pypi
[conda] triton                    3.2.0                    pypi_0    pypi
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.8.5
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
        GPU0    GPU1    GPU2    GPU3    GPU4    GPU5    GPU6    GPU7    NIC0    NIC1    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NODE    NODE    NODE    SYS     SYS     SYS     SYS     NODE    NODE    0-55,112-167    0               N/A
GPU1    NODE     X      NODE    NODE    SYS     SYS     SYS     SYS     NODE    NODE    0-55,112-167    0               N/A
GPU2    NODE    NODE     X      NODE    SYS     SYS     SYS     SYS     NODE    NODE    0-55,112-167    0               N/A
GPU3    NODE    NODE    NODE     X      SYS     SYS     SYS     SYS     NODE    NODE    0-55,112-167    0               N/A
GPU4    SYS     SYS     SYS     SYS      X      NODE    NODE    NODE    SYS     SYS     56-111,168-223  1               N/A
GPU5    SYS     SYS     SYS     SYS     NODE     X      NODE    NODE    SYS     SYS     56-111,168-223  1               N/A
GPU6    SYS     SYS     SYS     SYS     NODE    NODE     X      NODE    SYS     SYS     56-111,168-223  1               N/A
GPU7    SYS     SYS     SYS     SYS     NODE    NODE    NODE     X      SYS     SYS     56-111,168-223  1               N/A
NIC0    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS      X      PIX
NIC1    NODE    NODE    NODE    NODE    SYS     SYS     SYS     SYS     PIX      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

OMP_NUM_THREADS=200
MKL_NUM_THREADS=200
NCCL_CUMEM_ENABLE=0
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
CUDA_MODULE_LOADING=LAZY

🐛 Describe the bug

当我运行如下代码时：

outputs = llm.generate(
prompts=batch_prompt_list, # Pass the list of prompts
sampling_params=sampling_params,
use_tqdm=True # Enable tqdm for batch processing (optional, set False if not desired)
)
报错，报错信息如下：
Processing Batch: 7/104
Processed prompts:   3%|▎         | 170/5000 [00:08<03:58, 20.26it/s, est. speed input: 506.79 toks/s, output: 1667.23 toks/s]Traceback (most recent call last):
  File "/tmp/ipykernel_59475/3501300393.py", line 15, in <module>
    outputs = llm.generate(
              ^^^^^^^^^^^^^
  File "/root/autodl-tmp/conda/envs/vllm/lib/python3.12/site-packages/vllm/utils.py", line 1196, in inner
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/root/autodl-tmp/conda/envs/vllm/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 473, in generate
    outputs = self._run_engine(use_tqdm=use_tqdm)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/autodl-tmp/conda/envs/vllm/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 1423, in _run_engine
    step_outputs = self.llm_engine.step()
                   ^^^^^^^^^^^^^^^^^^^^^^
  File "/root/autodl-tmp/conda/envs/vllm/lib/python3.12/site-packages/vllm/v1/engine/llm_engine.py", line 221, in step
    processed_outputs = self.output_processor.process_outputs(
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/autodl-tmp/conda/envs/vllm/lib/python3.12/site-packages/vllm/v1/engine/output_processor.py", line 344, in process_outputs
    stop_string = req_state.detokenizer.update(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/autodl-tmp/conda/envs/vllm/lib/python3.12/site-packages/vllm/v1/engine/detokenizer.py", line 106, in update
    self.output_text += self.decode_next(new_token_id)
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/autodl-tmp/conda/envs/vllm/lib/python3.12/site-packages/vllm/v1/engine/detokenizer.py", line 201, in decode_next
    token = self.stream.step(self.tokenizer, next_token_id)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Exception: Invalid prefix encountered
Processed prompts:   3%|▎         | 170/5000 [00:09<04:20, 18.53it/s, est. speed input: 506.79 toks/s, output: 1667.23 toks/s]
!!!!!!!! An exception occurred during llm.generate() for batch 7 !!!!!!!!
Error Type: <class 'Exception'>
Error Details: Invalid prefix encountered
似乎只llm.generate的内部错误。

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.

[BooMinSeong]

I have a same issue, did you solve the problem?

[shintaro329]

pip install tokenizers==0.21.0 seems to be a solution

[bpopeters]

pip install tokenizers==0.21.0 seems to be a solution

I have encountered the same bug when decoding into Chinese and Korean with this mistral-based model. This solution appears to work for me.

[Cui-Peng-624]

vllm = v0.8.4
tokenizers==0.21.0
can solve this problem.

[cieske]

vllm = v0.8.4 tokenizers==0.21.0 can solve this problem.

Meet same error when running MITRE from CyberSecEval using Phi-3.5-MoE-Instruct w/ vllm==v0.8.5, downgrade to these version works

[fpaupier]

Meeting same error with Google Gemma3-1b-it on vllm==0.8.5.post1, which is incompatible with tokenizers==0.21.0
@DarkLight1337 could we re-open this issue ?

[DarkLight1337]

cc @njhill who upgraded the dependency

[DarkLight1337]

I think this should be opened as a separate issue since it's not really a bug.

[Cppowboy]

vllm = v0.8.4 tokenizers==0.21.0 can solve this problem.

It works for me to use tokenizers==0.21.0 and vllm==0.8.5.post1.

[njhill]

This does appear to be a legitimate unicode-related issue with the new detokenization. It would be very helpful if someone could provide a model and precise input parameters + prompt which triggers this!

[fpaupier]

Hello @njhill
Here are reproduce steps for the error

Host

• single nvidia GPU

Running env:

• vLLM engine v1 activated
• Python 3.12.3
• vllm==0.8.5.post1
• tokenizers==0.21.1

Mode used

• google/gemma-3-1b-it - https://huggingface.co/google/gemma-3-1b-it

To replicate the bug, you can run this snippet https://gist.github.com/fpaupier/0ed1375bd7633c5be6c894b1c7ac1be3
Make sure to set HF_TOKEN in your shell running this script to access the gated model

You will see an exception raised by the detokenizer in the V1 engine

vllm/vllm/v1/engine/detokenizer.py

Line 200 in d981396

def decode_next(self, next_token_id: int) -> str:

Complete log trace observed with the provided snippet.

INFO 05-20 07:40:35 [gpu_model_runner.py:1347] Model loading took 1.9150 GiB and 0.380501 seconds
INFO 05-20 07:40:41 [backends.py:420] Using cache directory: /home/ubuntu/.cache/vllm/torch_compile_cache/549ac03e3f/rank_0_0 for vLLM's torch.compile
INFO 05-20 07:40:41 [backends.py:430] Dynamo bytecode transform time: 6.74 s
INFO 05-20 07:40:45 [backends.py:118] Directly load the compiled graph(s) for shape None from the cache, took 3.712 s
INFO 05-20 07:40:49 [monitor.py:33] torch.compile takes 6.74 s in total
INFO 05-20 07:40:50 [kv_cache_utils.py:634] GPU KV cache size: 2,384,960 tokens
INFO 05-20 07:40:50 [kv_cache_utils.py:637] Maximum concurrency for 32,768 tokens per request: 72.78x
INFO 05-20 07:41:08 [gpu_model_runner.py:1686] Graph capturing finished in 18 secs, took 0.57 GiB
INFO 05-20 07:41:08 [core.py:159] init engine (profile, create kv cache, warmup model) took 33.18 seconds
INFO 05-20 07:41:08 [core_client.py:439] Core engine process 0 ready.
Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]Traceback (most recent call last):
  File "/home/ubuntu/vLLM/tests/issue.py", line 27, in <module>
    outputs = llm.generate(prompt, sampling_params)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/vLLM/tests/.venv/lib/python3.12/site-packages/vllm/utils.py", line 1196, in inner
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/vLLM/tests/.venv/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 473, in generate
    outputs = self._run_engine(use_tqdm=use_tqdm)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/vLLM/tests/.venv/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 1423, in _run_engine
    step_outputs = self.llm_engine.step()
                   ^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/vLLM/tests/.venv/lib/python3.12/site-packages/vllm/v1/engine/llm_engine.py", line 221, in step
    processed_outputs = self.output_processor.process_outputs(
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/vLLM/tests/.venv/lib/python3.12/site-packages/vllm/v1/engine/output_processor.py", line 344, in process_outputs
    stop_string = req_state.detokenizer.update(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/vLLM/tests/.venv/lib/python3.12/site-packages/vllm/v1/engine/detokenizer.py", line 106, in update
    self.output_text += self.decode_next(new_token_id)
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ubuntu/vLLM/tests/.venv/lib/python3.12/site-packages/vllm/v1/engine/detokenizer.py", line 201, in decode_next
    token = self.stream.step(self.tokenizer, next_token_id)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Exception: Invalid prefix encountered
Processed prompts:   0%|          | 0/1 [00:00<?, ?it/s, est. speed input: 0.00 toks/s, output: 0.00 toks/s]

Process finished with exit code 1

I noticed that adding a try/catch block in the detokenizer and returning "" in case of an error acted as a bandaid but is of course not an appropriate fix. happy to help in addressing this issue further if you need a patch.

[htdung167work0]

tokenizers==0.21.0 and vllm==0.8.5.post1.

tokenizers==0.21.0 and vllm==0.8.5.post1 work for me!!!