[Bug]: GuidedDecodingParams choice - Request-level structured output backend must match engine-level backend

[nrv]

Your current environment

The output of `python collect_env.py`

INFO 04-16 19:43:43 [__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: AlmaLinux 9.5 (Teal Serval) (x86_64)
GCC version: (GCC) 11.5.0 20240719 (Red Hat 11.5.0-2)
Clang version: Could not collect
CMake version: version 3.26.5
Libc version: glibc-2.34

Python version: 3.12.0 (main, Apr 16 2025, 17:47:26) [GCC 11.5.0 20240719 (Red Hat 11.5.0-2)] (64-bit runtime)
Python platform: Linux-5.14.0-503.16.1.el9_5.x86_64-x86_64-with-glibc2.34
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: 
GPU 0: NVIDIA H100 NVL
GPU 1: NVIDIA H100 NVL
GPU 2: NVIDIA H100 NVL
GPU 3: NVIDIA H100 NVL

Nvidia driver version: 565.57.01
cuDNN version: Could not collect
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):                               96
On-line CPU(s) list:                  0-95
Vendor ID:                            AuthenticAMD
Model name:                           AMD EPYC 9274F 24-Core Processor
CPU family:                           25
Model:                                17
Thread(s) per core:                   2
Core(s) per socket:                   24
Socket(s):                            2
Stepping:                             1
Frequency boost:                      enabled
CPU(s) scaling MHz:                   100%
CPU max MHz:                          4050.0000
CPU min MHz:                          1500.0000
BogoMIPS:                             8087.50
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 amd_lbr_v2 nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba perfmon_v2 ibrs ibpb stibp ibrs_enhanced vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local avx512_bf16 clzero irperf xsaveerptr rdpru wbnoinvd amd_ppin cppc arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif x2avic v_spec_ctrl vnmi avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq la57 rdpid overflow_recov succor smca fsrm flush_l1d debug_swap
Virtualization:                       AMD-V
L1d cache:                            1.5 MiB (48 instances)
L1i cache:                            1.5 MiB (48 instances)
L2 cache:                             48 MiB (48 instances)
L3 cache:                             512 MiB (16 instances)
NUMA node(s):                         2
NUMA node0 CPU(s):                    0-23,48-71
NUMA node1 CPU(s):                    24-47,72-95
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 Reg file data sampling: Not affected
Vulnerability Retbleed:               Not affected
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; Enhanced / Automatic IBRS; IBPB conditional; STIBP always-on; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds:                  Not affected
Vulnerability Tsx async abort:        Not affected

Versions of relevant libraries:
[pip3] numpy==2.2.4
[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] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.8.4
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
	GPU0	GPU1	GPU2	GPU3	CPU Affinity	NUMA Affinity	GPU NUMA ID
GPU0	 X 	NV12	SYS	SYS	0-23,48-71	0		N/A
GPU1	NV12	 X 	SYS	SYS	0-23,48-71	0		N/A
GPU2	SYS	SYS	 X 	NV12	24-47,72-95	1		N/A
GPU3	SYS	SYS	NV12	 X 	24-47,72-95	1		N/A

Legend:

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

NCCL_CUMEM_ENABLE=0
PYTORCH_NVML_BASED_CUDA_CHECK=1
TORCHINDUCTOR_COMPILE_THREADS=1
CUDA_MODULE_LOADING=LAZY

🐛 Describe the bug

When using a GuidedDecodingParams with choice option, there is a bug on the second call to generate(). The first one is always OK and the second one always produces the same error. Also tested by fixing the backend option but the error is still here.

from vllm import LLM, SamplingParams
from vllm.sampling_params import GuidedDecodingParams

guided_decoding_params_choice = GuidedDecodingParams(choice=["Positive", "Negative"])
sampling_params_choice = SamplingParams(guided_decoding=guided_decoding_params_choice)

prompt_choice_1 = "Classify this sentiment: vLLM is wonderful"
prompt_choice_2 = "Classify this sentiment: vLLM is really awful !"

llm = LLM(model="mistralai/Mistral-7B-Instruct-v0.1", max_model_len=100)

outputs_1 = llm.generate(prompts=prompt_choice_1, sampling_params=sampling_params_choice)
print(f"prompt_choice_1 : {outputs_1[0].outputs[0].text}")

outputs_2 = llm.generate(prompts=prompt_choice_2, sampling_params=sampling_params_choice)
print(f"prompt_choice_2 : {outputs_2[0].outputs[0].text}")

Output :

[...]
Processed prompts: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:00<00:00,  3.34it/s, est. speed input: 36.74 toks/s, output: 13.36 toks/s]
prompt_choice_1 : Negative
Traceback (most recent call last):
  File "/mnt/rex/home17/nherve/workspace/vllm-batch/vllm_batch_bug.py", line 15, in <module>
    outputs_2 = llm.generate(prompts=prompt_choice_2, sampling_params=sampling_params_choice)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/utils.py", line 1134, in inner
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 462, in generate
    self._validate_and_add_requests(
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 1342, in _validate_and_add_requests
    self._add_request(
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 1360, in _add_request
    self.llm_engine.add_request(
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/v1/engine/llm_engine.py", line 186, in add_request
    request = self.processor.process_inputs(request_id, prompt, params,
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 212, in process_inputs
    self._validate_params(params)
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 136, in _validate_params
    self._validate_sampling_params(params)
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 79, in _validate_sampling_params
    self._validate_structured_output(params)
  File "/home/nherve/.pyenv/versions/vllm/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 158, in _validate_structured_output
    raise ValueError("Request-level structured output backend "
ValueError: Request-level structured output backend must match engine-level backend. xgrammar != auto

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.

[rangehow]

Not sure if this is a related issue. I and @nrv are using similar calling methods, but I get an error the first time I generate in vllm 0.8.5 beta ( built from source yesterday ) . The error message is as follows:

  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/mtr/model_choice/answerable.py", line 26, in generate
    outputs = llm.generate(prompt_token_ids=dataset['input_ids'], sampling_params=sampling_params)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/utils.py", line 1184, in inner
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 462, in generate
    self._validate_and_add_requests(
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 1356, in _validate_and_add_requests
    self._add_request(
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/entrypoints/llm.py", line 1374, in _add_request
    self.llm_engine.add_request(
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/v1/engine/llm_engine.py", line 186, in add_request
    request = self.processor.process_inputs(request_id, prompt, params,
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 209, in process_inputs
    self._validate_params(params)
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 136, in _validate_params
    self._validate_sampling_params(params)
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 79, in _validate_sampling_params
    self._validate_structured_output(params)
  File "/mnt/dolphinfs/ssd_pool/docker/user/hadoop-aipnlp/INS/ruanjunhao04/miniforge3/envs/sglang/lib/python3.12/site-packages/vllm/v1/engine/processor.py", line 158, in _validate_structured_output
    raise ValueError(
ValueError: Request-level structured output backend selection is no longer supported. The request specified 'xgrammar', but vLLM was initialised with 'auto'. This error can be resolved by removing backend selection from the request.

[yliu2702]

ValueError: Request-level structured output backend must match engine-level backend. xgrammar != auto. So how to solve? I want to get a list of json as output

raw_schema = EntityList.model_json_schema()
        guided_decoding_params = GuidedDecodingParams(json=raw_schema, backend="xgrammar")
        sampling_params = SamplingParams(
                    temperature=temperature,
                    top_p=1,
                    max_tokens=max_tokens,
                    guided_decoding=guided_decoding_params)
llm = LLM(model = "Qwen/Qwen2.5-7B-Instruct",
            tokenizer = "Qwen/Qwen2.5-7B-Instruct",
            dtype='bfloat16',
            distributed_executor_backend="mp", 
            tensor_parallel_size= 2,
            gpu_memory_utilization = 0.95,
            max_model_len = 32768,
            max_num_seqs=64)

But get the same error. Where is the solution?

[DarkLight1337]

You should set the backend to xgrammar at engine startup (guided_decoding_backend="xgrammar" when creating LLM), rather than setting it in GuidedDecodingParams