vllm
vllm copied to clipboard
vllm-project
[Bug]: ChatCompletionRequest rejects its own defaults
Your current environment
The output of `python collect_env.py`
Collecting environment information...
PyTorch version: 2.5.1+cu124
Is debug build: False
CUDA used to build PyTorch: 12.4
ROCM used to build PyTorch: N/A
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: Could not collect
Libc version: glibc-2.39
Python version: 3.10.16 (main, Jan 5 2025, 05:32:43) [Clang 19.1.6 ] (64-bit runtime)
Python platform: Linux-5.15.167.4-microsoft-standard-WSL2-x86_64-with-glibc2.39
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
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: 46 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 14
On-line CPU(s) list: 0-13
Vendor ID: GenuineIntel
Model name: Intel(R) Core(TM) Ultra 7 165U
CPU family: 6
Model: 170
Thread(s) per core: 2
Core(s) per socket: 7
Socket(s): 1
Stepping: 4
BogoMIPS: 5376.00
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology tsc_reliable nonstop_tsc cpuid pni pclmulqdq vmx ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves avx_vnni umip waitpkg gfni vaes vpclmulqdq rdpid movdiri movdir64b fsrm md_clear serialize flush_l1d arch_capabilities
Virtualization: VT-x
Hypervisor vendor: Microsoft
Virtualization type: full
L1d cache: 336 KiB (7 instances)
L1i cache: 448 KiB (7 instances)
L2 cache: 14 MiB (7 instances)
L3 cache: 12 MiB (1 instance)
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: Mitigation; Enhanced IBRS
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 / Automatic IBRS; IBPB conditional; RSB filling; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Not affected
Versions of relevant libraries:
[pip3] mypy==1.12.0
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.26.4
[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-nccl-cu12==2.21.5
[pip3] nvidia-nvjitlink-cu12==12.4.127
[pip3] nvidia-nvtx-cu12==12.4.127
[pip3] onnxruntime-gpu==1.20.1
[pip3] pyzmq==26.2.0
[pip3] torch==2.5.1
[pip3] torchaudio==2.5.1
[pip3] torchvision==0.20.1
[pip3] transformers==4.48.3
[pip3] triton==3.1.0
[conda] Could not collect
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.7.3
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
Could not collect
NCCL_CUMEM_ENABLE=0
TORCHINDUCTOR_COMPILE_THREADS=1
🐛 Describe the bug
The pydantic model ChatCompletionRequest found in vllm.entrypoints.openai.protocol rejects its own default configuration of logprobs and top_logprobs.
Example to reproduce the problem:
from vllm.entrypoints.openai.protocol import ChatCompletionRequest
request = ChatCompletionRequest(model="model", messages=[{"role": "user", "content": "content"}])
ChatCompletionRequest.model_validate(request.model_dump())
Output:
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/.../.venv/lib/python3.10/site-packages/pydantic/main.py", line 596, in model_validate
return cls.__pydantic_validator__.validate_python(
pydantic_core._pydantic_core.ValidationError: 1 validation error for ChatCompletionRequest
Value error, when using `top_logprobs`, `logprobs` must be set to true. [type=value_error, input_value={'messages': [{'content':...ogits_processors': None}, input_type=dict]
For further information visit https://errors.pydantic.dev/2.9/v/value_error
This becomes a problem when an instance of the class is instantiated at some point (and consequently filled with its defaults if no values for logprobs and top_logprobs are given), then converted to json/ a dict and later reinstantiated from this dump. (Also just generally: Either the defaults are valid or they are not (in which case it should not be possible to create an instance that has the default values by not passing the values explicitly).)
To fix this the validator should change as follows:
@model_validator(mode="before")
@classmethod
def check_logprobs(cls, data):
if (prompt_logprobs := data.get("prompt_logprobs")) is not None:
if data.get("stream") and prompt_logprobs > 0:
raise ValueError(
"`prompt_logprobs` are not available when `stream=True`.")
if prompt_logprobs < 0:
raise ValueError("`prompt_logprobs` must be a positive value.")
if (top_logprobs := data.get("top_logprobs")) is not None:
if top_logprobs < 0:
raise ValueError("`top_logprobs` must be a positive value.")
if top_logprobs > 0 and not data.get("logprobs"):
raise ValueError(
"when using `top_logprobs`, `logprobs` must be set to true."
)
return data
Where the line if top_logprobs > 0 and not data.get("logprobs"): contains the relevant fix. I.e. logprobs must only be set to True if we actually request any top_logprobs.
Before submitting a new issue...
- [x] 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.
@hmellor might this be caused by the change to the priority of generation_config.json?
Are you referring to https://github.com/vllm-project/vllm/pull/12622? If so, it has not been merged yet
Is the issue here that top_logprobs is explicitly 0, which is triggering the error because:
- it is not
None - and
logprobsis false
@hmellor Exactly. When we explicitly set these values (instead of not passing any values which still resolves to False and 0) the validator gives an error. I'd argue that:
- If I want
top_logprobs=0logprobsshould be allowed to beFalsebecause I don't need to return any anyway and more importantly - I should be allowed to initialize this class with its default values. I don't see a logical reason why I should get an error when I try to instantiate the class and use the "suggested" values for all properties.
Alternative way to change the validator would be
@model_validator(mode="before")
@classmethod
def check_logprobs(cls, data):
if (prompt_logprobs := data.get("prompt_logprobs")) is not None:
if data.get("stream") and prompt_logprobs > 0:
raise ValueError(
"`prompt_logprobs` are not available when `stream=True`.")
if prompt_logprobs < 0:
raise ValueError("`prompt_logprobs` must be a positive value.")
if (top_logprobs := data.get("top_logprobs")):
if top_logprobs < 0:
raise ValueError("`top_logprobs` must be a positive value.")
if not data.get("logprobs"):
raise ValueError(
"when using `top_logprobs`, `logprobs` must be set to true."
)
return data
So that the first condition is not fulfilled for both top_logprobs=None and top_logprobs=0.
