[Bug] set logprobs = true and top_logprobs = 5 in restful server. The number of top logrobs is 4 which is unexpected.

[zhulinJulia24]

Checklist

• 1. I have searched related issues but cannot get the expected help.
• 2. The bug has not been fixed in the latest version.

Describe the bug

After set logprobs = true and top_logprobs = 5. The number of top logrobs response is not correct, only 4 for each token. I suppose it should be 5 for each token.

Reproduction

1. start a api_server such as internlm2-chat-20b, such as lmdeploy serve api_server /nvme/qa_test_models/internlm/internlm2-chat-20b --tp 2
2. open swagger and send request to /v1/chat/completions like this:

{
  "model": "internlm2",
  "messages": [
    {
      "content": "Shanghai is",
      "role": "user"
    }
  ],
  "logprobs": true,
  "top_logprobs": 5,
  "max_tokens": 20
}

check the response, only 4 top probs return. I want 5 top probs

3. And I cannot get logprobs return by using /v1/chat/completions api, the script is:

from lmdeploy.serve.openai.api_client import APIClient

api_client = APIClient('http://localhost:23333')
for output in api_client.chat_completions_v1(model='internlm2',
                                            messages='Shanghai is',
                                            logprobs=True,
                                            top_logprobs=5,
                                            max_tokens=20):
    continue
print(output)

logprobs in response is none, it's unexpected.

{'id': '1', 'object': 'chat.completion', 'created': 1714991049, 'model': 'internlm2', 'choices': [{'index': 0, 'message': {'role': 'assistant', 'content': ' a city of contrasts. The city is both ancient and modern, traditional and innovative, and a world'}, 'logprobs': None, 'finish_reason': 'length'}], 'usage': {'prompt_tokens': 4, 'total_tokens': 25, 'completion_tokens': 21}}

Environment

sys.platform: linux
Python: 3.10.14 (main, Mar 21 2024, 16:24:04) [GCC 11.2.0]
CUDA available: True
MUSA available: False
numpy_random_seed: 2147483648
GPU 0,1,2,3,4,5,6,7: NVIDIA A100-SXM4-80GB
CUDA_HOME: /usr/local/cuda-11.7
NVCC: Cuda compilation tools, release 11.7, V11.7.64
GCC: gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-44)
PyTorch: 2.1.0+cu118
PyTorch compiling details: PyTorch built with:
  - GCC 9.3
  - C++ Version: 201703
  - Intel(R) oneAPI Math Kernel Library Version 2022.2-Product Build 20220804 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v3.1.1 (Git Hash 64f6bcbcbab628e96f33a62c3e975f8535a7bde4)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - LAPACK is enabled (usually provided by MKL)
  - NNPACK is enabled
  - CPU capability usage: AVX512
  - CUDA Runtime 11.8
  - NVCC architecture flags: -gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_90,code=sm_90
  - CuDNN 8.7
  - Magma 2.6.1
  - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.8, CUDNN_VERSION=8.7.0, CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=0 -fabi-version=11 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOROCTRACER -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -Wno-invalid-partial-specialization -Wno-unused-private-field -Wno-aligned-allocation-unavailable -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_DISABLE_GPU_ASSERTS=ON, TORCH_VERSION=2.1.0, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=1, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, 

TorchVision: 0.16.0+cu118
LMDeploy: 0.4.0+
transformers: 4.40.1
gradio: 4.28.0
fastapi: 0.110.2
pydantic: 2.7.0
triton: 2.1.0

Error traceback

No response

[irexyc]

topk, topp操作会减少在做sample 之前 vocab的候选词的数量。当候选词减少到4个的时候，我觉得没必要非得输出5个词的概率（vllm那边会输出，不过概率是-inf)。如果输出的token不在前5的话，那么输出的top logrobs可能会有6个。

[zhulinJulia24]

topk, topp操作会减少在做sample 之前 vocab的候选词的数量。当候选词减少到4个的时候，我觉得没必要非得输出5个词的概率（vllm那边会输出，不过概率是-inf)。如果输出的token不在前5的话，那么输出的top logrobs可能会有6个。

是的，但在这种情况下我试过top > 5还是能输出>=5的数字，所以我理解topk, topp操作后候选词应该是多于5个，那么top=5的话应该输出>=5个？

[irexyc]

最少输出一个，最多输出top+1个(所选token不在top里面)