[Bug] lmdeploy 加载qwen1.5-32b-chat后，使用chat接口调用生成，后台模型服务中断

WCwalker commented 5 months ago

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

Producer process has been terminated before all shared CUDA tensors released. S 1713420898115(1)

Reproduction

lmdeploy serve api_server /home/mingqiang/model/model_file/origin_model/Qwen1.5-32B-Chat/ --cache-max-entry-count 0.1 --tp 4 --model-name qwen-32b --server-port 10001

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: NVIDIA A800 80GB PCIe
CUDA_HOME: /usr/local/cuda
NVCC: Cuda compilation tools, release 12.1, V12.1.105
GCC: gcc (Debian 10.2.1-6) 10.2.1 20210110
PyTorch: 2.1.2+cu121
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 12.1
  - 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_90,code=sm_90
  - CuDNN 8.9.2
  - Magma 2.6.1
  - Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=12.1, CUDNN_VERSION=8.9.2, 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=2.1.2, 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, 

LMDeploy: 0.3.0+
transformers: 4.38.1
gradio: Not Found
fastapi: 0.110.1
pydantic: 2.7.0
triton: 2.2.0

Error traceback

No response

WCwalker commented 5 months ago

另外补充一下：我在lmdeploy 0.2.4+环境可以运行qwen1.5-32B-chat，但是model name要设置为qwen-14；但是在运行期间如果传入tokens 大于一定数量(没有做统计，可能就3k)，容易出现"length"错误，返回空值，就算是session len 设置为8192 也是一样，好像没有起作用

zhyncs commented 4 months ago

另外补充一下：我在lmdeploy 0.2.4+环境可以运行qwen1.5-32B-chat，但是model name要设置为qwen-14；但是在运行期间如果传入tokens 大于一定数量(没有做统计，可能就3k)，容易出现"length"错误，返回空值，就算是session len 设置为8192 也是一样，好像没有起作用

Hi @WCwalker May you try the latest version v0.4.0?

InternLM / lmdeploy