容器内pytorch无法运行报错Unexpected error from cudaGetDeviceCount()

容器内pytorch无法运行报错Unexpected error from cudaGetDeviceCount()



报错如下: root@9e039aacfc5e:/opt/video-subtitle-generator/py38# python Python 3.8.19 (default, Sep 5 2024, 00:27:22) [GCC 12.2.0] on linux Type “help”, “copyright”, “credits” or “license” for more information.

import torch torch.cuda.is_available() /usr/local/lib/python3.8/site-packages/torch/cuda/init.py:128: UserWarning: CUDA initialization: Unexpected error from cudaGetDeviceCount(). Did you run some cuda functions before calling NumCudaDevices() that might have already set an error? Error 500: named symbol not found (Triggered internally at ../c10/cuda/CUDAFunctions.cpp:108.) return torch._C._cuda_getDeviceCount() > 0 False 原因: 在容器内运行 CUDA 应用程序时,带有 WSL2 的 NVIDIA 驱动程序 555.85 返回错误 500“找不到命名符号” ·问题 #520 ·NVIDIA/nvidia-容器工具包 解决: 安装新版的:nvidia-container-toolkit 仓库安装有很多教程 下面记录手动安装: 下载地址:Releases · NVIDIA/nvidia-container-toolkit

下载后直接解压: 解压后如下: 安装: #执行以下命令 dpkg -i ./lib* ./nvidia* root@DESKTOP-1M3C1UI:/mnt/c/pytorch/release-v1.16.2-stable/packages/ubuntu18.04/amd64# dpkg -i ./lib* ./nvidia* Selecting previously unselected package libnvidia-container-dev:amd64. (Reading database … 69861 files and directories currently installed.) Preparing to unpack …/libnvidia-container-dev_1.16.2-1_amd64.deb … Unpacking libnvidia-container-dev:amd64 (1.16.2-1) … Preparing to unpack …/libnvidia-container-tools_1.16.2-1_amd64.deb … Unpacking libnvidia-container-tools (1.16.2-1) over (1.13.5-1) … Selecting previously unselected package libnvidia-container1-dbg:amd64. Preparing to unpack …/libnvidia-container1-dbg_1.16.2-1_amd64.deb … Unpacking libnvidia-container1-dbg:amd64 (1.16.2-1) … Preparing to unpack …/libnvidia-container1_1.16.2-1_amd64.deb … Unpacking libnvidia-container1:amd64 (1.16.2-1) over (1.13.5-1) … Preparing to unpack …/nvidia-container-toolkit-base_1.16.2-1_amd64.deb … Unpacking nvidia-container-toolkit-base (1.16.2-1) over (1.13.5-1) … dpkg: warning: unable to delete old directory ‘/etc/nvidia-container-runtime’: Directory not empty Selecting previously unselected package nvidia-container-toolkit-operator-extensions.

Preparing to unpack …/nvidia-container-toolkit-operator-extensions_1.16.2-1_amd64.deb … Unpacking nvidia-container-toolkit-operator-extensions (1.16.2-1) … Preparing to unpack …/nvidia-container-toolkit_1.16.2-1_amd64.deb … Unpacking nvidia-container-toolkit (1.16.2-1) over (1.16.2-1) … Setting up libnvidia-container1:amd64 (1.16.2-1) … Setting up nvidia-container-toolkit-base (1.16.2-1) … Setting up nvidia-container-toolkit-operator-extensions (1.16.2-1) … Setting up libnvidia-container-dev:amd64 (1.16.2-1) … Setting up libnvidia-container-tools (1.16.2-1) … Setting up libnvidia-container1-dbg:amd64 (1.16.2-1) … Setting up nvidia-container-toolkit (1.16.2-1) … Processing triggers for libc-bin (2.35-0ubuntu3.8) …

page1

page2

page2

page2