错误：Torch 未在启用 CUDA 的情况下编译

Question

当我运行“torch.rand(10).to(“cuda”)”时，我遇到“错误：Torch 未在启用 CUDA 的情况下编译”

GPU：Nvidia RTX 3080 Ti

$nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.85.05    Driver Version: 525.85.05    CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+

$nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85

$conda list

cudatoolkit   version  :  11.0.221

$conda info


active environment : sw
    active env location : /mnt/user2/.conda/envs/sw
            shell level : 1
       user config file : /mnt/user2/.condarc
 populated config files :
          conda version : 23.1.0
    conda-build version : not installed
         python version : 3.10.9.final.0
       virtual packages : __archspec=1=x86_64
                          __glibc=2.27=0
                          __linux=5.4.0=0
                          __unix=0=0
       base environment : /opt/conda  (read only)
      conda av data dir : /opt/conda/etc/conda
  conda av metadata url : None
           channel URLs : https://repo.anaconda.com/pkgs/main/linux-64
                          https://repo.anaconda.com/pkgs/main/noarch
                          https://repo.anaconda.com/pkgs/r/linux-64
                          https://repo.anaconda.com/pkgs/r/noarch
          package cache : /opt/conda/pkgs
                          /mnt/user2/.conda/pkgs
       envs directories : /mnt/user2/.conda/envs
                          /opt/conda/envs
               platform : linux-64
             user-agent : conda/23.1.0 requests/2.28.1 CPython/3.10.9 Linux/5.4.0-144-generic ubuntu/18.04.6 glibc/2.27
                UID:GID : 1002:1001
             netrc file : None
           offline mode : False

火炬版本：2.0.0

我下载的是pytorch

$ conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia

我很困惑识别 cuda 版本。nvidia-smi 说 cuda 是 12.0，但 conda 列表中的表说 cuda 是 11.0.221，但 nvcc-V 说 cuda 9.1。

如果您需要更多信息，请评论。

Searching google to solve the problem but didn't work.

Answer 1

对我有帮助的是使用卸载火炬并使用而不是pip再次安装它。pipconda

pip uninstall torch
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118