torch环境安装/问题整理

HanC_A1

已于 2024-04-11 09:32:21 修改

阅读量341

点赞数 3

分类专栏：车道线提取文章标签：深度学习

于 2024-04-09 13:37:59 首次发布

本文链接：https://blog.csdn.net/chenghan_/article/details/137539827

版权

车道线提取专栏收录该内容

5 篇文章 0 订阅

订阅专栏

1、创建conda虚拟环境：

1.1 创建命令

conda create -n torch1121+cuda10.2 python==3.8

1.2 激活命令

conda activate torch1121+cuda10.2

问题1：

Collecting package metadata (current_repodata.json): failed

UnavailableInvalidChannel: The channel is not accessible or is invalid.
  channel name: anaconda/pkgs/main
  channel url: http://mirrors.ustc.edu.cn/anaconda/pkgs/main
  error code: 404

You will need to adjust your conda configuration to proceed.
Use `conda config --show channels` to view your configuration's current state,
and use `conda config --show-sources` to view config file locations.


(base) [chenghan@localhost ~]$ conda create -n torch1121+cuda10.2 python==3.9
Collecting package metadata (current_repodata.json): failed

UnavailableInvalidChannel: The channel is not accessible or is invalid.
  channel name: anaconda/pkgs/main
  channel url: http://mirrors.ustc.edu.cn/anaconda/pkgs/main
  error code: 404

You will need to adjust your conda configuration to proceed.
Use `conda config --show channels` to view your configuration's current state,
and use `conda config --show-sources` to view config file locations.

解决方法1：

输入以下命令：

conda config --remove-key channels

2、查看cuda版本

2.1 输入以下命令查看所有的cuda版本

ls -l /usr/local | grep cuda

lrwxrwxrwx   1 root root  20 3月   3 2021 cuda -> /usr/local/cuda-10.0
drwxr-xr-x. 19 root root 274 3月   3 2021 cuda-10.0
drwxr-xr-x  15 root root 288 2月  22 2021 cuda-10.2
drwxr-xr-x. 18 root root 249 8月   7 2020 cuda-9.0
drwxr-xr-x. 18 root root 249 9月  27 2020 cuda-9.2

2.2 输入以下命令查看当前的cuda版本：

nvcc -V

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_19:24:38_PDT_2019
Cuda compilation tools, release 10.2, V10.2.89

3、安装pytorch
根据自己的cuda版本以及pytorch在torch官网上找对应的安装命令：pytorch官网：

conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0 cudatoolkit=10.2 -c pytorch

3、安装mmcv-full

问题1

利用pip install mmcv-full==1.7.0以下命令出现以下问题：

 /home/chenghan/anaconda3/envs/torch1121+cuda10.2/lib/python3.9/site-packages/torch/utils/cpp_extension.py:820: UserWarning: There are no g++ version bounds defined for CUDA version 10.0
        warnings.warn(f'There are no {compiler_name} version bounds defined for CUDA version {cuda_str_version}')
      building 'mmcv._ext' extension
      creating build/temp.linux-x86_64-cpython-39
      creating build/temp.linux-x86_64-cpython-39/mmcv
      creating build/temp.linux-x86_64-cpython-39/mmcv/ops
      creating build/temp.linux-x86_64-cpython-39/mmcv/ops/csrc
      creating build/temp.linux-x86_64-cpython-39/mmcv/ops/csrc/pytorch
      creating build/temp.linux-x86_64-cpython-39/mmcv/ops/csrc/pytorch/cpu
      creating build/temp.linux-x86_64-cpython-39/mmcv/ops/csrc/pytorch/cuda
      gcc -pthread -B /home/chenghan/anaconda3/envs/torch1121+cuda10.2/compiler_compat -Wl,--sysroot=/ -Wno-unused-result -Wsign-compare -DNDEBUG -O2 -Wall -fPIC -O2 -isystem /home/chenghan/anaconda3/envs/torch1121+cuda10.2/include -fPIC -O2 -isystem /home/chenghan/anaconda3/envs/torch1121+cuda10.2/include -fPIC -DMMCV_WITH_CUDA -I/tmp/pip-install-3lhqn1mt/mmcv-full_c677e90658ad44a9b18bb7f3de410076/mmcv/ops/csrc/pytorch -I/tmp/pip-install-3lhqn1mt/mmcv-full_c677e90658ad44a9b18bb7f3de410076/mmcv/ops/csrc/common -I/tmp/pip-install-3lhqn1mt/mmcv-full_c677e90658ad44a9b18bb7f3de410076/mmcv/ops/csrc/common/cuda -I/home/chenghan/anaconda3/envs/torch1121+cuda10.2/lib/python3.9/site-packages/torch/include -I/home/chenghan/anaconda3/envs/torch1121+cuda10.2/lib/python3.9/site-packages/torch/include/torch/csrc/api/include -I/home/chenghan/anaconda3/envs/torch1121+cuda10.2/lib/python3.9/site-packages/torch/include/TH -I/home/chenghan/anaconda3/envs/torch1121+cuda10.2/lib/python3.9/site-packages/torch/include/THC -I/usr/local/cuda/include -I/home/chenghan/anaconda3/envs/torch1121+cuda10.2/include/python3.9 -c ./mmcv/ops/csrc/pytorch/active_rotated_filter.cpp -o build/temp.linux-x86_64-cpython-39/./mmcv/ops/csrc/pytorch/active_rotated_filter.o -std=c++14 -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE=\"_gcc\" -DPYBIND11_STDLIB=\"_libstdcpp\" -DPYBIND11_BUILD_ABI=\"_cxxabi1011\" -DTORCH_EXTENSION_NAME=_ext -D_GLIBCXX_USE_CXX11_ABI=0
      gcc: 错误：unrecognized command line option ‘-std=c++14’
      error: command '/usr/bin/gcc' failed with exit code 1
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for mmcv-full
  Running setup.py clean for mmcv-full
Failed to build mmcv-full
ERROR: Could not build wheels for mmcv-full, which is required to install pyproject.toml-based projects
(torch1121+cuda10.2) [chenghan@localhost ADNet-master]$ pip install mmcv-full==1.4.0 -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.10/index.html
Looking in links: https://download.openmmlab.com/mmcv/dist/cu102/torch1.10/index.html
^CERROR: Operation cancelled by user

解决方法1

修改命令行中的cu102和torch1.12版本号：

pip install mmcv-full==1.7.0 -f https://download.openmmlab.com/mmcv/dist/cu102/torch1.12/index.html

4、pip install 镜像源地址

-i https://pypi.tuna.tsinghua.edu.cn/simple

5、gcc版本问题

问题1：

gcc: error: unrecognized command line option ‘-std=c++14’ error: command ’

解决方法：

升级gcc版本，无root权限Linux升级gcc版本

下载并解压所需版本（可直接改自己所需要的版本）

wget http://ftp.gnu.org/gnu/gcc/gcc-7.4.0/gcc-7.4.0.tar.gz
tar -zxvf gcc-7.4.0.tar.gz
cd gcc-7.4.0
./contrib/download_prerequisites

编译，安装（修改–prefix=/home/username的地址，以及版本号）

cd gcc-5.4.0
mkdir objdir  
cd objdir
../configure --disable-checking --enable-languages=c,c++ --disable-multilib --prefix=/home/username/gcc-7.4 --enable-threads=posix
make
make install

设置环境变量(修改两个地址，与编译、安装地址保持一致)

vim ~/.bashrc
export PATH=/home/username/gcc-5.4/bin:/home/username/gcc-5.4/lib64:$PATH
export LD_LIBRARY_PATH=/home/username/gcc-5.4/lib/:$LD_LIBRARY_PATH
source ~/.bashrc

vim的简单用法：
shift+i:可编辑状态
Esc：为编辑完之后进入可读模式
shift+q:退出状态
wq：表示保存退出