mmpretrain安装教程（踩雷记录）

目录

ModuleNotFoundError: No module named ‘mmcv‘包命令pkgutil报错

OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root.

error: Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/visual-cpp-build-tools/

Git 配置代理，解决 "Failed to connect to github.com port 443" 及 "Recv failure: Connection was reset" 等问题

从源安装 mmpretrain指令：

---

ModuleNotFoundError: No module named ‘mmcv‘包命令pkgutil报错

解决方案：ModuleNotFoundError: No module named ‘mmcv‘包命令pkgutil报错-CSDN博客

OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root.

解决方案： anaconda虚拟环境中报错：OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA-CSDN博客

error: Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/visual-cpp-build-tools/

解决方案：老实安装Microsoft C++ 生成工具 - Visual Studio，大概6.6G

Git 配置代理，解决 "Failed to connect to github.com port 443" 及 "Recv failure: Connection was reset" 等问题

解决方案：三种方法解决git拒绝连接问题fatal: unable to connect to git | Go 技术论坛

取消代理后就可以git clone了

git config --global http.proxy
git config --global --unset http.proxy

从源安装 mmpretrain指令：

git clone https://github.com/open-mmlab/mmpretrain.git
cd mmpretrain
pip install -U openmim && mim install -e .  -f https://download.pytorch.org/whl/lts/1.8/torch_lts.html  -i http://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com

如果在mmcv卡住，可以试试看用pip指定mmcv版本安装

Installation — mmcv 2.1.0 documentation

即执行命令

pip install mmcv==2.1.0 -f https://download.openmmlab.com/mmcv/dist/cu121/torch2.1/index.html

我的版本没有匹配的，所以根据官方文档，从source开始编译。

Build MMCV from source — mmcv 2.1.0 文档

需要注意增加环境变量TORCH_CUDA_ARCH_LIST

我的取值是8.9，具体可以运行以下代码查询capability：

python -c "import torch; print(torch.cuda.get_device_capability())"