配置服务器过程

安装nvidia驱动

sudo sh ./NVIDIA-Linux-x86_64-525.89.02.run --no-kernel-module

安装Anaconda

https://blog.csdn.net/wyf2017/article/details/118676765

最后一步需要进行激活：

 source ~/.bashrc

安装cuda和cudnn

https://blog.csdn.net/h3c4lenovo/article/details/119003405

注意：安装cudnn需要按如下命令

sudo cp cuda/include/cudnn* /usr/local/cuda-10.1/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda-10.1/lib64
sudo chmod a+r /usr/local/cuda-11.3/include/cudnn* 
sudo chmod a+r /usr/local/cuda-11.3/lib64/libcudnn*

配置cuda环境

export CUDA_HOME=/usr/local/cuda-11.3/
export PATH=$PATH:/usr/local/cuda-11.3/bin/
export LD_LIBRARY_PATH="/usr/local/cuda-11.3/lib:$LD_LIBRARY_PATH"

安装Anaconda虚拟环境

创建虚拟环境

conda create -n env_name python=X.X

删除虚拟环境

conda remove -n your_env_name --all

安装pytorch和torchvision

对应版本号

NVIDIA GeForce RTX 3090 Ti with CUDA capability sm_86对应pytorch1.8.0--torchvision0.9.0---cuda--11.1

安装pytorch和torchvision

pip install torch-1.8.0+cu111-cp37-cp37m-linux_x86_64.whl
pip install torchvision-0.9.0+cu111-cp37-cp37m-linux_x86_64.whl

配置mmdetection

安装程序包

pip install requirements/*.txt

安装mmcv-full

python -m pip install  mmcv-full=={mmcv_version} -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.8.0/index.html

进行编译

pip install -v -e .