![[每周知识碎片] 3](http://pic.xiahunao.cn/yaotu/[每周知识碎片] 3)
conda配置源conda config --show channels conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ conda config --set show_channel_urls yes # 或者手动 vim .condarcanaconda安装chmod x Anaconda3-5.3.1-Linux-x86_64.sh ./Anaconda3-5.3.1-Linux-x86_64.sh # 安装路径默认为用户目录(可以自己指定)最后需要确认将路径加入用户的.bashrc中。conda 安装 pytorch1.2# 先给conda换源否则会很慢 # conda会自动安装cudatoolkit和cudnn这点相比pip方式更人性化 conda install pytorch torchvision cudatoolkit10.0 -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ # -c 指定channelarch linux中安装cudnn的deb包# deb本质上也是一种压缩包所以直接解压就行 # cudnn7.4(for cuda10.0)下载 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libcudnn7_7.4.1.5-1cuda10.0_amd64.deb mkdir libcudnn7 cd libcudnn7 ar xv libcudnn7_7.4.1.5-1cuda10.0_amd64.deb #得到 control.tar.xz data.tar.xz debian- tar xf data.tar.xz cp -a usr/lib/x86_64-linux-gnu/* ${pkgdir}/usr/lib/ # cudnn7.4 dev(for cuda10.0) 下载 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libcudnn7-dev_7.4.1.5-1cuda10.0_amd64.deb mkdir libcudnn7-dev cd libcudnn7-dev ar xv ${srcdir}/libcudnn7-dev_${pkgver}-1cuda${_cudaver}.0_amd64.deb tar xf data.tar.xz cp -a usr/include/x86_64-linux-gnu/* ${pkgdir}/usr/include/ cp -a usr/lib/x86_64-linux-gnu/* ${pkgdir}/usr/lib/archlinux下非root用户安装单独的cuda注意不安装显卡驱动只安装toolkit和cudnn# cuda9.0下载地址 https://developer.nvidia.com/compute/cuda/9.0/Prod/local_installers/cuda_9.0.176_384.81_linux-run chmod x cuda_9.0.176_384.81_linux-run ./cuda_9.0.176_384.81_linux-run # cudnn7.0 for cuda9.0 下载地址 https://developer.download.nvidia.com/compute/redist/cudnn/v7.0.5/cudnn-9.0-linux-x64-v7.tgz tar -zvxf cudnn-9.0-linux-x64-v7.tgz cp include/cudnn.h /xxx/cuda-9.0/include/ cp lib64/libcudnn* /xxx/cuda-9.0/lib64/ chmod ar include/cudnn.h lib64/libcudnn* # cuda10.0下载地址 https://developer.nvidia.com/compute/cuda/10.0/Prod/local_installers/cuda_10.0.130_410.48_linux # cudnn7.4 for cuda10.0下载地址见上注意下载地址是从archlinux pkgs官网找到的历史版本最后修改一下启动设置vim ~/.bashrc # 将下述内容添加到.bashrc文件末尾 export CUDA_HOME/xxx/cuda-9.0 export PATH$PATH:$CUDA_HOME/bin export LD_LIBRARY_PATH$LD_LIBRARY_PATH:$CUDA_HOME/lib64参考1 参考2
conda配置源conda config --show channels conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/ conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ conda config --set show_channel_urls yes # 或者手动 vim .condarcanaconda安装chmod x Anaconda3-5.3.1-Linux-x86_64.sh ./Anaconda3-5.3.1-Linux-x86_64.sh # 安装路径默认为用户目录(可以自己指定)最后需要确认将路径加入用户的.bashrc中。conda 安装 pytorch1.2# 先给conda换源否则会很慢 # conda会自动安装cudatoolkit和cudnn这点相比pip方式更人性化 conda install pytorch torchvision cudatoolkit10.0 -c https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/ # -c 指定channelarch linux中安装cudnn的deb包# deb本质上也是一种压缩包所以直接解压就行 # cudnn7.4(for cuda10.0)下载 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libcudnn7_7.4.1.5-1cuda10.0_amd64.deb mkdir libcudnn7 cd libcudnn7 ar xv libcudnn7_7.4.1.5-1cuda10.0_amd64.deb #得到 control.tar.xz data.tar.xz debian- tar xf data.tar.xz cp -a usr/lib/x86_64-linux-gnu/* ${pkgdir}/usr/lib/ # cudnn7.4 dev(for cuda10.0) 下载 https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/libcudnn7-dev_7.4.1.5-1cuda10.0_amd64.deb mkdir libcudnn7-dev cd libcudnn7-dev ar xv ${srcdir}/libcudnn7-dev_${pkgver}-1cuda${_cudaver}.0_amd64.deb tar xf data.tar.xz cp -a usr/include/x86_64-linux-gnu/* ${pkgdir}/usr/include/ cp -a usr/lib/x86_64-linux-gnu/* ${pkgdir}/usr/lib/archlinux下非root用户安装单独的cuda注意不安装显卡驱动只安装toolkit和cudnn# cuda9.0下载地址 https://developer.nvidia.com/compute/cuda/9.0/Prod/local_installers/cuda_9.0.176_384.81_linux-run chmod x cuda_9.0.176_384.81_linux-run ./cuda_9.0.176_384.81_linux-run # cudnn7.0 for cuda9.0 下载地址 https://developer.download.nvidia.com/compute/redist/cudnn/v7.0.5/cudnn-9.0-linux-x64-v7.tgz tar -zvxf cudnn-9.0-linux-x64-v7.tgz cp include/cudnn.h /xxx/cuda-9.0/include/ cp lib64/libcudnn* /xxx/cuda-9.0/lib64/ chmod ar include/cudnn.h lib64/libcudnn* # cuda10.0下载地址 https://developer.nvidia.com/compute/cuda/10.0/Prod/local_installers/cuda_10.0.130_410.48_linux # cudnn7.4 for cuda10.0下载地址见上注意下载地址是从archlinux pkgs官网找到的历史版本最后修改一下启动设置vim ~/.bashrc # 将下述内容添加到.bashrc文件末尾 export CUDA_HOME/xxx/cuda-9.0 export PATH$PATH:$CUDA_HOME/bin export LD_LIBRARY_PATH$LD_LIBRARY_PATH:$CUDA_HOME/lib64参考1 参考2
