linux下的CUDA安装最终效果需要能够实现三个命令nvidia-smi看到显卡nvcc--version看到CUDA版本通常是12.1python -PY import torch print(torch.cuda.is_available()) print(torch.cuda.get_device_name(0)) PY输出为True 你的 NVIDIA 显卡型号1查显卡型号在window下方法一打开任务管理器性能 → GPU方法二PowerShell 里面执行nvidia-smi如果 NVIDIA 驱动正常会显示类似---------------------------------------------------------------------------------------|NVIDIA-SMI552.22Driver Version:552.22CUDA Version:12.4||GPU Name...|0NVIDIA GeForce RTX5060... ---------------------------------------------------------------------------------------你要记三个东西GPU Name Driver Version CUDA Version2在ubuntu安装2.1先确定ubuntu能识别到显卡lspci|grep-invidia如果有.01:00.0 VGA compatible controller: NVIDIA Corporation GA106[GeForce RTX3060]说明 Ubuntu 识别到硬件。如果没输出可能是BIOS 没开独显笔记本混合显卡设置问题驱动还没装但一般 lspci 仍应能看到2.2 安装驱动方案A自动安装驱动先能拉下载的驱动包sudoaptupdatesudoaptupgrade-ysudoaptinstall-yubuntu-drivers-common可以查看推荐的驱动ubuntu-drivers devices然后自动安装sudoubuntu-drivers autoinstall安装完重启sudoreboot重启后检查nvidia-smi正常应该显示NVIDIA-SMI xxx.xx Driver Version: xxx.xx CUDA Version:12.x GPU Name: NVIDIA GeForce...如果nvidia-smi报错了NVIDIA-SMI has failed because it couldnt communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.大概率是驱动没加载常见原因是 Secure Boot。重启电脑进 BIOS把Secure Boot设置为Disabled方案B根据型号安装1.鄙人的在window里面查的信息为NVIDIA GeForce RTX5060Laptop GPU Windows 驱动582.05 Windows 显示 CUDA Version13.0需要在双系统中安装NVIDIA Driver580 CUDA Toolkit12.82.进入ubuntu先查系统和显卡lspci|grep-invidia根据型号找到需要对应安装的驱动3.安装NVIDIA驱动更新系统sudoaptupdatesudoaptupgrade-y添加 NVIDIA 驱动 PPA因为 RTX 5060 很新建议加 PPA 获取更新驱动sudoadd-apt-repository-yppa:graphics-drivers/ppasudoaptupdate查看系统推荐驱动ubuntu-drivers devices你要优先装带 recommended 的那个根据推荐结果安装驱动情况 A看到 nvidia-driver-580 推荐sudoaptinstall-ynvidia-driver-580情况 B看到 nvidia-driver-580-open 推荐sudoaptinstall-ynvidia-driver-580-open重启并检查驱动在这里插入代码片sudoreboot重启进入 Ubuntu 后执行nvidia-smi笔记本双显卡设置系统里可能还有 Intel 核显如果 nvidia-smi 正常但 Python 里找不到 GPU可以安装并设置 primesudoaptinstall-ynvidia-prime切到 NVIDIA 模式sudoprime-select nvidiasudoreboot安装 CUDA Toolkit根据Ubuntu二选一注意根据自己的CUDA来选择cuda-toolkit-12-8Ubuntu 22.04 安装 CUDA 12.8cd~wgethttps://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.debsudodpkg-icuda-keyring_1.1-1_all.debsudoaptupdatesudoaptinstall-ycuda-toolkit-12-8Ubuntu 24.04 安装 CUDA 12.8cd~wgethttps://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.debsudodpkg-icuda-keyring_1.1-1_all.debsudoaptupdatesudoaptinstall-ycuda-toolkit-12-8配置 CUDA 环境变量echoexport CUDA_HOME/usr/local/cuda-12.8~/.bashrcechoexport PATH/usr/local/cuda-12.8/bin:$PATH~/.bashrcechoexport LD_LIBRARY_PATH/usr/local/cuda-12.8/lib64:$LD_LIBRARY_PATH~/.bashrcsource~/.bashrc检查nvcc--version正常应该看到Cuda compilation tools, release12.8
【 linux下的CUDA安装】
linux下的CUDA安装最终效果需要能够实现三个命令nvidia-smi看到显卡nvcc--version看到CUDA版本通常是12.1python -PY import torch print(torch.cuda.is_available()) print(torch.cuda.get_device_name(0)) PY输出为True 你的 NVIDIA 显卡型号1查显卡型号在window下方法一打开任务管理器性能 → GPU方法二PowerShell 里面执行nvidia-smi如果 NVIDIA 驱动正常会显示类似---------------------------------------------------------------------------------------|NVIDIA-SMI552.22Driver Version:552.22CUDA Version:12.4||GPU Name...|0NVIDIA GeForce RTX5060... ---------------------------------------------------------------------------------------你要记三个东西GPU Name Driver Version CUDA Version2在ubuntu安装2.1先确定ubuntu能识别到显卡lspci|grep-invidia如果有.01:00.0 VGA compatible controller: NVIDIA Corporation GA106[GeForce RTX3060]说明 Ubuntu 识别到硬件。如果没输出可能是BIOS 没开独显笔记本混合显卡设置问题驱动还没装但一般 lspci 仍应能看到2.2 安装驱动方案A自动安装驱动先能拉下载的驱动包sudoaptupdatesudoaptupgrade-ysudoaptinstall-yubuntu-drivers-common可以查看推荐的驱动ubuntu-drivers devices然后自动安装sudoubuntu-drivers autoinstall安装完重启sudoreboot重启后检查nvidia-smi正常应该显示NVIDIA-SMI xxx.xx Driver Version: xxx.xx CUDA Version:12.x GPU Name: NVIDIA GeForce...如果nvidia-smi报错了NVIDIA-SMI has failed because it couldnt communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.大概率是驱动没加载常见原因是 Secure Boot。重启电脑进 BIOS把Secure Boot设置为Disabled方案B根据型号安装1.鄙人的在window里面查的信息为NVIDIA GeForce RTX5060Laptop GPU Windows 驱动582.05 Windows 显示 CUDA Version13.0需要在双系统中安装NVIDIA Driver580 CUDA Toolkit12.82.进入ubuntu先查系统和显卡lspci|grep-invidia根据型号找到需要对应安装的驱动3.安装NVIDIA驱动更新系统sudoaptupdatesudoaptupgrade-y添加 NVIDIA 驱动 PPA因为 RTX 5060 很新建议加 PPA 获取更新驱动sudoadd-apt-repository-yppa:graphics-drivers/ppasudoaptupdate查看系统推荐驱动ubuntu-drivers devices你要优先装带 recommended 的那个根据推荐结果安装驱动情况 A看到 nvidia-driver-580 推荐sudoaptinstall-ynvidia-driver-580情况 B看到 nvidia-driver-580-open 推荐sudoaptinstall-ynvidia-driver-580-open重启并检查驱动在这里插入代码片sudoreboot重启进入 Ubuntu 后执行nvidia-smi笔记本双显卡设置系统里可能还有 Intel 核显如果 nvidia-smi 正常但 Python 里找不到 GPU可以安装并设置 primesudoaptinstall-ynvidia-prime切到 NVIDIA 模式sudoprime-select nvidiasudoreboot安装 CUDA Toolkit根据Ubuntu二选一注意根据自己的CUDA来选择cuda-toolkit-12-8Ubuntu 22.04 安装 CUDA 12.8cd~wgethttps://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.debsudodpkg-icuda-keyring_1.1-1_all.debsudoaptupdatesudoaptinstall-ycuda-toolkit-12-8Ubuntu 24.04 安装 CUDA 12.8cd~wgethttps://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.debsudodpkg-icuda-keyring_1.1-1_all.debsudoaptupdatesudoaptinstall-ycuda-toolkit-12-8配置 CUDA 环境变量echoexport CUDA_HOME/usr/local/cuda-12.8~/.bashrcechoexport PATH/usr/local/cuda-12.8/bin:$PATH~/.bashrcechoexport LD_LIBRARY_PATH/usr/local/cuda-12.8/lib64:$LD_LIBRARY_PATH~/.bashrcsource~/.bashrc检查nvcc--version正常应该看到Cuda compilation tools, release12.8
