使用CUDA进行深度学习(pytorch)导航

三种方法,任选其一,推荐方法1、2

方法1、CUDA,pip/conda pytorch

CUDA

先去NVIDIA官网找CUDA,安装CUDA,再去pytorch官网找pytorch,安装pytorch
最新版CUDA https://developer.nvidia.com/cuda-downloads
历史版本CUDA https://developer.nvidia.com/cuda-toolkit-archive
根据需要选择

pip/conda pytorch

pytorch https://pytorch.org/get-started/locally/
历史版本pytorch https://pytorch.org/get-started/previous-versions/
根据需要选择

方法3、NVIDIA Driver,pip/conda pytorch

NVIDIA Driver

Windows驱动 https://www.nvidia.cn/Download/index.aspx?lang=cn
ubuntu 驱动

sudo apt search nvidia-driver | grep nvidia-driver

找一个合适的且较新的即可,还是根据需求选择

pip/conda pytorch

pytorch https://pytorch.org/get-started/locally/
历史版本pytorch https://pytorch.org/get-started/previous-versions/
根据需要选择

方法3、NVIDIA Driver,NVIDIA Docker

NVIDIA Driver

Windows驱动 https://www.nvidia.cn/Download/index.aspx?lang=cn
Linux 驱动 ,以Ubuntu举例

# 命令行输入
sudo apt search nvidia-driver | grep nvidia-driver

找一个合适的且较新的即可,还是根据需求选择

NVIDIA Docker

NVIDIA Docker指南 https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#docker

猜你喜欢

转载自blog.csdn.net/CSDN_Ethan2086/article/details/131260130