通过滴滴云内网下载、安装Nvidia GPU驱动、CUDA Toolkit、cuDNN

滴滴云技术支持发表于:2020年01月02日 17:25:53

国内从NVIDIA官网下载GPU驱动、CUDA Toolkit、cuDNN通常比较慢,滴滴云提供了这些软件包的内网链接,在滴滴云GPU云服务器上通过内网链接下载会极大加快下载速度。

一、软件包内网链接地址

软件包名版本内网链接
NVIDIA GPU 驱动Repo Filehttp://mirrors.intra.didiyun.com/gpu_resource/centos_cuda.repo

NVIDIA-Linux-x86_64-384.66.runhttp://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-384.66.run

NVIDIA-Linux-x86_64-390.30.runhttp://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-390.30.run

NVIDIA-Linux-x86_64-410.79.runhttp://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-410.79.run

NVIDIA-Linux-x86_64-410.104.runhttp://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-410.104.run

NVIDIA-Linux-x86_64-418.56.runhttp://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-418.56.run

NVIDIA-Linux-x86_64-430.34.runhttp://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-430.34.run

NVIDIA-Linux-x86_64-440.36.runhttp://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-440.36.run
CUDA Toolkit10.1.105(Linux)http://mirrors.intra.didiyun.com/gpu_resource/cuda_10.1.105_418.39_linux.run

10.1.168(Windows)http://mirrors.intra.didiyun.com/gpu_resource/cuda_10.1.168_425.25_win10.exe

10.0.130(Linux)http://mirrors.intra.didiyun.com/gpu_resource/cuda_10.0.130_410.48_linux.run

9.1.85(Linux)http://mirrors.intra.didiyun.com/gpu_resource/cuda_9.1.85_387.26_linux.run

9.1.85(Windows)http://mirrors.intra.didiyun.com/gpu_resource/cuda_9.1.85_windows.exe

9.0.176(Linux)http://mirrors.intra.didiyun.com/gpu_resource/cuda_9.0.176_384.81_linux-run

8.0.61(Linux)http://mirrors.intra.didiyun.com/gpu_resource/cuda_8.0.61_375.26_linux-run
cuDNN7.6.5.32(CUDA10.0)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-10.0-linux-x64-v7.6.5.32.tgz

7.6.1.34(CUDA10.1)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-10.1-linux-x64-v7.6.1.34.tgz

7.6.1.34(CUDA10.0)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-10.0-linux-x64-v7.6.1.34.tgz

7.4.2.24(CUDA10.0)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-10.0-linux-x64-v7.4.2.24.tgz

7.0.5(CUDA9.1)(Windows)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-9.1-windows7-x64-v7.zip

7.0.5(CUDA9.1)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-9.1-linux-x64-v7.tgz

7.0.5(CUDA9.0)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-9.0-linux-x64-v7.tgz

7.0.1(CUDA8.0)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-8.0-linux-x64-v7.tgz

5.1.10(CUDA8.0)http://mirrors.intra.didiyun.com/gpu_resource/cudnn-8.0-linux-x64-v5.1.tgz

二、软件包安装方法

下面以某个版本举例说明NVIDIA GPU驱动、CUDA Toolkit、cuDNN的安装方法。

2.1 安装 GPU 驱动


apt install -y gcc g++ build-essential linux-headers-$(uname -r)

wget http://mirrors.intra.didiyun.com/gpu_resource/NVIDIA-Linux-x86_64-430.34.run

chmod 755 NVIDIA-Linux-x86_64-430.34.run

./NVIDIA-Linux-x86_64-430.34.run

nvidia-smi -pm 1


2.2 安装 CUDA Toolkit


wget http://mirrors.intra.didiyun.com/gpu_resource/cuda_10.1.105_418.39_linux.run

chmod 755 cuda_10.1.105_418.39_linux.run

./cuda_10.1.105_418.39_linux.run

注意上一步已经安装了 GPU 驱动,这一步安装时要跳过驱动安装。


2.3 安装 cuDNN:


wget http://mirrors.intra.didiyun.com/gpu_resource/cudnn-10.1-linux-x64-v7.6.1.34.tgz

tar zxvf cudnn-10.1-linux-x64-v7.6.1.34.tgz -C /usr/local/