深度学习服务器环境配置-tensorflow

 

1、cudnn版本查询

[root@65-35-GPU Tools]# cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2

2、cuda版本查询

[root@65-35-GPU Tools]# cat /usr/local/cuda/version.txt

3、cudnn、cuda、python、tensorflow-gpu、nvidia驱动版本、glibc对应关系（建议先更新驱动，否则会出现libc.so.6删除的错误）

https://www.tensorflow.org/install/source#common_installation_problems

本人配置(官网下载需注册，而且很慢，所有下载后的工具包过段时间整理后放网盘，到时补充链接)：

Anaconda3-5.3.1-Linux-x86_64.sh
cuda_10.0.130_410.48_linux.run
cudnn-10.0-linux-x64-v7.4.2.24.tgz
glibc-2.23.tar.gz
NVIDIA-Linux-x86_64-410.129-diagnostic.run
tensorflow_gpu-1.14.0-cp37-cp37m-manylinux1_x86_64.whl

4、cuda 10.0安装

(1)安装

[root@65-35-GPU Tools]# chmod +x ~/cuda_10.0.130_410.48_linux.run
[root@65-35-GPU Tools]# ./cuda_10.0.130_410.48_linux.run
Do you accept the previously read EULA?
accept/decline/quit: accept

Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 410.48?
(y)es/(n)o/(q)uit: no

Install the CUDA 10.0 Toolkit?
(y)es/(n)o/(q)uit: yes

Enter Toolkit Location
 [ default is /usr/local/cuda-10.0 ]: 

Do you want to install a symbolic link at /usr/local/cuda?
(y)es/(n)o/(q)uit: no

Install the CUDA 10.0 Samples?
(y)es/(n)o/(q)uit: no

（2）添加环境变量（~/.bashrc）

[root@65-35-GPU Tools]# export PATH="/usr/local/cuda-10.0/bin:$PATH"
[root@65-35-GPU Tools]# export LD_LIBRARY_PATH="/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH"

(3)安装验证

[root@65-35-GPU Tools]# nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130

5、cudnn 7.4.2安装

（1）下载解压

（2）文件替换

[root@65-35-GPU Tools]# cp cuda/include/cudnn.h /usr/local/cuda/include/

[root@65-35-GPU Tools]# cp cuda/lib64/libcudnn* /usr/local/cuda/lib64/

6、NVIDIA驱动更新

[root@65-35-GPU Tools]# ./NVIDIA-Linux-x86_64-410.129-diagnostic.run