cuda+cuDNN和tensorflow安装

ERROR: An NVIDIA kernel module 'nvidia-drm' appears to already be loaded in your kernel.

进命令行，关闭图形化界面，ctrl+alt+f4命令行界面，ctrl+alt+f1恢复，需要重启sudo

 

cuda+cuDNN版本需要对应

主要参考：ubuntu 18.04 安装GPU +CUDA9.1+cuDNNv7.0.5（包含gcc，g++降级，英伟达驱动安装）

https://blog.youkuaiyun.com/u010801439/article/details/80483036

 

protoc: error while loading shared libraries: libprotoc.so.14: cannot open shared object file: No such file or directory

https://stackoverflow.com/questions/25518701/protobuf-cannot-find-shared-libraries

 

版本支持文件

https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#axzz3x7mwvZrG

 

Ubuntu下查看显卡型号及NVIDIA驱动版本

https://blog.youkuaiyun.com/jyl1999xxxx/article/details/78871622

 

环境变量设置

https://www.cnblogs.com/shrimp-can/p/5253672.html

 

cuDNN安装完成的验证方法和显卡安装方法（perform mok management）

https://blog.youkuaiyun.com/cskywit/article/details/81146429

CUDA Runtime Version = 9.0不能高于CUDA Driver Version = 9.1

 

tensorflow failed call to cuInit: CUDA_ERROR_UNKNOWN

https://stackoverflow.com/questions/43992230/tensorflow-in-nvidia-docker-failed-call-to-cuinit-cuda-error-unknown/51120389#51120389

 

cuDNN下载链接

https://developer.nvidia.com/rdp/cudnn-download

cuda下载链接

https://developer.nvidia.com/cuda-downloads

tensorflow安装

https://tensorflow.google.cn/install/pip