cuDNN Installation Error : Cuda Failurer version : GCC 5.4.0

I would try to install tensorflow.
I installed Cuda Toolkit 9.1.

I tried to install cuDNN properly :
cuDNN v7.0.5 Runtime Library for Ubuntu16.04 (Deb)
cuDNN v7.0.5 Developer Library for Ubuntu16.04 (Deb)
cuDNN v7.0.5 Code Samples and User Guide for Ubuntu16.04 (Deb)

$ ./mnistCUDNN

cudnnGetVersion() : 7005 , CUDNN_VERSION from cudnn.h : (7005) (7.0.5)
Cuda failurer version : GCC 5.4.0
Error : unknown error
error_util.h:93
Aborting…

So I think cuDNN isn’t installed properly.
How can it be solved ?

Ubuntu 16.04
GPU GTX 1080 ti
64 bit
NVIDIA Driver 390.25 : NVIDIA-Linux-x86_64-390.25.run
CUDA Toolkit 9.1
Python3.6.4

Unfortunately the requirement is gcc 5.3.1 I believe. So you have to downgrade your gcc to 5.3.1 or lesser. I am planning on going to 4.9 because it is easier to install. I’ll tell you if that makes a difference

Yes, I think it maybe related to gcc version.
Gcc5.3.1 is supported one. I have gcc5.4.0.
So how can it be updated to old one ?

$ apt policy gcc
gcc : Installed : 4:5.3.1-1ubuntu1
Candidate : 4:5.3.1-1ubuntu1
Version table :
*** 4:5.3.1-1ubuntu1 500
500 http://tr.achive.ubuntu.com/ubuntu xenial/main amd64 Packages
100 /var/lib/dpkg/status

$ inxi -S
No command ‘inxi’ found

$ apt-apt-cache policy gcc-5
gcc-5:
Installed:5.4.0-6ubuntu1~16.04.9
Candidate:5.4.0-6ubuntu1~16.04.9
Version table:
*** 5.4.0-6ubuntu1~16.04.9 500
500 http://tr.archive.ubuntu.com/ubuntu xenial-updates/main amd64 Packages
500 http://security.ubuntu.com/ubuntu xenial-security/main amd64 Packages
100 /var/lib/dpkg/status
5.3.1-14ubuntu2 500
500 http://tr.archive.ubuntu.com/ubuntu xenial-updates/main amd64 Packages

http://docs.nvidia.com/cuda/cuda-installation-guide-linux/

$ ./mnistCUDNN

cudnnGetVersion() : 7005 , CUDNN_VERSION from cudnn.h : (7005) (7.0.5)
Cuda failurer version : GCC 5.4.0
Error : unknown error
error_util.h:93
Aborting…

I try do run cuda.
How will I make version correct ?

Hi:

Did you fix your problem with downgrading gcc?

I’ve got a same problem but I think as long as the compilation is finished gcc should be fine.