files under /usr/local/cuda-10.0 direcotry..


I have both /usr/local/cuda-10.0 and /usr/local/cuda-9.0 directories.
When I do ls under /usr/local/cuda-10.0/lib64 directory, I get libnppist_static.a
libcudadevrt.a libnppc_static.a libnppisu_static.a
libcudart_static.a libnppial_static.a libnppitc_static.a
libcufft_static.a libnppicc_static.a
libcufft_static_nocallback.a libnpps_static.a

I think I should have files and links.
and I thought I installed cudnn but can’t find those files there.
I need to run tensorflow (which expects CUDA 10.0), do I have re-install cuda-10.0?
(I tested with and it seems running ok with GPUs but complains it cannot find files. (so tensorflow is looking for cuda 10.0)
Is this normal? or something is messed?
I’m considering re-installing cuda-10.0 and cudnn version for that.
Please give me some advices. Thank you!


the 10.1 stuff should not be in cuda-10.0 directory. That is wrong, not sure how that happened.

If it were me, I would delete /usr/local/cuda-10.0 before doing the reinstall.

I removed cuda and nvidia related package using dpkg --purge commands.
Then I rebooted the computer (ubuntu 16.04), and I tried the cuda 10.0 install commands below.

`sudo dpkg -i cuda-repo-ubuntu1604-10-0-local-10.0.130-410.48_1.0-1_amd64.deb`
`sudo apt-key add /var/cuda-repo-<version>/`
`sudo apt-get update`
`sudo apt-get install cuda`

at the last command, I get

ckim@etri-supermicro:~/Downloads/CUDA10.0$ sudo apt-get install cuda
[sudo] password for ckim:
Reading package lists… Done
Building dependency tree
Reading state information… Done
You might want to run ‘apt-get -f install’ to correct these:
The following packages have unmet dependencies:
cuda : Depends: cuda-10-0 (>= 10.0.130) but it is not going to be installed
libcuinj64-7.5 : Depends: libcuda1 (>= 352.39) but it is not installable or
E: Unmet dependencies. Try ‘apt-get -f install’ with no packages (or specify a solution).

Could you recommend what I should do?