Error in cuda 9.0 installation


I have a system with 3D controller: NVIDIA Corporation GV100 [Tesla V100 SXM2] (rev a1)

I was able to successfully install cuda 9.2 and cudnn 7.1.4
But then after banging my head around for an entire day trying to install/build tensorflow, I conceded to downgrading my cuda to 9.0 as the pre compiled tf binaries seem happy with them.

Please find attached error log :

Installing the NVIDIA display driver...
The driver installation is unable to locate the kernel source. Please make sure that the kernel source packages are installed and set up correctly.
If you know that the kernel source packages are installed and set up correctly, you may pass the location of the kernel source with the '--kernel-source-path' flag.

= Summary =

Driver:   Installation Failed
Toolkit:  Installation skipped
Samples:  Installation skipped

complete log :

The driver that you have (presumably) successfully installed for CUDA 9.2 will work just fine with your install of CUDA 9.0 (in fact, you can easily install CUDA 9.0 side-by-side with CUDA 9.2, and be able to use them both, with just a change of environment variables).

Since you don’t need to replace the driver, just install the CUDA 9.0 toolkit using the runfile installer as you are demonstrating, but select “no” when prompted to install the display driver.

This assumes, of course, that you are starting with a proper functioning CUDA 9.2 installation, as you have indicated you were successful with.

Also, NGC offers you the possibility to run well-maintained TF (and other!) containers on your existing setup, without having to downgrade your install of CUDA 9.2 to CUDA 9.0:

You can follow the instructions given here:

to prep your machine for use of NGC containers.

Thanks a lot. This solved my problem.

I have cuda 9.2 driver and toolkit currently also installed only the cuda 9.0 toolkit and updating

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-9.0/lib64 >> ~/.bashrc

runs the tf 1.8 pre-compiled binaries available.