Thanks Nvidia for the helpful error message.

Oh yeah and this is the actual output on the console:

@ng1905

Don't even get me started on all the bugs with cudnn

NVIDIA is so much easier to work with on Windows, seriously. I wonder why their Linux support is so lackluster, given that a lot of work in HPC is on Linux.

fuck you, nvidia

@ng1905 I see you are using Pop!_OS. Welcome to this side of the world!

@RememberMe exactly my sentiments. I am effectively back to Window because of that.

@kortnevdmitriy Installed it yesterday and I am already a fan of it 😁

They realized that average user won't read the error message anyway🤔😂.

Just, be careful!

@kortnevdmitriy ayy I use PopOS as well, its pretty damn good imo

The 39x didn't work for me, but the 385 works perfectly with TF 1.7, cuDNN 7.0.5 and CUDA 9

Is this pop os ?? I have nvidia GPU and if it got bugs I would switch to Ubuntu !

@ShivamRawat0l The bug also exists in ubuntu, it's a problem with the Linux kernel. You can update the kernel (I used 4.16.0-rc7) and it will work perfectly.

@ng1905 how did you update the kernel

@ShivamRawat0l You can get all linux kernels at https://kernel.org (if the version you want is not provided by your package manager)

@ShivamRawat0l If you prefer .deb files for easier installation, see http://kernel.ubuntu.com/~kernel-pp...

@ng1905 yeah I would first install pop os then try these meathods

because you are using GTX card, this problem happens much less often for scientific cards on cluster environments...you should thank Nvidia for actually provide you proprietary drivers, the default nouveau driver on Linux was probably outdated since Bush presidency.