Nvidia fuck you and your neural meme networks
Literally all that machine learning don't work. I have experience, i am using linux for 3 years as main system.
But installing:
>cuda toolkit
is hard and often fucks up, even on ubuntu
>cunn + torch
Cuda neutral networks i tried 4 fucking times, even bother to install it clean fedora, not a single chance.
>nvidia-docker
docker is kinda broken, but this fucking thing is just unbelievable it is beyond mess.
I just lost about 20 hours. Only to be able to run deepdream on nvidia-docker, neural-style don't work.
Good luck with those self driving cars
>>58046494
no post of course, i forgot that all nvidia /g/ users are gaymers.
And i bought 1060 most for the cuda capability
>>58046494
I use nvidia-docker without any problem at all. What the fuck are you doing wrong?
>using ubuntu repository software
Well, found your problem. Uninstall and install driver + CUDA from latest release.
>>58046494
>I'm shit at technology therefore the technology is shit
What I just typed up is the modus operandi of /g/.
OP IS A FAGGOT
>>58046494
>Muh neural networks
>Muh machine learning
How about you stop being a degenerate retard and actually learn how to use CUDA.
>CUDA toolkit from ubuntu
Kek
http://www.netmux.com/blog/how-to-build-a-password-cracking-rig
>I never could have imagined the hardest part was getting Nvidia drivers to install properly on Ubuntu
2017 is going to be truly linux year
>>58047334
I don't know how incompetent these retards are, just download the fucking driver from Nvidia's own website, like Windows users do.
>>58047350
Are you sure that 4 cards in sli + for parallel computing is as easy to set up on linux as on windows?
>>58047334
>placing graphic cards on both sides of the QPI
Holy shit what the fuck is that retard doing?
is there some linux distro with cuda and all the neural networks shit?
Just linux + CUDA out of the box
>>58046494
t. poojeet
>>58047386
>sli
>computing
They are not put in SLI (not the lack of SLI bridge). Not to mention that SLI will not have any effect because it basically just extends the frame buffer.
Also, yes, I run experiments on the time on a GPU cluster at work.
>>58046494
thumbs up if you read it "INDIA FUCK YOU" lol XDD
Are you the guy trying to run tensor flow in nvidia-docker and currently spamming the email-list, OP?
>>58047509
>>58047285
>>58047350
no i actually i got tensorflow and deepdream gpu image working, but anything else just don't work
i tried reinstalls (fedora/ubuntu), 2 different versions 7.5 and 8.
But all that lua, torch7, cudnn is absolutly insane, i don't know, maybe i am fucking idiot , but please explain me how to easily got it working