|
Centripetal Horse posted:I think he just means something like Amazon EC2. I have a GPU 2xlarge instance running, right now, training a recurrent neural network to spew humorous text. You can get them for less than $.15 per hour if you use spot instances. Just set your maximum bid up near the actual per-hour price, and you'll usually be good for hours. You can also get the 8xlarge, which has monster CUDA power. So, I got all the technical stuff, I just need some help understanding how to work with Amazon.
|
# ? Mar 30, 2016 17:30 |
|
|
# ? Apr 18, 2024 15:48 |
|
Centripetal Horse posted:I only did 500 iterations, because it's sharing the GPU with my RNN project: Nice! Will you be willing to take over requests in the thread?
|
# ? Mar 30, 2016 17:31 |
|
Are we allowed to double-dip requests? Because I got a great idea for whomever can mash these two togethers
|
# ? Mar 30, 2016 17:33 |
KoRMaK posted:Can you help me out and point me in the direction to get that stuff setup? I have this thing running on a local VM, and its chugging along but Its taking all my processing power so I gotta turn it off. It looks like it will take me 4 hours to do 100 iterations. Paladinus posted:Nice! Will you be willing to take over requests in the thread? Sure. I can help with both of those things later on tonight.
|
|
# ? Mar 30, 2016 17:34 |
|
with with with with
|
# ? Mar 30, 2016 17:37 |
|
Centripetal Horse posted:Sure. I can help with both of those things later on tonight. No problem! I can manage this for a while longer.
|
# ? Mar 30, 2016 17:41 |
|
Centripetal Horse posted:Sure. I can help with both of those things later on tonight.
|
# ? Mar 30, 2016 17:43 |
|
Paladinus posted:
Sweeeet. Spoonbill in assorted Japanese fabrics.
|
# ? Mar 30, 2016 17:55 |
|
Centripetal Horse posted:I only did 500 iterations, because it's sharing the GPU with my RNN project: Was it easy to setup? edit: that is to say, were you able just to follow the step by step and it worked? I dont have a server but I have a OC'd skylake + 980ti since it apparently uses cuda penus penus penus fucked around with this message at 18:04 on Mar 30, 2016 |
# ? Mar 30, 2016 18:00 |
|
THE DOG HOUSE posted:Was it easy to setup? I'm going through it right now. The linux part is super easy, took me like 10 minutes of copy/pasting some commands. What is holding me up now is setting up Amazon EC2. I've never done it before, so I have a lot of questions with everything I'm selecting. Like, for instance, should I first create an AMI and then use that? If so, how the hell do I do that. e: To your "e" question - yes, its dirt simple.
|
# ? Mar 30, 2016 18:03 |
|
KoRMaK posted:I'm going through it right now. The linux part is super easy, took me like 10 minutes of copy/pasting some commands. Sweet the results are so insanely interesting I'm pretty excited to try it But I've never used linux so this is probably going to be a major headache fumbling around
|
# ? Mar 30, 2016 18:05 |
|
I really liked how this one turned out, even though its not that crazy, it looks like a screen shot from some sort of weird Dune adventure game
|
# ? Mar 30, 2016 18:12 |
|
i used the one website someone linked earlier to make this one
|
# ? Mar 30, 2016 18:16 |
KoRMaK posted:I'm going through it right now. The linux part is super easy, took me like 10 minutes of copy/pasting some commands. Use a g2.2xlarge instance. Select CentOS 7.2 with cloud init from the community AMIs. The same picture as before, but done with 2,000 iterations. Edit: The Amazon instance costs me like $20/day, so I won't be able to take huge numbers of requests.
|
|
# ? Mar 30, 2016 18:20 |
|
|
# ? Mar 30, 2016 18:25 |
|
Paladinus posted:Here's your monster dog. drat Thats straight out of the movie "The Thing"
|
# ? Mar 30, 2016 18:26 |
|
Centripetal Horse posted:Use a g2.2xlarge instance. Select CentOS 7.2 with cloud init from the community AMIs. Oh man, but I'm only use to ubuntu Will it be mostly the same? e: there's two AMIs like that, hvm and para - I'm guessing I should go with a HVM if I want to take advantage of CUDA? KoRMaK fucked around with this message at 18:30 on Mar 30, 2016 |
# ? Mar 30, 2016 18:27 |
|
Oh boy. Can't wait for pages of reactions from this.
|
# ? Mar 30, 2016 18:32 |
|
Huh.
|
# ? Mar 30, 2016 18:37 |
|
KoRMaK posted:Oh man, but I'm only use to ubuntu Will it be mostly the same? they should be effectively the same for what you need. there are different commands to install packages
|
# ? Mar 30, 2016 18:38 |
I am doing this one. Can't promise it will be real soon. KoRMaK posted:Oh man, but I'm only use to ubuntu Will it be mostly the same? The package names will be different, since CentOS uses yum instead of apt-get, but it's similar. I had some issues using GPU with Ubuntu on Amazon, so I stick with CentOS. Edit: Yes, HVM.
|
|
# ? Mar 30, 2016 18:43 |
|
Gatekeeper posted:ahh drat I meant for the order to be switched. still magnificent tho, thanks!! Here you go.
|
# ? Mar 30, 2016 19:04 |
|
i seriously dont even want to see this one
|
# ? Mar 30, 2016 19:20 |
|
Use goatse as the first pic.
|
# ? Mar 30, 2016 19:26 |
|
traumatize me lord internet
|
# ? Mar 30, 2016 19:39 |
|
If this ends up being less than stellar, please Goatse the seed pods.
|
# ? Mar 30, 2016 19:55 |
|
|
# ? Mar 30, 2016 19:56 |
|
Well for a little trip report for anybody who cares I just installed Ubuntu for the first time ever (followed the step by step on their website) and am pretty much finished with the installation of the ostagram crap not knowing anything at all about linux so if youre interested dont be afraid
|
# ? Mar 30, 2016 20:00 |
|
THE DOG HOUSE posted:Well for a little trip report for anybody who cares I just installed Ubuntu for the first time ever (followed the step by step on their website) and am pretty much finished with the installation of the ostagram crap not knowing anything at all about linux so if youre interested dont be afraid Are you using virtual box or Amazon? I'd like to try it out but last I checked you can't use CUDA through virtual box
|
# ? Mar 30, 2016 20:04 |
|
This one's positively magical...
|
# ? Mar 30, 2016 20:07 |
|
gibb3h posted:Are you using virtual box or Amazon? I'd like to try it out but last I checked you can't use CUDA through virtual box I am using virtual box here at work just to give it a little test run but im expecting performance to be awful of course. When I get home im going to just install it as dual boot on my computer so I get all the sweet performance
|
# ? Mar 30, 2016 20:07 |
|
E2: Actually this is a better fit for Meatship Fender Anarchist fucked around with this message at 20:22 on Mar 30, 2016 |
# ? Mar 30, 2016 20:11 |
|
EorayMel posted:Are we allowed to double-dip requests? Because I got a great idea for whomever can mash these two togethers
|
# ? Mar 30, 2016 20:32 |
It's a pity the one us plebs use takes hours cause I feel like I could spend some time finding the right giger picture and giant robot to combine.
|
|
# ? Mar 30, 2016 20:33 |
|
Paladinus posted:No problem! I can manage this for a while longer. Are you the goon that hired c-horse godbless
|
# ? Mar 30, 2016 20:44 |
|
Sheep-Goats posted:Are you the goon that hired c-horse godbless Most definitely not, whatever c-horse is.
|
# ? Mar 30, 2016 20:50 |
|
Even better than I expected, thanks.
|
# ? Mar 30, 2016 20:56 |
THE DOG HOUSE posted:i seriously dont even want to see this one It came out pretty goofy. I think the pictures being very different in dimensions has something to do with it.
|
|
# ? Mar 30, 2016 21:51 |
|
gibb3h posted:
gibb3h posted:
|
# ? Mar 30, 2016 22:04 |
|
|
# ? Apr 18, 2024 15:48 |
|
Centripetal Horse posted:It came out pretty goofy. I think the pictures being very different in dimensions has something to do with it. Jesus Christ. I've added this to the queue, so we'll be able to compare the result. I suspect they run less than 500 iterations on the website.
|
# ? Mar 30, 2016 22:05 |