Re:Facebook open-sources deep-learning modules
by Houshalter on Friday January 16, 2015 @09:36PM
(#8903123)
Deep neural networks are usually trained on GPUs which can provide a huge speed up, and is much cheaper. However they can easily be run on CPUs after being trained. I don't see why they wouldn't have access to GPUs in production though.
Re:Facebook open-sources deep-learning modules
by raverbashing on Friday January 16, 2015 @03:36PM
(#8901514)
Well, it depends, you can definitely have servers with NVidia hardware, and you don't need to have all of them with it. (Amazon offers GPGPU servers on EC2 for example)
Re:Facebook open-sources deep-learning modules
by bhouston on Friday January 16, 2015 @07:00PM
(#8902594)
Amazon only offers very costly Tesla and Quadro cards on their GPGPUs. Super expensive. For scale, if you do not need the extra memory on the Tesla cards, what you want is commodity NVIDIA GeForce cards that are a fraction of the cost.
Re:Facebook open-sources deep-learning modules
by polskibus on Saturday January 17, 2015 @02:33AM
(#8903781)
Standard cards are handicapped for double precision operations though (1/8 speed afaik).
Re:Facebook open-sources deep-learning modules
by p1esk on Saturday January 17, 2015 @04:48AM
(#8903984)
You don't need double precision for typical DL tasks, such as training a convnet.
Re:Facebook open-sources deep-learning modules
by raverbashing on Saturday January 17, 2015 @05:50AM
(#8904060)
Double precision is overrated :) (at least for ML)
I think it is more likely that this is just research software.