How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu

#1

Post by Aryal007 » Thu, May 31, 2018, 6:06 AM UTC

How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu
python36.com

Re: How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu

#2

Post by vivaan » Thu, May 31, 2018, 7:53 AM UTC

sounds like big news

Re: How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu

#3

Post by dfee » Thu, May 31, 2018, 8:03 AM UTC

Ah yes, python36.com. Now that’s not Python 3.6, that’s a domain registered by someone who wanted to hikack some seo juice (joose?).

It looks like they didn’t register python37.com though, which is probably a big deal as Python 3.7 is in beta.

Re: How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu

#4

Post by yonkshi » Thu, May 31, 2018, 8:34 AM UTC

To add on top of the tutorial, it's recommended to compile with AVX, SSE and FMA instructions enabled if you are using a modern Intel chipset. It has a pretty big boost for calculations that needs to be done on the CPU.

The pip version of TF does not come with AVX and FMA for some reason, so this is one of perks from compiling from source

Re: How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu

#5

Post by SebNag_ » Thu, May 31, 2018, 8:44 AM UTC

vivaan wrote: post #2

sounds like big news

Contains some buzzwords though

Re: How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu

#6

Post by arunmandal53 » Thu, May 31, 2018, 11:58 AM UTC

yonkshi wrote: post #4

To add on top of the tutorial, it's recommended to compile with AVX, SSE and FMA instructions enabled if you are using a modern Intel chipset. It has a pretty big boost for calculations that needs to be done on the CPU. The pip version of TF does not come with AVX and FMA for some reason, so this is one of perks from compiling from source

It seems like Tensorflow 1.8 with CUDA 9.2 performs up to 37% faster when compared to earlier versions of Tensorflow as described in the post. (link below) http://www.python36.com/benchmark-tensorflow-on-cifar10/

Re: How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu

#7

Post by madmax108 » Thu, May 31, 2018, 7:41 PM UTC

arunmandal53 wrote: post #6

yonkshi wrote: post #4

To add on top of the tutorial, it's recommended to compile with AVX, SSE and FMA instructions enabled if you are using a modern Intel chipset. It has a pretty big boost for calculations that needs to be done on the CPU. The pip version of TF does not come with AVX and FMA for some reason, so this is one of perks from compiling from source

It seems like Tensorflow 1.8 with CUDA 9.2 performs up to 37% faster when compared to earlier versions of Tensorflow as described in the post. (link below) http://www.python36.com/benchmark-tensorflow-on-cifar10/

I've been playing around with a couple DL frameworks recently, so was wondering, what is the performance tradeoff between what you mention and PyTorch? Is it significantly different? Because I enjoy the pyhonic "style" of PyTorch way more than the graph creation/precomputation method of TensorFlow.