you can play with no GPU. Van Gogh painting: Ubuntu TensorFlow CPU Edition
July Online Development/marketing team Xiao Zhe, Li Wei, July
Date: September 27, 2016
First, preface
September 22, our development/marketing team of two colleagues using DL study Van Gogh painting, Installation Cuda 8.0 times countless pits, many friends seek refuge from the pit. Therefore, 3 days later, September 25, the tutorial will teach you from start to finish using DL study Van Gogh painting: GTX 1070 Cuda 8.0 tensorflow GPU version released. But it's embarrassing that many students don't have a GTX 1070, or even a decent GPU.
For friends in Beijing, we can provide 1070 of machines for free to experiment, but for more friends who are not in Beijing? In order for everyone to have a play, we hereby release the CPU version of this tensorflow painting tutorial. You'll see that building content is a lot simpler compared to GPU versions, but the downside is that there's no GPU, and the computational process is very tough, i7-6700 about 1.5h (you know, after the GPU is done, the final calculation process is a matter of minutes).
But anyway, before you said no GTX 1070 so did not go, and now there is a CPU can run, not limited by the hardware conditions, why not try it now? I hope more friends will play more interesting and fun experiments with us.
Second, installation2.1, Installation TensorFlow
Open Browser under Windows , type:https://github.com/tensorflow/tensorflow
Download Linux Cpu-only:python2 For example, the back Steps 3.1 This file will be used
2.2 Installing Neural-style
Then enter in the Address bar:https://github.com/anishathalye/neural-style
2.3 Installing Vgg19
Then enter in the Address bar:http://www.vlfeat.org/matconvnet/models/beta16/imagenet-vgg-verydeep-19.mat
a download dialog box pops up and downloads directly.
The downloaded file is placed in the root directory of the neural-style folder just below the line , that is, I downloaded the copy good file .
PS: Remember to put all of the above folders on your own U- disk for use under Ubuntu .
2.4, Installation Ubuntu14.04
installed Ubuntu16.04, the CPU is not required for Ubuntu versions.
Download Ubuntu14.04:http://www.ubuntu.com/download/alternative-downloads(select 64-bit download)
UltraISO Soft Disc pass:
http://cn.ultraiso.net/xiazai.html()
Next, install directly:
http://jingyan.baidu.com/article/eb9f7b6d8536a8869364e813.html
If you encounter problems, see the GPU version for instructions on installing Ubuntu. ^__^
Third, the experiment 3.1 preparatory work
Open the Ubuntu Command window , type:sudo–i
(Your password)
Apt-get Update( update all sources)
sudo apt-get install python-pip python-dev python-scipy git
(Above command is Download Some necessary libraries for TensorFlow )
After you enter the above command, insert The U-disk that just saved the download file, and then copy the file to the Download directory (the other catalogs are also available, familiar Linux students do it by themselves ok! )
Then enter:cd/home/*** (your user name )/download/ Enter
Input: Pip install TENSORFLOW-0.10.0-CP27-NONE-LINUX_X86_64.WHL// This file is downloaded in 2.1 step
3.2 Starting the experiment
after successful input : CD Neural-style-neural-style-master into the Neural folder directory
Input :python neural_style.py--content ./examples/1-content.jpg --styles ./examples/ 1-style.jpg --output ./examples/julyedu.jpg Enter
And then wait.
Waiting for the class to finish
Waiting for school.
Waiting for the game of childhood
...
just wait (you'll see what we said before: The process of waiting is like having a baby), the calculation process can take two hours, possibly longer (look at CPU configuration and picture size).
But until the child is born, you will be doubly fond of it.
PostScript
Nothing else to say is that we will do more experiments and will take you to more experiments in the September ml algorithm class.
July Online Development/marketing team Xiao Zhe, Li Wei, July, September 27, 2016.
You can play with no GPU. Van Gogh painting: Ubuntu TensorFlow CPU Edition