DL-Environment-Autosetup

This repository includes this readme and a Makefile. The Makefile allows you to easily setup a deep learning platform using keras, tensorflow and theano. The testing OS is Ubuntu 14.04.5 LST 64bit, and hardware environmnets are X79 Extream 9 with Core i7, 32Gb memory and GTX Titan Black. All processes can be done just to follow as below.

1. Make a directory to keep the Makefile and download the Makefile from this repository. Let’s assume that we keep it in “DL_folder” under your user accuount folder.

• This Makefile uses a relative path. So, YOU MAST KEEP THE MAKEFILE IN A FOLDER JUST UNDER YOUR USER ACCOUNT.

2. Open terminal and move to “DL_folder”. (E.g. cd ./DL_folder)

3. Type sudo make basic in terminal.

• This will get g++ and git libraries installed on your machine. If you already have those ones, instalations will be skipped.

4. Type sudo make blacklist in termianl.

• This will stop nouveau display driver, which is an open source display driver, but not originally from nvidia.

5. Type sudo make texton in terminal.

• This will force you to reboot as CUI mode to install a nvidia original display driver.

6. Now you’re in CUI, right? Try to login your account.

• Only text massage will appear on the screen.

7. Move to “DL_folder” and type make cudainstall in CUI.

• It will take sometime to download and install cuda. Cuda is about 1.4Gb, so be patient.
• Just before the instalation process starts, you need to type in your superuser password.
• To install cuda, you need to follow the screen guides and basically type in as below:
  • accept > y > y > y > y > Enter > y > y > Enter > “Instalation starts”

8. Type sudo make cudaecho soon after you install cuda.

• This is to configure a path for cuda.

9. Type sudo make textoff to return GUI.

• Now the computer starts rebooting.

10. Open terminal and move to “DL_folder” again.

11. Type sudo make cudnn in terminal.

• This will install cuDNN library.

12. Type make anainstall in terminal

• This step is to download and install Anaconda. Anaconda is ~500Mb.
• To install anaconda, you need to follow guides.
  • Enter > yes > Enter > yes
• If you don’t mind, better select yes to set a path to Anaconda’s python.

13. Type make pip in terminal.

• Now you install all required packages for deep learning.

14. Type make git in terminal.

• Now you can test your platform. (E.g. python ../keras/example/mnist_cnn.py)

This document and links in the document are confirmed on 22nd March 2017.

• Software versions used in this process
  • “cuDNN v5.1 Runtime Library for Ubuntu 14.04 (Deb)”
  • “cuda_8.0.61_375.26_linux.run”
  • “Anaconda3-4.3.0-Linux-x86_64.sh”