Python, Keras, and Tensorflow have made neural networks easy and accessable to everyone. I personally have had a lot of trouble finding a nice and easy guide detailing how to set up all three on a system. This guide contains simple, step-by-step instructions on how to install these three things.
Anaconda is a python package manager that does a lot of stuff for you. It contains many packages including pip, numpy, scipy, etc.
Step one is to install it. Do that by going here, downloading the package for your OS, and installing it.
This step may not be necessary in the future, but for now, it is required.
open powershell or terminal and run the following command:
conda install python=3.5.2
Note, for your system to actually use the GPU, it must have a Compute Capability >= to 3.0
Go to this website and download CUDA for your OS.
Windows:
- double-click the executable and follow setup instructions
Linux:
- follow the instructions here
You will need to make a NVIDIA developer account to get these files.
Go here, login, and download CUDNN for your OS
Windows:
- Extract the folder
- Copy the files in the 3 folders to the same folders in C:\Program_Files\NVIDIA CUDA Toolkit (This requires admin permissions)
Ubuntu:
- Extract the folder
- run
sudo nautilusto open the file manager as root - copy the files from the 2 extracted folers to the same-named folders in /usr/local/CUDA
This is the last step in system setup. Just open powershell or terminal and run one of the following commands.
With GPU:
pip install tensorflow-gpu keras
Without GPU:
pip install tensorflow keras