LiTeFlow is a Lightweight TensorFlow extension library. It collects a bunch of functions that can help you building and testing some complex neural network layers.
LiTeFlow is not intended to be a super-layer like Keras or Lasagne: it's just a collection of handy off-the-shelf code snippets that can make us save some time when implementing some common operations. As soon as the official TensorFlow project will be provided with higher level APIs, thisl library will become useless -- I know, with such a cool name it's kind of a waste, but building something more is way beyond the scope of my work and my abilities. Said that: have fun!
Do you run a GPU or CPU? Whatever, you need to install the proper tensorflow version
before installing LiTeFlow. To use LiTeFlow, you can install it via pip running:
$ pip install git+https://github.com/petrux/LiTeFlow.git@master
If you wan to contribute to LiTeFlow or just dig a bit deeper into it, clone the repository:
$ git clone https://github.com/petrux/LiTeFlow.git
then get into the directory and run one of the script in the bin directory to create
a virtual environent to work with:
$ cd LiTeFlow
$ ./bin/py3venv.sh
We strongly recommend to use Python 3 and to run a virtual environment.
The ./bin/py3venv.sh script will create a .py3venv directory with the virtual
environment. Now, you just need to activate it and install the proper version of
TensorFlow for your machine (GPU or CPU).
$ source .py3venv/bin/activate
$ pip install tensorflow-gpu