What I cannot create, I do not understand. -- Feynman
This repo contains TensorFlow re-implementation of interesting models from random papers I read. I use the low-level API as much as possible since research models almost never fit into the standard API.
Personally I use editable installs to install the modules locally in a virtual environment created with virtualenvwrapper since the changes made locally will be readily available in local environment without re-installing the package.
python install -e path/to/tfmodelsIt should be also be easy to integrate code from this repo to your project simply by copy-and-paste.
- Text classification. Word-level CNN models, char-level model.