This is an implementation of the paper "Text Understanding from Scratch" in Keras 2.x.
The model can be found as the class crepeK2 in crepe.py. This contains wrappers in order to fit, evaluate and predict.
A companion Jupyter Notebook file is provided showing an example of using the model for the AG News Dataset. The AG News Dataset is included within the git repository. Other datasets can be found on this Google Drive: (https://drive.google.com/drive/folders/0Bz8a_Dbh9Qhbfll6bVpmNUtUcFdjYmF2SEpmZUZUcVNiMUw1TWN6RDV3a0JHT3kxLVhVR2M)