/dynamic-coattention-network

Tensorflow implementation of Dynamic Coattention Networks for Question Answering.

Primary LanguagePythonMIT LicenseMIT

Dynamic Coattention Networks For Question Answering

Tensorflow implementaion of Dynamic Coattention Networks.

This repo contains work on progress.

Requirements

  • Python 3.5.2
  • TensorFlow r0.11
  • spaCy 1.2.0

Data

Retrieve json training files from SQuAD.

dataset.py contains code to create a dataset and vocab for training network.

Training the network

In order to train the network, execute

python main.py --mode train

Credit

Modified codes are originally from textsum.