PyTorch implementation of Hierarchical Attention Network.
A GRU layer that takes in batches of words and encodes them
Attention Layer that takes in encoded words and generates Sentence Vectors
GRU layer that accepts documents as sentences and encodes the sentences
Attention Layer that takes in encoded sentences and generate document vectors
Monolithic class that takes in documents, and trains the network
Download the data from https://www.kaggle.com/utathya/imdb-review-dataset
Prepare the data for network using: python imdb_data.py imdb_master.csv
python train.py