Torch implementation of recurrent network architecture that predicts the number of upvotes a reddit comment will get.
The architecture has an encode and decode stage. The encode stage takes in char values. The decode stage is prompted by an EOS tag (-1 in the current implementation) and outputs a number representing the predicted votes.
Note: This project is for learning purposes. The code is heavily influenced and at many points adapted (especially the util directory) from Andrej Karpathy's char-rnn.