Debug current code

[AngelinaZhai]

• Created one-hot encoding of vocab since prebuilt libraries (ie. GloVe, word2vec) didn't work
• Patched up dimension errors
• Added GPU support in training loop
• Combined all RNN type models into one with more configurable parameters
• Need to fix graphing (from data point per iteration to data point per epoch)
• Need to incorporate BERT embeddings for better word representation