In this project we build and trained an RNN and an LSTM from scratch in PyTorch on Shakespeare poems in order to be able to generate poems. We then compared the LSTM with the RNN. In addition, we used Word2vec for token embeddings, as well as Byte-Pair Encoding (BPE) for tokenisation fo the text.
This project was a part of the course DD2424 (Deep Learning in Data Science) at KTH.