The idea comes from this scratch version of LSTM: https://www.kaggle.com/rayjan/lstm-neural-network-from-scratch-106ffb/edit The main goal of this subject is to learn the concept of LSTM with PyTorch that you can also compare with the scratch version on Kaggle link above. This toy code doesn't carefully setted up with exception, it can cause error if you change some hyperparameters such as BATCH_SIZE.
