1.0.0

• Added method for finding the most salient words of an input sentence and added example notebook. (#17)
• Fixed some bugs with mask_zero and enforced keys in vocab_map.
• Refactored get_embedding_weights and made it so adapt_embeddings is only an argument for the model constructor.

0.1.1

New features:

• Implemented basic SWEM-max model (#1)
• Made it so dropout is applied to the embeddings rather than the output of the max-pooling layer (#6)
• Added option to initialize the model with pre-trained GloVe embeddings and adapt them (#7)
• Added option to freeze the embedding layer (#15)
• Implemented global explainability for word embedding components, as proposed in section 4.1.1 of the original paper (#2)