why does the variable 'vocab' there have the property - 'vectors'?

[HandanDiana]

Hi, I have some confusions about the variable **'vocab'** there. I'm the beginner in this field, so maybe these are some silly questions. But hopefully, you guys can help to give me some suggestions. Thanks!

Firstly, whether the variable 'vocab' was generated by CharVocab.from_data function as mentioned in README.md. Additionally, if so, I couldn't find a property called 'vectors'.
What the exactly type of the variable 'vocab' there? Whether I missed other key points on 'vocab'?

[davidegraff]

vectors is an attribute of a OneHotVocabulary, a subclass of CharVocab. It's uncessary though- the purpose of the class is really to initialize the embedding layer (nn.Embedding) with the identity matrix (L18). Two notes:

1. the weights of the embedding layer can be learned and there's nothing special about the identity matrix
2. the embedding size does not need to be the same as the vocab size. many VAE implementations project the vocabulary into a higher-dimensional embedding space (i.e., d_emb > n_vocab) with this typically improving performance