Search code examples
pythongensim

Index2word in Gensim's Doc2vec raises an Attribute error

I trained a doc2vec (gensim.models.Doc2Vec) model and now I'm using this line:

print(dict([(model.index2word[i], similarity) for i, similarity in enumerate(model.similar_by_word('igdumd32.dll@0x', topn=False))])['igdumd64.dll@0x'])

but it yields this error: AttributeError: 'Doc2Vec' object has no attribute 'index2word'

I am using gensim 1.0.1

Can you help?

Solution

  • The index2word list of word-vectors has moved to the wv property of the model in recent gensim versions, so where you would say model.index2word you must now use model.wv.index2word.

    (Note that this is still just word-vectors, which are only trained by the "DM" dm=1 Doc2Vec modes. Doc-vectors are in the model.docvecs object, and you can see a list of the string tags to which doc-vectors may be associated in model.docvecs.offset2doctag.)