wmdistance returns inf for single words in vocabulary

[Tixierae]

import gensim
# loading Google News vectors
model = gensim.models.word2vec.Word2Vec.load_word2vec_format('E:\\GoogleNews-vectors-negative300.bin.gz', binary=True)

The following:

# computing the Word Mover's Distance of an in-vocab word to itself
model.wmdistance(['obama'],['obama'])

returns inf, when obviously it should be zero (according to the documentation, float(inf) is only returned for out-of-vocab words).

Interestingly, model.wmdistance('obama','obama') does return zero, and model.wmdistance(['obama'],['obama','illinois']), model.wmdistance(['obama'],['illinois']) both work. The problem only seems to arise for single identical words.

[tmylk]

Hmm... There was a fix for when the dictionary size is 1 by @rbahumi and it should have covered this case.

[Tixierae]

Thanks! I will update my installation of the module and try again.

[tmylk]

@Tixierae did an update fix the issue?

[tmylk]

Closing as abandoned.

[Tixierae]

Sorry for the delay in response. Updating indeed fixed the issue. Thanks!