word2vec.py - reset_from bug

[bruderjakob12]

Description

gensim.models.word2vec: reset_from seems to use the wrong variables for vocab and index2word

Steps/Code/Corpus to Reproduce

from gensim.models import word2vec

sentences = ['human', 'machine']
model = word2vec.Word2Vec()
model.build_vocab(sentences)
new_model = word2vec.Word2Vec()
new_model.reset_from(model)

Expected Results

new_model initialized with settings from model

Actual Results

Traceback (most recent call last):
File "test.py", line 8, in
new_model.reset_from(model)
File "/usr/local/lib/python2.7/dist-packages/gensim/models/word2vec.py", line 734, in reset_from
self.wv.vocab = other_model.vocab
AttributeError: 'Word2Vec' object has no attribute 'vocab'

Versions

gensim 1.0.1

Fix

gensim/models/word2vec.py

Line 734 change
self.wv.vocab = other_model.vocab
to
self.wv.vocab = other_model.wv.vocab

Line 735 change
self.wv.index2word = other_model.index2word
to
self.wv.index2word = other_model.wv.index2word

[piskvorky]

Thank you for reporting!

@tmylk needs unit tests -- or was this solved in #1230?

[bruderjakob12]

I guess it is fixed by #1230