gensim word2vec - array dimensions in updating with online word embedding

Question

Word2Vec from gensim 0.13.4.1 to update the word vectors on the fly does not work.

model.build_vocab(sentences, update=False)

works fine; however,

model.build_vocab(sentences, update=True)

does not.

---

I am using this website to try and emulate what they have done; hence I use the following script at some point:

model = gensim.models.Word2Vec()
sentences = gensim.models.word2vec.LineSentence("./text8/text8")
model.build_vocab(sentences, keep_raw_vocab=False, trim_rule=None, progress_per=10000, update=False)
model.train(sentences)

However while this runs with update=False, using update=True gives me the following traceback:

Traceback (most recent call last):
  File "word2vecAttempt.py", line 34, in <module>
    model.build_vocab(sentences, progress_per=10000, update=True)
  File "/home/brownc/anaconda3/lib/python3.5/site-packages/gensim/models/word2vec.py", line 535, in build_vocab
    self.finalize_vocab(update=update)  # build tables & arrays
  File "/home/brownc/anaconda3/lib/python3.5/site-packages/gensim/models/word2vec.py", line 708, in finalize_vocab
    self.update_weights()
  File "/home/brownc/anaconda3/lib/python3.5/site-packages/gensim/models/word2vec.py", line 1070, in update_weights
    self.wv.syn0 = vstack([self.wv.syn0, newsyn0])
  File "/home/brownc/anaconda3/lib/python3.5/site-packages/numpy/core/shape_base.py", line 230, in vstack
    return _nx.concatenate([atleast_2d(_m) for _m in tup], 0)
ValueError: all the input array dimensions except for the concatenation axis must match exactly

Answer 1

I was able to reproduce your error. I think you're calling update=True when the model is not trained yet. You should only call it when it has been pre-trained.

This works:

import gensim

model = gensim.models.Word2Vec()
sentences = gensim.models.word2vec.LineSentence("text8")
model.build_vocab(sentences, update=False)
model.train(sentences)

model.build_vocab(sentences, update=True)
model.train(sentences)

But this will fail:

import gensim

model = gensim.models.Word2Vec()
sentences = gensim.models.word2vec.LineSentence("text8")
model.build_vocab(sentences, update=True)
model.train(sentences)

ValueError: all the input array dimensions except for the concatenation axis must match exactly

Using the latest version of gensim 0.13.4.1.