gensim version

[engineersunny]

Hi, I'm using gensim 3.8.0 but still gets this error when I run extract_features().

'Doc2Vec' object has no attribute 'neg_labels'

Is there a way to avoid this?

[lowecg]

That's strange.

I found this quote this on an SO post:

that error resembles a very-old bug which only showed up if Gensim was not fully installed to have the necessary Cython-optimized routines for fast training/inference operations. (That caused some older, seldom-run code to be run that had a dependency on the missing neg_labels. Newer versions of Gensim have eliminated that slow code-path entirely.)

Would you mind bumping the dependency version to gensim 3.8.3 and trying again? According to the release notes, there's a particular fix, "Fix missing C extensions", which might help given the above quote.

[lowecg]

I just tested gensim 3.8.3

I had to edit sherlock/features/paragraph_vectors.py and comment out the line:

assert gensim.models.doc2vec.FAST_VERSION > -1, "This will be painfully slow otherwise"

I ran the notebook 01-data-preprocessing.ipynb from a fresh clone of the repo. It ran cleanly and there was no observable impact on the performance despite the scary message from the assertion we just disabled.

[madelonhulsebos]

Thanks for reporting this issue @engineersunny, and testing and sharing your solution @lowecg. @engineersunny did this solution work for you?

@lowecg I did not encounter this issue myself, do you know if it this is a general issue (e.g. it happens in the extract_features_to_csv() function as well)? Do you think we should comment this line by default?

[engineersunny]

@madelonhulsebos I got that assertion error each time when I ran the code as well so I commented out earlier but still getting the error message above. It might be a gensim version issue as I couldn't install the old version on my machine.