It should be clearer that Vectors.from_glove doesn't use the binary glove creates

[hoagy-davis-digges]

Glove generates a vectors.bin file and a vocab.txt file. It is not clear from the documentation that the files expected by the Vectors.from_glove method are different formats to the ones glove creates.

Which page or section is this issue related to?

https://spacy.io/api/vectors#from_glove

[adrianeboyd]

You're right that this isn't able to load the format produced by glove. I'm not entirely sure at this point, but I think this is older code that may be intended to read an internal numpy binary format rather than the one produced by glove.

Instead of this method, I'd recommend using spacy init-model to load vectors from a text file:

https://spacy.io/api/cli#init-model

You'll need to make sure the dimensions are provided in the first line (-write-header 1 option with glove). This will create and save a spacy model that you can load with spacy.load() that contains the vectors as part of the vocab (nlp.vocab.vectors).

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.