Sofie Van Landeghem

Because the pip-installed version of spaCy >= 2.2.0 did not distribute the `pxd` files which neuralcoref builds against. This will be fixed in spaCy in the new versions. [EDIT: this...

We're currently working hard on getting spaCy v.3 out, which will use the revamped [thinc](https://github.com/explosion/thinc.ai). Once all that is done, the next priority will be to get `neuralcoref` integrated with...

> Hi, it seems to work fine with spaCy 2.3.0 without building spaCy from source. > > ``` > # installed spacy 2.3.0 then > git clone https://github.com/huggingface/neuralcoref.git > cd...

Hm, that's odd. Did you build neuralcoref from source after installing spaCy 2.3.2 ?

Yea, that explains the segfault. You'll need to build neuralcoref from source if you want to use spacy > 2.1.

Hi @fhamborg! Since my original post in September 2019, quite a bit has changed. We initially intended to bring this code base up-to-date with the latest spaCy, but then we...

>Or are there much bigger changes to spacy that will require a rewrite of sense2vec? Yep, this is the case ;-) There have been quite some refactors, new & moved...

> This PR makes the `CategoricalCrossentropy` loss more strict only allowing `guesses` and `truths` that represent exclusive classes Why can't / shouldn't it (also) support multi-label classification?

Thanks for the additional explanation, and agreed!

I don't think it's being used in spaCy right now, but it should. And when we do, I'm proposing we work via specific versions, e.g. ``` loss_function = registry.get("losses", "CategoricalCrossentropy.v3")...