Closed jonringer closed 5 years ago
Hi @jonringer , As you've noticed, textacy
isn't compatible with this v0.21 of scikit-learn
. APIs and dependencies change, and that means you can't always support the full range of versions. Currently, textacy supports v0.18 through v0.20 — see here — but I'll keep this issue in mind when the package is in a position to update its support.
Sounds good, it just looks like you defaulted the value, and I'm not sure how import n_topics
is to what problem it's trying to solve. But if the scikit-learn community removed it, then my assumption is the argument had limited impact.
But, it's on your radar, so deal with it when appropriate :). This seems to be the only blocker to adopting scikit-learn 0.21, according to the test suite.
The bigger issue with v0.21 is that it's PY3-only, and I'm still supporting PY2.7. Happily, that's going to change soon, so I may bump the scikit-learn
version requirements when I transition to a PY3-only codebase. :+1:
steps to reproduce
expected vs. actual behavior
possible solution?
just remove the optional argument and don't pass it through to the other constructor?
context
passes the argument
n_topics
to LDA, however, the argument was removedhttps://github.com/scikit-learn/scikit-learn/blob/7b346e784b1391815bd17f64fd239d3fbd78dd5a/sklearn/decomposition/online_lda.py#L277
environment
spacy
version:spacy
models:textacy
version: 0.8