Using a mixin to cache a transform method call in scikit-learn

cached_transform_mixin.py

from sklearn.feature_extraction.text import CountVectorizer
from joblib import Memory
from sklearn.base import clone
from sklearn.datasets import fetch_20newsgroups


class CachedTransformMixin:
    memory = Memory('/tmp/cache')

    def transform(self, *args, **kwargs):
        return self.memory.cache(super(CachedTransformMixin, self).transform)(*args, **kwargs)


class CachedCountVectorizer(CachedTransformMixin, CountVectorizer):
    pass


X = fetch_20newsgroups().data

est = CachedCountVectorizer().fit(X)
%time Xt = est.transform(X)
%time Xt = est.transform(X)
est = clone(est).fit(X)
%time Xt = est.transform(X)