Closed salilgupta1 closed 4 years ago
The fix looks good.
For posterity, I think this is the analogous part of joblib that checks for None for the multiprocessing backend.
Are there already tests running through this method? If so we might want to add this test case.
I didn't find any tests for this function.
sklearn==0.22.0
sets the default forn_jobs
toNone
forLogisticRegression
models https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html#sklearn.linear_model.LogisticRegressionThis is a breaking change from
0.19.x
which set it to 1. Thejoblib
library accounts for this already in a recent bug fix but we don't in our custom backend. Here's the fix.