I'm trying to take advantage of the parallelizing options available in sklearn, but setting n_jobs!=1 doesn't seem to start any parallel process... When looking at the computer performance, only 12.5% of CPU capacity is being used (I've got an Intel Core i7, 4 cores, 8 threads; that is, it's using just one thread...).
I've tried with different functions of sklearn and importing my function (the one that uses the sklearn function) from a separate module, but still the same. I also found many other answers but were specifically for Linux, while I'm using Windows 10. In addition, I've tried this n_jobs functionality on Linux with no errors, so I guess it must be something between joblib and Windows/Intel, but I have no idea of what it may be... I read once something like joblib had some problems handling Intel virtual cores on Windows, but not sure if, being that the case, there's already a solution or not.
The particular line of code I'm using is:
labels = MeanShift(bandwidth=25,bin_seeding=True,n_jobs=4).fit_predict(X)
where X is a numpy matrix of 6557436x3.
I guess CPU processing capacity should be 50% when executing the code, but, as I said, it is only 12.5%...
Thanks in advance!