Hi, I am just passing this error along to the joblib devs.

I am passing large numpy arrays to a memoized function. My guess is it taking a long time to hash those arrays. Maybe there is a staged way to hash them so that you can make the comparison faster?