How to overcome joblib's "TypeError: can't pickle instancemethod objects" on class method?
scikit-learn
Solution
It is probably safer to only pass Python stateless functions to `joblib.Parallel`. You can try to `import dill` to register pickling support for instance methods: https://pypi.python.org/pypi/dill (your mileage may vary though).
Problem
Given a class and 2 methods in it A, B. I am getting a "TypeError: can't pickle instancemethod objects" when trying to execute the following command while in A: ``` joblib.Parallel(n_jobs=-1)(joblib.delayed(self.B)(arg1, arg2, arg3) for arg1 in arg1_values_list) ``` I found several threads on Stackoverflow regarding instancemethod and part of them are even related to joblib, but I still didn't manage to overcome this error. EDIT: The following code: ``` from sklearn.externals.joblib import Parallel, delayed def a(self, x): print x return x*10 class c(object): def b(self): Parallel(n_jobs=-1)( delayed(a)(self, i) for i in range(10)) test_c = c() test_c.b() ``` runs ok when normally ran. but fails when trying to profile using "python -m cProfile"