Incompatible with dask in processes scheduler

from pyforest import *
import dask

def inc1(a):
    return a + 1

def inc2(a):
    return np.add(a, 1)

print(dask.compute(map(dask.delayed(inc1), [1,2,3]), scheduler='processes')[0])
print(dask.compute(map(dask.delayed(inc2), [1,2,3]), scheduler='processes')[0])

inc1 work, inc2 don’t

[RafeyIqbalRahman]

@lumyuto I ran your code on Google Colab and got the following error:

PicklingError: Could not pickle object as excessively deep recursion required.

This error is related to the Dask library itself.

[FlorianWetschoreck]

Thank you for the bug report. We discussed this internally but had no idea on how to even start working on this problem. If you can come up with a solution that would be great.

So long, we will flag this as “help wanted” and will see if someone comes up with a solution.

In the mean time, you can also just import np explicitely because it will overwrite pyforest. So, you should be able to use

from pyforest import *
import numpy as np

Does this work for you?