Defining inside main() in multiprocess will report error

[hainingpan]

Can I define the function inside main() when using MPIPoolExecutor()?
Here is a MWE:

import adaptive
import numpy as np
from mpi4py.futures import MPIPoolExecutor


def main():    
    offsets=np.linspace(0,1,2);
    def func(xy,offsets=offsets):
        x,y=xy
        return np.sin(-np.array([(x-offset)**2+(y-offset)**2 for offset in offsets]))
        
    loss=adaptive.learner.learner2D.resolution_loss_function(min_distance=0.01,max_distance=1)
    learner=adaptive.Learner2D(func,bounds=[(-1,1),(-1,1)],loss_per_triangle=loss)
    runner=adaptive.BlockingRunner(learner,goal=lambda l:l.loss()<0.01,executor=MPIPoolExecutor(),shutdown_executor=True)
    

if __name__=="__main__":
	main()                

I used mpiexec -n 4 python -m mpi4py.futures MWE.py to run the code. But it always prints errors like:

AttributeError: Can't pickle local object 'main.<locals>.func'

Since lambda function cannot be pickled 206 ,can I define a function that has parameters which cannot be determined in adavnce? I mean, can I def a func(xy, parameters) outside main() and call with func2=lambda xy: func(xy,parameters=para) inside main(), here para cannot be determined before running?

[hainingpan]

Well, I found this.
It says def func() must be on the top level in order to be pickled.
But then the problem will be if the parameter in func(xy, parameters) cannot be determined before, how can I pass the parameter into func and is it possible to use adaptive with multiprocessing?

[basnijholt]

You can use functools.partial(func, parameters=...) like in the tutorial here.

That should be pickleble.

[hainingpan]

Thank you!