Python MPI pool distrubuted training model

Hi there i have question.

How to fine max accuracy with python in distributed mpi pool function?

i try to create global dict or named tuple filled by train function with key value which is also iterable and get with max () key value but i have some issue:

  • after function finished the dictionary is empty i try with local instance and global variable to be accessible any difference variables but i get an error

Please propose me a better way to do that

BR
George