Using ALPSO in parallel mode (ALPSO: IndexError: too many indices for array) about pyoptsparse HOT 1 CLOSED

Hello @kmodexc, thanks for sharing a MWE, I could replicate your error.
I initially thought this was a bug, but if you diff alpso.py and alpso_ext (the latter is the one imported when "parallelType": "EXT") you will see how the parallel version expects a different objFunc signature that returns a vector f of len(swarsize) instead of a scalar.
I am not familiar with this PSO algorithm , but in the default mode alpso loops over every particle of the swarm and evaluates the (scalar) objfunc at that point. In the parallel version, there is no such loop. We do not have examples afaik, but it looks like alpso_ext passes an array with all the DVs of all the particles at a given iteration, and expects a vector f with all the function values of the particles at the same iteration.

In short, if you remove the "parallelType": "EXT" option the optimizer will work. To run the optimizer in parallel mode, it is on the user to provide an objfunc that runs in parallel and returns a vector of the correct size.

from pyoptsparse.