I see the possibility of using GAs with torchGA to optimize neural networks using genetic algorithms, but is it possible to use Pytorch instead of numpy arrays to speed up the computation time of GAs?