Utilities
parallel_map learned to retain to feeder thread and workers until iterator completion via “master stop” mechanic. This allows the mechanism to now support PyTorch tensors in multiprocessing mode.
parallel_map