pylibwholegraph.torch.comm.get_global_communicator#

pylibwholegraph.torch.comm.get_global_communicator(distributed_backend='nccl')#

Get the global communicator of this job :return: WholeMemoryCommunicator that has all GPUs in it.