PyTorch is adding native MoE token dispatch and combine

Torch distributed TokenSwitch backed by NCCL-EP brings expert-parallel MoE dispatch and combine closer to a standard PyTorch primitive

cool find this morning: pytorch is moving toward native MoE token dispatch/combine via torch.distributed.TokenSwitch, backed by NCCL-EP pretty cool since performant EP dispatch/combine has historically lived more in specialized/custom com