torch.distributed

API名称

是否支持

限制与说明

torch.distributed._backend

是

  

torch.distributed.group

是

  

torch.distributed.GroupMember

是

  

torch.distributed.destroy_process_group

是

  

torch.distributed.all_reduce_coalesced

是

  

torch.distributed.all_gather_coalesced

是

  

torch.distributed.Reducer

是

  

torch.distributed.torch.distributed._DEFAULT_FIRST_BUCKET_BYTES

是

  

torch.distributed.Logger

是

  

torch.distributed.all_gather_togather

是

  

torch.distributed.is_available

是

  

torch.distributed.init_process_group

是

  

torch.distributed.is_initialized

是

  

torch.distributed.is_mpi_available

是

  

torch.distributed.is_nccl_available

是

  

torch.distributed.is_torchelastic_launched

否

  

torch.distributed.Backend

是

  

torch.distributed.Backend.register_backend

是

  

torch.distributed.get_backend

是

  

torch.distributed.get_rank

是

  

torch.distributed.get_world_size

是

  

torch.distributed.Store

否

  

torch.distributed.TCPStore

否

  

torch.distributed.HashStore

否

  

torch.distributed.FileStore

否

  

torch.distributed.PrefixStore

否

  

torch.distributed.Store.set

否

  

torch.distributed.Store.get

否

  

torch.distributed.Store.add

否

  

torch.distributed.Store.compare_set

否

  

torch.distributed.Store.wait

否

  

torch.distributed.Store.num_keys

否

  

torch.distributed.Store.delete_key

否

  

torch.distributed.Store.set_timeout

否

  

torch.distributed.new_group

是

  

torch.distributed.send

是

  

torch.distributed.recv

是

  

torch.distributed.isend

否

  

torch.distributed.irecv

否

  

torch.distributed.broadcast

是

  

torch.distributed.broadcast_object_list

是

  

torch.distributed.all_reduce

是

  

torch.distributed.reduce

否

  

torch.distributed.all_gather

是

  

torch.distributed.all_gather_object

是

  

torch.distributed.gather

否

  

torch.distributed.gather_object

否

  

torch.distributed.scatter

否

  

torch.distributed.scatter_object_list

否

  

torch.distributed.reduce_scatter

是

  

torch.distributed.all_to_all

是

  

torch.distributed.barrier

是

  

torch.distributed.monitored_barrier

否

  

torch.distributed.ReduceOp

是

  

torch.distributed.reduce_op

是

  

torch.distributed.broadcast_multigpu

否

  

torch.distributed.all_reduce_multigpu

否

  

torch.distributed.reduce_multigpu

否

  

torch.distributed.all_gather_multigpu

否

  

torch.distributed.reduce_scatter_multigpu

否

  

torch.distributed.launch

是

  

torch.multiprocessing.spawn

是

  

is_completed

是

  

wait

是

  

get_future

否