Reorder torch.distributed.rpc.init_rpc docstring arguments (pytorch#5…

…0419)

Summary: Pull Request resolved: pytorch#50419

Test Plan: Imported from OSS

Reviewed By: glaringlee

Differential Revision: D25911561

Pulled By: pbelevich

fbshipit-source-id: 62c9a5c3f5ec5eddcbd149821ebdf484ff392158

torch/distributed/rpc/__init__.py

@@ -86,15 +86,15 @@ def init_rpc(
         process ready to send and receive RPCs.
 
         Args:
+            name (str): a globally unique name of this node. (e.g.,
+                ``Trainer3``, ``ParameterServer2``, ``Master``, ``Worker1``)
+                Name can only contain number, alphabet, underscore, colon,
+                and/or dash, and must be shorter than 128 characters.
             backend (BackendType, optional): The type of RPC backend
                 implementation. Supported values include
                 ``BackendType.TENSORPIPE`` (the default) and
                 ``BackendType.PROCESS_GROUP``. See :ref:`rpc-backends` for more
                 information.
-            name (str): a globally unique name of this node. (e.g.,
-                ``Trainer3``, ``ParameterServer2``, ``Master``, ``Worker1``)
-                Name can only contain number, alphabet, underscore, colon,
-                and/or dash, and must be shorter than 128 characters.
             rank (int): a globally unique id/rank of this node.
             world_size (int): The number of workers in the group.
             rpc_backend_options (RpcBackendOptions, optional): The options