Module `agents`¶

Exposed Classes¶

Actor (`agents.Actor`)¶

Learner (`agents.Learner`)¶

Parent classes¶

RpcCallee (`agents.RpcCallee`)¶

RpcCaller (`agents.RpcCaller`)¶

class pytorch_seed_rl.agents.rpc_caller.RpcCaller(rank: int, callee_rref: torch.distributed.rpc.RRef)[source]¶

Bases: object

RPC object that handles communication with an assigned RPC callee.

Parameters

rank (int) – Rank given by the RPC group on initiation (as in torch.distributed.rpc.init_rpc()).
callee_rref (torch.distributed.rpc.RRef) – RRef referencing a remote RpcCallee.

loop()[source]¶

Main loop function of an RpcCaller.

Checks in with assigned RpcCallee.
Loops _loop() until self.shutdown is set True.
Checks out with assigned RpcCallee.
Calls _cleanup()

batched_rpc(*args, **kwargs) → torch.futures.Future [source]¶: Wrap for batched async RPC ran on remote callee.