def __init__(self, batch_sampler: torch.utils.data.BatchSampler,
num_workers: int, rank: int) -> None:
check.gt(rank, -1, "rank must be non-negative")
check.gt(num_workers, 0, "num_workers must be positive")
check.lt(rank, num_workers, "rank must be less than num_workers")
self.batch_sampler = batch_sampler
self.num_workers = num_workers
self.rank = rank
def receive_non_blocking(
self, send_rank: int, deadline: Optional[float] = None
) -> Tuple[bool, Any]:
check.lt(send_rank, len(self.sockets))
timeout = 1000 if not deadline else int(deadline - time.time()) * 1000
timeout = max(timeout, 100)
if self.sockets[send_rank].poll(timeout) == 0: # type: ignore
return False, None
message = self.sockets[send_rank].recv_pyobj() # type: ignore
return True, message
def _bind_to_random_ports(self, port_range: Tuple[int, int],
num_connections: int) -> None:
check.lt(num_connections, port_range[1] - port_range[0])
for _ in range(num_connections):
socket = self.context.socket(zmq.REP)
try:
selected_port = socket.bind_to_random_port(
addr="tcp://*",
min_port=port_range[0],
max_port=port_range[1])
self.ports.append(selected_port)
except ZMQBindError as e:
raise det.errors.InternalException(
f"Failed to bind to port range {port_range}.") from e
self.sockets.append(socket)
def receive_blocking(self, send_rank: int) -> Any:
check.lt(send_rank, len(self.sockets))
message = self.sockets[send_rank].recv_pyobj() # type: ignore
return message
