up

Signed-off-by: Rui Qiao <ruisearch42@gmail.com>

Author: ruisearch42

vllm/v1/engine/core.py

@@ -431,7 +431,7 @@ def _perform_handshake(
 
             yield addresses
 
-            # Send ready message
+            # Send ready message.
             num_gpu_blocks = vllm_config.cache_config.num_gpu_blocks
             handshake_socket.send(
                 msgspec.msgpack.encode({

vllm/v1/engine/core_client.py

@@ -11,7 +11,6 @@
 from concurrent.futures import Future
 from dataclasses import dataclass
 from threading import Thread
-from types import NoneType
 from typing import Any, Callable, Optional, TypeVar, Union
 
 import msgspec.msgpack
@@ -492,7 +491,7 @@ def _wait_for_engine_startup(self, handshake_socket: zmq.Socket,
                 coordinator.get_engine_socket_addresses())
 
         proc_manager = self.resources.engine_manager
-        assert isinstance(proc_manager, (NoneType, CoreEngineProcManager)), (
+        assert isinstance(proc_manager, (type(None), CoreEngineProcManager)), (
             "_wait_for_engine_startup should only be "
             "called with CoreEngineProcManager")