fix merge

Signed-off-by: Bill Nell <bnell@redhat.com>

Author: bnellnm

tests/kernels/moe/deepep_utils.py

@@ -5,7 +5,9 @@
 import dataclasses
 import importlib
 import os
+import socket
 import traceback
+from contextlib import closing
 from typing import Callable, Optional
 
 import torch
@@ -14,8 +16,6 @@
     spawn)  # pyright: ignore[reportPrivateImportUsage]
 from typing_extensions import Concatenate, ParamSpec
 
-from vllm.model_executor.layers.fused_moe.utils import find_free_port
-
 has_deep_ep = importlib.util.find_spec("deep_ep") is not None
 if has_deep_ep:
     from vllm.model_executor.layers.fused_moe.deepep_ht_prepare_finalize import (  # noqa: E501
@@ -82,6 +82,13 @@ def _worker_parallel_launch(
         torch.distributed.destroy_process_group()
 
 
+def find_free_port():
+    with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
+        s.bind(('', 0))
+        s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
+        return s.getsockname()[1]
+
+
 def parallel_launch(
     world_size: int,
     worker: Callable[Concatenate[ProcessGroupInfo, P], None],

vllm/model_executor/layers/fused_moe/utils.py

@@ -1,7 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-import socket
-from contextlib import closing
 from math import prod
 from typing import Optional
 
@@ -101,13 +99,6 @@ def _fp8_perm(m: torch.Tensor, idx: torch.Tensor) -> torch.Tensor:
         return m[idx, ...]
 
 
-def find_free_port():
-    with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
-        s.bind(('', 0))
-        s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
-        return s.getsockname()[1]
-
-
 # TODO(bnell): better name
 def maybe_fix_scales(scales: Optional[torch.Tensor],
                      num_experts: int) -> Optional[torch.Tensor]: