[CI] Try introducing isort. (#3495)

vllm/executor/ray_gpu_executor.py

@@ -1,20 +1,20 @@
 import asyncio
 import copy
-from collections import defaultdict
 import os
 import pickle
+from collections import defaultdict
 from typing import TYPE_CHECKING, Any, Dict, List, Optional
 
-from vllm.config import (CacheConfig, DeviceConfig, ModelConfig,
-                         ParallelConfig, SchedulerConfig, LoRAConfig)
+from vllm.config import (CacheConfig, DeviceConfig, LoRAConfig, ModelConfig,
+                         ParallelConfig, SchedulerConfig)
 from vllm.engine.ray_utils import RayWorkerVllm, ray
 from vllm.executor.executor_base import ExecutorAsyncBase, ExecutorBase
 from vllm.executor.utils import check_block_size_valid
 from vllm.logger import init_logger
 from vllm.lora.request import LoRARequest
 from vllm.sequence import SamplerOutput, SequenceGroupMetadata
-from vllm.utils import (set_cuda_visible_devices, get_ip, get_open_port,
-                        get_distributed_init_method, make_async)
+from vllm.utils import (get_distributed_init_method, get_ip, get_open_port,
+                        make_async, set_cuda_visible_devices)
 
 if ray is not None:
     from ray.util.scheduling_strategies import PlacementGroupSchedulingStrategy
@@ -343,7 +343,7 @@ class RayGPUExecutor(ExecutorBase):
             raise ValueError(f"Ray version {required_version} or greater is "
                              f"required, but found {current_version}")
 
-        from ray.dag import MultiOutputNode, InputNode
+        from ray.dag import InputNode, MultiOutputNode
         assert self.parallel_config.worker_use_ray
 
         # Right now, compiled DAG requires at least 1 arg. We send