[CI] Try introducing isort. (#3495)

vllm/utils.py

+import asyncio
 import enum
+import gc
 import os
 import socket
 import subprocess
 import uuid
-import gc
+import warnings
+from collections import OrderedDict
+from functools import lru_cache, partial
 from platform import uname
-from typing import List, Tuple, Union, Generic
-from packaging.version import parse, Version
+from typing import (Any, Awaitable, Callable, Generic, Hashable, List,
+                    Optional, Tuple, TypeVar, Union)
 
 import psutil
 import torch
-import asyncio
-from functools import partial, lru_cache
-from typing import (
-    Awaitable,
-    Callable,
-    TypeVar,
-)
-from collections import OrderedDict
-from typing import Any, Hashable, Optional
+from packaging.version import Version, parse
 
 from vllm.logger import init_logger
-import warnings
 
 T = TypeVar("T")
 logger = init_logger(__name__)

vllm/worker/cache_engine.py

@@ -6,7 +6,7 @@ import torch
 from vllm.attention import get_attn_backend
 from vllm.config import CacheConfig, ModelConfig, ParallelConfig
 from vllm.logger import init_logger
-from vllm.utils import is_pin_memory_available, STR_DTYPE_TO_TORCH_DTYPE
+from vllm.utils import STR_DTYPE_TO_TORCH_DTYPE, is_pin_memory_available
 
 logger = init_logger(__name__)
 

vllm/worker/model_runner.py

 import contextlib
 import time
-from typing import Dict, List, Optional, Tuple, Set
+from typing import Dict, List, Optional, Set, Tuple
 
 import numpy as np
 import torch
 import torch.nn as nn
 
 from vllm.attention import AttentionMetadata, get_attn_backend
-from vllm.config import (DeviceConfig, ModelConfig, LoRAConfig, ParallelConfig,
+from vllm.config import (DeviceConfig, LoRAConfig, ModelConfig, ParallelConfig,
                          SchedulerConfig)
 from vllm.logger import init_logger
+from vllm.lora.layers import LoRAMapping
+from vllm.lora.request import LoRARequest
+from vllm.lora.worker_manager import LRUCacheWorkerLoRAManager
 from vllm.model_executor import SamplingMetadata
 from vllm.model_executor.model_loader import get_model
-from vllm.model_executor.parallel_utils import cupy_utils
+from vllm.model_executor.parallel_utils import cupy_utils, custom_all_reduce
 from vllm.model_executor.parallel_utils.communication_op import (
     broadcast_tensor_dict)
 from vllm.model_executor.parallel_utils.parallel_state import (
     with_cupy_nccl_for_all_reduce)
-from vllm.model_executor.parallel_utils import custom_all_reduce
 from vllm.sampling_params import SamplingParams, SamplingType
 from vllm.sequence import SamplerOutput, SequenceData, SequenceGroupMetadata
-from vllm.lora.worker_manager import LRUCacheWorkerLoRAManager
-from vllm.lora.layers import LoRAMapping
-from vllm.lora.request import LoRARequest
-from vllm.utils import (async_tensor_h2d, CudaMemoryProfiler,
+from vllm.utils import (CudaMemoryProfiler, async_tensor_h2d,
                         is_pin_memory_available, make_tensor_with_pad,
                         maybe_expand_dim)
 

vllm/worker/worker.py

 """A GPU worker class."""
 import gc
 import os
-from typing import Dict, List, Tuple, Set, Optional
+from typing import Dict, List, Optional, Set, Tuple
 
 import torch
 import torch.distributed
 
-from vllm.config import (CacheConfig, DeviceConfig, ModelConfig,
-                         ParallelConfig, SchedulerConfig, LoRAConfig)
+from vllm.config import (CacheConfig, DeviceConfig, LoRAConfig, ModelConfig,
+                         ParallelConfig, SchedulerConfig)
+from vllm.lora.request import LoRARequest
 from vllm.model_executor import set_random_seed
 from vllm.model_executor.parallel_utils import cupy_utils
 from vllm.model_executor.parallel_utils.communication_op import (
@@ -18,7 +19,6 @@ from vllm.model_executor.parallel_utils.parallel_state import (
 from vllm.sequence import SamplerOutput, SequenceGroupMetadata
 from vllm.worker.cache_engine import CacheEngine
 from vllm.worker.model_runner import ModelRunner
-from vllm.lora.request import LoRARequest
 
 
 class Worker: