[Chore] Separate out system utilities from vllm.utils (vllm-project#27201)

dongbo910220 · DarkLight1337 · usberkeley · commit f8f1a8896e18 · 2025-10-23T21:20:04.000+08:00
Signed-off-by: dongbo910220 &lt;1275604947@qq.com&gt;
Co-authored-by: Cyrus Leung &lt;tlleungac@connect.ust.hk&gt;
diff --git a/tests/compile/test_async_tp.py b/tests/compile/test_async_tp.py
@@ -25,7 +25,7 @@
     initialize_model_parallel,
 )
 from vllm.platforms import current_platform
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 from ..models.registry import HF_EXAMPLE_MODELS
 from ..utils import (
diff --git a/tests/compile/test_fusion_all_reduce.py b/tests/compile/test_fusion_all_reduce.py
@@ -31,7 +31,7 @@
     GroupShape,
 )
 from vllm.platforms import current_platform
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 from ..utils import has_module_attribute, multi_gpu_test
 from .backend import TestBackend
diff --git a/tests/compile/test_sequence_parallelism.py b/tests/compile/test_sequence_parallelism.py
@@ -29,7 +29,7 @@
 from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.quantization.utils.w8a8_utils import Fp8LinearOp
 from vllm.platforms import current_platform
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 from ..utils import multi_gpu_test
 from .backend import TestBackend
diff --git a/tests/distributed/test_eplb_execute.py b/tests/distributed/test_eplb_execute.py
@@ -15,7 +15,7 @@
     get_tp_group,
     init_distributed_environment,
 )
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 
 def distributed_run(fn, world_size):
diff --git a/tests/distributed/test_nccl_symm_mem_allreduce.py b/tests/distributed/test_nccl_symm_mem_allreduce.py
@@ -23,7 +23,7 @@
     initialize_model_parallel,
 )
 from vllm.platforms import current_platform
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 torch.manual_seed(42)
 random.seed(44)
diff --git a/tests/distributed/test_pynccl.py b/tests/distributed/test_pynccl.py
@@ -18,7 +18,7 @@
     graph_capture,
     init_distributed_environment,
 )
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 
 def distributed_run(fn, world_size):
diff --git a/tests/distributed/test_shm_broadcast.py b/tests/distributed/test_shm_broadcast.py
@@ -10,8 +10,8 @@
 
 from vllm.distributed.device_communicators.shm_broadcast import MessageQueue
 from vllm.distributed.utils import StatelessProcessGroup
-from vllm.utils import update_environment_variables
 from vllm.utils.network_utils import get_open_port
+from vllm.utils.system_utils import update_environment_variables
 
 
 def get_arrays(n: int, seed: int = 0) -> list[np.ndarray]:
diff --git a/tests/distributed/test_symm_mem_allreduce.py b/tests/distributed/test_symm_mem_allreduce.py
@@ -23,7 +23,7 @@
 from vllm.engine.arg_utils import EngineArgs
 from vllm.engine.llm_engine import LLMEngine
 from vllm.platforms import current_platform
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 torch.manual_seed(42)
 random.seed(44)
diff --git a/tests/distributed/test_utils.py b/tests/distributed/test_utils.py
@@ -10,8 +10,8 @@
 import vllm.envs as envs
 from vllm.distributed.device_communicators.pynccl import PyNcclCommunicator
 from vllm.distributed.utils import StatelessProcessGroup
-from vllm.utils import update_environment_variables
 from vllm.utils.network_utils import get_open_port
+from vllm.utils.system_utils import update_environment_variables
 from vllm.utils.torch_utils import cuda_device_count_stateless
 
 from ..utils import multi_gpu_test
diff --git a/tests/kernels/mamba/test_mamba_mixer2.py b/tests/kernels/mamba/test_mamba_mixer2.py
@@ -13,7 +13,7 @@
 )
 from vllm.model_executor.layers.mamba.mamba_mixer2 import Mixer2RMSNormGated
 from vllm.platforms import current_platform
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 
 
 @multi_gpu_test(num_gpus=2)
diff --git a/tests/models/test_vision.py b/tests/models/test_vision.py
@@ -19,8 +19,8 @@
     run_dp_sharded_vision_model,
 )
 from vllm.platforms import current_platform
-from vllm.utils import update_environment_variables
 from vllm.utils.network_utils import get_open_port
+from vllm.utils.system_utils import update_environment_variables
 
 pytestmark = pytest.mark.cpu_test
 
diff --git a/tests/utils_/test_system_utils.py b/tests/utils_/test_system_utils.py
@@ -0,0 +1,19 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+
+import tempfile
+from pathlib import Path
+
+from vllm.utils.system_utils import unique_filepath
+
+
+def test_unique_filepath():
+    temp_dir = tempfile.mkdtemp()
+    path_fn = lambda i: Path(temp_dir) / f"file_{i}.txt"
+    paths = set()
+    for i in range(10):
+        path = unique_filepath(path_fn)
+        path.write_text("test")
+        paths.add(path)
+    assert len(paths) == 10
+    assert len(list(Path(temp_dir).glob("*.txt"))) == 10
diff --git a/tests/utils_/test_utils.py b/tests/utils_/test_utils.py
@@ -19,7 +19,6 @@
 from vllm.utils import (
     FlexibleArgumentParser,
     bind_kv_cache,
-    unique_filepath,
 )
 from ..utils import create_new_process_for_each_test, flat_product
 
@@ -466,18 +465,6 @@ def test_load_config_file(tmp_path):
     os.remove(str(config_file_path))
 
 
-def test_unique_filepath():
-    temp_dir = tempfile.mkdtemp()
-    path_fn = lambda i: Path(temp_dir) / f"file_{i}.txt"
-    paths = set()
-    for i in range(10):
-        path = unique_filepath(path_fn)
-        path.write_text("test")
-        paths.add(path)
-    assert len(paths) == 10
-    assert len(list(Path(temp_dir).glob("*.txt"))) == 10
-
-
 def test_flat_product():
     # Check regular itertools.product behavior
     result1 = list(flat_product([1, 2, 3], ["a", "b"]))
diff --git a/tests/v1/worker/test_gpu_model_runner.py b/tests/v1/worker/test_gpu_model_runner.py
@@ -21,8 +21,8 @@
 from vllm.model_executor.layers.mamba.mamba_mixer2 import MambaMixer2
 from vllm.platforms import current_platform
 from vllm.sampling_params import SamplingParams
-from vllm.utils import update_environment_variables
 from vllm.utils.mem_constants import GiB_bytes
+from vllm.utils.system_utils import update_environment_variables
 from vllm.v1.core.kv_cache_utils import estimate_max_model_len, get_kv_cache_configs
 from vllm.v1.core.sched.output import CachedRequestData, NewRequestData, SchedulerOutput
 from vllm.v1.kv_cache_interface import (
diff --git a/vllm/compilation/pass_manager.py b/vllm/compilation/pass_manager.py
@@ -8,7 +8,7 @@
 from vllm.config import VllmConfig, set_current_vllm_config
 from vllm.logger import init_logger
 from vllm.platforms import current_platform
-from vllm.utils import set_env_var
+from vllm.utils.system_utils import set_env_var
 
 from .post_cleanup import PostCleanupPass
 from .vllm_inductor_pass import VllmInductorPass
diff --git a/vllm/compilation/vllm_inductor_pass.py b/vllm/compilation/vllm_inductor_pass.py
@@ -114,7 +114,7 @@ def dump_patterns(self, config: VllmConfig, pm_pass: PatternMatcherPass):
 
         debug_dump_path.mkdir(parents=True, exist_ok=True)
 
-        from vllm.utils import unique_filepath
+        from vllm.utils.system_utils import unique_filepath
 
         file_path = unique_filepath(
             lambda i: debug_dump_path / f"patterns.{self.pass_name}.{i}.py"
diff --git a/vllm/distributed/device_communicators/all_reduce_utils.py b/vllm/distributed/device_communicators/all_reduce_utils.py
@@ -22,7 +22,7 @@
 from vllm.model_executor.layers.batch_invariant import (
     vllm_is_batch_invariant,
 )
-from vllm.utils import update_environment_variables
+from vllm.utils.system_utils import update_environment_variables
 from vllm.utils.torch_utils import cuda_device_count_stateless
 
 logger = init_logger(__name__)
diff --git a/vllm/entrypoints/cli/serve.py b/vllm/entrypoints/cli/serve.py
@@ -18,12 +18,9 @@
 from vllm.entrypoints.utils import VLLM_SUBCMD_PARSER_EPILOG
 from vllm.logger import init_logger
 from vllm.usage.usage_lib import UsageContext
-from vllm.utils import (
-    FlexibleArgumentParser,
-    decorate_logs,
-    set_process_title,
-)
+from vllm.utils import FlexibleArgumentParser
 from vllm.utils.network_utils import get_tcp_uri
+from vllm.utils.system_utils import decorate_logs, set_process_title
 from vllm.v1.engine.core import EngineCoreProc
 from vllm.v1.engine.utils import CoreEngineProcManager, launch_core_engines
 from vllm.v1.executor import Executor
diff --git a/vllm/entrypoints/openai/api_server.py b/vllm/entrypoints/openai/api_server.py
@@ -108,13 +108,9 @@
 from vllm.logger import init_logger
 from vllm.reasoning import ReasoningParserManager
 from vllm.usage.usage_lib import UsageContext
-from vllm.utils import (
-    Device,
-    FlexibleArgumentParser,
-    decorate_logs,
-    set_ulimit,
-)
+from vllm.utils import Device, FlexibleArgumentParser, set_ulimit
 from vllm.utils.network_utils import is_valid_ipv6_address
+from vllm.utils.system_utils import decorate_logs
 from vllm.v1.engine.exceptions import EngineDeadError
 from vllm.v1.metrics.prometheus import get_prometheus_registry
 from vllm.version import __version__ as VLLM_VERSION
diff --git a/vllm/utils/__init__.py b/vllm/utils/__init__.py
diff --git a/vllm/utils/system_utils.py b/vllm/utils/system_utils.py
diff --git a/vllm/v1/engine/coordinator.py b/vllm/v1/engine/coordinator.py
diff --git a/vllm/v1/engine/core.py b/vllm/v1/engine/core.py
diff --git a/vllm/v1/executor/multiproc_executor.py b/vllm/v1/executor/multiproc_executor.py
diff --git a/vllm/v1/worker/worker_base.py b/vllm/v1/worker/worker_base.py

Original file line number	Diff line number	Diff line change
`@@ -25,7 +25,7 @@`
`25`	`25`	`initialize_model_parallel,`
`26`	`26`	`)`
`27`	`27`	`from vllm.platforms import current_platform`
`28`		`-from vllm.utils import update_environment_variables`
	`28`	`+from vllm.utils.system_utils import update_environment_variables`
`29`	`29`
`30`	`30`	`from ..models.registry import HF_EXAMPLE_MODELS`
`31`	`31`	`from ..utils import (`
Original file line number	Diff line number	Diff line change
`@@ -31,7 +31,7 @@`
`31`	`31`	`GroupShape,`
`32`	`32`	`)`
`33`	`33`	`from vllm.platforms import current_platform`
`34`		`-from vllm.utils import update_environment_variables`
	`34`	`+from vllm.utils.system_utils import update_environment_variables`
`35`	`35`
`36`	`36`	`from ..utils import has_module_attribute, multi_gpu_test`
`37`	`37`	`from .backend import TestBackend`
Original file line number	Diff line number	Diff line change
`@@ -15,7 +15,7 @@`
`15`	`15`	`get_tp_group,`
`16`	`16`	`init_distributed_environment,`
`17`	`17`	`)`
`18`		`-from vllm.utils import update_environment_variables`
	`18`	`+from vllm.utils.system_utils import update_environment_variables`
`19`	`19`
`20`	`20`
`21`	`21`	`def distributed_run(fn, world_size):`
Original file line number	Diff line number	Diff line change
`@@ -23,7 +23,7 @@`
`23`	`23`	`initialize_model_parallel,`
`24`	`24`	`)`
`25`	`25`	`from vllm.platforms import current_platform`
`26`		`-from vllm.utils import update_environment_variables`
	`26`	`+from vllm.utils.system_utils import update_environment_variables`
`27`	`27`
`28`	`28`	`torch.manual_seed(42)`
`29`	`29`	`random.seed(44)`
Original file line number	Diff line number	Diff line change
`@@ -18,7 +18,7 @@`
`18`	`18`	`graph_capture,`
`19`	`19`	`init_distributed_environment,`
`20`	`20`	`)`
`21`		`-from vllm.utils import update_environment_variables`
	`21`	`+from vllm.utils.system_utils import update_environment_variables`
`22`	`22`
`23`	`23`
`24`	`24`	`def distributed_run(fn, world_size):`
Original file line number	Diff line number	Diff line change
`@@ -13,7 +13,7 @@`
`13`	`13`	`)`
`14`	`14`	`from vllm.model_executor.layers.mamba.mamba_mixer2 import Mixer2RMSNormGated`
`15`	`15`	`from vllm.platforms import current_platform`
`16`		`-from vllm.utils import update_environment_variables`
	`16`	`+from vllm.utils.system_utils import update_environment_variables`
`17`	`17`
`18`	`18`
`19`	`19`	`@multi_gpu_test(num_gpus=2)`