Remove Python 3.10 support now that it is EOL (#60402)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

.pre-commit-config.yaml

@@ -199,29 +199,13 @@ repos:
     additional_dependencies: [PyYAML, regex]
     verbose: true # full-scan `note:` lines are advisory, so the hook passes
     require_serial: true # one full scan, not one per pre-commit file batch
-  - id: mypy-3.10 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
+  - id: mypy # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
-    name: Run mypy for Python 3.10
+    name: Run mypy for the oldest supported Python version
-    entry: python tools/pre_commit/mypy.py "3.10"
-    <<: &mypy_common
-      language: python
-      types_or: [python, pyi]
-      require_serial: true
-      additional_dependencies: ["mypy==1.20.2", regex, types-cachetools, types-setuptools, types-PyYAML, types-requests, types-torch, pydantic]
-  - id: mypy-3.11 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
-    name: Run mypy for Python 3.11
     entry: python tools/pre_commit/mypy.py "3.11"
-    <<: *mypy_common
+    language: python
-    stages: [manual] # Only run in CI
+    types_or: [python, pyi]
-  - id: mypy-3.12 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
+    require_serial: true
-    name: Run mypy for Python 3.12
+    additional_dependencies: ["mypy==1.20.2", regex, types-cachetools, types-setuptools, types-PyYAML, types-requests, types-torch, pydantic]
-    entry: python tools/pre_commit/mypy.py "3.12"
-    <<: *mypy_common
-    stages: [manual] # Only run in CI
-  - id: mypy-3.13 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
-    name: Run mypy for Python 3.13
-    entry: python tools/pre_commit/mypy.py "3.13"
-    <<: *mypy_common
-    stages: [manual] # Only run in CI
   - id: png-lint
     name: Lint PNG exports from excalidraw
     entry: tools/pre_commit/png-lint.sh

AGENTS.md

@@ -114,9 +114,6 @@ pre-commit run --all-files
 
 # Run a specific hook:
 pre-commit run ruff-check --all-files
-
-# Run mypy as it is in CI:
-pre-commit run mypy-3.12 --all-files --hook-stage manual
 ```
 
 The line length limit for Python code is 88 characters. If you are not sure, use pre-commit to check.

CMakeLists.txt

@@ -46,7 +46,7 @@ install(CODE "set(CMAKE_INSTALL_LOCAL_ONLY TRUE)" ALL_COMPONENTS)
 # Supported python versions.  These versions will be searched in order, the
 # first match will be selected.  These should be kept in sync with setup.py.
 #
-set(PYTHON_SUPPORTED_VERSIONS "3.10" "3.11" "3.12" "3.13" "3.14")
+set(PYTHON_SUPPORTED_VERSIONS "3.11" "3.12" "3.13" "3.14")
 
 # Supported AMD GPU architectures.
 set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1250;gfx1030;gfx1100;gfx1101;gfx1102;gfx1103;gfx1150;gfx1151;gfx1152;gfx1153;gfx1200;gfx1201")
@@ -134,40 +134,34 @@ endif()
 #
 # spinloop extension (pure CXX; must stay above the non-CUDA device branch so
 # CPU builds define the target before the early return)
-# This extension requires SABI 3.11 since it relies on Py_buffer support. Loading
+# This extension requires SABI 3.11 since it relies on Py_buffer support.
-# failure is handled gracefully on vLLM side for lower Python versions.
-# Skip the target entirely on Python < 3.11 so the build doesn't break.
 #
-if(Python_VERSION VERSION_GREATER_EQUAL "3.11")
+set(VLLM_SPINLOOP_EXT_SRC "csrc/spinloop.cpp")
-  set(VLLM_SPINLOOP_EXT_SRC "csrc/spinloop.cpp")
+set(SPINLOOP_COMPILE_FLAGS "")
-  set(SPINLOOP_COMPILE_FLAGS "")
+if(CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64|amd64")
-  if(CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64|amd64")
+  list(APPEND SPINLOOP_COMPILE_FLAGS "-mmwaitx")
-    list(APPEND SPINLOOP_COMPILE_FLAGS "-mmwaitx")
-  endif()
-  define_extension_target(
-    spinloop
-    DESTINATION vllm
-    LANGUAGE CXX
-    SOURCES ${VLLM_SPINLOOP_EXT_SRC}
-    COMPILE_FLAGS ${SPINLOOP_COMPILE_FLAGS}
-    USE_SABI 3.11
-    WITH_SOABI)
 endif()
+define_extension_target(
+  spinloop
+  DESTINATION vllm
+  LANGUAGE CXX
+  SOURCES ${VLLM_SPINLOOP_EXT_SRC}
+  COMPILE_FLAGS ${SPINLOOP_COMPILE_FLAGS}
+  USE_SABI 3.11
+  WITH_SOABI)
 
 #
 # fs_io extension (pure CXX; must stay above the non-CUDA device branch
 # so CPU builds define the target before the early return).
 # GIL-releasing filesystem helpers for FileSystemTierManager.
 #
-if(Python_VERSION VERSION_GREATER_EQUAL "3.11")
+define_extension_target(
-  define_extension_target(
+  fs_io_C
-    fs_io_C
+  DESTINATION vllm
-    DESTINATION vllm
+  LANGUAGE CXX
-    LANGUAGE CXX
+  SOURCES csrc/fs_io.cpp
-    SOURCES csrc/fs_io.cpp
+  USE_SABI 3.11
-    USE_SABI 3.11
+  WITH_SOABI)
-    WITH_SOABI)
-endif()
 
 #
 # Forward the non-CUDA device extensions to external CMake scripts.

docker/Dockerfile.cpu

@@ -13,7 +13,7 @@
 #   vllm-dev: used for development
 #
 # Build arguments:
-#   PYTHON_VERSION=3.13|3.12 (default)|3.11|3.10
+#   PYTHON_VERSION=3.13|3.12 (default)|3.11
 #   VLLM_CPU_X86=false (default)|true (for cross-compilation)
 #   VLLM_CPU_ARM_BF16=false (default)|true (for cross-compilation)
 #

docs/contributing/README.md

@@ -74,7 +74,7 @@ For an optimized workflow when iterating on C++/CUDA kernels, see the [Increment
 For JIT kernel warmup conventions, see [JIT Kernel Warmup](./jit_kernel_warmup.md).
 
 !!! tip
-    vLLM is compatible with Python versions 3.10 to 3.13. However, vLLM's default [Dockerfile](../../docker/Dockerfile) ships with Python 3.12 and tests in CI (except `mypy`) are run with Python 3.12.
+    vLLM is compatible with Python versions 3.11 to 3.14. However, vLLM's default [Dockerfile](../../docker/Dockerfile) ships with Python 3.12 and tests in CI (except `mypy`) are run with Python 3.12.
 
     Therefore, we recommend developing with Python 3.12 to minimise the chance of your local environment clashing with our CI environment.
 
@@ -102,14 +102,6 @@ integration, use the configuration files in `tools/pre_commit/`:
     pre-commit run -a  # runs on all files (short for --all-files)
     ```
 
-    ---
-
-    Some `pre-commit` hooks only run in CI. If you need to, you can run them locally with:
-
-    ```bash
-    pre-commit run --hook-stage manual mypy-3.11
-    ```
-
 ### Documentation
 
 MkDocs is a fast, simple and downright gorgeous static site generator that's geared towards building project documentation. Documentation source files are written in Markdown, and configured with a single YAML configuration file, [mkdocs.yaml](../../mkdocs.yaml).

docs/deployment/frameworks/skypilot.md

@@ -37,7 +37,7 @@ See the vLLM SkyPilot YAML for serving, [serving.yaml](https://github.com/skypil
       HF_TOKEN: <your-huggingface-token>  # Change to your own huggingface token, or use --env to pass.
 
     setup: |
-      conda create -n vllm python=3.10 -y
+      conda create -n vllm python=3.12 -y
       conda activate vllm
 
       pip install vllm==0.4.0.post1
@@ -136,7 +136,7 @@ SkyPilot can scale up the service to multiple service replicas with built-in aut
       HF_TOKEN: <your-huggingface-token>  # Change to your own huggingface token, or use --env to pass.
 
     setup: |
-      conda create -n vllm python=3.10 -y
+      conda create -n vllm python=3.12 -y
       conda activate vllm
 
       pip install vllm==0.4.0.post1
@@ -248,7 +248,7 @@ This will scale the service up to when the QPS exceeds 2 for each replica.
       HF_TOKEN: <your-huggingface-token>  # Change to your own huggingface token, or use --env to pass.
 
     setup: |
-      conda create -n vllm python=3.10 -y
+      conda create -n vllm python=3.12 -y
       conda activate vllm
 
       pip install vllm==0.4.0.post1
@@ -293,7 +293,7 @@ It is also possible to access the Llama-3 service with a separate GUI frontend,
       cpus: 2
 
     setup: |
-      conda create -n vllm python=3.10 -y
+      conda create -n vllm python=3.12 -y
       conda activate vllm
 
       # Install Gradio for web UI.

docs/getting_started/installation/cpu.md

@@ -30,7 +30,7 @@ When open a Github issue about the CPU backend, please add `[CPU Backend]` in th
 
 ## Requirements
 
-- Python: 3.10 -- 3.13
+- Python: 3.11 -- 3.14
 
 === "Intel/AMD x86"
 

docs/getting_started/installation/gpu.md

@@ -25,7 +25,7 @@ vLLM is a Python library that supports the following GPU variants. Select your G
 ## Requirements
 
 - OS: Linux
-- Python: 3.10 -- 3.13
+- Python: 3.11 -- 3.14
 
 !!! note
     vLLM does not support Windows natively. To run vLLM on Windows, you can use the Windows Subsystem for Linux (WSL) with a compatible Linux distribution, or use some community-maintained forks, e.g. [https://github.com/SystemPanic/vllm-windows](https://github.com/SystemPanic/vllm-windows).

docs/getting_started/quickstart.md

@@ -8,7 +8,7 @@ This guide will help you quickly get started with vLLM to perform:
 ## Prerequisites
 
 - OS: Linux
-- Python: 3.10 -- 3.13
+- Python: 3.11 -- 3.14
 
 !!! note
     vLLM also works on macOS with [vLLM-Metal](https://github.com/vllm-project/vllm-metal) for Apple Silicon GPU acceleration. See the [GPU installation guide](installation/gpu.md) and select the "Apple Silicon" tab.

examples/features/structured_outputs/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "examples-online-structured-outputs"
-requires-python = ">=3.10, <3.14"
+requires-python = ">=3.11, <3.15"
 dependencies = ["openai==1.78.1", "pydantic==2.11.4"]
 version = "0.0.0"
 

pyproject.toml

@@ -21,7 +21,6 @@ license-files = ["LICENSE"]
 readme = "README.md"
 description = "A high-throughput and memory-efficient inference and serving engine for LLMs"
 classifiers = [
-    "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
     "Programming Language :: Python :: 3.13",
@@ -32,7 +31,7 @@ classifiers = [
     "Topic :: Scientific/Engineering :: Artificial Intelligence",
     "Topic :: Scientific/Engineering :: Information Analysis",
 ]
-requires-python = ">=3.10,<3.15"
+requires-python = ">=3.11,<3.15"
 dynamic = [ "version", "dependencies", "optional-dependencies"]
 
 [project.urls]
@@ -103,6 +102,8 @@ ignore = [
     "B007",
     # f-string format
     "UP032",
+    # `StrEnum` changes `str()` and `format()` output of members
+    "UP042",
 ]
 
 [tool.ruff.lint.pydocstyle]

requirements/cpu.txt

@@ -22,7 +22,6 @@ torchcodec >= 0.14; platform_machine != "riscv64" and platform_machine != "ppc64
 intel-openmp==2024.2.1; platform_machine == "x86_64"
 
 # triton-cpu, only for x86_64 CPUs
-triton @ https://github.com/triton-lang/triton-cpu/releases/download/v3.8.0/triton-3.8.0+cpu-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl ; platform_machine == "x86_64" and python_version == "3.10"
 triton @ https://github.com/triton-lang/triton-cpu/releases/download/v3.8.0/triton-3.8.0+cpu-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl ; platform_machine == "x86_64" and python_version == "3.11"
 triton @ https://github.com/triton-lang/triton-cpu/releases/download/v3.8.0/triton-3.8.0+cpu-cp312-abi3-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl ; platform_machine == "x86_64" and python_version >= "3.12"
 

setup.py

@@ -268,8 +268,8 @@ class cmake_build_ext(build_ext):
         cfg = envs.CMAKE_BUILD_TYPE or default_cfg
 
         cmake_args = [
-            "-DCMAKE_BUILD_TYPE={}".format(cfg),
+            f"-DCMAKE_BUILD_TYPE={cfg}",
-            "-DVLLM_TARGET_DEVICE={}".format(VLLM_TARGET_DEVICE),
+            f"-DVLLM_TARGET_DEVICE={VLLM_TARGET_DEVICE}",
         ]
 
         verbose = envs.VERBOSE
@@ -293,7 +293,7 @@ class cmake_build_ext(build_ext):
 
         # Pass the python executable to cmake so it can find an exact
         # match.
-        cmake_args += ["-DVLLM_PYTHON_EXECUTABLE={}".format(sys.executable)]
+        cmake_args += [f"-DVLLM_PYTHON_EXECUTABLE={sys.executable}"]
 
         # Pass the python path to cmake so it can reuse the build dependencies
         # on subsequent calls to python.
@@ -305,7 +305,7 @@ class cmake_build_ext(build_ext):
         # To override this, set the FETCHCONTENT_BASE_DIR environment variable.
         fc_base_dir = os.path.join(ROOT_DIR, ".deps")
         fc_base_dir = os.environ.get("FETCHCONTENT_BASE_DIR", fc_base_dir)
-        cmake_args += ["-DFETCHCONTENT_BASE_DIR={}".format(fc_base_dir)]
+        cmake_args += [f"-DFETCHCONTENT_BASE_DIR={fc_base_dir}"]
 
         #
         # Setup parallelism and build tool
@@ -313,13 +313,13 @@ class cmake_build_ext(build_ext):
         num_jobs, nvcc_threads = self.compute_num_jobs()
 
         if nvcc_threads:
-            cmake_args += ["-DNVCC_THREADS={}".format(nvcc_threads)]
+            cmake_args += [f"-DNVCC_THREADS={nvcc_threads}"]
 
         if is_ninja_available():
             build_tool = ["-G", "Ninja"]
             cmake_args += [
                 "-DCMAKE_JOB_POOL_COMPILE:STRING=compile",
-                "-DCMAKE_JOB_POOLS:STRING=compile={}".format(num_jobs),
+                f"-DCMAKE_JOB_POOLS:STRING=compile={num_jobs}",
             ]
         else:
             # Default build tool to whatever cmake picks.
@@ -1374,7 +1374,7 @@ if _is_cuda() or _is_hip():
     # copying the relevant .py files from the source repository.
     ext_modules.append(CMakeExtension(name="vllm.triton_kernels", optional=True))
 
-if not _is_xpu() and sys.version_info >= (3, 11):
+if not _is_xpu():
     ext_modules.append(CMakeExtension(name="vllm.spinloop"))
     ext_modules.append(CMakeExtension(name="vllm.fs_io_C"))
 

tests/entrypoints/launchers/test_dp_supervisor.py

@@ -630,7 +630,7 @@ async def _poll_until_api_server_running(
                     timeout=request_timeout,
                 ):
                     return
-            except (aiohttp.ClientError, asyncio.TimeoutError) as exc:
+            except (TimeoutError, aiohttp.ClientError) as exc:
                 last_exc = exc
                 print("Test detected not started yet, sleeping for 1s")
                 remaining_s = deadline - time.monotonic()

tests/entrypoints/launchers/test_shutdown.py

@@ -285,7 +285,7 @@ async def test_wait_timeout_completes_requests():
             sigterm_sent.set()
 
             await asyncio.wait_for(request_task, timeout=_SHUTDOWN_DETECTION_TIMEOUT)
-        except asyncio.TimeoutError:
+        except TimeoutError:
             pass
         finally:
             state.stop_requesting = True
@@ -463,7 +463,7 @@ async def test_abort_timeout_fails_inflight_requests():
 
         try:
             await asyncio.wait_for(request_task, timeout=5)
-        except asyncio.TimeoutError:
+        except TimeoutError:
             pass
         finally:
             state.stop_requesting = True
@@ -592,7 +592,7 @@ async def test_multi_api_server_shutdown():
 
         try:
             await asyncio.wait_for(request_task, timeout=_SHUTDOWN_DETECTION_TIMEOUT)
-        except asyncio.TimeoutError:
+        except TimeoutError:
             pass
         finally:
             state.stop_requesting = True

tests/model_executor/model_loader/tensorizer_loader/test_tensorizer.py

@@ -579,7 +579,7 @@ async def test_serialize_and_serve_entrypoints(tmp_path):
 
     try:
         await asyncio.wait_for(fut, 180)
-    except asyncio.TimeoutError:
+    except TimeoutError:
         pytest.fail("Server did not start successfully")
     finally:
         proc.terminate()

tests/multimodal/media/test_connector.py

@@ -1,7 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
-import asyncio
 import mimetypes
 import os
 import shutil
@@ -292,7 +291,7 @@ async def test_fetch_video_http(video_url: str, num_frames: int):
     try:
         video_sync, metadata_sync = connector.fetch_video(video_url)
         video_async, metadata_async = await connector.fetch_video_async(video_url)
-    except (TimeoutError, asyncio.TimeoutError) as e:
+    except TimeoutError as e:
         pytest.skip(f"Timeout fetching video (CI network flakiness): {e}")
 
     assert np.array_equal(video_sync, video_async)
@@ -324,7 +323,7 @@ async def test_fetch_video_http_with_dynamic_loader(
         try:
             video_sync, metadata_sync = connector.fetch_video(video_url)
             video_async, metadata_async = await connector.fetch_video_async(video_url)
-        except (TimeoutError, asyncio.TimeoutError) as e:
+        except TimeoutError as e:
             pytest.skip(f"Timeout fetching video (CI network flakiness): {e}")
 
         assert np.array_equal(video_sync, video_async)
@@ -402,7 +401,7 @@ async def test_allowed_media_domains(video_url: str, num_frames: int):
     try:
         video_sync, metadata_sync = connector.fetch_video(video_url)
         video_async, metadata_async = await connector.fetch_video_async(video_url)
-    except (TimeoutError, asyncio.TimeoutError) as e:
+    except TimeoutError as e:
         pytest.skip(f"Timeout fetching video (CI network flakiness): {e}")
 
     assert np.array_equal(video_sync, video_async)

tests/v1/distributed/test_async_llm_dp.py

@@ -719,7 +719,7 @@ async def test_dp_pause_barrier_request_deadlock():
         # Drive the staggered barrier.  Old code deadlocks here.
         try:
             await asyncio.wait_for(client.call_utility_async("barrier"), timeout=30)
-        except asyncio.TimeoutError:
+        except TimeoutError:
             for t in mid_barrier_tasks:
                 t.cancel()
             pytest.fail(

tools/ci/kernel_symbol_map.py

@@ -255,7 +255,7 @@ def main() -> int:
         or os.environ.get("VLLM_BUILD_COMMIT")
         or "",
         "cuda": cuda_release(nvcc),
-        "generated_at": dt.datetime.now(dt.timezone.utc).isoformat(timespec="seconds"),
+        "generated_at": dt.datetime.now(dt.UTC).isoformat(timespec="seconds"),
         "build_dirs": [],
         "incomplete": False,
         "errors": [],

tools/pre_commit/check_forbidden_imports.py

@@ -88,7 +88,11 @@ CHECK_IMPORTS = {
         allowed_pattern=re.compile(r"^\s*import\s+pybase64(\s*|\s+as\s+base64\s*)$"),
     ),
     "re": ForbiddenImport(
-        pattern=r"^\s*(?:import\s+re(?:$|\s|,)|from\s+re\s+import)",
+        # regex has no equivalent of the private sre modules
+        pattern=(
+            r"^\s*(?:import\s+re(?:$|\s|,)"
+            r"|from\s+re\s+import(?!(?:\s*,?\s*_(?:constants|parser))+\s*(?:#.*)?$))"
+        ),
         tip="Replace 'import re' with 'import regex as re' or 'import regex'.",
         allowed_pattern=re.compile(r"^\s*import\s+regex(\s*|\s+as\s+re\s*)$"),
         allowed_files={"setup.py"},

tools/pre_commit/mypy.py

@@ -10,7 +10,7 @@ Usage:
     python tools/pre_commit/mypy.py <python_version> <changed_files...>
 
 Args:
-    python_version: Python version to use (e.g., "3.10") or "local" to use
+    python_version: Python version to use (e.g., "3.11") or "local" to use
         the local Python version.
     changed_files: List of changed files to check.
 
@@ -152,7 +152,7 @@ def mypy(
 
     Args:
         targets: List of files or directories to check.
-        python_version: Python version to use (e.g., "3.10") or None to use
+        python_version: Python version to use (e.g., "3.11") or None to use
             the default mypy version.
         follow_imports: Value for the --follow-imports option or None to use
             the default mypy behavior.

vllm/benchmarks/serve.py

@@ -253,7 +253,7 @@ async def fetch_spec_decode_metrics(
                 num_accepted_tokens=num_accepted_tokens,
                 accepted_per_pos=accepted_per_pos,
             )
-    except (aiohttp.ClientError, asyncio.TimeoutError):
+    except (TimeoutError, aiohttp.ClientError):
         return None
 
 
@@ -315,7 +315,7 @@ async def fetch_diffusion_metrics(
                 num_canvas_positions=num_canvas_positions,
                 num_committed_tokens=num_committed_tokens,
             )
-    except (aiohttp.ClientError, asyncio.TimeoutError):
+    except (TimeoutError, aiohttp.ClientError):
         return None
 
 

vllm/benchmarks/sweep/plot.py

@@ -8,9 +8,9 @@ from dataclasses import dataclass
 from functools import partial
 from pathlib import Path
 from types import TracebackType
-from typing import TYPE_CHECKING, ClassVar
+from typing import TYPE_CHECKING, ClassVar, Self
 
-from typing_extensions import Self, override
+from typing_extensions import override
 
 from vllm.utils.argparse_utils import FlexibleArgumentParser
 from vllm.utils.collection_utils import full_groupby

vllm/benchmarks/sweep/serve_workload.py

@@ -4,10 +4,9 @@ import argparse
 import math
 from dataclasses import dataclass, fields
 from pathlib import Path
-from typing import ClassVar, Literal, get_args
+from typing import ClassVar, Literal, assert_never, get_args
 
 import numpy as np
-from typing_extensions import assert_never
 
 from vllm.benchmarks.datasets import DEFAULT_NUM_PROMPTS
 from vllm.utils.argparse_utils import FlexibleArgumentParser

vllm/benchmarks/sweep/server.py

@@ -6,9 +6,9 @@ import signal
 import subprocess
 import time
 from types import TracebackType
+from typing import Self
 
 import requests
-from typing_extensions import Self
 
 
 class ServerProcess:

vllm/config/engram.py

@@ -1,10 +1,9 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Self
 
 from pydantic import model_validator
-from typing_extensions import Self
 
 from vllm.config.utils import config, get_hash_factors, hash_factors
 

vllm/config/lora.py

@@ -2,11 +2,10 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 import os
-from typing import TYPE_CHECKING, Any, Literal
+from typing import TYPE_CHECKING, Any, Literal, Self
 
 import torch
 from pydantic import ConfigDict, Field, model_validator
-from typing_extensions import Self
 
 from vllm import envs
 from vllm.config.utils import config

vllm/config/parallel.py

@@ -5,13 +5,12 @@ import os
 import socket
 from collections.abc import Callable
 from datetime import timedelta
-from typing import TYPE_CHECKING, Any, Literal, overload
+from typing import TYPE_CHECKING, Any, Literal, Self, overload
 
 import regex as re
 import torch
 from pydantic import Field, field_validator, model_validator
 from torch.distributed import ProcessGroup, ReduceOp, Store
-from typing_extensions import Self
 
 import vllm.envs as envs
 from vllm.config.fault_tolerance import FaultToleranceConfig

vllm/config/profiler.py

@@ -2,10 +2,9 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 import os
-from typing import Any, Literal
+from typing import Any, Literal, Self
 
 from pydantic import Field, model_validator
-from typing_extensions import Self
 
 from vllm.config.utils import config
 from vllm.logger import init_logger

vllm/config/scheduler.py

@@ -3,10 +3,9 @@
 
 from collections.abc import Callable
 from dataclasses import InitVar
-from typing import TYPE_CHECKING, Any, ClassVar, Literal, cast
+from typing import TYPE_CHECKING, Any, ClassVar, Literal, Self, cast
 
 from pydantic import Field, field_validator
-from typing_extensions import Self
 
 from vllm.config.utils import config
 from vllm.logger import init_logger

vllm/config/speculative.py

@@ -4,10 +4,9 @@
 import copy
 import functools
 from collections.abc import Callable, Mapping
-from typing import TYPE_CHECKING, Any, Literal, get_args
+from typing import TYPE_CHECKING, Any, Literal, Self, get_args
 
 from pydantic import Field, SkipValidation, field_validator, model_validator
-from typing_extensions import Self
 
 from vllm.config import LoadConfig
 from vllm.config.cache import CacheDType

vllm/config/structured_outputs.py

@@ -1,10 +1,9 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
-from typing import Any, Literal
+from typing import Any, Literal, Self
 
 from pydantic import model_validator
-from typing_extensions import Self
 
 from vllm.config.utils import config
 from vllm.utils.hashing import safe_hash

vllm/config/watermarking.py

@@ -2,10 +2,9 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 import hashlib
-from typing import Literal
+from typing import Literal, Self
 
 from pydantic import Field, model_validator
-from typing_extensions import Self
 
 from vllm.config.utils import config
 from vllm.logger import init_logger

vllm/distributed/utils.py

@@ -38,9 +38,7 @@ logger = init_logger(__name__)
 # We prefer to use os.sched_yield as it results in tighter polling loops,
 # measured to be around 3e-7 seconds. However on earlier versions of Python
 # os.sched_yield() does not release the GIL, so we fall back to time.sleep(0)
-USE_SCHED_YIELD = (sys.version_info[:3] >= (3, 11, 1)) or (
+USE_SCHED_YIELD = sys.version_info[:3] >= (3, 11, 1)
-    sys.version_info[:2] == (3, 10) and sys.version_info[2] >= 8
-)
 
 
 def sched_yield():
@@ -52,8 +50,8 @@ def sched_yield():
 
 def ensure_divisibility(numerator, denominator):
     """Ensure that numerator is divisible by the denominator."""
-    assert numerator % denominator == 0, "{} is not divisible by {}".format(
+    assert numerator % denominator == 0, (
-        numerator, denominator
+        f"{numerator} is not divisible by {denominator}"
     )
 
 
@@ -146,9 +144,7 @@ def get_pp_indices(
         try:
             partitions = [int(layer) for layer in partition_list_str.split(",")]
         except ValueError as err:
-            raise ValueError(
+            raise ValueError(f"Invalid partition string: {partition_list_str}") from err
-                "Invalid partition string: {}".format(partition_list_str)
-            ) from err
         if len(partitions) != pp_size:
             raise ValueError(f"{len(partitions)=} does not match {pp_size=}.")
         if sum(partitions) != num_hidden_layers:

vllm/distributed/weight_transfer/base.py

@@ -11,12 +11,12 @@ from typing import (
     ClassVar,
     Generic,
     Protocol,
+    Self,
     TypeVar,
     runtime_checkable,
 )
 
 import torch
-from typing_extensions import Self
 
 if TYPE_CHECKING:
     from vllm.config import VllmConfig

vllm/distributed/weight_transfer/ipc_engine.py

@@ -4,12 +4,11 @@
 
 import pickle
 from dataclasses import asdict, dataclass
-from typing import TYPE_CHECKING, Any, ClassVar
+from typing import TYPE_CHECKING, Any, ClassVar, Self
 
 import pybase64 as base64
 import torch
 from torch.multiprocessing.reductions import rebuild_cuda_tensor, reduce_tensor
-from typing_extensions import Self
 
 from vllm import envs
 from vllm.config.weight_transfer import WeightTransferConfig

vllm/distributed/weight_transfer/m2n_trainer.py

@@ -15,10 +15,9 @@ in every caller.
 
 from concurrent.futures import Future, ThreadPoolExecutor
 from dataclasses import dataclass, field
-from typing import TYPE_CHECKING, Any, ClassVar, cast
+from typing import TYPE_CHECKING, Any, ClassVar, Self, cast
 
 import torch
-from typing_extensions import Self
 
 from vllm.distributed.weight_transfer.base import (
     TrainerInitInfo,

vllm/distributed/weight_transfer/nccl_engine.py

@@ -5,10 +5,9 @@
 from collections.abc import Iterator
 from concurrent.futures import ThreadPoolExecutor
 from dataclasses import asdict, dataclass
-from typing import TYPE_CHECKING, ClassVar
+from typing import TYPE_CHECKING, ClassVar, Self
 
 import torch
-from typing_extensions import Self
 
 if TYPE_CHECKING:
     from vllm.config import VllmConfig

vllm/distributed/weight_transfer/sharded_rdt_trainer.py

@@ -24,7 +24,7 @@ import uuid
 from collections.abc import Callable, Collection
 from concurrent.futures import ThreadPoolExecutor
 from dataclasses import asdict, dataclass, field
-from typing import Any, ClassVar
+from typing import Any, ClassVar, Self
 
 import ray
 import torch
@@ -33,7 +33,6 @@ from torch.multiprocessing.reductions import (
     rebuild_cuda_tensor,
     reduce_tensor,
 )
-from typing_extensions import Self
 
 from vllm.distributed.weight_transfer.base import (
     ParamMeta,

vllm/distributed/weight_transfer/sparse_nccl_engine.py

@@ -11,10 +11,9 @@ from collections.abc import Iterable
 from concurrent.futures import ThreadPoolExecutor
 from contextlib import nullcontext
 from dataclasses import asdict, dataclass
-from typing import TYPE_CHECKING, ClassVar
+from typing import TYPE_CHECKING, ClassVar, Self
 
 import torch
-from typing_extensions import Self
 
 if TYPE_CHECKING:
     from vllm.config import VllmConfig

vllm/entrypoints/chat_utils.py

@@ -17,6 +17,7 @@ from typing import (
     Final,
     Generic,
     Literal,
+    Required,
     TypeAlias,
     TypeVar,
     Union,
@@ -48,7 +49,7 @@ from PIL import Image
 from pydantic import BaseModel, ConfigDict, TypeAdapter
 
 # pydantic needs the TypedDict from typing_extensions
-from typing_extensions import Required, TypedDict, override
+from typing_extensions import TypedDict, override
 
 from vllm import envs
 from vllm.config import ModelConfig

vllm/entrypoints/launchers/dp_supervisor.py

@@ -200,7 +200,7 @@ async def _probe_endpoint(
                 # vLLM returns 503 on EngineDeadError, so we should return
                 # immediately if vLLM responds with a non-200 status code.
                 return response.status == HTTPStatus.OK
-        except (aiohttp.ClientError, asyncio.TimeoutError) as e:
+        except (TimeoutError, aiohttp.ClientError) as e:
             # Allow retry of connection errors.
             logger.debug(
                 "Probe attempt %d/%d failed on port %d: %r",

vllm/entrypoints/llm.py

@@ -3,13 +3,12 @@
 
 from collections.abc import Callable, Sequence
 from pathlib import Path
-from typing import TYPE_CHECKING, Any
+from typing import TYPE_CHECKING, Any, overload
 
 import cloudpickle
 import torch.nn as nn
 from pydantic import ValidationError
 from tqdm.auto import tqdm
-from typing_extensions import overload
 
 from vllm.config import (
     AttentionConfig,

vllm/entrypoints/pooling/embed/serving.py

@@ -1,9 +1,8 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-from typing import TypeAlias, cast
+from typing import TypeAlias, assert_never, cast
 
 from fastapi.responses import JSONResponse, Response, StreamingResponse
-from typing_extensions import assert_never
 
 from vllm.logger import init_logger
 from vllm.outputs import PoolingRequestOutput

vllm/entrypoints/pooling/pooling/serving.py

@@ -1,9 +1,8 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
-from typing import cast
+from typing import assert_never, cast
 
 from fastapi.responses import JSONResponse, Response, StreamingResponse
-from typing_extensions import assert_never
 
 from vllm.exceptions import VLLMValidationError
 from vllm.logger import init_logger

vllm/entrypoints/pooling/scoring/typing.py

@@ -1,9 +1,9 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from dataclasses import dataclass
-from typing import TypeAlias
+from typing import Required, TypeAlias
 
-from typing_extensions import Required, TypedDict
+from typing_extensions import TypedDict
 
 from vllm.entrypoints.chat_utils import (
     ChatCompletionContentPartImageEmbedsParam,

vllm/entrypoints/serve/tokenize/api_router.py

@@ -3,10 +3,10 @@
 
 
 from http import HTTPStatus
+from typing import assert_never
 
 from fastapi import APIRouter, Depends, FastAPI, Request
 from fastapi.responses import JSONResponse
-from typing_extensions import assert_never
 
 from vllm.entrypoints.serve.engine.protocol import ErrorResponse
 from vllm.entrypoints.serve.tokenize.protocol import (

vllm/inputs/engine.py

@@ -3,9 +3,9 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from collections.abc import Mapping, Sequence
-from typing import TYPE_CHECKING, Literal, TypeAlias
+from typing import TYPE_CHECKING, Literal, NotRequired, TypeAlias, assert_never
 
-from typing_extensions import NotRequired, TypedDict, assert_never
+from typing_extensions import TypedDict
 
 from vllm.exceptions import VLLMValidationError
 

vllm/inputs/llm.py

@@ -3,9 +3,9 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from collections.abc import Mapping, Sequence
-from typing import TYPE_CHECKING, Any, TypeAlias, TypeVar, final
+from typing import TYPE_CHECKING, Any, NotRequired, TypeAlias, TypeVar, final
 
-from typing_extensions import NotRequired, TypedDict
+from typing_extensions import TypedDict
 
 if TYPE_CHECKING:
     import torch

vllm/model_executor/kernels/linear/base.py

@@ -3,10 +3,9 @@
 
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
-from typing import Any, ClassVar, Generic, TypeVar
+from typing import Any, ClassVar, Generic, Self, TypeVar
 
 import torch
-from typing_extensions import Self
 
 from vllm.model_executor.layers.quantization.utils.quant_utils import QuantKey
 

vllm/model_executor/kernels/linear/scaled_mm/BlockScaledMMLinearKernel.py

@@ -3,10 +3,9 @@
 
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
-from typing import ClassVar
+from typing import ClassVar, Self
 
 import torch
-from typing_extensions import Self
 
 from vllm.model_executor.layers.quantization.input_quant_fp8 import QuantFP8
 from vllm.model_executor.layers.quantization.utils.fp8_utils import (

vllm/model_executor/model_loader/utils.py

@@ -6,11 +6,10 @@ import inspect
 import time
 import warnings
 from contextlib import contextmanager, nullcontext
-from typing import Any
+from typing import Any, assert_never
 
 import torch
 from torch import nn
-from typing_extensions import assert_never
 
 import vllm.envs as envs
 from vllm.config import (

vllm/model_executor/models/interfaces.py

@@ -21,6 +21,7 @@ from typing import (
     ClassVar,
     Literal,
     Protocol,
+    Self,
     TypeAlias,
     overload,
     runtime_checkable,
@@ -31,7 +32,7 @@ import torch
 import torch.nn as nn
 from torch import Tensor
 from transformers.models.whisper.tokenization_whisper import LANGUAGES
-from typing_extensions import Self, TypeIs
+from typing_extensions import TypeIs
 
 from vllm.logger import init_logger
 from vllm.model_executor.layers.quantization import QuantizationConfig

vllm/models/deepseek_v4/common/mm_preprocess.py

@@ -22,13 +22,12 @@ reference's out-of-vocab scheme.
 
 import math
 from collections.abc import Mapping, Sequence
-from typing import Any
+from typing import Any, assert_never
 
 import numpy as np
 import torch
 from PIL import Image, ImageOps
 from transformers import BatchFeature
-from typing_extensions import assert_never
 
 from vllm.config.multimodal import MultiModalDummyOptions
 from vllm.inputs import MultiModalDataDict

vllm/multimodal/parse.py

@@ -13,11 +13,11 @@ from typing import (
     TypeAlias,
     TypeGuard,
     TypeVar,
+    assert_never,
 )
 
 import numpy as np
 import torch
-from typing_extensions import assert_never
 
 from vllm.inputs import ModalityData, MultiModalDataDict, MultiModalUUIDDict
 from vllm.utils.collection_utils import is_list_of

vllm/multimodal/processing/processor.py

@@ -13,10 +13,11 @@ from typing import (
     NamedTuple,
     Protocol,
     TypeAlias,
+    assert_never,
 )
 
 import torch
-from typing_extensions import TypeVar, assert_never
+from typing_extensions import TypeVar
 
 from vllm.config import SchedulerConfig
 from vllm.inputs import (

vllm/renderers/hf.py

@@ -1000,7 +1000,7 @@ class HfRenderer(BaseRenderer[HfTokenizer]):
             return await coro
         try:
             return await asyncio.wait_for(coro, timeout=timeout)
-        except asyncio.TimeoutError:
+        except TimeoutError:
             self._replace_executor()
             raise TimeoutError(
                 f"Chat template rendering timed out after "

vllm/snapshot/manifest.py

@@ -6,7 +6,7 @@ import math
 import os
 import stat
 from pathlib import Path
-from typing import Annotated, Any, Literal
+from typing import Annotated, Any, Literal, Self
 
 from pydantic import (
     BaseModel,
@@ -19,7 +19,6 @@ from pydantic import (
     field_validator,
     model_validator,
 )
-from typing_extensions import Self
 
 
 class SnapshotCompatibilityError(RuntimeError):

vllm/snapshot/runtime.py

@@ -24,7 +24,7 @@ import urllib.request
 from collections.abc import Iterable
 from contextlib import suppress
 from dataclasses import dataclass
-from datetime import datetime, timezone
+from datetime import UTC, datetime
 from pathlib import Path
 from typing import NoReturn
 
@@ -872,7 +872,7 @@ class LocalSnapshotTools:
         return SnapshotManifest(
             schema_version=1,
             boundary="post-engine-init-reloadable-state-released",
-            created_at=datetime.now(timezone.utc).isoformat(),
+            created_at=datetime.now(UTC).isoformat(),
             artifact_bytes=self._artifact_bytes(workdir),
             **identity.model_dump(),
             model=source_model,

vllm/tokenizers/registry.py

@@ -4,10 +4,10 @@ import contextlib
 from dataclasses import dataclass, field
 from functools import lru_cache
 from pathlib import Path
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, assert_never
 
 import huggingface_hub
-from typing_extensions import TypeVar, assert_never
+from typing_extensions import TypeVar
 
 from vllm import envs
 from vllm.logger import init_logger

vllm/usage/usage_lib.py

@@ -69,7 +69,7 @@ def is_usage_stats_enabled():
 
 
 def _get_current_timestamp_ns() -> int:
-    return int(datetime.datetime.now(datetime.timezone.utc).timestamp() * 1e9)
+    return int(datetime.datetime.now(datetime.UTC).timestamp() * 1e9)
 
 
 def _detect_cloud_provider() -> str:

vllm/utils/collection_utils.py

@@ -8,9 +8,9 @@ This is similar in concept to the `collections` module.
 import math
 from collections import defaultdict
 from collections.abc import Callable, Generator, Hashable, Iterable, Mapping, Sequence
-from typing import Generic, Literal, TypeVar
+from typing import Generic, Literal, TypeVar, assert_never, overload
 
-from typing_extensions import TypeIs, assert_never, overload
+from typing_extensions import TypeIs
 
 T = TypeVar("T")
 

vllm/utils/import_utils.py

@@ -12,10 +12,9 @@ import os
 import sys
 from functools import cache
 from types import ModuleType
-from typing import Any
+from typing import Any, Never
 
 import regex as re
-from typing_extensions import Never
 
 from vllm.logger import init_logger
 

vllm/v1/kv_cache_interface.py

@@ -11,10 +11,9 @@ from enum import Enum, IntEnum
 from fractions import Fraction
 from functools import cached_property
 from math import prod
-from typing import TYPE_CHECKING, TypeVar
+from typing import TYPE_CHECKING, Self, TypeVar
 
 import torch
-from typing_extensions import Self
 
 from vllm.logger import init_logger
 from vllm.utils.math_utils import cdiv, round_up

vllm/v1/metrics/perf.py

@@ -10,12 +10,11 @@ import time
 from abc import ABC, abstractmethod
 from collections.abc import Iterable
 from dataclasses import asdict, dataclass, field
-from typing import Any, Protocol
+from typing import Any, Protocol, Self
 
 import prometheus_client
 import torch
 from pydantic import BaseModel, Field, ValidationError, model_validator
-from typing_extensions import Self
 
 import vllm.envs as envs
 from vllm.config import VllmConfig

vllm/v1/structured_output/backend_outlines.py

@@ -4,10 +4,9 @@
 from __future__ import annotations
 
 import ast
-import importlib
 import json
-import sys
 from dataclasses import dataclass, field
+from re import _constants, _parser  # type: ignore[attr-defined]
 from typing import TYPE_CHECKING
 
 import torch
@@ -36,19 +35,6 @@ else:
     oc = LazyLoader("oc", globals(), "outlines_core")
     json_schema = LazyLoader("json_schema", globals(), "outlines_core.json_schema")
 
-# Python 3.11+ sre_parse and sre_constants
-# are deprecated, so we must import them from re
-if sys.version_info >= (3, 11):
-    # Hack to get around pre-commit regex module rule
-    # because going through re is the only way to get sre_parse
-    # and sre_constants in Python 3.11+
-    _re = importlib.import_module("re")
-    sre_parse = _re._parser
-    sre_constants = _re._constants
-else:
-    import sre_constants
-    import sre_parse
-
 
 @dataclass
 class OutlinesBackend(StructuredOutputBackend):
@@ -261,20 +247,20 @@ def _prefix_needs_context(parsed) -> bool:
         tokens = parsed.data if hasattr(parsed, "data") else parsed
         for ttype, tval in tokens:
             # literal, character class, or dot always consumes
-            if ttype in (sre_parse.LITERAL, sre_parse.IN, sre_parse.ANY):
+            if ttype in (_parser.LITERAL, _parser.IN, _parser.ANY):
                 return True
             # quantified subpattern: check inner pattern
-            elif ttype == sre_parse.MAX_REPEAT:
+            elif ttype == _parser.MAX_REPEAT:
                 _, mx, sub = tval
                 if mx != 0 and subpattern_consumes(sub):
                     return True
             # alternation: if any branch consumes, the whole does
-            elif ttype == sre_parse.BRANCH:
+            elif ttype == _parser.BRANCH:
                 _, branches = tval
                 if any(subpattern_consumes(br) for br in branches):
                     return True
             # grouped subpattern: recurse into its contents
-            elif ttype == sre_parse.SUBPATTERN and subpattern_consumes(tval[3]):
+            elif ttype == _parser.SUBPATTERN and subpattern_consumes(tval[3]):
                 return True
         # No consumers, return False
         return False
@@ -282,14 +268,11 @@ def _prefix_needs_context(parsed) -> bool:
     tokens = parsed.data if hasattr(parsed, "data") else parsed
     for ttype, tval in tokens:
         # Direct anchors or look-around
-        if ttype == sre_parse.AT or ttype in (
+        if ttype == _parser.AT or ttype in (_constants.ASSERT, _constants.ASSERT_NOT):
-            sre_constants.ASSERT,
-            sre_constants.ASSERT_NOT,
-        ):
             return True
 
         # Nested subpattern: check
-        if ttype == sre_parse.SUBPATTERN:
+        if ttype == _parser.SUBPATTERN:
             # tval: (group, add_flags, del_flags, subpattern)
             if _prefix_needs_context(tval[3]):
                 return True
@@ -298,7 +281,7 @@ def _prefix_needs_context(parsed) -> bool:
 
         # if any branch has a prefix anchor => True,
         # else if at least one branch consumes => prefix ends => False
-        elif ttype == sre_parse.BRANCH:
+        elif ttype == _parser.BRANCH:
             saw_consumer = False
             for br in tval[1]:
                 if _prefix_needs_context(br):
@@ -309,11 +292,11 @@ def _prefix_needs_context(parsed) -> bool:
                 return False
 
         # Immediate consumer tokens
-        elif ttype in (sre_parse.LITERAL, sre_parse.IN, sre_parse.ANY):
+        elif ttype in (_parser.LITERAL, _parser.IN, _parser.ANY):
             return False
 
         # if subpattern has anchor => True, if it can consume => stop
-        elif ttype == sre_parse.MAX_REPEAT:
+        elif ttype == _parser.MAX_REPEAT:
             if _prefix_needs_context(tval[2]):
                 return True
             if subpattern_consumes(tval[2]):
@@ -327,25 +310,25 @@ def _check_unsupported(parsed) -> None:
     tokens = parsed.data if hasattr(parsed, "data") else parsed
     for ttype, tval in tokens:
         # backreference
-        if ttype in (sre_parse.GROUPREF, sre_parse.GROUPREF_EXISTS):
+        if ttype in (_parser.GROUPREF, _parser.GROUPREF_EXISTS):
             raise ValueError("Backreferences are unsupported.")
 
         # look-around assertion
-        elif ttype in (sre_constants.ASSERT, sre_constants.ASSERT_NOT):
+        elif ttype in (_constants.ASSERT, _constants.ASSERT_NOT):
             raise ValueError("Look-Around assertion are unsupported.")
 
         # unicode word boundaries
-        elif ttype == sre_parse.AT:
+        elif ttype == _parser.AT:
-            if tval in (sre_constants.AT_BOUNDARY, sre_constants.AT_NON_BOUNDARY):
+            if tval in (_constants.AT_BOUNDARY, _constants.AT_NON_BOUNDARY):
                 raise ValueError("Unicode word boundaries are unsupported.")
 
-        elif ttype == sre_parse.BRANCH:
+        elif ttype == _parser.BRANCH:
             # tval is (None, branches)
             for branch in tval[1]:
                 _check_unsupported(branch)
 
         # tval is (min, max, subpattern)
-        elif ttype == sre_parse.MAX_REPEAT:
+        elif ttype == _parser.MAX_REPEAT:
             _check_unsupported(tval[2])
 
 
@@ -357,9 +340,9 @@ def validate_regex_is_buildable(pattern: str) -> None:
     https://docs.rs/regex-automata/latest/regex_automata/dfa/trait.Automaton.html#method.universal_start_state
     """
     try:
-        parsed = sre_parse.parse(pattern)
+        parsed = _parser.parse(pattern)
 
-    except sre_constants.error as e:
+    except _constants.error as e:
         raise VLLMValidationError(f"Error parsing regex: {e}") from e
 
     try:

vllm/v1/worker/workspace.py

@@ -229,12 +229,7 @@ class WorkspaceManager:
                 # Walk up the stack skipping WorkspaceManager frames
                 curr_frame = curr_frame.f_back
                 while curr_frame is not None:
-                    # TODO: This only catches instance methods (self), missing
+                    if curr_frame.f_code.co_qualname.startswith("WorkspaceManager."):
-                    # classmethods and staticmethods. Once Python 3.11+ is the
-                    # minimum supported version, use co_qualname instead:
-                    #   qualname = curr_frame.f_code.co_qualname
-                    #   if qualname.startswith("WorkspaceManager."):
-                    if isinstance(curr_frame.f_locals.get("self"), WorkspaceManager):
                         curr_frame = curr_frame.f_back
                         continue
                     filename = os.path.basename(curr_frame.f_code.co_filename)