Remove Python 3.9 support ahead of PyTorch 2.9 in v0.11.1 (#26416)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

.pre-commit-config.yaml

@@ -55,11 +55,6 @@ repos:
       types_or: [python, pyi]
       require_serial: true
       additional_dependencies: [mypy==1.11.1, regex, types-cachetools, types-setuptools, types-PyYAML, types-requests, types-torch, pydantic]
-  - id: mypy-3.9 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
-    name: Run mypy for Python 3.9
-    entry: python tools/pre_commit/mypy.py 1 "3.9"
-    <<: *mypy_common
-    stages: [manual] # Only run in CI
   - id: mypy-3.10 # TODO: Use https://github.com/pre-commit/mirrors-mypy when mypy setup is less awkward
     name: Run mypy for Python 3.10
     entry: python tools/pre_commit/mypy.py 1 "3.10"
@@ -75,6 +70,11 @@ repos:
     entry: python tools/pre_commit/mypy.py 1 "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 1 "3.13"
+    <<: *mypy_common
+    stages: [manual] # Only run in CI
   - id: shellcheck
     name: Lint shell scripts
     entry: tools/shellcheck.sh

CMakeLists.txt

@@ -34,7 +34,7 @@ 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 "3.9" "3.10" "3.11" "3.12" "3.13")
+set(PYTHON_SUPPORTED_VERSIONS "3.10" "3.11" "3.12" "3.13")
 
 # Supported AMD GPU architectures.
 set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1030;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151")

benchmarks/multi_turn/benchmark_serving_multi_turn.py

@@ -13,7 +13,7 @@ from datetime import datetime
 from enum import Enum
 from http import HTTPStatus
 from statistics import mean
-from typing import NamedTuple, Optional, Union
+from typing import NamedTuple, Union
 
 import aiohttp  # type: ignore
 import numpy as np  # type: ignore
@@ -46,9 +46,9 @@ class ConversationSampling(str, Enum):
 
 class ClientArgs(NamedTuple):
     seed: int
-    max_num_requests: Optional[int]
+    max_num_requests: int | None
     skip_first_turn: bool
-    max_turns: Optional[int]
+    max_turns: int | None
     max_active_conversations: int
     verbose: bool
     print_content: bool
@@ -109,9 +109,9 @@ class RequestStats(NamedTuple):
 
 class MetricStats:
     def __init__(self) -> None:
-        self.min: Optional[float] = None
-        self.max: Optional[float] = None
-        self.avg: Optional[float] = None
+        self.min: float | None = None
+        self.max: float | None = None
+        self.avg: float | None = None
         self.sum = 0.0
         self.count = 0
 
@@ -143,7 +143,7 @@ class MovingAverage:
         self.index = 0
         self.sum = 0.0
         self.count = 0
-        self.avg: Optional[float] = None
+        self.avg: float | None = None
 
     def update(self, new_value: float) -> None:
         if self.count < self.window_size:
@@ -198,14 +198,6 @@ class DebugStats:
         self.logger.info("-" * 50)
 
 
-# Must support Python 3.8, we can't use str.removeprefix(prefix)
-# introduced in Python 3.9
-def remove_prefix(text: str, prefix: str) -> str:
-    if text.startswith(prefix):
-        return text[len(prefix) :]
-    return text
-
-
 def nanosec_to_millisec(value: float) -> float:
     return value / 1000000.0
 
@@ -220,8 +212,8 @@ async def send_request(
     chat_url: str,
     model: str,
     stream: bool = True,
-    min_tokens: Optional[int] = None,
-    max_tokens: Optional[int] = None,
+    min_tokens: int | None = None,
+    max_tokens: int | None = None,
 ) -> ServerResponse:
     payload = {
         "model": model,
@@ -250,9 +242,9 @@ async def send_request(
     timeout = aiohttp.ClientTimeout(total=timeout_sec)
 
     valid_response = True
-    ttft: Optional[float] = None
+    ttft: float | None = None
     chunk_delay: list[int] = []
-    latency: Optional[float] = None
+    latency: float | None = None
     first_chunk = ""
     generated_text = ""
 
@@ -269,7 +261,7 @@ async def send_request(
                 if not chunk_bytes:
                     continue
 
-                chunk = remove_prefix(chunk_bytes.decode("utf-8"), "data: ")
+                chunk = chunk_bytes.decode("utf-8").removeprefix("data: ")
                 if chunk == "[DONE]":
                     # End of stream
                     latency = time.perf_counter_ns() - start_time
@@ -364,7 +356,7 @@ async def send_turn(
     req_args: RequestArgs,
     verbose: bool,
     verify_output: bool,
-) -> Optional[RequestStats]:
+) -> RequestStats | None:
     assert messages_to_use > 0
     assert messages_to_use <= len(conversation_messages)
 
@@ -769,7 +761,7 @@ def get_client_config(
             "Number of conversations must be equal or larger than the number of clients"
         )
 
-    max_req_per_client: Optional[int] = None
+    max_req_per_client: int | None = None
     if args.max_num_requests is not None:
         # Max number of requests per client
         req_per_client = args.max_num_requests // args.num_clients
@@ -1032,7 +1024,7 @@ def process_statistics(
     warmup_percentages: list[float],
     test_params: dict,
     verbose: bool,
-    gen_conv_args: Optional[GenConvArgs] = None,
+    gen_conv_args: GenConvArgs | None = None,
     excel_output: bool = False,
 ) -> None:
     if len(client_metrics) == 0:

docker/Dockerfile.cpu

@@ -13,7 +13,7 @@
 #   vllm-dev: used for development
 #
 # Build arguments:
-#   PYTHON_VERSION=3.12 (default)|3.11|3.10|3.9
+#   PYTHON_VERSION=3.13|3.12 (default)|3.11|3.10
 #   VLLM_CPU_DISABLE_AVX512=false (default)|true
 #   VLLM_CPU_AVX512BF16=false (default)|true
 #   VLLM_CPU_AVX512VNNI=false (default)|true

docs/contributing/README.md

@@ -54,7 +54,7 @@ For more details about installing from source and installing for other hardware,
 For an optimized workflow when iterating on C++/CUDA kernels, see the [Incremental Compilation Workflow](./incremental_build.md) for recommendations.
 
 !!! tip
-    vLLM is compatible with Python versions 3.9 to 3.12. However, vLLM's default [Dockerfile](gh-file: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 3.10 to 3.13. However, vLLM's default [Dockerfile](gh-file: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.
 
@@ -83,7 +83,7 @@ vLLM's `pre-commit` hooks will now run automatically every time you commit.
 
     ```bash
     pre-commit run --hook-stage manual markdownlint
-    pre-commit run --hook-stage manual mypy-3.9
+    pre-commit run --hook-stage manual mypy-3.10
     ```
 
 ### Documentation

docs/getting_started/installation/cpu.md

@@ -20,7 +20,7 @@ vLLM is a Python library that supports the following CPU variants. Select your C
 
 ## Requirements
 
-- Python: 3.9 -- 3.12
+- Python: 3.10 -- 3.13
 
 === "Intel/AMD x86"
 

docs/getting_started/installation/gpu.md

@@ -17,7 +17,7 @@ vLLM is a Python library that supports the following GPU variants. Select your G
 ## Requirements
 
 - OS: Linux
-- Python: 3.9 -- 3.12
+- Python: 3.10 -- 3.13
 
 !!! 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: 3.9 -- 3.13
+- Python: 3.10 -- 3.13
 
 ## Installation
 

examples/online_serving/structured_outputs/pyproject.toml

 [project]
 name = "examples-online-structured-outputs"
-requires-python = ">=3.9, <3.13"
+requires-python = ">=3.10, <3.14"
 dependencies = ["openai==1.78.1", "pydantic==2.11.4"]
 version = "0.0.0"
 

pyproject.toml

@@ -20,7 +20,6 @@ license-files = ["LICENSE"]
 readme = "README.md"
 description = "A high-throughput and memory-efficient inference and serving engine for LLMs"
 classifiers = [
-    "Programming Language :: Python :: 3.9",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
@@ -31,7 +30,7 @@ classifiers = [
     "Topic :: Scientific/Engineering :: Artificial Intelligence",
     "Topic :: Scientific/Engineering :: Information Analysis",
 ]
-requires-python = ">=3.9,<3.14"
+requires-python = ">=3.10,<3.14"
 dynamic = [ "version", "dependencies", "optional-dependencies"]
 
 [project.urls]
@@ -79,12 +78,18 @@ ignore = [
     "F405", "F403",
     # lambda expression assignment
     "E731",
+    # zip without `strict=`
+    "B905",
     # Loop control variable not used within loop body
     "B007",
     # f-string format
     "UP032",
     # Can remove once 3.10+ is the minimum Python version
     "UP007",
+    "UP027",
+    "UP035",
+    "UP038",
+    "UP045",
 ]
 
 [tool.ruff.format]

requirements/cpu.txt

 # Common dependencies
 -r common.txt
 
-numba == 0.60.0; python_version == '3.9' and platform_machine != "s390x" # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
-numba == 0.61.2; python_version > '3.9' and platform_machine != "s390x"
+numba == 0.61.2; platform_machine != "s390x" # Required for N-gram speculative decoding
 
 # Dependencies for CPUs
 packaging>=24.2

requirements/cuda.txt

 # Common dependencies
 -r common.txt
 
-numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
-numba == 0.61.2; python_version > '3.9'
+numba == 0.61.2 # Required for N-gram speculative decoding
 
 # Dependencies for NVIDIA GPUs
 ray[cgraph]>=2.48.0 # Ray Compiled Graph, required for pipeline parallelism in V1.

requirements/nightly_torch_test.txt

@@ -40,8 +40,7 @@ buildkite-test-collector==0.1.9
 genai_perf==0.0.8
 tritonclient==2.51.0
 
-numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
-numba == 0.61.2; python_version > '3.9'
+numba == 0.61.2 # Required for N-gram speculative decoding
 numpy
 runai-model-streamer[s3,gcs]==0.14.0
 fastsafetensors>=0.1.10

requirements/rocm.txt

 # Common dependencies
 -r common.txt
 
-numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
-numba == 0.61.2; python_version > '3.9'
+numba == 0.61.2 # Required for N-gram speculative decoding
 
 # Dependencies for AMD GPUs
 datasets

requirements/test.in

@@ -48,8 +48,7 @@ buildkite-test-collector==0.1.9
 genai_perf==0.0.8
 tritonclient==2.51.0
 
-numba == 0.60.0; python_version == '3.9' # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
-numba == 0.61.2; python_version > '3.9'
+numba == 0.61.2 # Required for N-gram speculative decoding
 numpy
 runai-model-streamer[s3,gcs]==0.14.0
 fastsafetensors>=0.1.10

requirements/xpu.txt

@@ -9,7 +9,7 @@ setuptools>=77.0.3,<80.0.0
 wheel
 jinja2>=3.1.6
 datasets # for benchmark scripts
-numba == 0.60.0 # v0.61 doesn't support Python 3.9. Required for N-gram speculative decoding
+numba == 0.61.2 # Required for N-gram speculative decoding
 nixl==0.3.0 # for PD disaggregation
 torch==2.8.0+xpu
 torchaudio

tools/validate_config.py

@@ -8,6 +8,7 @@ and that each field has a docstring.
 import ast
 import inspect
 import sys
+from itertools import pairwise
 
 import regex as re
 
@@ -20,19 +21,6 @@ def get_attr_docs(cls_node: ast.ClassDef) -> dict[str, str]:
     https://davidism.com/mit-license/
     """
 
-    def pairwise(iterable):
-        """
-        Manually implement https://docs.python.org/3/library/itertools.html#itertools.pairwise
-
-        Can be removed when Python 3.9 support is dropped.
-        """
-        iterator = iter(iterable)
-        a = next(iterator, None)
-
-        for b in iterator:
-            yield a, b
-            a = b
-
     out = {}
 
     # Consider each pair of nodes.

vllm/config/utils.py

@@ -7,6 +7,7 @@ import inspect
 import textwrap
 from collections.abc import Iterable
 from dataclasses import MISSING, Field, field, fields, is_dataclass, replace
+from itertools import pairwise
 from typing import TYPE_CHECKING, Any, Protocol, TypeVar
 
 import regex as re
@@ -102,19 +103,6 @@ def get_attr_docs(cls: type[Any]) -> dict[str, str]:
     https://davidism.com/mit-license/
     """
 
-    def pairwise(iterable):
-        """
-        Manually implement https://docs.python.org/3/library/itertools.html#itertools.pairwise
-
-        Can be removed when Python 3.9 support is dropped.
-        """
-        iterator = iter(iterable)
-        a = next(iterator, None)
-
-        for b in iterator:
-            yield a, b
-            a = b
-
     try:
         cls_node = ast.parse(textwrap.dedent(inspect.getsource(cls))).body[0]
     except (OSError, KeyError, TypeError):

vllm/plugins/__init__.py

@@ -15,11 +15,6 @@ plugins_loaded = False
 
 
 def load_plugins_by_group(group: str) -> dict[str, Callable[[], Any]]:
-    import sys
-
-    if sys.version_info < (3, 10):
-        from importlib_metadata import entry_points
-    else:
     from importlib.metadata import entry_points
 
     allowed_plugins = envs.VLLM_PLUGINS

vllm/v1/sample/logits_processor/__init__.py

@@ -55,11 +55,6 @@ BUILTIN_LOGITS_PROCESSORS: list[type[LogitsProcessor]] = [
 def _load_logitsprocs_plugins() -> list[type[LogitsProcessor]]:
     """Load all installed logit processor plugins"""
 
-    import sys
-
-    if sys.version_info < (3, 10):
-        from importlib_metadata import entry_points
-    else:
     from importlib.metadata import entry_points
 
     installed_logitsprocs_plugins = entry_points(group=LOGITSPROCS_GROUP)