Fix and simplify deprecated=True CLI kwarg (#17781)

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

examples/offline_inference/basic/chat.py

@@ -7,9 +7,8 @@ from vllm.utils import FlexibleArgumentParser
 def create_parser():
     parser = FlexibleArgumentParser()
     # Add engine args
-    engine_group = parser.add_argument_group("Engine arguments")
+    EngineArgs.add_cli_args(parser)
-    EngineArgs.add_cli_args(engine_group)
+    parser.set_defaults(model="meta-llama/Llama-3.2-1B-Instruct")
-    engine_group.set_defaults(model="meta-llama/Llama-3.2-1B-Instruct")
     # Add sampling params
     sampling_group = parser.add_argument_group("Sampling parameters")
     sampling_group.add_argument("--max-tokens", type=int)

examples/offline_inference/basic/generate.py

@@ -7,9 +7,8 @@ from vllm.utils import FlexibleArgumentParser
 def create_parser():
     parser = FlexibleArgumentParser()
     # Add engine args
-    engine_group = parser.add_argument_group("Engine arguments")
+    EngineArgs.add_cli_args(parser)
-    EngineArgs.add_cli_args(engine_group)
+    parser.set_defaults(model="meta-llama/Llama-3.2-1B-Instruct")
-    engine_group.set_defaults(model="meta-llama/Llama-3.2-1B-Instruct")
     # Add sampling params
     sampling_group = parser.add_argument_group("Sampling parameters")
     sampling_group.add_argument("--max-tokens", type=int)

vllm/engine/arg_utils.py

@@ -833,6 +833,7 @@ class EngineArgs:
         parser.add_argument('--use-v2-block-manager',
                             action='store_true',
                             default=True,
+                            deprecated=True,
                             help='[DEPRECATED] block manager v1 has been '
                             'removed and SelfAttnBlockSpaceManager (i.e. '
                             'block manager v2) is now the default. '

vllm/utils.py

@@ -41,7 +41,6 @@ from collections.abc import (AsyncGenerator, Awaitable, Generator, Hashable,
 from concurrent.futures.process import ProcessPoolExecutor
 from dataclasses import dataclass, field
 from functools import cache, lru_cache, partial, wraps
-from gettext import gettext as _gettext
 from types import MappingProxyType
 from typing import (TYPE_CHECKING, Any, Callable, Generic, Literal, NamedTuple,
                     Optional, Sequence, Tuple, Type, TypeVar, Union, cast,
@@ -1333,31 +1332,10 @@ class SortedHelpFormatter(ArgumentDefaultsHelpFormatter):
         super().add_arguments(actions)
 
 
-class _FlexibleArgumentGroup(_ArgumentGroup):
-
-    def __init__(self, parser: FlexibleArgumentParser, *args, **kwargs):
-        self._parser = parser
-        super().__init__(*args, **kwargs)
-
-    def add_argument(self, *args: Any, **kwargs: Any):
-        if sys.version_info < (3, 13):
-            deprecated = kwargs.pop('deprecated', False)
-            action = super().add_argument(*args, **kwargs)
-            object.__setattr__(action, 'deprecated', deprecated)
-            if deprecated and action.dest not in \
-                    self._parser.__class__._deprecated:
-                self._parser._deprecated.add(action)
-            return action
-
-        # python>3.13
-        return super().add_argument(*args, **kwargs)
-
-
 class FlexibleArgumentParser(ArgumentParser):
     """ArgumentParser that allows both underscore and dash in names."""
 
     _deprecated: set[Action] = set()
-    _seen: set[str] = set()
 
     def __init__(self, *args, **kwargs):
         # Set the default 'formatter_class' to SortedHelpFormatter
@@ -1366,39 +1344,36 @@ class FlexibleArgumentParser(ArgumentParser):
         super().__init__(*args, **kwargs)
 
     if sys.version_info < (3, 13):
+        # Enable the deprecated kwarg for Python 3.12 and below
 
-        def parse_known_args(  # type: ignore[override]
+        def parse_known_args(self, args=None, namespace=None):
-            self,
-            args: Sequence[str] | None = None,
-            namespace: Namespace | None = None,
-        ) -> tuple[Namespace | None, list[str]]:
             namespace, args = super().parse_known_args(args, namespace)
             for action in FlexibleArgumentParser._deprecated:
-                if action.dest not in FlexibleArgumentParser._seen and getattr(
+                if (hasattr(namespace, dest := action.dest)
-                        namespace, action.dest,
+                        and getattr(namespace, dest) != action.default):
-                        None) != action.default:  # noqa: E501
+                    logger.warning_once("argument '%s' is deprecated", dest)
-                    self._warning(
-                        _gettext("argument '%(argument_name)s' is deprecated")
-                        % {'argument_name': action.dest})
-                    FlexibleArgumentParser._seen.add(action.dest)
             return namespace, args
 
-        def add_argument(self, *args: Any, **kwargs: Any):
+        def add_argument(self, *args, **kwargs):
-            # add a deprecated=True compatibility
+            deprecated = kwargs.pop("deprecated", False)
-            # for python < 3.13
-            deprecated = kwargs.pop('deprecated', False)
             action = super().add_argument(*args, **kwargs)
-            object.__setattr__(action, 'deprecated', deprecated)
+            if deprecated:
-            if deprecated and \
+                FlexibleArgumentParser._deprecated.add(action)
-                action not in FlexibleArgumentParser._deprecated:
-                self._deprecated.add(action)
-
             return action
 
-        def _warning(self, message: str):
+        class _FlexibleArgumentGroup(_ArgumentGroup):
-            self._print_message(
+
-                _gettext('warning: %(message)s\n') % {'message': message},
+            def add_argument(self, *args, **kwargs):
-                sys.stderr)
+                deprecated = kwargs.pop("deprecated", False)
+                action = super().add_argument(*args, **kwargs)
+                if deprecated:
+                    FlexibleArgumentParser._deprecated.add(action)
+                return action
+
+        def add_argument_group(self, *args, **kwargs):
+            group = self._FlexibleArgumentGroup(self, *args, **kwargs)
+            self._action_groups.append(group)
+            return group
 
     def parse_args(  # type: ignore[override]
         self,
@@ -1575,15 +1550,6 @@ class FlexibleArgumentParser(ArgumentParser):
 
         return processed_args
 
-    def add_argument_group(
-        self,
-        *args: Any,
-        **kwargs: Any,
-    ) -> _FlexibleArgumentGroup:
-        group = _FlexibleArgumentGroup(self, self, *args, **kwargs)
-        self._action_groups.append(group)
-        return group
-
 
 async def _run_task_with_lock(task: Callable, lock: asyncio.Lock, *args,
                               **kwargs):