A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 13:30 UTC
Scheduled removal of `override_pooler_config` and `disable_log_requests` (#29402)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Harry Mellor committed
Nov 25, 2025 at 16:08 UTC
a1f267687956ae1f0b7c1668b48020055d764619
5 files changed
+1
-65
vllm/config/model.py
-16
@@ -299,9 +299,6 @@ class ModelConfig:
299
pooler_config: PoolerConfig | None = None
300
"""Pooler config which controls the behaviour of output pooling in pooling
301
models."""
302
- override_pooler_config: dict | PoolerConfig | None = None
303
- """[DEPRECATED] Use `pooler_config` instead. This field will be removed in
304
- v0.12.0 or v1.0.0, whichever is sooner."""
302
303
# Multimodal config and init vars
304
multimodal_config: MultiModalConfig | None = None
@@ -359,7 +356,6 @@ class ModelConfig:
356
"logits_processors",
357
"io_processor_plugin",
358
"pooler_config",
362
- "override_pooler_config",
359
"multimodal_config",
360
"limit_mm_per_prompt",
361
"media_io_kwargs",
@@ -648,18 +644,6 @@ class ModelConfig:
644
645
# Init pooler config if needed
646
if self.runner_type == "pooling":
651
- if self.override_pooler_config is not None:
652
- logger.warning_once(
653
- "`override_pooler_config` is deprecated and will be "
654
- "removed in v0.12.0 or v1.0.0, whichever is sooner. "
655
- "Please use `pooler_config` instead."
656
- )
657
-
658
- if isinstance(self.override_pooler_config, dict):
659
- self.pooler_config = PoolerConfig(**self.override_pooler_config)
660
- else:
661
- self.pooler_config = self.override_pooler_config
662
-
647
if self.pooler_config is None:
648
self.pooler_config = PoolerConfig()
649
vllm/engine/arg_utils.py
+1
-28
@@ -29,7 +29,7 @@ import regex as re
29
import torch
30
from pydantic import TypeAdapter, ValidationError
31
from pydantic.fields import FieldInfo
32
-from typing_extensions import TypeIs, deprecated
32
+from typing_extensions import TypeIs
33
34
import vllm.envs as envs
35
from vllm.attention.backends.registry import AttentionBackendEnum
@@ -520,9 +520,6 @@ class EngineArgs:
520
scheduler_cls: str | type[object] | None = SchedulerConfig.scheduler_cls
521
522
pooler_config: PoolerConfig | None = ModelConfig.pooler_config
523
- override_pooler_config: dict | PoolerConfig | None = (
524
- ModelConfig.override_pooler_config
525
- )
523
compilation_config: CompilationConfig = get_field(VllmConfig, "compilation_config")
524
worker_cls: str = ParallelConfig.worker_cls
525
worker_extension_cls: str = ParallelConfig.worker_extension_cls
@@ -659,11 +656,6 @@ class EngineArgs:
656
)
657
model_group.add_argument("--hf-overrides", **model_kwargs["hf_overrides"])
658
model_group.add_argument("--pooler-config", **model_kwargs["pooler_config"])
662
- model_group.add_argument(
663
- "--override-pooler-config",
664
- **model_kwargs["override_pooler_config"],
665
- deprecated=True,
666
- )
659
model_group.add_argument(
660
"--logits-processor-pattern", **model_kwargs["logits_processor_pattern"]
661
)
@@ -1243,7 +1235,6 @@ class EngineArgs:
1235
mm_encoder_tp_mode=self.mm_encoder_tp_mode,
1236
mm_encoder_attn_backend=self.mm_encoder_attn_backend,
1237
pooler_config=self.pooler_config,
1246
- override_pooler_config=self.override_pooler_config,
1238
logits_processor_pattern=self.logits_processor_pattern,
1239
generation_config=self.generation_config,
1240
override_generation_config=self.override_generation_config,
@@ -2047,24 +2038,6 @@ class AsyncEngineArgs(EngineArgs):
2038
2039
enable_log_requests: bool = False
2040
2050
- @property
2051
- @deprecated(
2052
- "`disable_log_requests` is deprecated and has been replaced with "
2053
- "`enable_log_requests`. This will be removed in v0.12.0. Please use "
2054
- "`enable_log_requests` instead."
2055
- )
2056
- def disable_log_requests(self) -> bool:
2057
- return not self.enable_log_requests
2058
-
2059
- @disable_log_requests.setter
2060
- @deprecated(
2061
- "`disable_log_requests` is deprecated and has been replaced with "
2062
- "`enable_log_requests`. This will be removed in v0.12.0. Please use "
2063
- "`enable_log_requests` instead."
2064
- )
2065
- def disable_log_requests(self, value: bool):
2066
- self.enable_log_requests = not value
2067
-
2041
@staticmethod
2042
def add_cli_args(
2043
parser: FlexibleArgumentParser, async_args_only: bool = False
vllm/entrypoints/llm.py
-5
@@ -174,9 +174,6 @@ class LLM:
174
For example, for Phi-3-Vision: `{"num_crops": 4}`.
175
pooler_config: Initialize non-default pooling config for the pooling
176
model. e.g. `PoolerConfig(pooling_type="mean", normalize=False)`.
177
- override_pooler_config: [DEPRECATED] Use `pooler_config` instead. This
178
- argument is deprecated and will be removed in v0.12.0 or v1.0.0,
179
- whichever is sooner.
177
compilation_config: Either an integer or a dictionary. If it is an
178
integer, it is used as the mode of compilation optimization. If it
179
is a dictionary, it can specify the full compilation configuration.
@@ -214,7 +211,6 @@ class LLM:
211
hf_overrides: HfOverrides | None = None,
212
mm_processor_kwargs: dict[str, Any] | None = None,
213
pooler_config: PoolerConfig | None = None,
217
- override_pooler_config: PoolerConfig | None = None,
214
structured_outputs_config: dict[str, Any]
215
| StructuredOutputsConfig
216
| None = None,
@@ -330,7 +326,6 @@ class LLM:
326
hf_overrides=hf_overrides,
327
mm_processor_kwargs=mm_processor_kwargs,
328
pooler_config=pooler_config,
333
- override_pooler_config=override_pooler_config,
329
structured_outputs_config=structured_outputs_instance,
330
compilation_config=compilation_config_instance,
331
logits_processors=logits_processors,
vllm/utils/argparse_utils.py
-8
@@ -73,14 +73,6 @@ class FlexibleArgumentParser(ArgumentParser):
73
# Enable the deprecated kwarg for Python 3.12 and below
74
75
def parse_known_args(self, args=None, namespace=None):
76
- if args is not None and "--disable-log-requests" in args:
77
- # Special case warning because the warning below won't trigger
78
- # if –-disable-log-requests because its value is default.
79
- logger.warning_once(
80
- "argument '--disable-log-requests' is deprecated and "
81
- "replaced with '--enable-log-requests'. This will be "
82
- "removed in v0.12.0."
83
- )
76
namespace, args = super().parse_known_args(args, namespace)
77
for action in FlexibleArgumentParser._deprecated:
78
if (
vllm/v1/engine/async_llm.py
-8
@@ -31,7 +31,6 @@ from vllm.transformers_utils.tokenizer import AnyTokenizer, init_tokenizer_from_
31
from vllm.usage.usage_lib import UsageContext
32
from vllm.utils.async_utils import cancel_task_threadsafe
33
from vllm.utils.collection_utils import as_list
34
-from vllm.utils.func_utils import deprecate_kwargs
34
from vllm.utils.math_utils import cdiv
35
from vllm.v1.engine import EngineCoreRequest
36
from vllm.v1.engine.core_client import EngineCoreClient
@@ -195,12 +194,6 @@ class AsyncLLM(EngineClient):
194
self.profiler = None
195
196
@classmethod
198
- @deprecate_kwargs(
199
- "disable_log_requests",
200
- additional_message=(
201
- "This argument will have no effect. Use `enable_log_requests` instead."
202
- ),
203
- )
197
def from_vllm_config(
198
cls,
199
vllm_config: VllmConfig,
@@ -213,7 +206,6 @@ class AsyncLLM(EngineClient):
206
client_addresses: dict[str, str] | None = None,
207
client_count: int = 1,
208
client_index: int = 0,
216
- disable_log_requests: bool = True, # Deprecated, will be removed
209
) -> "AsyncLLM":
210
# Create the LLMEngine.
211
return cls(