biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Scheduled removal of guided_* config fields (#29326)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-11-25 05:24:05 +00:00
committed by GitHub
parent 2d9ee28cab
commit 316c8492bf
5 changed files with 43 additions and 262 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
vllm/sampling_params.py
View File

@@ -3,7 +3,6 @@
"""Sampling parameters for text generation."""
import copy
import warnings
from dataclasses import field
from enum import Enum, IntEnum
from functools import cached_property
@@ -100,19 +99,6 @@ class StructuredOutputsParams:
)
@dataclass
class GuidedDecodingParams(StructuredOutputsParams):
def __post_init__(self):
warnings.warn(
"GuidedDecodingParams is deprecated. This will be removed in "
"v0.12.0 or v1.0.0, which ever is soonest. Please use "
"StructuredOutputsParams instead.",
DeprecationWarning,
stacklevel=2,
)
return super().__post_init__()
class RequestOutputKind(Enum):
# Return entire output so far in every RequestOutput
CUMULATIVE = 0
@@ -234,8 +220,6 @@ class SamplingParams(
# Fields used to construct logits processors
structured_outputs: StructuredOutputsParams | None = None
"""Parameters for configuring structured outputs."""
guided_decoding: GuidedDecodingParams | None = None
"""Deprecated alias for structured_outputs."""
logit_bias: dict[int, float] | None = None
"""If provided, the engine will construct a logits processor that applies
these logit biases."""
@@ -283,7 +267,6 @@ class SamplingParams(
truncate_prompt_tokens: Annotated[int, msgspec.Meta(ge=-1)] | None = None,
output_kind: RequestOutputKind = RequestOutputKind.CUMULATIVE,
structured_outputs: StructuredOutputsParams | None = None,
guided_decoding: GuidedDecodingParams | None = None,
logit_bias: dict[int, float] | dict[str, float] | None = None,
allowed_token_ids: list[int] | None = None,
extra_args: dict[str, Any] | None = None,
@@ -295,16 +278,6 @@ class SamplingParams(
int(token): min(100.0, max(-100.0, bias))
for token, bias in logit_bias.items()
}
if guided_decoding is not None:
warnings.warn(
"guided_decoding is deprecated. This will be removed in "
"v0.12.0 or v1.0.0, which ever is soonest. Please use "
"structured_outputs instead.",
DeprecationWarning,
stacklevel=2,
)
structured_outputs = guided_decoding
guided_decoding = None
return SamplingParams(
n=1 if n is None else n,
@@ -387,17 +360,6 @@ class SamplingParams(
# eos_token_id is added to this by the engine
self._all_stop_token_ids.update(self.stop_token_ids)
if self.guided_decoding is not None:
warnings.warn(
"guided_decoding is deprecated. This will be removed in "
"v0.12.0 or v1.0.0, which ever is soonest. Please use "
"structured_outputs instead.",
DeprecationWarning,
stacklevel=2,
)
self.structured_outputs = self.guided_decoding
self.guided_decoding = None
if self.skip_reading_prefix_cache is None:
# If prefix caching is enabled,
# the output of prompt logprobs may less than n_prompt_tokens,