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

Simplify (and fix) passing of guided decoding backend options (#17008)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-04-29 20:02:23 +01:00
committed by GitHub
parent 2fa2a50bf9
commit a6977dbd15
17 changed files with 309 additions and 217 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
vllm/engine/arg_utils.py
View File

@@ -18,9 +18,9 @@ from vllm import version
from vllm.config import (BlockSize, CacheConfig, CacheDType, CompilationConfig,
ConfigFormat, ConfigType, DecodingConfig, Device,
DeviceConfig, DistributedExecutorBackend,
GuidedDecodingBackendV1, HfOverrides,
KVTransferConfig, LoadConfig, LoadFormat, LoRAConfig,
ModelConfig, ModelImpl, MultiModalConfig,
GuidedDecodingBackend, GuidedDecodingBackendV1,
HfOverrides, KVTransferConfig, LoadConfig, LoadFormat,
LoRAConfig, ModelConfig, ModelImpl, MultiModalConfig,
ObservabilityConfig, ParallelConfig, PoolerConfig,
PrefixCachingHashAlgo, PromptAdapterConfig,
SchedulerConfig, SchedulerPolicy, SpeculativeConfig,
@@ -317,7 +317,12 @@ class EngineArgs:
bool] = SchedulerConfig.enable_chunked_prefill
disable_chunked_mm_input: bool = SchedulerConfig.disable_chunked_mm_input
guided_decoding_backend: str = DecodingConfig.guided_decoding_backend
guided_decoding_backend: GuidedDecodingBackend = DecodingConfig.backend
guided_decoding_disable_fallback: bool = DecodingConfig.disable_fallback
guided_decoding_disable_any_whitespace: bool = \
DecodingConfig.disable_any_whitespace
guided_decoding_disable_additional_properties: bool = \
DecodingConfig.disable_additional_properties
logits_processor_pattern: Optional[str] = None
speculative_config: Optional[Dict[str, Any]] = None
@@ -498,9 +503,17 @@ class EngineArgs:
title="DecodingConfig",
description=DecodingConfig.__doc__,
)
guided_decoding_group.add_argument("--guided-decoding-backend",
**guided_decoding_kwargs["backend"])
guided_decoding_group.add_argument(
'--guided-decoding-backend',
**guided_decoding_kwargs["guided_decoding_backend"])
"--guided-decoding-disable-fallback",
**guided_decoding_kwargs["disable_fallback"])
guided_decoding_group.add_argument(
"--guided-decoding-disable-any-whitespace",
**guided_decoding_kwargs["disable_any_whitespace"])
guided_decoding_group.add_argument(
"--guided-decoding-disable-additional-properties",
**guided_decoding_kwargs["disable_additional_properties"])
guided_decoding_group.add_argument(
"--reasoning-parser",
# This choices is a special case because it's not static
@@ -1244,7 +1257,11 @@ class EngineArgs:
if self.enable_prompt_adapter else None
decoding_config = DecodingConfig(
guided_decoding_backend=self.guided_decoding_backend,
backend=self.guided_decoding_backend,
disable_fallback=self.guided_decoding_disable_fallback,
disable_any_whitespace=self.guided_decoding_disable_any_whitespace,
disable_additional_properties=\
self.guided_decoding_disable_additional_properties,
reasoning_backend=self.reasoning_parser
if self.enable_reasoning else None,
)
@@ -1335,9 +1352,8 @@ class EngineArgs:
recommend_to_remove=True)
return False
# remove backend options when doing this check
if self.guided_decoding_backend.split(':')[0] \
not in get_args(GuidedDecodingBackendV1):
if self.guided_decoding_backend not in get_args(
GuidedDecodingBackendV1):
_raise_or_fallback(
feature_name=
f"--guided-decoding-backend={self.guided_decoding_backend}",