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 5, 04:45 UTC
[Reasoning][Frontend] Add model config to adjust_request in reasoning parser (#37848)
Signed-off-by: rishitdholakia13 <rishit+github@cohere.com> Signed-off-by: rishitdholakia13 <123388671+rishitdholakia13@users.noreply.github.com> Signed-off-by: Aaron Pham <contact@aarnphm.xyz> Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> Co-authored-by: Aaron Pham <contact@aarnphm.xyz>
rishitdholakia13 committed
Apr 15, 2026 at 00:29 UTC
70ed01550c2a47d1888450935f30b8cf2f55d6bd
2 files changed
+7
-1
vllm/entrypoints/serve/render/serving.py
+3
-1
@@ -548,7 +548,9 @@ class OpenAIServingRender:
548
549
if reasoning_parser is not None:
550
tokenizer = renderer.get_tokenizer()
551
- request = reasoning_parser(tokenizer).adjust_request(request=request)
551
+ request = reasoning_parser(
552
+ tokenizer, model_config=self.model_config
553
+ ).adjust_request(request=request)
554
555
# tool parsing is done only if a tool_parser has been set and if
556
# tool_choice is not "none" (if tool_choice is "none" but a tool_parser
vllm/reasoning/abs_reasoning_parsers.py
+4
@@ -14,6 +14,7 @@ from vllm.utils.collection_utils import is_list_of
14
from vllm.utils.import_utils import import_from_path
15
16
if TYPE_CHECKING:
17
+ from vllm.config import ModelConfig
18
from vllm.entrypoints.openai.chat_completion.protocol import ChatCompletionRequest
19
from vllm.entrypoints.openai.engine.protocol import DeltaMessage
20
from vllm.entrypoints.openai.responses.protocol import ResponsesRequest
@@ -32,6 +33,9 @@ class ReasoningParser:
33
34
def __init__(self, tokenizer: "TokenizerLike", *args, **kwargs):
35
self.model_tokenizer = tokenizer
36
+ # Optional vLLM ModelConfig from the server. Use get (not pop) so composite
37
+ # parsers can forward **kwargs to nested parsers.
38
+ self._model_config: ModelConfig | None = kwargs.get("model_config")
39
40
@cached_property
41
def vocab(self) -> dict[str, int]: