vllm/vllm/reasoning/identity_reasoning_parser.py
Harry Mellor d9ab1ad9d1
reasoning_content -> reasoning (#27752)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
2025-11-08 12:15:08 +00:00

59 lines
1.9 KiB
Python

# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from collections.abc import Sequence
from transformers import PreTrainedTokenizerBase
from vllm.entrypoints.openai.protocol import ChatCompletionRequest, DeltaMessage
from vllm.logger import init_logger
from vllm.reasoning import ReasoningParser
logger = init_logger(__name__)
class IdentityReasoningParser(ReasoningParser):
"""
Identity reasoning parser.
This parser does not attempt to parse or strip out reasoning tokens.
It treats the entire model output as content and ignores reasoning.
"""
def __init__(self, tokenizer: PreTrainedTokenizerBase, *args, **kwargs):
super().__init__(tokenizer, *args, **kwargs)
if not self.model_tokenizer:
raise ValueError(
"The model tokenizer must be passed to the ReasoningParser "
"constructor during construction."
)
def is_reasoning_end(self, input_ids: list[int]) -> bool:
# Always return True, since we never treat reasoning specially
return True
def extract_content_ids(self, input_ids: list[int]) -> list[int]:
# Identity: return all tokens as content
return input_ids
def extract_reasoning_streaming(
self,
previous_text: str,
current_text: str,
delta_text: str,
previous_token_ids: Sequence[int],
current_token_ids: Sequence[int],
delta_token_ids: Sequence[int],
) -> DeltaMessage | None:
# Just wrap delta_text as content, ignore reasoning
if delta_text:
return DeltaMessage(content=delta_text)
return None
def extract_reasoning(
self, model_output: str, request: ChatCompletionRequest
) -> tuple[str | None, str | None]:
# No reasoning separation: return None for reasoning,
# and full model_output as content
return None, model_output