mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2026-01-15 19:54:30 +08:00
68 lines
2.0 KiB
Python
68 lines
2.0 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
|
|
|
from collections.abc import Sequence
|
|
|
|
from vllm.entrypoints.openai.protocol import (
|
|
ChatCompletionRequest,
|
|
DeltaMessage,
|
|
ResponsesRequest,
|
|
)
|
|
from vllm.logger import init_logger
|
|
from vllm.reasoning.abs_reasoning_parsers import ReasoningParser
|
|
from vllm.reasoning.basic_parsers import BaseThinkingReasoningParser
|
|
from vllm.transformers_utils.tokenizer import AnyTokenizer
|
|
|
|
logger = init_logger(__name__)
|
|
|
|
|
|
class MiniMaxM2ReasoningParser(BaseThinkingReasoningParser):
|
|
"""
|
|
Reasoning parser for MiniMax M2 model.
|
|
"""
|
|
|
|
@property
|
|
def start_token(self) -> str:
|
|
"""The token that starts reasoning content."""
|
|
return "<think>"
|
|
|
|
@property
|
|
def end_token(self) -> str:
|
|
"""The token that ends reasoning content."""
|
|
return "</think>"
|
|
|
|
|
|
class MiniMaxM2AppendThinkReasoningParser(ReasoningParser):
|
|
"""
|
|
Reasoning parser for MiniMax M2 model.
|
|
"""
|
|
|
|
def __init__(self, tokenizer: AnyTokenizer, *args, **kwargs):
|
|
super().__init__(tokenizer, *args, **kwargs)
|
|
self.end_token_id = self.vocab.get("</think>")
|
|
|
|
def is_reasoning_end(self, input_ids: list[int]) -> bool:
|
|
end_token_id = self.end_token_id
|
|
return any(input_id == end_token_id for input_id in reversed(input_ids))
|
|
|
|
def extract_content_ids(self, input_ids: list[int]) -> list[int]:
|
|
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:
|
|
if len(previous_token_ids) == 0:
|
|
delta_text = "<think>" + delta_text
|
|
return DeltaMessage(content=delta_text)
|
|
|
|
def extract_reasoning(
|
|
self, model_output: str, request: ChatCompletionRequest | ResponsesRequest
|
|
) -> tuple[str | None, str | None]:
|
|
return None, "<think>" + model_output
|