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

Convert formatting to use ruff instead of yapf + isort (#26247)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-10-05 15:06:22 +01:00
committed by GitHub
parent 17edd8a807
commit d6953beb91
1508 changed files with 115244 additions and 94146 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
vllm/reasoning/step3_reasoning_parser.py
View File

@@ -7,8 +7,7 @@ from typing import Optional, Union
import regex as re
from transformers import PreTrainedTokenizerBase
from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
DeltaMessage)
from vllm.entrypoints.openai.protocol import ChatCompletionRequest, DeltaMessage
from vllm.logger import init_logger
from vllm.reasoning import ReasoningParser, ReasoningParserManager
@@ -20,7 +19,7 @@ class Step3ReasoningParser(ReasoningParser):
"""
Reasoning parser for Step3 model.
The Step3 model uses </think> token to denote the end of reasoning
The Step3 model uses </think> token to denote the end of reasoning
text. This parser extracts all content before </think> as reasoning content.
"""
@@ -28,19 +27,20 @@ class Step3ReasoningParser(ReasoningParser):
super().__init__(tokenizer, *args, **kwargs)
self.think_end_token = "</think>"
self.reasoning_regex = re.compile(rf"(.*?){self.think_end_token}",
re.DOTALL)
self.reasoning_regex = re.compile(rf"(.*?){self.think_end_token}", re.DOTALL)
if not self.model_tokenizer:
raise ValueError(
"The model tokenizer must be passed to the ReasoningParser "
"constructor during construction.")
"constructor during construction."
)
self.think_end_token_id = self.vocab.get(self.think_end_token)
if self.think_end_token_id is None:
raise RuntimeError(
"Step3 reasoning parser could not locate think end "
"token in the tokenizer!")
"token in the tokenizer!"
)
def extract_reasoning_content_streaming(
self,
@@ -60,17 +60,18 @@ class Step3ReasoningParser(ReasoningParser):
- 'xyz' goes to content
"""
# Skip single special token
if len(delta_token_ids
) == 1 and delta_token_ids[0] == self.think_end_token_id:
if len(delta_token_ids) == 1 and delta_token_ids[0] == self.think_end_token_id:
return None
if self.think_end_token_id in delta_token_ids:
# </think> in delta, extract reasoning content and remaining content
end_index = delta_text.find(self.think_end_token)
reasoning_content = delta_text[:end_index]
content = delta_text[end_index + len(self.think_end_token):]
return DeltaMessage(reasoning_content=reasoning_content,
content=content if content else None)
content = delta_text[end_index + len(self.think_end_token) :]
return DeltaMessage(
reasoning_content=reasoning_content,
content=content if content else None,
)
elif self.think_end_token_id in previous_token_ids:
# </think> already seen in previous text, everything is content
return DeltaMessage(content=delta_text)
@@ -79,9 +80,8 @@ class Step3ReasoningParser(ReasoningParser):
return DeltaMessage(reasoning_content=delta_text)
def extract_reasoning_content(
self, model_output: str, request: ChatCompletionRequest
self, model_output: str, request: ChatCompletionRequest
) -> tuple[Optional[str], Optional[str]]:
# Check if the model output contains the </think> token
if self.think_end_token not in model_output:
# If no </think> token, everything is reasoning content
@@ -92,7 +92,7 @@ class Step3ReasoningParser(ReasoningParser):
reasoning_content = model_output[:end_index]
# Content after </think> token
content = model_output[end_index + len(self.think_end_token):]
content = model_output[end_index + len(self.think_end_token) :]
if len(content) == 0:
content = None
@@ -106,4 +106,4 @@ class Step3ReasoningParser(ReasoningParser):
if self.think_end_token_id not in input_ids[:-1]:
return []
else:
return input_ids[input_ids.index(self.think_end_token_id) + 1:]
return input_ids[input_ids.index(self.think_end_token_id) + 1 :]