apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] add tool parser (#3483)

Browse Source 
* add tool parser

* add x1 enable_thinking

* restart ci

* fix vl reasoning parser

* modify call style

* modify call style

* add offline enablethinking

* fix completion

* fix

* fix unit test

* fix unit test

* fix unit test

* fix vl reasoning parser

* fix vl reasoning parser
This commit is contained in:
luukunn
2025-08-21 17:25:44 +08:00
committed by GitHub
parent 466cbb5a99
commit 371fb3f853
14 changed files with 197 additions and 222 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
fastdeploy/reasoning/qwen3_reasoning_parsers.py
View File

@@ -48,6 +48,9 @@ class Qwen3ReasoningParser(ReasoningParser):
if self.think_end_token_id is None:
raise RuntimeError("Qwen3 reasoning parser could not locate think end " "tokens in the tokenizer!")
def is_reasoning_end(self, input_ids: list[int]) -> bool:
return self.think_end_token_id in input_ids
def extract_reasoning_content_streaming(
self,
previous_text: str,
@@ -66,7 +69,7 @@ class Qwen3ReasoningParser(ReasoningParser):
- 'xyz' goes to content
"""
if len(delta_token_ids) == 1 and (delta_token_ids[0] in [self.think_start_token_id, self.think_end_token_id]):
return "", ""
return None
# </think> in delta
if self.think_end_token_id in delta_token_ids:
@@ -76,28 +79,28 @@ class Qwen3ReasoningParser(ReasoningParser):
end_index = delta_token_ids.find(self.think_end_token)
reasoning_content = delta_text[start_index + len(self.think_start_token) : end_index]
content = delta_text[end_index + len(self.think_end_token) :]
return reasoning_content, content
return DeltaMessage(reasoning_content=reasoning_content, content=content)
# <think> in previous, </think> in delta,
else:
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) :]
content = content if content else None
return reasoning_content, content
return DeltaMessage(reasoning_content=reasoning_content, content=content)
# </think> in previous reasoning content continues
elif self.think_end_token_id in previous_token_ids:
return "", delta_text
return DeltaMessage(content=delta_text)
# <think> in previous
elif self.think_start_token_id in previous_token_ids:
return delta_text, ""
return DeltaMessage(reasoning_content=delta_text)
# <think> in delta
elif self.think_start_token_id in delta_token_ids:
start_index = delta_text.find(self.think_start_token)
reasoning_content = delta_text[start_index + len(self.think_start_token) :]
content = ""
return reasoning_content, content
return DeltaMessage(reasoning_content=reasoning_content, content=content)
else:
return delta_text, ""
return DeltaMessage(reasoning_content=delta_text)
def extract_reasoning_content(
self, model_output: str, request: ChatCompletionRequest