apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 09:07:10 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] add tool parser (#3518)

Browse Source 
* [Feature] Pass through the `chat_template_kwargs` to the data processing module (#3421)

* fix chat_template_args

* fix args

* add offline

* add offline

* fix

* fix

* fix default enable_thinking value

* fix default enable_thinking value

* modify condition

* Revert "modify condition"

This reverts commit 26430bdeb1.

* fix unit test

* add Tool Parser (#3272)

* add tool-parser

* add tool-parser

* add tool parser

* add tool parser

* fix

* add offline

* add offline

* fix

* parsers:tool&reasoning

* 修改tool parser名称·

* update

* fix reasoning-parser

* add requirements

* fix finish reason

* fix

* fix reasoning-parser

* fix

* fix

* fix

* fix

* fix

---------

Co-authored-by: zhuzixuan <zhuzixuan@baidu.com>

* [Feature] add tool parser (#3483)

* 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

* fix unit test

---------

Co-authored-by: zhuzixuan <zhuzixuan@baidu.com>
This commit is contained in:
luukunn
2025-08-22 11:14:35 +08:00
committed by GitHub
parent d97aab25bc
commit 4a9c04a746
31 changed files with 1289 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
fastdeploy/input/text_processor.py
View File

@@ -148,7 +148,7 @@ class BaseDataProcessor(ABC):
class DataProcessor(BaseDataProcessor):
def __init__(self, model_name_or_path, reasoning_parser_obj=None):
def __init__(self, model_name_or_path, reasoning_parser_obj=None, tool_parser_obj=None):
"""
Initializes the DecodeStatus object.
@@ -168,6 +168,7 @@ class DataProcessor(BaseDataProcessor):
self._init_config()
self.decode_status = dict()
self.tool_parser_dict = dict()
self.tokenizer = self._load_tokenizer()
data_processor_logger.info(
f"tokenizer information: bos_token is {self.tokenizer.bos_token}, {self.tokenizer.bos_token_id}, \
@@ -180,6 +181,7 @@ class DataProcessor(BaseDataProcessor):
self.eos_token_id_len = len(self.eos_token_ids)
self.pad_token_id = self.get_pad_id()
self.reasoning_parser = None
self.tool_parser_obj = tool_parser_obj
if reasoning_parser_obj:
self.reasoning_parser = reasoning_parser_obj(self.tokenizer)
self.tokenizer.pad_token_id = self.pad_token_id
@@ -345,6 +347,12 @@ class DataProcessor(BaseDataProcessor):
else:
# 模型不支持思考,并且没单独设置enable_thinking为false
response_dict.outputs.text = full_text
if self.tool_parser_obj:
tool_parser = self.tool_parser_obj(self.tokenizer)
tool_call_info = tool_parser.extract_tool_calls(full_text, response_dict)
if tool_call_info.tools_called:
response_dict.outputs.tool_calls = tool_call_info.tool_calls
response_dict.outputs.text = tool_call_info.content
data_processor_logger.info(f"req_id:{req_id}, token)ids: {token_ids}")
return response_dict
@@ -376,6 +384,12 @@ class DataProcessor(BaseDataProcessor):
response_dict["outputs"]["reasoning_content"] = reasoning_content
else:
response_dict["outputs"]["text"] = full_text
if self.tool_parser_obj:
tool_parser = self.tool_parser_obj(self.tokenizer)
tool_call_info = tool_parser.extract_tool_calls(full_text, response_dict)
if tool_call_info.tools_called:
response_dict["outputs"]["tool_call"] = tool_call_info.tool_calls
response_dict["outputs"]["text"] = tool_call_info.content
data_processor_logger.info(f"req_id:{req_id}, decode_status: {self.decode_status[req_id]}")
del self.decode_status[req_id]
return response_dict
@@ -400,8 +414,10 @@ class DataProcessor(BaseDataProcessor):
token_ids = token_ids[:-1]
delta_text, previous_token_ids, previous_texts = self.ids2tokens(token_ids, req_id)
response_dict["outputs"]["raw_prediction"] = delta_text
if enable_thinking and self.reasoning_parser:
reasoning_content, text = self.reasoning_parser.extract_reasoning_content_streaming(
if self.reasoning_parser and (
enable_thinking or self.reasoning_parser.__class__.__name__ == "ErnieX1ReasoningParser"
):
reasoning_delta_message = self.reasoning_parser.extract_reasoning_content_streaming(
previous_texts,
previous_texts + delta_text,
delta_text,
@@ -409,13 +425,28 @@ class DataProcessor(BaseDataProcessor):
previous_token_ids + token_ids,
token_ids,
)
response_dict["outputs"]["text"] = text
response_dict["outputs"]["reasoning_content"] = reasoning_content
else:
response_dict["outputs"]["text"] = delta_text
response_dict["outputs"]["delta_message"] = reasoning_delta_message
if self.tool_parser_obj:
if req_id not in self.tool_parser_dict:
self.tool_parser_dict[req_id] = self.tool_parser_obj(self.tokenizer)
tool_parser = self.tool_parser_dict[req_id]
tool_call = tool_parser.extract_tool_calls_streaming(
previous_texts,
previous_texts + delta_text,
delta_text,
previous_token_ids,
previous_token_ids + token_ids,
token_ids,
response_dict,
)
if tool_call is None or tool_call.tool_calls:
response_dict["outputs"]["delta_message"] = tool_call
response_dict["outputs"]["text"] = delta_text
if is_end:
data_processor_logger.info(f"req_id:{req_id}, decode_status: {self.decode_status[req_id]}")
del self.decode_status[req_id]
if req_id in self.tool_parser_dict:
del self.tool_parser_dict[req_id]
return response_dict
def process_response_dict(self, response_dict, **kwargs):