feat(deepdata): 引入修改密码、文件预览组件与清洗算子增强，deepflow 改名为 deep-flow
本次同步 09-12 至 09-28 的代码变更，主要内容： 新增功能 - 修改密码：后端新增 ChangePasswordRequest DTO 与UserController/UserService 对应接口，前端新增 ChangePasswordDialog 并在 Header 接入入口 - 文件预览组件重构：file-preview 目录拆分出 VideoPreview（支持 blobUrl 直传）、 ImagePreview、AudioPreview，index.tsx 统一调度各类型预览器 - 新增 my_ops 算子模块：markitdown_parser、mineru_pdf_parser、text_quality_evaluator - 新增 tools/minio-file-server MinIO 文件服务 功能增强 - 数据清洗：cleaning_task_service/scheduler/routes 重构， data_juicer_executor 与 wrapper 增强，OperatorTable/LogsTable/TaskDetail 同步调整 - 数据标注：annotation sync.py、project.py 接口与schema 更新 - 数据采集、算子市场、首页与菜单路由相关页面调整 - DataJuicer 算子接入（useOperatorOperations、dj_ops_schema） 结构调整 - deepflow/ 更名为 deep-flow/（419 个文件重命名） - 新增 helm 部署编排、GitHub Actions 之外的自建镜像构建脚本 未纳入本次提交 - scripts-backup/、runtime/python-executor-backup/、components_backup/（备份目录） - *.backup*、__pycache__ 产物 - zh_core_web_sm-3.8.0-py3-none-any.whl（46MB 构建产物） - DataX/（内嵌独立仓库，远程亦未收录其源码）
This commit is contained in:
沃巴蒂
@@ -1,98 +1,99 @@
|
||||
# -- encoding: utf-8 --
|
||||
|
||||
"""
|
||||
Description: 基于LLM通过用户设置维度和相应描述进行QA对评估
|
||||
Create: 2023/11/7 9:26
|
||||
"""
|
||||
import json
|
||||
import re
|
||||
import time
|
||||
from pathlib import Path
|
||||
from typing import List, Dict, Any
|
||||
|
||||
from loguru import logger
|
||||
|
||||
|
||||
from deepdata.core.base_op import LLM
|
||||
|
||||
|
||||
class QAConditionEvaluator(LLM):
|
||||
def __init__(self, *args, **kwargs):
|
||||
super(QAConditionEvaluator, self).__init__(*args, **kwargs)
|
||||
self.pattern = r'结果[:：] ?[YN]'
|
||||
self.template_path = Path(__file__).parent / "resources/template.txt"
|
||||
self.examples_path = Path(__file__).parent / "resources/examples.json"
|
||||
self.task_id = kwargs.get("taskId", "default_id")
|
||||
self.dimensions = kwargs.get("dimension", [
|
||||
{
|
||||
"dimension": "回答是否有针对性",
|
||||
"description": "回答应对问题中的所有疑问点提供正面、直接的回答，"
|
||||
"不应引起疑惑。同时，答案不应有任何内容的遗漏，需构成一个完整的陈述。"
|
||||
},
|
||||
{
|
||||
"dimension": "问题是否独立",
|
||||
"description": "仅分析问题，问题的主体和客体都比较明确，即使有省略，也符合语言习惯。"
|
||||
"在不需要补充其他信息的情况下不会引起疑惑。"
|
||||
},
|
||||
{
|
||||
"dimension": "语法是否错误",
|
||||
"description": "问题为疑问句，答案为陈述句; 不存在词语搭配不当的情况;连接词和标点符号不存在错用情况；"
|
||||
"逻辑混乱的情况不存在；语法结构都正确且完整;"
|
||||
}
|
||||
])
|
||||
|
||||
self.llm = self.get_llm(*args, **kwargs)
|
||||
self.prompts = self.build_llm_prompt(*args, **kwargs)
|
||||
|
||||
@staticmethod
|
||||
def _process_examples(dimension_example: List) -> str:
|
||||
if not dimension_example:
|
||||
return "\n"
|
||||
res = "\n以下是一些案例供你参考："
|
||||
for single_example in dimension_example:
|
||||
res += (f"\n问题：{single_example['question']}"
|
||||
f"\n回答：{single_example['answer']}"
|
||||
f"\n分析思路：{single_example['evaluate']}"
|
||||
f"\n结果：{single_example['result']}\n")
|
||||
return res
|
||||
|
||||
def execute(self, sample: Dict[str, Any]) -> Dict[str, Any]:
|
||||
start = time.time()
|
||||
qas = json.loads(sample[self.text_key])
|
||||
single_content_res = []
|
||||
for qa in qas:
|
||||
single_qa_res = []
|
||||
for dimension, prompt in self.prompts.items():
|
||||
local_result = self._llm_call_parse(qa, prompt, retry=2)
|
||||
single_qa_res.append({"dimension": dimension, "result": local_result})
|
||||
qa_response = {"qaId": qa["qaId"], "result": single_qa_res}
|
||||
single_content_res.append(qa_response)
|
||||
|
||||
sample[self.text_key] = "Sucess"
|
||||
self.save_sample(single_content_res, sample)
|
||||
cost_time = time.time() - start
|
||||
logger.info(f"task id: {self.task_id}, method: QAConditionEvaluator costs {cost_time:.6f} s")
|
||||
return sample
|
||||
|
||||
def build_llm_prompt(self, *args, **kwargs) -> Dict:
|
||||
templates = self.template_path.read_text(encoding="utf-8")
|
||||
examples_dict = json.loads(self.examples_path.read_text(encoding="utf-8"))
|
||||
prompts_dict = {}
|
||||
for dimension in self.dimensions:
|
||||
name, des = dimension["dimension"], dimension["description"]
|
||||
dimension_example = self._process_examples(examples_dict.get(name))
|
||||
dimension_prompt = templates.format(criterion=des, examples=dimension_example, question="{question}",
|
||||
answer="{answer}")
|
||||
prompts_dict[name] = dimension_prompt
|
||||
return prompts_dict
|
||||
|
||||
def _llm_call_parse(self, data: Dict, prompt: str, retry: int = 2):
|
||||
try:
|
||||
for _ in range(retry):
|
||||
response = self.llm(prompt.format(question=data["question"], answer=data["answer"]))
|
||||
result = re.findall(self.pattern, response)
|
||||
if result:
|
||||
return "Y" in result[0]
|
||||
except RuntimeError as e:
|
||||
logger.error(f"method: QAConditionEvaluator execution error, cause by {e}")
|
||||
return False
|
||||
# -- encoding: utf-8 --
|
||||
|
||||
"""
|
||||
Description: 基于LLM通过用户设置维度和相应描述进行QA对评估
|
||||
Create: 2023/11/7 9:26
|
||||
"""
|
||||
import json
|
||||
import re
|
||||
import time
|
||||
from pathlib import Path
|
||||
from typing import List, Dict, Any
|
||||
|
||||
from loguru import logger
|
||||
|
||||
|
||||
from deepdata.core.base_op import LLM
|
||||
|
||||
|
||||
class QAConditionEvaluator(LLM):
|
||||
def __init__(self, *args, **kwargs):
|
||||
super(QAConditionEvaluator, self).__init__(*args, **kwargs)
|
||||
self.pattern = r'结果[:：] ?[YN]'
|
||||
self.template_path = Path(__file__).parent / "resources/template.txt"
|
||||
self.examples_path = Path(__file__).parent / "resources/examples.json"
|
||||
self.task_id = kwargs.get("taskId", "default_id")
|
||||
self.dimensions = kwargs.get("dimension", [
|
||||
{
|
||||
"dimension": "回答是否有针对性",
|
||||
"description": "回答应对问题中的所有疑问点提供正面、直接的回答，"
|
||||
"不应引起疑惑。同时，答案不应有任何内容的遗漏，需构成一个完整的陈述。"
|
||||
},
|
||||
{
|
||||
"dimension": "问题是否独立",
|
||||
"description": "仅分析问题，问题的主体和客体都比较明确，即使有省略，也符合语言习惯。"
|
||||
"在不需要补充其他信息的情况下不会引起疑惑。"
|
||||
},
|
||||
{
|
||||
"dimension": "语法是否错误",
|
||||
"description": "问题为疑问句，答案为陈述句; 不存在词语搭配不当的情况;连接词和标点符号不存在错用情况；"
|
||||
"逻辑混乱的情况不存在；语法结构都正确且完整;"
|
||||
}
|
||||
])
|
||||
|
||||
self.llm = self.get_llm(*args, **kwargs)
|
||||
self.prompts = self.build_llm_prompt(*args, **kwargs)
|
||||
|
||||
@staticmethod
|
||||
def _process_examples(dimension_example: List) -> str:
|
||||
if not dimension_example:
|
||||
return "\n"
|
||||
res = "\n以下是一些案例供你参考："
|
||||
for single_example in dimension_example:
|
||||
res += (f"\n问题：{single_example['question']}"
|
||||
f"\n回答：{single_example['answer']}"
|
||||
f"\n分析思路：{single_example['evaluate']}"
|
||||
f"\n结果：{single_example['result']}\n")
|
||||
return res
|
||||
|
||||
def execute(self, sample: Dict[str, Any]) -> Dict[str, Any]:
|
||||
start = time.time()
|
||||
qas = json.loads(sample[self.text_key])
|
||||
single_content_res = []
|
||||
for qa in qas:
|
||||
single_qa_res = []
|
||||
for dimension, prompt in self.prompts.items():
|
||||
local_result = self._llm_call_parse(qa, prompt, retry=2)
|
||||
single_qa_res.append({"dimension": dimension, "result": local_result})
|
||||
qa_response = {"qaId": qa["qaId"], "result": single_qa_res}
|
||||
single_content_res.append(qa_response)
|
||||
|
||||
# 评估结果已保存到文件，标记任务完成
|
||||
sample[self.text_key] = ""
|
||||
self.save_sample(single_content_res, sample)
|
||||
cost_time = time.time() - start
|
||||
logger.info(f"task id: {self.task_id}, method: QAConditionEvaluator costs {cost_time:.6f} s")
|
||||
return sample
|
||||
|
||||
def build_llm_prompt(self, *args, **kwargs) -> Dict:
|
||||
templates = self.template_path.read_text(encoding="utf-8")
|
||||
examples_dict = json.loads(self.examples_path.read_text(encoding="utf-8"))
|
||||
prompts_dict = {}
|
||||
for dimension in self.dimensions:
|
||||
name, des = dimension["dimension"], dimension["description"]
|
||||
dimension_example = self._process_examples(examples_dict.get(name))
|
||||
dimension_prompt = templates.format(criterion=des, examples=dimension_example, question="{question}",
|
||||
answer="{answer}")
|
||||
prompts_dict[name] = dimension_prompt
|
||||
return prompts_dict
|
||||
|
||||
def _llm_call_parse(self, data: Dict, prompt: str, retry: int = 2):
|
||||
try:
|
||||
for _ in range(retry):
|
||||
response = self.llm(prompt.format(question=data["question"], answer=data["answer"]))
|
||||
result = re.findall(self.pattern, response)
|
||||
if result:
|
||||
return "Y" in result[0]
|
||||
except RuntimeError as e:
|
||||
logger.error(f"method: QAConditionEvaluator execution error, cause by {e}")
|
||||
return False
|
||||
|
||||
@@ -1,113 +1,114 @@
|
||||
# -- encoding: utf-8 --
|
||||
|
||||
"""
|
||||
Description: 基于LLM通过用户设置维度和相应描述进行文本质量评估
|
||||
Create: 2025/3/14 11:00
|
||||
"""
|
||||
import re
|
||||
import time
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
from functools import partial
|
||||
from typing import Dict, Any
|
||||
|
||||
from loguru import logger
|
||||
|
||||
from deepdata.common.utils.text_splitter import TextSplitter
|
||||
from deepdata.core.base_op import LLM
|
||||
from .constant import EVAL_DIMENSION_MAP, BUSINESS_EVAL_DIMENSION_MAP
|
||||
from .prompt_config import TEXT_QUALITY_EVALUATE_TEMPLATE
|
||||
|
||||
CHUNK_SIZE = 4000
|
||||
CHUNK_OVERLAP = 0
|
||||
|
||||
|
||||
class TextQualityEvaluation(LLM):
|
||||
def __init__(self, *args, **kwargs):
|
||||
super(TextQualityEvaluation, self).__init__(*args, **kwargs)
|
||||
self.total_length = 0
|
||||
self.text_list = []
|
||||
self.total_scores = [0, 0, 0, 0, 0, 0]
|
||||
self.text_splitter = TextSplitter(1024 * 1024, CHUNK_SIZE, CHUNK_OVERLAP)
|
||||
self.pattern = r'\d+\.\d+'
|
||||
self.task_id = kwargs.get("taskId", "default_id")
|
||||
|
||||
self.llm = self.get_llm(*args, **kwargs)
|
||||
|
||||
def execute(self, sample: Dict[str, Any]) -> Dict[str, Any]:
|
||||
start = time.time()
|
||||
tmp_text_list = self.text_splitter.split_text(sample[self.text_key])
|
||||
logger.info(f"task id: {self.task_id}, the length of chunks: {len(tmp_text_list)}")
|
||||
self.text_list = tmp_text_list
|
||||
text_res = {}
|
||||
self._evaluate_concurrently_text(text_res)
|
||||
|
||||
sample[self.text_key] = "Success"
|
||||
self.save_sample([text_res], sample)
|
||||
cost_time = time.time() - start
|
||||
logger.info(f"task id: {self.task_id}, method: TextQualityEvaluation costs {cost_time:.6f} s")
|
||||
self.text_list = []
|
||||
return sample
|
||||
|
||||
def _evaluate_concurrently_text(self, text_res, max_workers: int = 5):
|
||||
for eval_dimension in EVAL_DIMENSION_MAP + BUSINESS_EVAL_DIMENSION_MAP:
|
||||
text_res[eval_dimension["score_name"]] = 0
|
||||
self.total_scores = [0, 0, 0, 0, 0, 0]
|
||||
self.total_length = 0
|
||||
with ThreadPoolExecutor(max_workers=max_workers) as executor:
|
||||
# 使用 partial 绑定多参数
|
||||
future_to_params = {
|
||||
executor.submit(
|
||||
partial(self.get_current_score_concurrently, text)): text
|
||||
for text in self.text_list
|
||||
}
|
||||
for future in as_completed(future_to_params):
|
||||
self.parse_execute_result(future, future_to_params)
|
||||
for _, eval_dimension in enumerate(EVAL_DIMENSION_MAP + BUSINESS_EVAL_DIMENSION_MAP):
|
||||
total_score = self.total_scores[_]
|
||||
text_res[eval_dimension["score_name"]] = 0
|
||||
if self.total_length > 0:
|
||||
text_res[eval_dimension["score_name"]] = total_score / self.total_length
|
||||
|
||||
def parse_execute_result(self, future, future_to_params):
|
||||
text = future_to_params[future]
|
||||
try:
|
||||
scores = future.result()
|
||||
if scores and len(scores) == len(self.total_scores):
|
||||
self.total_length += len(text)
|
||||
for _, score in enumerate(scores):
|
||||
self.total_scores[_] = self.total_scores[_] + score * len(text)
|
||||
except Exception as e:
|
||||
logger.error(f"Evaluate error, error details: {e}")
|
||||
|
||||
def get_current_score_concurrently(self, text, retry: int = 2):
|
||||
dimension_list = []
|
||||
for eval_dimension in EVAL_DIMENSION_MAP + BUSINESS_EVAL_DIMENSION_MAP:
|
||||
dimension = eval_dimension["dimension"] + ":" + eval_dimension["description"]
|
||||
dimension_list.append(dimension)
|
||||
prompt = TEXT_QUALITY_EVALUATE_TEMPLATE.format(context=text, dimension0=dimension_list[0],
|
||||
dimension1=dimension_list[1], dimension2=dimension_list[2],
|
||||
dimension3=dimension_list[3], dimension4=dimension_list[4],
|
||||
dimension5=dimension_list[5])
|
||||
retry_time = 0
|
||||
while True:
|
||||
try:
|
||||
return self.get_scores(prompt)
|
||||
except RuntimeError as e:
|
||||
if retry_time < retry:
|
||||
retry_time += 1
|
||||
else:
|
||||
logger.warning(f"Request LLM error, details: {e}")
|
||||
return []
|
||||
|
||||
def get_scores(self, prompt):
|
||||
response = self.llm(prompt)
|
||||
scores_str_list = response.split(",")
|
||||
scores = []
|
||||
for scores_str in scores_str_list:
|
||||
decimals = re.findall(self.pattern, scores_str)
|
||||
if decimals:
|
||||
score = float(decimals[-1])
|
||||
if 0 <= score <= 1:
|
||||
scores.append(score)
|
||||
logger.info(f"current evaluate scores: {scores}")
|
||||
return scores
|
||||
# -- encoding: utf-8 --
|
||||
|
||||
"""
|
||||
Description: 基于LLM通过用户设置维度和相应描述进行文本质量评估
|
||||
Create: 2025/3/14 11:00
|
||||
"""
|
||||
import re
|
||||
import time
|
||||
from concurrent.futures import ThreadPoolExecutor, as_completed
|
||||
from functools import partial
|
||||
from typing import Dict, Any
|
||||
|
||||
from loguru import logger
|
||||
|
||||
from deepdata.common.utils.text_splitter import TextSplitter
|
||||
from deepdata.core.base_op import LLM
|
||||
from .constant import EVAL_DIMENSION_MAP, BUSINESS_EVAL_DIMENSION_MAP
|
||||
from .prompt_config import TEXT_QUALITY_EVALUATE_TEMPLATE
|
||||
|
||||
CHUNK_SIZE = 4000
|
||||
CHUNK_OVERLAP = 0
|
||||
|
||||
|
||||
class TextQualityEvaluation(LLM):
|
||||
def __init__(self, *args, **kwargs):
|
||||
super(TextQualityEvaluation, self).__init__(*args, **kwargs)
|
||||
self.total_length = 0
|
||||
self.text_list = []
|
||||
self.total_scores = [0, 0, 0, 0, 0, 0]
|
||||
self.text_splitter = TextSplitter(1024 * 1024, CHUNK_SIZE, CHUNK_OVERLAP)
|
||||
self.pattern = r'\d+\.\d+'
|
||||
self.task_id = kwargs.get("taskId", "default_id")
|
||||
|
||||
self.llm = self.get_llm(*args, **kwargs)
|
||||
|
||||
def execute(self, sample: Dict[str, Any]) -> Dict[str, Any]:
|
||||
start = time.time()
|
||||
tmp_text_list = self.text_splitter.split_text(sample[self.text_key])
|
||||
logger.info(f"task id: {self.task_id}, the length of chunks: {len(tmp_text_list)}")
|
||||
self.text_list = tmp_text_list
|
||||
text_res = {}
|
||||
self._evaluate_concurrently_text(text_res)
|
||||
|
||||
# 评估结果已保存到文件，标记任务完成
|
||||
sample[self.text_key] = ""
|
||||
self.save_sample([text_res], sample)
|
||||
cost_time = time.time() - start
|
||||
logger.info(f"task id: {self.task_id}, method: TextQualityEvaluation costs {cost_time:.6f} s")
|
||||
self.text_list = []
|
||||
return sample
|
||||
|
||||
def _evaluate_concurrently_text(self, text_res, max_workers: int = 5):
|
||||
for eval_dimension in EVAL_DIMENSION_MAP + BUSINESS_EVAL_DIMENSION_MAP:
|
||||
text_res[eval_dimension["score_name"]] = 0
|
||||
self.total_scores = [0, 0, 0, 0, 0, 0]
|
||||
self.total_length = 0
|
||||
with ThreadPoolExecutor(max_workers=max_workers) as executor:
|
||||
# 使用 partial 绑定多参数
|
||||
future_to_params = {
|
||||
executor.submit(
|
||||
partial(self.get_current_score_concurrently, text)): text
|
||||
for text in self.text_list
|
||||
}
|
||||
for future in as_completed(future_to_params):
|
||||
self.parse_execute_result(future, future_to_params)
|
||||
for _, eval_dimension in enumerate(EVAL_DIMENSION_MAP + BUSINESS_EVAL_DIMENSION_MAP):
|
||||
total_score = self.total_scores[_]
|
||||
text_res[eval_dimension["score_name"]] = 0
|
||||
if self.total_length > 0:
|
||||
text_res[eval_dimension["score_name"]] = total_score / self.total_length
|
||||
|
||||
def parse_execute_result(self, future, future_to_params):
|
||||
text = future_to_params[future]
|
||||
try:
|
||||
scores = future.result()
|
||||
if scores and len(scores) == len(self.total_scores):
|
||||
self.total_length += len(text)
|
||||
for _, score in enumerate(scores):
|
||||
self.total_scores[_] = self.total_scores[_] + score * len(text)
|
||||
except Exception as e:
|
||||
logger.error(f"Evaluate error, error details: {e}")
|
||||
|
||||
def get_current_score_concurrently(self, text, retry: int = 2):
|
||||
dimension_list = []
|
||||
for eval_dimension in EVAL_DIMENSION_MAP + BUSINESS_EVAL_DIMENSION_MAP:
|
||||
dimension = eval_dimension["dimension"] + ":" + eval_dimension["description"]
|
||||
dimension_list.append(dimension)
|
||||
prompt = TEXT_QUALITY_EVALUATE_TEMPLATE.format(context=text, dimension0=dimension_list[0],
|
||||
dimension1=dimension_list[1], dimension2=dimension_list[2],
|
||||
dimension3=dimension_list[3], dimension4=dimension_list[4],
|
||||
dimension5=dimension_list[5])
|
||||
retry_time = 0
|
||||
while True:
|
||||
try:
|
||||
return self.get_scores(prompt)
|
||||
except RuntimeError as e:
|
||||
if retry_time < retry:
|
||||
retry_time += 1
|
||||
else:
|
||||
logger.warning(f"Request LLM error, details: {e}")
|
||||
return []
|
||||
|
||||
def get_scores(self, prompt):
|
||||
response = self.llm(prompt)
|
||||
scores_str_list = response.split(",")
|
||||
scores = []
|
||||
for scores_str in scores_str_list:
|
||||
decimals = re.findall(self.pattern, scores_str)
|
||||
if decimals:
|
||||
score = float(decimals[-1])
|
||||
if 0 <= score <= 1:
|
||||
scores.append(score)
|
||||
logger.info(f"current evaluate scores: {scores}")
|
||||
return scores
|
||||
|
||||
Reference in New Issue
Block a user