4.7.1-alpha2 (#1153)

Co-authored-by: UUUUnotfound <31206589+UUUUnotfound@users.noreply.github.com> Co-authored-by: Hexiao Zhang <731931282qq@gmail.com> Co-authored-by: heheer <71265218+newfish-cmyk@users.noreply.github.com>
2025-07-24 05:23:57 +00:00 · 2024-04-08 21:17:33 +08:00
parent 3b0b2d68cc
commit 1fbc407ecf
84 changed files with 1773 additions and 715 deletions
--- a/python/bge-rerank/README.md
+++ b/python/bge-rerank/README.md
@@ -0,0 +1,114 @@
+# 接入 bge-rerank 重排模型
+
+## 不同模型推荐配置
+
+推荐配置如下：
+
+| 模型名           | 内存  | 显存  | 硬盘空间 | 启动命令      |
+| ---------------- | ----- | ----- | -------- | ------------- |
+| bge-rerank-base  | >=4GB | >=4GB | >=8GB    | python app.py |
+| bge-rerank-large | >=8GB | >=8GB | >=8GB    | python app.py |
+| bge-rerank-v2-m3 | >=8GB | >=8GB | >=8GB    | python app.py |
+
+## 源码部署
+
+### 1. 安装环境
+
+- Python 3.9, 3.10
+- CUDA 11.7
+- 科学上网环境
+
+### 2. 下载代码
+
+3 个模型代码分别为：
+
+1. [https://github.com/labring/FastGPT/tree/main/python/reranker/bge-reranker-base](https://github.com/labring/FastGPT/tree/main/python/reranker/bge-reranker-base)
+2. [https://github.com/labring/FastGPT/tree/main/python/reranker/bge-reranker-large](https://github.com/labring/FastGPT/tree/main/python/reranker/bge-reranker-large)
+3. [https://github.com/labring/FastGPT/tree/main/python/reranker/bge-rerank-v2-m3](https://github.com/labring/FastGPT/tree/main/python/reranker/bge-rerank-v2-m3)
+
+### 3. 安装依赖
+
+```sh
+pip install -r requirements.txt
+```
+
+### 4. 下载模型
+
+3个模型的 huggingface 仓库地址如下：
+
+1. [https://huggingface.co/BAAI/bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base)
+2. [https://huggingface.co/BAAI/bge-reranker-large](https://huggingface.co/BAAI/bge-reranker-large)
+3. [https://huggingface.co/BAAI/bge-rerank-v2-m3](https://huggingface.co/BAAI/bge-rerank-v2-m3)
+
+在对应代码目录下 clone 模型。目录结构：
+
+```
+bge-reranker-base/
+app.py
+Dockerfile
+requirements.txt
+```
+
+### 5. 运行代码
+
+```bash
+python app.py
+```
+
+启动成功后应该会显示如下地址：
+
+![](./rerank1.png)
+
+> 这里的 `http://0.0.0.0:6006` 就是请求地址。
+
+## docker 部署
+
+**镜像名分别为:**
+
+1. registry.cn-hangzhou.aliyuncs.com/fastgpt/bge-rerank-base:v0.1
+2. registry.cn-hangzhou.aliyuncs.com/fastgpt/bge-rerank-large:v0.1
+3. registry.cn-hangzhou.aliyuncs.com/fastgpt/bge-rerank-v2-m3:v0.1
+
+**端口**
+
+6006
+
+**环境变量**
+
+```
+ACCESS_TOKEN=访问安全凭证，请求时，Authorization: Bearer ${ACCESS_TOKEN}
+```
+
+**运行命令示例**
+
+```sh
+# auth token 为mytoken
+docker run -d --name reranker -p 6006:6006 -e ACCESS_TOKEN=mytoken --gpus all registry.cn-hangzhou.aliyuncs.com/fastgpt/bge-rerank-base:v0.1
+```
+
+**docker-compose.yml示例**
+
+```
+version: "3"
+services:
+  reranker:
+    image: registry.cn-hangzhou.aliyuncs.com/fastgpt/rerank:v0.2
+    container_name: reranker
+    # GPU运行环境，如果宿主机未安装，将deploy配置隐藏即可
+    deploy:
+      resources:
+        reservations:
+          devices:
+          - driver: nvidia
+            count: all
+            capabilities: [gpu]
+    ports:
+      - 6006:6006
+    environment:
+      - ACCESS_TOKEN=mytoken
+
+```
+
+## 接入 FastGPT
+
+参考 [ReRank模型接入](https://doc/fastai.site/docs/development/configuration/#rerank-接入)
--- a/python/bge-rerank/bge-reranker-base/Dockerfile
+++ b/python/bge-rerank/bge-reranker-base/Dockerfile
--- a/python/bge-rerank/bge-reranker-base/app.py
+++ b/python/bge-rerank/bge-reranker-base/app.py
@@ -17,20 +17,9 @@ from FlagEmbedding import FlagReranker
 from pydantic import Field, BaseModel, validator
 from typing import Optional, List

-def response(code, msg, data=None):
-    time = str(datetime.datetime.now())
-    if data is None:
-        data = []
-    result = {
-        "code": code,
-        "message": msg,
-        "data": data,
-        "time": time
-    }
-    return result
-
-def success(data=None, msg=''):
-    return
+app = FastAPI()
+security = HTTPBearer()
+env_bearer_token = 'ACCESS_TOKEN'

 class QADocs(BaseModel):
    query: Optional[str]
@@ -46,42 +35,35 @@ class Singleton(type):

 RERANK_MODEL_PATH = os.path.join(os.path.dirname(__file__), "bge-reranker-base")

-class Reranker(metaclass=Singleton):
+class ReRanker(metaclass=Singleton):
    def __init__(self, model_path):
-        self.reranker = FlagReranker(model_path,
-                                     use_fp16=False)
+        self.reranker = FlagReranker(model_path, use_fp16=False)

    def compute_score(self, pairs: List[List[str]]):
        if len(pairs) > 0:
-            result = self.reranker.compute_score(pairs)
+            result = self.reranker.compute_score(pairs, normalize=True)
            if isinstance(result, float):
                result = [result]
            return result
        else:
            return None

-
 class Chat(object):
    def __init__(self, rerank_model_path: str = RERANK_MODEL_PATH):
-        self.reranker = Reranker(rerank_model_path)
+        self.reranker = ReRanker(rerank_model_path)

    def fit_query_answer_rerank(self, query_docs: QADocs) -> List:
        if query_docs is None or len(query_docs.documents) == 0:
            return []
-        new_docs = []
-        pair = []
-        for answer in query_docs.documents:
-            pair.append([query_docs.query, answer])
-        scores = self.reranker.compute_score(pair)
-        for index, score in enumerate(scores):
-            new_docs.append({"index": index, "text": query_docs.documents[index], "score": 1 / (1 + np.exp(-score))})
-        #results = [{"document": {"text": documents["text"]}, "index": documents["index"], "relevance_score": documents["score"]} for documents in list(sorted(new_docs, key=lambda x: x["score"], reverse=True))]
-        results = [{"index": documents["index"], "relevance_score": documents["score"]} for documents in list(sorted(new_docs, key=lambda x: x["score"], reverse=True))]
-        return {"results": results}

-app = FastAPI()
-security = HTTPBearer()
-env_bearer_token = 'ACCESS_TOKEN'
+        pair = [[query_docs.query, doc] for doc in query_docs.documents]
+        scores = self.reranker.compute_score(pair)
+
+        new_docs = []
+        for index, score in enumerate(scores):
+            new_docs.append({"index": index, "text": query_docs.documents[index], "score": score})
+        results = [{"index": documents["index"], "relevance_score": documents["score"]} for documents in list(sorted(new_docs, key=lambda x: x["score"], reverse=True))]
+        return results

@app.post('/v1/rerank')
 async def handle_post_request(docs: QADocs, credentials: HTTPAuthorizationCredentials = Security(security)):
@@ -89,8 +71,12 @@ async def handle_post_request(docs: QADocs, credentials: HTTPAuthorizationCreden
    if env_bearer_token is not None and token != env_bearer_token:
        raise HTTPException(status_code=401, detail="Invalid token")
    chat = Chat()
-    qa_docs_with_rerank = chat.fit_query_answer_rerank(docs)
-    return response(200, msg="重排成功", data=qa_docs_with_rerank)
+    try:
+        results = chat.fit_query_answer_rerank(docs)
+        return {"results": results}
+    except Exception as e:
+        print(f"报错：\n{e}")
+        return {"error": "重排出错"}

 if __name__ == "__main__":
    token = os.getenv("ACCESS_TOKEN")
--- a/python/bge-rerank/bge-reranker-base/requirements.txt
+++ b/python/bge-rerank/bge-reranker-base/requirements.txt
@@ -1,6 +1,6 @@
 fastapi==0.104.1
 transformers[sentencepiece]
-FlagEmbedding==1.1.5
+FlagEmbedding==1.2.8
 pydantic==1.10.13
 uvicorn==0.17.6
 itsdangerous
--- a/python/bge-rerank/bge-reranker-large/Dockerfile
+++ b/python/bge-rerank/bge-reranker-large/Dockerfile
@@ -0,0 +1,12 @@
+FROM pytorch/pytorch:2.0.1-cuda11.7-cudnn8-runtime
+
+# please download the model from https://huggingface.co/BAAI/bge-reranker-large and put it in the same directory as Dockerfile
+COPY ./bge-reranker-large ./bge-reranker-large
+
+COPY requirements.txt .
+
+RUN python3 -m pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+COPY app.py Dockerfile .
+
+ENTRYPOINT python3 app.py
--- a/python/bge-rerank/bge-reranker-large/app.py
+++ b/python/bge-rerank/bge-reranker-large/app.py
@@ -0,0 +1,88 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+@Time: 2023/11/7 22:45
+@Author: zhidong
+@File: reranker.py
+@Desc:
+"""
+import os
+import numpy as np
+import logging
+import uvicorn
+import datetime
+from fastapi import FastAPI, Security, HTTPException
+from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
+from FlagEmbedding import FlagReranker
+from pydantic import Field, BaseModel, validator
+from typing import Optional, List
+
+app = FastAPI()
+security = HTTPBearer()
+env_bearer_token = 'ACCESS_TOKEN'
+
+class QADocs(BaseModel):
+    query: Optional[str]
+    documents: Optional[List[str]]
+
+
+class Singleton(type):
+    def __call__(cls, *args, **kwargs):
+        if not hasattr(cls, '_instance'):
+            cls._instance = super().__call__(*args, **kwargs)
+        return cls._instance
+
+
+RERANK_MODEL_PATH = os.path.join(os.path.dirname(__file__), "bge-reranker-large")
+
+class ReRanker(metaclass=Singleton):
+    def __init__(self, model_path):
+        self.reranker = FlagReranker(model_path, use_fp16=False)
+
+    def compute_score(self, pairs: List[List[str]]):
+        if len(pairs) > 0:
+            result = self.reranker.compute_score(pairs, normalize=True)
+            if isinstance(result, float):
+                result = [result]
+            return result
+        else:
+            return None
+
+class Chat(object):
+    def __init__(self, rerank_model_path: str = RERANK_MODEL_PATH):
+        self.reranker = ReRanker(rerank_model_path)
+
+    def fit_query_answer_rerank(self, query_docs: QADocs) -> List:
+        if query_docs is None or len(query_docs.documents) == 0:
+            return []
+
+        pair = [[query_docs.query, doc] for doc in query_docs.documents]
+        scores = self.reranker.compute_score(pair)
+
+        new_docs = []
+        for index, score in enumerate(scores):
+            new_docs.append({"index": index, "text": query_docs.documents[index], "score": score})
+        results = [{"index": documents["index"], "relevance_score": documents["score"]} for documents in list(sorted(new_docs, key=lambda x: x["score"], reverse=True))]
+        return results
+
+@app.post('/v1/rerank')
+async def handle_post_request(docs: QADocs, credentials: HTTPAuthorizationCredentials = Security(security)):
+    token = credentials.credentials
+    if env_bearer_token is not None and token != env_bearer_token:
+        raise HTTPException(status_code=401, detail="Invalid token")
+    chat = Chat()
+    try:
+        results = chat.fit_query_answer_rerank(docs)
+        return {"results": results}
+    except Exception as e:
+        print(f"报错：\n{e}")
+        return {"error": "重排出错"}
+
+if __name__ == "__main__":
+    token = os.getenv("ACCESS_TOKEN")
+    if token is not None:
+        env_bearer_token = token
+    try:
+        uvicorn.run(app, host='0.0.0.0', port=6006)
+    except Exception as e:
+        print(f"API启动失败！\n报错：\n{e}")
--- a/python/bge-rerank/bge-reranker-large/requirements.txt
+++ b/python/bge-rerank/bge-reranker-large/requirements.txt
@@ -0,0 +1,7 @@
+fastapi==0.104.1
+transformers[sentencepiece]
+FlagEmbedding==1.2.8
+pydantic==1.10.13
+uvicorn==0.17.6
+itsdangerous
+protobuf
--- a/python/bge-rerank/bge-reranker-v2-m3/Dockerfile
+++ b/python/bge-rerank/bge-reranker-v2-m3/Dockerfile
@@ -0,0 +1,12 @@
+FROM pytorch/pytorch:2.0.1-cuda11.7-cudnn8-runtime
+
+# please download the model from https://huggingface.co/BAAI/bge-reranker-v2-m3 and put it in the same directory as Dockerfile
+COPY ./bge-reranker-v2-m3 ./bge-reranker-v2-m3
+
+COPY requirements.txt .
+
+RUN python3 -m pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
+
+COPY app.py Dockerfile .
+
+ENTRYPOINT python3 app.py
--- a/python/bge-rerank/bge-reranker-v2-m3/app.py
+++ b/python/bge-rerank/bge-reranker-v2-m3/app.py
@@ -0,0 +1,88 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+@Time: 2023/11/7 22:45
+@Author: zhidong
+@File: reranker.py
+@Desc:
+"""
+import os
+import numpy as np
+import logging
+import uvicorn
+import datetime
+from fastapi import FastAPI, Security, HTTPException
+from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
+from FlagEmbedding import FlagReranker
+from pydantic import Field, BaseModel, validator
+from typing import Optional, List
+
+app = FastAPI()
+security = HTTPBearer()
+env_bearer_token = 'ACCESS_TOKEN'
+
+class QADocs(BaseModel):
+    query: Optional[str]
+    documents: Optional[List[str]]
+
+
+class Singleton(type):
+    def __call__(cls, *args, **kwargs):
+        if not hasattr(cls, '_instance'):
+            cls._instance = super().__call__(*args, **kwargs)
+        return cls._instance
+
+
+RERANK_MODEL_PATH = os.path.join(os.path.dirname(__file__), "bge-reranker-v2-m3")
+
+class ReRanker(metaclass=Singleton):
+    def __init__(self, model_path):
+        self.reranker = FlagReranker(model_path, use_fp16=False)
+
+    def compute_score(self, pairs: List[List[str]]):
+        if len(pairs) > 0:
+            result = self.reranker.compute_score(pairs, normalize=True)
+            if isinstance(result, float):
+                result = [result]
+            return result
+        else:
+            return None
+
+class Chat(object):
+    def __init__(self, rerank_model_path: str = RERANK_MODEL_PATH):
+        self.reranker = ReRanker(rerank_model_path)
+
+    def fit_query_answer_rerank(self, query_docs: QADocs) -> List:
+        if query_docs is None or len(query_docs.documents) == 0:
+            return []
+
+        pair = [[query_docs.query, doc] for doc in query_docs.documents]
+        scores = self.reranker.compute_score(pair)
+
+        new_docs = []
+        for index, score in enumerate(scores):
+            new_docs.append({"index": index, "text": query_docs.documents[index], "score": score})
+        results = [{"index": documents["index"], "relevance_score": documents["score"]} for documents in list(sorted(new_docs, key=lambda x: x["score"], reverse=True))]
+        return results
+
+@app.post('/v1/rerank')
+async def handle_post_request(docs: QADocs, credentials: HTTPAuthorizationCredentials = Security(security)):
+    token = credentials.credentials
+    if env_bearer_token is not None and token != env_bearer_token:
+        raise HTTPException(status_code=401, detail="Invalid token")
+    chat = Chat()
+    try:
+        results = chat.fit_query_answer_rerank(docs)
+        return {"results": results}
+    except Exception as e:
+        print(f"报错：\n{e}")
+        return {"error": "重排出错"}
+
+if __name__ == "__main__":
+    token = os.getenv("ACCESS_TOKEN")
+    if token is not None:
+        env_bearer_token = token
+    try:
+        uvicorn.run(app, host='0.0.0.0', port=6006)
+    except Exception as e:
+        print(f"API启动失败！\n报错：\n{e}")
--- a/python/bge-rerank/bge-reranker-v2-m3/requirements.txt
+++ b/python/bge-rerank/bge-reranker-v2-m3/requirements.txt
@@ -0,0 +1,7 @@
+fastapi==0.104.1
+transformers[sentencepiece]
+FlagEmbedding==1.2.8
+pydantic==1.10.13
+uvicorn==0.17.6
+itsdangerous
+protobuf
--- a/python/bge-rerank/rerank1.png
+++ b/python/bge-rerank/rerank1.png
--- a/python/reranker/bge-reranker-base/README.md
+++ b/python/reranker/bge-reranker-base/README.md
@@ -1,48 +0,0 @@
-
-## 推荐配置
-
-推荐配置如下：
-
-{{< table "table-hover table-striped-columns" >}}
-| 类型 | 内存 | 显存 | 硬盘空间 | 启动命令 |
-|------|---------|---------|----------|--------------------------|
-| base | >=4GB | >=3GB | >=8GB | python app.py |
-{{< /table >}}
-
-## 部署
-
-### 环境要求
-
- Python 3.10.11
- CUDA 11.7
- 科学上网环境
-
-### 源码部署
-
-1. 根据上面的环境配置配置好环境，具体教程自行 GPT；
-2. 下载 [python 文件](app.py)
-3. 在命令行输入命令 `pip install -r requirments.txt`；
-4. 按照[https://huggingface.co/BAAI/bge-reranker-base](https://huggingface.co/BAAI/bge-reranker-base)下载模型仓库到app.py同级目录
-5. 添加环境变量 `export ACCESS_TOKEN=XXXXXX` 配置 token，这里的 token 只是加一层验证，防止接口被人盗用，默认值为 `ACCESS_TOKEN` ；
-6. 执行命令 `python app.py`。
-
-然后等待模型下载，直到模型加载完毕为止。如果出现报错先问 GPT。
-
-启动成功后应该会显示如下地址：
-
-![](/imgs/chatglm2.png)
-
-> 这里的 `http://0.0.0.0:6006` 就是连接地址。
-
-### docker 部署
-
-**镜像和端口**
-
-+ 镜像名: `registry.cn-hangzhou.aliyuncs.com/fastgpt/rerank:v0.2`  
-+ 端口号: 6006
-
-```
-# 设置安全凭证（即oneapi中的渠道密钥）
-通过环境变量ACCESS_TOKEN引入，默认值：ACCESS_TOKEN。
-有关docker环境变量引入的方法请自寻教程，此处不再赘述。
-```