分類目錄：《大模型從入門到應(yīng)用》總目錄

LangChain系列文章：

• 基礎(chǔ)知識(shí)
• 快速入門
  • 安裝與環(huán)境配置
  • 鏈（Chains）、代理（Agent:）和記憶（Memory）
  • 快速開發(fā)聊天模型
• 模型（Models）
  • 基礎(chǔ)知識(shí)
  • 大型語言模型（LLMs）
    • 基礎(chǔ)知識(shí)
    • LLM的異步API、自定義LLM包裝器、虛假LLM和人類輸入LLM（Human Input LLM）
    • 緩存LLM的調(diào)用結(jié)果
    • 加載與保存LLM類、流式傳輸LLM與Chat Model響應(yīng)和跟蹤tokens使用情況
  • 聊天模型（Chat Models）
    • 基礎(chǔ)知識(shí)
    • 使用少量示例和響應(yīng)流式傳輸
  • 文本嵌入模型
    • Aleph Alpha、Amazon Bedrock、Azure OpenAI、Cohere等
    • Embaas、Fake Embeddings、Google Vertex AI PaLM等
• 提示（Prompts）
  • 基礎(chǔ)知識(shí)
  • 提示模板
    • 基礎(chǔ)知識(shí)
    • 連接到特征存儲(chǔ)
    • 創(chuàng)建自定義提示模板和含有Few-Shot示例的提示模板
    • 部分填充的提示模板和提示合成
    • 序列化提示信息
  • 示例選擇器（Example Selectors）
  • 輸出解析器（Output Parsers）
• 記憶（Memory）
  • 基礎(chǔ)知識(shí)
  • 記憶的類型
    • 會(huì)話緩存記憶、會(huì)話緩存窗口記憶和實(shí)體記憶
    • 對(duì)話知識(shí)圖譜記憶、對(duì)話摘要記憶和會(huì)話摘要緩沖記憶
    • 對(duì)話令牌緩沖存儲(chǔ)器和基于向量存儲(chǔ)的記憶
  • 將記憶添加到LangChain組件中
  • 自定義對(duì)話記憶與自定義記憶類
  • 聊天消息記錄
  • 記憶的存儲(chǔ)與應(yīng)用
• 索引（Indexes）
  • 基礎(chǔ)知識(shí)
  • 文檔加載器（Document Loaders）
  • 文本分割器（Text Splitters）
  • 向量存儲(chǔ)器（Vectorstores）
  • 檢索器（Retrievers）
• 鏈（Chains）
  • 基礎(chǔ)知識(shí)
  • 通用功能
    • 自定義Chain和Chain的異步API
    • LLMChain和RouterChain
    • SequentialChain和TransformationChain
    • 鏈的保存（序列化）與加載（反序列化）
  • 鏈與索引
    • 文檔分析和基于文檔的聊天
    • 問答的基礎(chǔ)知識(shí)
    • 圖問答（Graph QA）和帶來源的問答（Q&A with Sources）
    • 檢索式問答
    • 文本摘要（Summarization）、HyDE和向量數(shù)據(jù)庫(kù)的文本生成
• 代理（Agents）
  • 基礎(chǔ)知識(shí)
  • 代理類型
  • 自定義代理（Custom Agent）
  • 自定義MRKL代理
  • 帶有ChatModel的LLM聊天自定義代理和自定義多操作代理（Custom MultiAction Agent）
  • 工具
    • 基礎(chǔ)知識(shí)
    • 自定義工具（Custom Tools）
    • 多輸入工具和工具輸入模式
    • 人工確認(rèn)工具驗(yàn)證和Tools作為OpenAI函數(shù)
  • 工具包（Toolkit）
  • 代理執(zhí)行器（Agent Executor）
    • 結(jié)合使用Agent和VectorStore
    • 使用Agents的異步API和創(chuàng)建ChatGPT克隆
    • 處理解析錯(cuò)誤、訪問中間步驟和限制最大迭代次數(shù)
    • 為代理程序設(shè)置超時(shí)時(shí)間和限制最大迭代次數(shù)和為代理程序和其工具添加共享內(nèi)存
  • 計(jì)劃與執(zhí)行
• 回調(diào)函數(shù)（Callbacks）

創(chuàng)建自定義提示模板

假設(shè)我們希望LLM根據(jù)函數(shù)名稱生成該函數(shù)的英文語言解釋。為了實(shí)現(xiàn)這個(gè)任務(wù)，我們將創(chuàng)建一個(gè)自定義的提示模板，以函數(shù)名稱作為輸入，并格式化提示模板以提供函數(shù)的源代碼。LangChain提供了一組默認(rèn)的提示模板，可用于生成各種任務(wù)的提示。但是，在某些情況下，默認(rèn)的提示模板可能無法滿足我們的需求。例如，我們可能希望創(chuàng)建一個(gè)具有特定動(dòng)態(tài)指令的提示模板，以適應(yīng)我們的語言模型。在這種情況下，我們可以創(chuàng)建自定義的提示模板。

有兩種不同的提示模板：

• 字符串提示模板：提供一個(gè)簡(jiǎn)單的字符串格式提示
• 聊天提示模板：生成一個(gè)更結(jié)構(gòu)化的聊天API使用的提示

在本文中，我們將使用字符串提示模板創(chuàng)建一個(gè)自定義提示。要?jiǎng)?chuàng)建自定義字符串提示模板，有兩個(gè)要求：

• 它具有input_variables屬性，用于公開提示模板期望的輸入變量
• 它公開一個(gè)format方法，該方法接受與預(yù)期的input_variables相對(duì)應(yīng)的關(guān)鍵字參數(shù)，并返回格式化的提示

我們將創(chuàng)建一個(gè)自定義的提示模板，它以函數(shù)名稱作為輸入，并格式化提示以提供函數(shù)的源代碼。為了實(shí)現(xiàn)這一點(diǎn)，讓我們首先創(chuàng)建一個(gè)函數(shù)，該函數(shù)將根據(jù)函數(shù)名稱返回函數(shù)的源代碼。

import inspect

def get_source_code(function_name):
    # Get the source code of the function
    return inspect.getsource(function_name)

接下來，我們將創(chuàng)建一個(gè)自定義的提示模板，該模板以函數(shù)名稱作為輸入，并格式化提示模板以提供函數(shù)的源代碼：

from langchain.prompts import StringPromptTemplate
from pydantic import BaseModel, validator

class FunctionExplainerPromptTemplate(StringPromptTemplate, BaseModel):
    """一個(gè)自定義的提示模板，接受函數(shù)名作為輸入，并格式化提示模板以提供函數(shù)的源代碼。"""

    @validator("input_variables")
    def validate_input_variables(cls, v):
        """驗(yàn)證輸入變量的正確性。"""
        if len(v) != 1 or "function_name" not in v:
            raise ValueError("function_name必須是唯一的輸入變量。")
        return v

    def format(self, **kwargs) -> str:
        # 獲取函數(shù)的源代碼
        source_code = get_source_code(kwargs["function_name"])

        # 生成要發(fā)送給語言模型的提示
        prompt = f"""
        給定函數(shù)名和源代碼，生成一個(gè)關(guān)于函數(shù)的英文語言解釋。
        函數(shù)名：{kwargs["function_name"].__name__}
        源代碼：
        {source_code}
        解釋：
        """
        return prompt
    
    def _prompt_type(self):
        return "function-explainer"

現(xiàn)在我們已經(jīng)創(chuàng)建了一個(gè)自定義的提示模板，我們可以使用它來生成我們?nèi)蝿?wù)的提示：

fn_explainer = FunctionExplainerPromptTemplate(input_variables=["function_name"])

# 為函數(shù)"get_source_code"生成一個(gè)提示
prompt = fn_explainer.format(function_name=get_source_code)
print(prompt)

輸出：

給定函數(shù)名和源代碼，生成一個(gè)關(guān)于函數(shù)的英文語言解釋。
函數(shù)名：get_source_code
源代碼：
def get_source_code(function_name):
    # Get the source code of the function
    return inspect.getsource(function_name)

解釋：

創(chuàng)建含有Few-Shot示例的提示模板

在下文中，我們將學(xué)習(xí)如何創(chuàng)建含有Few-Shot示例的提示模板。我們將使用FewShotPromptTemplate類來創(chuàng)建一個(gè)含有Few-Shot示例的提示模板。該類可以接受一組示例或者一個(gè)ExampleSelector對(duì)象。在下文中，我們將分別為自我提問與搜索配置Few-Shot示例討論這兩種選項(xiàng)。

使用示例集

首先，創(chuàng)建一個(gè)Few-Shot示例的列表。每個(gè)示例應(yīng)該是一個(gè)字典，其中鍵是輸入變量，值是這些輸入變量的值。

from langchain.prompts.few_shot import FewShotPromptTemplate
from langchain.prompts.prompt import PromptTemplate

examples = [
  {
    "question": "Who lived longer, Muhammad Ali or Alan Turing?",
    "answer": 
"""
Are follow up questions needed here: Yes.
Follow up: How old was Muhammad Ali when he died?
Intermediate answer: Muhammad Ali was 74 years old when he died.
Follow up: How old was Alan Turing when he died?
Intermediate answer: Alan Turing was 41 years old when he died.
So the final answer is: Muhammad Ali
"""
  },
  {
    "question": "When was the founder of craigslist born?",
    "answer": 
"""
Are follow up questions needed here: Yes.
Follow up: Who was the founder of craigslist?
Intermediate answer: Craigslist was founded by Craig Newmark.
Follow up: When was Craig Newmark born?
Intermediate answer: Craig Newmark was born on December 6, 1952.
So the final answer is: December 6, 1952
"""
  },
  {
    "question": "Who was the maternal grandfather of George Washington?",
    "answer":
"""
Are follow up questions needed here: Yes.
Follow up: Who was the mother of George Washington?
Intermediate answer: The mother of George Washington was Mary Ball Washington.
Follow up: Who was the father of Mary Ball Washington?
Intermediate answer: The father of Mary Ball Washington was Joseph Ball.
So the final answer is: Joseph Ball
"""
  },
  {
    "question": "Are both the directors of Jaws and Casino Royale from the same country?",
    "answer":
"""
Are follow up questions needed here: Yes.
Follow up: Who is the director of Jaws?
Intermediate Answer: The director of Jaws is Steven Spielberg.
Follow up: Where is Steven Spielberg from?
Intermediate Answer: The United States.
Follow up: Who is the director of Casino Royale?
Intermediate Answer: The director of Casino Royale is Martin Campbell.
Follow up: Where is Martin Campbell from?
Intermediate Answer: New Zealand.
So the final answer is: No
"""
  }
]

然后，我們可以為Few Shot示例創(chuàng)建格式化程序。配置一個(gè)將Few Shot示例格式化為字符串的格式化程序。該格式化程序應(yīng)該是一個(gè)PromptTemplate對(duì)象。

example_prompt = PromptTemplate(input_variables=["question", "answer"], template="Question: {question}\n{answer}")

print(example_prompt.format(**examples[0]))
Question: Who lived longer, Muhammad Ali or Alan Turing?

Are follow up questions needed here: Yes.
Follow up: How old was Muhammad Ali when he died?
Intermediate answer: Muhammad Ali was 74 years old when he died.
Follow up: How old was Alan Turing when he died?
Intermediate answer: Alan Turing was 41 years old when he died.
So the final answer is: Muhammad Ali

最后，創(chuàng)建一個(gè)FewShotPromptTemplate對(duì)象。該對(duì)象接受Few Shot示例和Few Shot示例的格式化程序作為輸入。

prompt = FewShotPromptTemplate(
    examples=examples, 
    example_prompt=example_prompt, 
    suffix="Question: {input}", 
    input_variables=["input"]
)

print(prompt.format(input="Who was the father of Mary Ball Washington?"))

輸出：

    Question: Who lived longer, Muhammad Ali or Alan Turing?
    
    Are follow up questions needed here: Yes.
    Follow up: How old was Muhammad Ali when he died?
    Intermediate answer: Muhammad Ali was 74 years old when he died.
    Follow up: How old was Alan Turing when he died?
    Intermediate answer: Alan Turing was 41 years old when he died.
    So the final answer is: Muhammad Ali
    
    
    Question: When was the founder of craigslist born?
    
    Are follow up questions needed here: Yes.
    Follow up: Who was the founder of craigslist?
    Intermediate answer: Craigslist was founded by Craig Newmark.
    Follow up: When was Craig Newmark born?
    Intermediate answer: Craig Newmark was born on December 6, 1952.
    So the final answer is: December 6, 1952
    
    
    Question: Who was the maternal grandfather of George Washington?
    
    Are follow up questions needed here: Yes.
    Follow up: Who was the mother of George Washington?
    Intermediate answer: The mother of George Washington was Mary Ball Washington.
    Follow up: Who was the father of Mary Ball Washington?
    Intermediate answer: The father of Mary Ball Washington was Joseph Ball.
    So the final answer is: Joseph Ball
    
    
    Question: Are both the directors of Jaws and Casino Royale from the same country?
    
    Are follow up questions needed here: Yes.
    Follow up: Who is the director of Jaws?
    Intermediate Answer: The director of Jaws is Steven Spielberg.
    Follow up: Where is Steven Spielberg from?
    Intermediate Answer: The United States.
    Follow up: Who is the director of Casino Royale?
    Intermediate Answer: The director of Casino Royale is Martin Campbell.
    Follow up: Where is Martin Campbell from?
    Intermediate Answer: New Zealand.
    So the final answer is: No
    
    
    Question: Who was the father of Mary Ball Washington?

使用示例選擇器

我們將重復(fù)使用上文中的示例集和格式化程序。但是，與其直接將示例輸入到FewShotPromptTemplate對(duì)象中，我們將把它們輸入到一個(gè)ExampleSelector對(duì)象中。在下文中，我們將使用SemanticSimilarityExampleSelector類。該類根據(jù)示例與輸入之間的相似度選擇Few-Shot示例。它使用嵌入模型計(jì)算輸入與Few-Shot示例之間的相似度，并使用向量存儲(chǔ)執(zhí)行最近鄰搜索。

from langchain.prompts.example_selector import SemanticSimilarityExampleSelector
from langchain.vectorstores import Chroma
from langchain.embeddings import OpenAIEmbeddings

example_selector = SemanticSimilarityExampleSelector.from_examples(
    # 這是可供選擇的示例列表。
    examples,
    # 這是用于生成嵌入的嵌入類，用于衡量語義相似度。
    OpenAIEmbeddings(),
    # 這是用于存儲(chǔ)嵌入并進(jìn)行相似度搜索的向量存儲(chǔ)類。
    Chroma,
    # 這是要生成的示例數(shù)量。
    k=1
)

# 選擇與輸入最相似的示例。
question = "Who was the father of Mary Ball Washington?"
selected_examples = example_selector.select_examples({"question": question})
print(f"Examples most similar to the input: {question}")
for example in selected_examples:
    print("\n")
    for k, v in example.items():
        print(f"{k}: {v}")

輸出：

Running Chroma using direct local API.
Using DuckDB in-memory for database. Data will be transient.
Examples most similar to the input: Who was the father of Mary Ball Washington?


question: Who was the maternal grandfather of George Washington?
answer: 
Are follow up questions needed here: Yes.
Follow up: Who was the mother of George Washington?
Intermediate answer: The mother of George Washington was Mary Ball Washington.
Follow up: Who was the father of Mary Ball Washington?
Intermediate answer: The father of Mary Ball Washington was Joseph Ball.
So the final answer is: Joseph Ball

我們還可以將示例選擇器應(yīng)用于FewShotPromptTemplate。創(chuàng)建一個(gè)FewShotPromptTemplate對(duì)象。該對(duì)象接收示例選擇器和用于Few-Shot示例的格式化程序：

prompt = FewShotPromptTemplate(
    example_selector=example_selector, 
    example_prompt=example_prompt, 
    suffix="Question: {input}", 
    input_variables=["input"]
)

print(prompt.format(input="Who was the father of Mary Ball Washington?"))

輸出：

Question: Who was the maternal grandfather of George Washington?

Are follow up questions needed here: Yes.
Follow up: Who was the mother of George Washington?
Intermediate answer: The mother of George Washington was Mary Ball Washington.
Follow up: Who was the father of Mary Ball Washington?
Intermediate answer: The father of Mary Ball Washington was Joseph Ball.
So the final answer is: Joseph Ball

Question: Who was the father of Mary Ball Washington?

參考文獻(xiàn)：
[1] LangChain ????? 中文網(wǎng)，跟著LangChain一起學(xué)LLM/GPT開發(fā)：https://www.langchain.com.cn/
[2] LangChain中文網(wǎng) - LangChain 是一個(gè)用于開發(fā)由語言模型驅(qū)動(dòng)的應(yīng)用程序的框架：http://www.cnlangchain.com/文章來源地址http://www.zghlxwxcb.cn/news/detail-622263.html

到了這里，關(guān)于自然語言處理從入門到應(yīng)用——LangChain：提示（Prompts）-[提示模板：創(chuàng)建自定義提示模板和含有Few-Shot示例的提示模板]的文章就介紹完了。如果您還想了解更多內(nèi)容，請(qǐng)?jiān)谟疑辖撬阉鱐OY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！