概述

这将帮助您开始使用 Outline LLM。有关所有 Outline 功能和配置的详细文档，请前往 API 参考。

Outline 是一个用于生成约束语言的库。它允许您将大型语言模型（LLM）与各种后端一起使用，同时对生成的输出应用约束。

概述

集成详细信息

类	包	本地化	序列化	JS 支持	软件包下载	最新包装
Outlines	langchain-community	✅	beta	❌

设置

要访问 Outline 模型，您需要有互联网连接才能从 huggingface 下载模型权重。根据后端，您需要安装所需的依赖项（参见概述 docs)

凭据

Outline 没有内置的身份验证机制。

安装

LangChain Outline 集成位于langchain-community包中，需要outlines库：

%pip install -qU langchain-community outlines

实例

现在我们可以实例化我们的 Model 对象并生成聊天补全：

from langchain_community.llms import Outlines

# For use with llamacpp backend
model = Outlines(model="microsoft/Phi-3-mini-4k-instruct", backend="llamacpp")

# For use with vllm backend (not available on Mac)
model = Outlines(model="microsoft/Phi-3-mini-4k-instruct", backend="vllm")

# For use with mlxlm backend (only available on Mac)
model = Outlines(model="microsoft/Phi-3-mini-4k-instruct", backend="mlxlm")

# For use with huggingface transformers backend
model = Outlines(
    model="microsoft/Phi-3-mini-4k-instruct"
)  # defaults to backend="transformers"

API 参考：概述

调用

model.invoke("Hello how are you?")

链接

from langchain_core.prompts import PromptTemplate

prompt = PromptTemplate.from_template("How to say {input} in {output_language}:\n")

chain = prompt | model
chain.invoke(
    {
        "output_language": "German",
        "input": "I love programming.",
    }
)

API 参考：PromptTemplate

流

Outline 支持令牌流式处理：

for chunk in model.stream("Count to 10 in French:"):
    print(chunk, end="", flush=True)

约束生成

Outline 允许您对生成的输出应用各种约束：

Regex 约束

model.regex = r"((25[0-5]|2[0-4]\d|[01]?\d\d?)\.){3}(25[0-5]|2[0-4]\d|[01]?\d\d?)"
response = model.invoke("What is the IP address of Google's DNS server?")

response

类型约束

model.type_constraints = int
response = model.invoke("What is the answer to life, the universe, and everything?")

JSON 架构

from pydantic import BaseModel

class Person(BaseModel):
    name: str

model.json_schema = Person
response = model.invoke("Who is the author of LangChain?")
person = Person.model_validate_json(response)

person

语法约束

model.grammar = """
?start: expression
?expression: term (("+" | "-") term)
?term: factor (("" | "/") factor)
?factor: NUMBER | "-" factor | "(" expression ")"
%import common.NUMBER
%import common.WS
%ignore WS
"""
response = model.invoke("Give me a complex arithmetic expression:")

response

API 参考

有关所有 ChatOutline 功能和配置的详细文档，请访问 API 参考：https://python.langchain.com/api_reference/community/chat_models/langchain_community.chat_models.outlines.ChatOutlines.html

概述文档：

https://dottxt-ai.github.io/outlines/latest/

LLM 概念指南
LLM 操作指南

概述