Anyscale

Anyscale 是一个完全托管的 Ray 平台，您可以在其上构建、部署和管理可扩展的 AI 和 Python 应用程序

此示例介绍了如何使用 LangChain 与 Anyscale 端点 进行交互。

##Installing the langchain packages needed to use the integration
%pip install -qU langchain-community

ANYSCALE_API_BASE = "..."
ANYSCALE_API_KEY = "..."
ANYSCALE_MODEL_NAME = "..."

import os

os.environ["ANYSCALE_API_BASE"] = ANYSCALE_API_BASE
os.environ["ANYSCALE_API_KEY"] = ANYSCALE_API_KEY

from langchain.chains import LLMChain
from langchain_community.llms import Anyscale
from langchain_core.prompts import PromptTemplate

API 参考：LLMChain | Anyscale | PromptTemplate

template = """Question: {question}

Answer: Let's think step by step."""

prompt = PromptTemplate.from_template(template)

llm = Anyscale(model_name=ANYSCALE_MODEL_NAME)

llm_chain = prompt | llm

question = "When was George Washington president?"

llm_chain.invoke({"question": question})

使用 Ray，我们可以分发查询而无需异步实现。这不仅适用于 Anyscale LLM 模型，也适用于任何其他未实现 _acall 或 _agenerate 的 Langchain LLM 模型。

prompt_list = [
    "When was George Washington president?",
    "Explain to me the difference between nuclear fission and fusion.",
    "Give me a list of 5 science fiction books I should read next.",
    "Explain the difference between Spark and Ray.",
    "Suggest some fun holiday ideas.",
    "Tell a joke.",
    "What is 2+2?",
    "Explain what is machine learning like I am five years old.",
    "Explain what is artifical intelligence.",
]

import ray


@ray.remote(num_cpus=0.1)
def send_query(llm, prompt):
    resp = llm.invoke(prompt)
    return resp


futures = [send_query.remote(llm, prompt) for prompt in prompt_list]
results = ray.get(futures)

相关

• LLM 概念指南
• LLM 操作指南