Skip to main content
Open In ColabOpen on GitHub

SageMaker 追踪

Amazon SageMaker 是一种完全托管的服务，可用于快速、轻松地构建、训练和部署机器学习（ML）模型。

Amazon SageMaker Experiments 是 Amazon SageMaker 的一项功能，可让您组织、跟踪、比较和评估机器学习实验及模型版本。

该笔记本展示了如何使用 LangChain 回调将提示和其他 LLM 超参数记录并跟踪到 SageMaker Experiments 中。在这里，我们使用不同的场景来展示其功能：

  • 场景1: 单个大语言模型（LLM） - 使用单个LLM模型根据给定提示生成输出的情况。
  • 场景 2：顺序链 - 使用两个大语言模型（LLM）的顺序链的情况。
  • 场景3：带工具的智能体（思维链） - 一种除了使用大语言模型（LLM）外，还结合多个工具（如搜索和数学计算）的情况。

在本笔记本中，我们将创建一个实验来记录每种场景中的提示。

安装与设置

%pip install --upgrade --quiet  sagemaker
%pip install --upgrade --quiet  langchain-openai
%pip install --upgrade --quiet  google-search-results

首先，设置所需的API密钥

import os

## Add your API keys below
os.environ["OPENAI_API_KEY"] = "<ADD-KEY-HERE>"
os.environ["SERPAPI_API_KEY"] = "<ADD-KEY-HERE>"
from langchain_community.callbacks.sagemaker_callback import SageMakerCallbackHandler
from langchain.agents import initialize_agent, load_tools
from langchain.chains import LLMChain, SimpleSequentialChain
from langchain_core.prompts import PromptTemplate
from langchain_openai import OpenAI
from sagemaker.analytics import ExperimentAnalytics
from sagemaker.experiments.run import Run
from sagemaker.session import Session

LLM 提示跟踪​

# LLM Hyperparameters
HPARAMS = {
    "temperature": 0.1,
    "model_name": "gpt-3.5-turbo-instruct",
}

# Bucket used to save prompt logs (Use `None` is used to save the default bucket or otherwise change it)
BUCKET_NAME = None

# Experiment name
EXPERIMENT_NAME = "langchain-sagemaker-tracker"

# Create SageMaker Session with the given bucket
session = Session(default_bucket=BUCKET_NAME)

场景1 - 大型语言模型​

RUN_NAME = "run-scenario-1"
PROMPT_TEMPLATE = "tell me a joke about {topic}"
INPUT_VARIABLES = {"topic": "fish"}
with Run(
    experiment_name=EXPERIMENT_NAME, run_name=RUN_NAME, sagemaker_session=session
) as run:
    # Create SageMaker Callback
    sagemaker_callback = SageMakerCallbackHandler(run)

    # Define LLM model with callback
    llm = OpenAI(callbacks=[sagemaker_callback], **HPARAMS)

    # Create prompt template
    prompt = PromptTemplate.from_template(template=PROMPT_TEMPLATE)

    # Create LLM Chain
    chain = LLMChain(llm=llm, prompt=prompt, callbacks=[sagemaker_callback])

    # Run chain
    chain.run(**INPUT_VARIABLES)

    # Reset the callback
    sagemaker_callback.flush_tracker()

场景2 - 顺序链​

RUN_NAME = "run-scenario-2"

PROMPT_TEMPLATE_1 = """You are a playwright. Given the title of play, it is your job to write a synopsis for that title.
Title: {title}
Playwright: This is a synopsis for the above play:"""
PROMPT_TEMPLATE_2 = """You are a play critic from the New York Times. Given the synopsis of play, it is your job to write a review for that play.
Play Synopsis: {synopsis}
Review from a New York Times play critic of the above play:"""

INPUT_VARIABLES = {
    "input": "documentary about good video games that push the boundary of game design"
}
with Run(
    experiment_name=EXPERIMENT_NAME, run_name=RUN_NAME, sagemaker_session=session
) as run:
    # Create SageMaker Callback
    sagemaker_callback = SageMakerCallbackHandler(run)

    # Create prompt templates for the chain
    prompt_template1 = PromptTemplate.from_template(template=PROMPT_TEMPLATE_1)
    prompt_template2 = PromptTemplate.from_template(template=PROMPT_TEMPLATE_2)

    # Define LLM model with callback
    llm = OpenAI(callbacks=[sagemaker_callback], **HPARAMS)

    # Create chain1
    chain1 = LLMChain(llm=llm, prompt=prompt_template1, callbacks=[sagemaker_callback])

    # Create chain2
    chain2 = LLMChain(llm=llm, prompt=prompt_template2, callbacks=[sagemaker_callback])

    # Create Sequential chain
    overall_chain = SimpleSequentialChain(
        chains=[chain1, chain2], callbacks=[sagemaker_callback]
    )

    # Run overall sequential chain
    overall_chain.run(**INPUT_VARIABLES)

    # Reset the callback
    sagemaker_callback.flush_tracker()

场景3 - 带工具的代理​

RUN_NAME = "run-scenario-3"
PROMPT_TEMPLATE = "Who is the oldest person alive? And what is their current age raised to the power of 1.51?"
with Run(
    experiment_name=EXPERIMENT_NAME, run_name=RUN_NAME, sagemaker_session=session
) as run:
    # Create SageMaker Callback
    sagemaker_callback = SageMakerCallbackHandler(run)

    # Define LLM model with callback
    llm = OpenAI(callbacks=[sagemaker_callback], **HPARAMS)

    # Define tools
    tools = load_tools(["serpapi", "llm-math"], llm=llm, callbacks=[sagemaker_callback])

    # Initialize agent with all the tools
    agent = initialize_agent(
        tools, llm, agent="zero-shot-react-description", callbacks=[sagemaker_callback]
    )

    # Run agent
    agent.run(input=PROMPT_TEMPLATE)

    # Reset the callback
    sagemaker_callback.flush_tracker()

加载日志数据​

提示记录后，我们可以轻松地将其加载并转换为Pandas DataFrame，如下所示。

# Load
logs = ExperimentAnalytics(experiment_name=EXPERIMENT_NAME)

# Convert as pandas dataframe
df = logs.dataframe(force_refresh=True)

print(df.shape)
df.head()

如上所示，该实验包含对应每个场景的三个运行（行）。每次运行都会将提示词及相关的大语言模型设置/超参数以json格式记录，并保存到s3存储桶中。您可以随时从每个json路径加载并查看日志数据。