Skip to main content
Open In ColabOpen on GitHub

Sagemaker 端点

Amazon SageMaker 是一个系统，可通过完全托管的基础设施、工具和工作流程，为任何使用场景构建、训练和部署机器学习（ML）模型。

此笔记本介绍了如何使用托管在 SageMaker endpoint 上的LLM。

!pip3 install langchain boto3

设置​

您必须设置以下 SagemakerEndpoint 调用所需的参数：

  • endpoint_name：来自已部署的 Sagemaker 模型的端点名称。 必须在 AWS 区域内唯一。
  • credentials_profile_name：在 ~/.aws/credentials 或 ~/.aws/config 文件中的配置文件名称， 该配置文件需指定访问密钥或角色信息。 如果未指定，则将使用默认的凭证配置文件；或者，如果在 EC2 实例上，则使用 IMDS 中的凭证。 参见：https://boto3.amazonaws.com/v1/documentation/api/latest/guide/credentials.html

示例​

from langchain_core.documents import Document
API 参考:文档
example_doc_1 = """
Peter and Elizabeth took a taxi to attend the night party in the city. While in the party, Elizabeth collapsed and was rushed to the hospital.
Since she was diagnosed with a brain injury, the doctor told Peter to stay besides her until she gets well.
Therefore, Peter stayed with her at the hospital for 3 days without leaving.
"""

docs = [
    Document(
        page_content=example_doc_1,
    )
]

使用外部 boto3 会话进行初始化的示例​

跨账户场景​

import json
from typing import Dict

import boto3
from langchain.chains.question_answering import load_qa_chain
from langchain_aws.llms import SagemakerEndpoint
from langchain_aws.llms.sagemaker_endpoint import LLMContentHandler
from langchain_core.prompts import PromptTemplate

query = """How long was Elizabeth hospitalized?
"""

prompt_template = """Use the following pieces of context to answer the question at the end.

{context}

Question: {question}
Answer:"""
PROMPT = PromptTemplate(
    template=prompt_template, input_variables=["context", "question"]
)

roleARN = "arn:aws:iam::123456789:role/cross-account-role"
sts_client = boto3.client("sts")
response = sts_client.assume_role(
    RoleArn=roleARN, RoleSessionName="CrossAccountSession"
)

client = boto3.client(
    "sagemaker-runtime",
    region_name="us-west-2",
    aws_access_key_id=response["Credentials"]["AccessKeyId"],
    aws_secret_access_key=response["Credentials"]["SecretAccessKey"],
    aws_session_token=response["Credentials"]["SessionToken"],
)


class ContentHandler(LLMContentHandler):
    content_type = "application/json"
    accepts = "application/json"

    def transform_input(self, prompt: str, model_kwargs: Dict) -> bytes:
        input_str = json.dumps({"inputs": prompt, "parameters": model_kwargs})
        return input_str.encode("utf-8")

    def transform_output(self, output: bytes) -> str:
        response_json = json.loads(output.read().decode("utf-8"))
        return response_json[0]["generated_text"]


content_handler = ContentHandler()

chain = load_qa_chain(
    llm=SagemakerEndpoint(
        endpoint_name="endpoint-name",
        client=client,
        model_kwargs={"temperature": 1e-10},
        content_handler=content_handler,
    ),
    prompt=PROMPT,
)

chain({"input_documents": docs, "question": query}, return_only_outputs=True)
import json
from typing import Dict

from langchain.chains.question_answering import load_qa_chain
from langchain_aws.llms import SagemakerEndpoint
from langchain_aws.llms.sagemaker_endpoint import LLMContentHandler
from langchain_core.prompts import PromptTemplate

query = """How long was Elizabeth hospitalized?
"""

prompt_template = """Use the following pieces of context to answer the question at the end.

{context}

Question: {question}
Answer:"""
PROMPT = PromptTemplate(
    template=prompt_template, input_variables=["context", "question"]
)


class ContentHandler(LLMContentHandler):
    content_type = "application/json"
    accepts = "application/json"

    def transform_input(self, prompt: str, model_kwargs: Dict) -> bytes:
        input_str = json.dumps({"inputs": prompt, "parameters": model_kwargs})
        return input_str.encode("utf-8")

    def transform_output(self, output: bytes) -> str:
        response_json = json.loads(output.read().decode("utf-8"))
        return response_json[0]["generated_text"]


content_handler = ContentHandler()

chain = load_qa_chain(
    llm=SagemakerEndpoint(
        endpoint_name="endpoint-name",
        credentials_profile_name="credentials-profile-name",
        region_name="us-west-2",
        model_kwargs={"temperature": 1e-10},
        content_handler=content_handler,
    ),
    prompt=PROMPT,
)

chain({"input_documents": docs, "question": query}, return_only_outputs=True)