跳到主要内容
Open In ColabOpen on GitHub

Beam

调用 Beam API 封装器,以在云部署中部署 gpt2 LLM 实例并进行后续调用。需要安装 Beam 库并注册 Beam 客户端 ID 和客户端密钥。通过调用封装器,模型实例将被创建并运行,返回与提示相关的文本。随后可以通过直接调用 Beam API 进行额外的调用。

创建账户,如果您还没有。从仪表板获取您的 API 密钥。

安装 Beam CLI

!curl https://raw.githubusercontent.com/slai-labs/get-beam/main/get-beam.sh -sSfL | sh

注册 API 密钥并设置您的 Beam 客户端 ID 和密钥环境变量

import os

beam_client_id = "<Your beam client id>"
beam_client_secret = "<Your beam client secret>"

# Set the environment variables
os.environ["BEAM_CLIENT_ID"] = beam_client_id
os.environ["BEAM_CLIENT_SECRET"] = beam_client_secret

# Run the beam configure command
!beam configure --clientId={beam_client_id} --clientSecret={beam_client_secret}

安装 Beam SDK

%pip install --upgrade --quiet  beam-sdk

直接从 LangChain 部署并调用 Beam!

请注意,冷启动可能需要几分钟才能返回响应,但后续调用会更快!

from langchain_community.llms.beam import Beam

llm = Beam(
model_name="gpt2",
name="langchain-gpt2-test",
cpu=8,
memory="32Gi",
gpu="A10G",
python_version="python3.8",
python_packages=[
"diffusers[torch]>=0.10",
"transformers",
"torch",
"pillow",
"accelerate",
"safetensors",
"xformers",
],
max_length="50",
verbose=False,
)

llm._deploy()

response = llm._call("Running machine learning on a remote GPU")

print(response)
API 参考:Beam