1. Amazon China准备EC2实例

a. AWS区域选择AWS China

b.  AMI选择Amazon Deep Learning AMI，具体为Deep Learning OSS Nvidia Driver AMI GPU PyTorch 2.6 (Amazon Linux 2023)

c. 实例类型选择t3.medium，实际使用可以选择更大规格

d. EC2实例安全组入站规则需要放开SSH 22端口和TCP 8888端口

2. EC2实例安装配置JupyterLab

a. 安装JupyterLab

pip install jupyterlab

b. 生成配置文件

jupyter lab --generate-config

# 预期结果

Writing default config to: /home/ec2-user/.jupyter/jupyter_lab_config.py

c. 进入Python命令行修改Jupyterlab密码

python

>>> from notebook.auth import passwd 

>>> passwd()

# 会输出一串很长的加密字符，记住它，后面修改配置会用到

>>> exit()

d. 修改配置文件

cd ~

mkdir jupyterlab_projects

sudo vi ~/.jupyter/jupyter_lab_config.py

# 添加如下配置：

c.NotebookApp.allow_remote_access = True #允许远程连接

c.NotebookApp.ip='*'                    # 设置所有ip皆可访问

c.NotebookApp.password = '*******'  # 上面的密钥

c.NotebookApp.open_browser = False      # 禁止自动打开浏览器

c.NotebookApp.port = 8888                # 设置打开端口,随便设置,保证不冲突即可

c.NotebookApp.notebook_dir = '/home/ec2-user/jupyterlab_projects'  #设置Notebook启动进入的目录

3. 设置为开机自启动

sudo vim /etc/systemd/system/jupyter.service

# 复制如下配置文件

[Unit]

Description=Jupyterlab

After=network.target

[Service]

Type=simple

ExecStart=/opt/pytorch/bin/jupyter-lab --config=/home/ec2-user/.jupyter/jupyter_lab_config.py --no-browser

User=ec2-user

Group=ec2-user

WorkingDirectory=/home/ec2-user/jupyterlab_projects/

Restart=always

RestartSec=10

[Install]

WantedBy=multi-user.target

设置为开机自启动服务

sudo systemctl enable jupyter

sudo systemctl start jupyter

sudo systemctl status jupyter

4. 测试验证

浏览器打开如下URL进行测试:

http://52.83.41.25:8888/lab

输入此前设置的密码，即可登录使用：