The Ultimate Guide For Basic Personal WSL DL Env Development

Simplified tensorflow gpu dev env

First, make sure NVIDIA driver is installed and works good, check through command nvidia-smi .

Then install wsl —> full doc link, here I use ubuntu.

# List of distros could be installed
wsl --list --online

# Install
wsl --install -d Ubuntu-24.04

In file explorer, find the WSL file system at \\wsl$\Ubuntu\home\USERNAME

Double ckeck nvidia-smi under WSL terminal session. Install uv and setting up tf-gpu as followed:

uv init
source .venv/bin/activate
uv add tensorflow[and-cuda]

Old version with SSH & network config

此刻是二零二四年三月一日凌晨一点四十分，我的心情无比沉重且轻盈，我将写下这篇文档以在将来帮助自己或有需要的人。

Fuck Windows, 灾难级。

配置 Windows

1. 重装电脑。

2. 主机基本配置

• 删除自带的第三方防火墙/网络安全设备
• 关闭Windows defender(防火墙)(或者在Windows defender下配置入展规则)
• (如需开启Windows主机的ssh server)在”设置->账户->你的信息”中选择本地账户登陆
• 在”设置->系统->电源和电池”中配置从不休眠电脑，(如果是笔记本)在控制面板中设置笔记本盒盖不采取任何操作
• (有代理需求的自行配置好代理)

3. 主机安装NVIDIA driver

• 在NVIDIA官网下载，选择，安装驱动 https://www.nvidia.com/download/index.aspx
• 执行 nvidia-smi 检查

配置 WSL

1. WSL基本配置(默认宿主机win11，WSL2；其他环境自行解决)

• 在 C:\Users\&Username&\ 新建.wslconfig文件并粘贴以下内容以设置wsl网络mirror

  experimental]<br>
  autoMemoryReclaim=gradual # 开启自动回收内存，可在 gradual, dropcache, disabled 之间选择<br>
  networkingMode=mirrored # 开启镜像网络<br>
  dnsTunneling=true # 开启 DNS Tunneling<br>
  firewall=true # 开启 Windows 防火墙<br>
  autoProxy=true # 开启自动同步代理<br>
  sparseVhd=true # 开启自动释放 WSL2 虚拟硬盘空间<br>

• 执行 wsl --install 安装WSL(需要WSL2，较新版本的Windows会直接下载WSL2，注意检查一下)
• 在Microsoft Store上下载 Ubuntu 22.04.3 LTS，并install。(如下载默认的Ubuntu 18*，有可能导致代理失效)
• 执行 sudo passwd root, su root, 以root用户登陆，默认的账户一直要sudo很烦且有可能碰到指令执行报错

2. WSL 配置DL开发环境，以安装tensorflow gpu版本为例

• 安装miniconda https://docs.anaconda.com/free/miniconda/index.html
• 安装gpu tensorflow https://www.tensorflow.org/install/pip

SSH

apt install openssh-server
apt install build-essential

VSCODE

安装vscode参考自文档：https://dev.to/abbazs/how-to-install-vscode-in-ubuntu-using-apt-get-2m8o

1. Download the Microsoft GPG key: Run the following command to download the Microsoft GPG key and save it as microsoft.gpg: curl https://packages.microsoft.com/keys/microsoft.asc | gpg --dearmor > microsoft.gpg

2. Install the GPG key: Run the following command to install the GPG key into the trusted keyring: sudo install -o root -g root -m 644 microsoft.gpg /etc/apt/trusted.gpg.d/

3. Add the VSCode and Microsoft Edge repositories: Run the following commands to add the VSCode and Microsoft Edge repositories to the system: sudo sh -c 'echo "deb [arch=amd64] https://packages.microsoft.com/repos/edge stable main" > /etc/apt/sources.list.d/microsoft-edge.list' sudo sh -c 'echo "deb [arch=amd64] https://packages.microsoft.com/repos/code stable main" > /etc/apt/sources.list.d/vscode.list'

4. Remove the downloaded GPG key: Run the following command to remove the microsoft.gpg file: sudo rm microsoft.gpg

5. Update package information: Run the following command to update the package information from the repositories: sudo apt update

6. Install VSCode: Run the following command to install Visual Studio Code: sudo apt install -y code

7. (Optional) Install with aptitude: If you use aptitude package manager, you can run the following commands instead of step 6 to install VSCode: sudo aptitude update sudo aptitude install -y code