Tensorflow GPU with Docker in WSL2 using VS Code Dev Containers

• Windows 11
• Windows Subsystem for Linux (WSL2)
• Latest NVIDIA Windows GPU Driver
• Docker Desktop WSL2
• Visual Studio Code with Dev Containers extension

• Create a new repository from the template or clone the code in WSL
• In Dockerfile:
  • Update the TensorFlow Docker image: run nvidia-smi command in your Terminal, and get the CUDA version. Then find the Tensorflow version compatible with the CUDA version here
  • Update USERNAME ARG to create a non-root user
  • Update USER_UID and USER_GID ARGs (use id command in your Terminal to get them)
  • Update required packages (optional)
• Update pip packages in requirements.txt (optional)
• Open your project in WSL with VS Code and reopen in Container: (F1) use the Dev Containers: Reopen in Container command

Run pyhton main.py to test it.