Skip to content

DOCKER_WINDOWS.md

Latest commit

 

History

History
221 lines (171 loc) · 5.84 KB

DOCKER_WINDOWS.md

File metadata and controls

221 lines (171 loc) · 5.84 KB

Docker Setup for Windows (Ubuntu 22.04 LTS)

This guide covers running LearnFlake (ROS2 Humble + RoboSuite) on Windows using Docker Desktop with WSL2.

Prerequisites

1. Windows Requirements

  • Windows 10 version 21H2+ or Windows 11
  • WSL2 enabled with a Linux distribution installed
  • Docker Desktop for Windows (WSL2 backend)

2. Install Docker Desktop

  1. Download from Docker Desktop
  2. During installation, ensure "Use WSL 2 instead of Hyper-V" is selected
  3. After installation, go to Settings > Resources > WSL Integration
  4. Enable integration with your WSL2 distro

3. (Optional) NVIDIA GPU Support

If you have an NVIDIA GPU and want CUDA acceleration:

  1. Install the latest NVIDIA drivers on Windows
  2. In WSL2, install NVIDIA Container Toolkit:
# Inside WSL2 terminal
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
  sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
  sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
sudo apt-get update
sudo apt-get install -y nvidia-container-toolkit
  1. Restart Docker Desktop

Quick Start

Build the Image

docker compose -f docker-compose.ubuntu.yml build

Run a Container

CPU only (recommended for most use cases):

docker compose -f docker-compose.ubuntu.yml run --rm rover_cpu bash

With GPU support:

docker compose -f docker-compose.ubuntu.yml run --rm rover_gpu bash

For RL training:

docker compose -f docker-compose.ubuntu.yml run --rm rover_rl bash

Available Services

Service Description GPU GUI
rover_base Minimal container, headless No No
rover_cpu CPU with WSLg GUI support No Yes
rover_rl RL training focused No Yes
rover_gpu NVIDIA GPU with CUDA Yes Yes
rover_dev Development with exposed ports No Yes

Common Workflows

Start a detached container

docker compose -f docker-compose.ubuntu.yml up -d rover_cpu

Enter a running container

docker compose -f docker-compose.ubuntu.yml exec rover_cpu bash

Stop all containers

docker compose -f docker-compose.ubuntu.yml down

Rebuild without cache

docker compose -f docker-compose.ubuntu.yml build --no-cache

Inside the Container

Source ROS2 environment (auto-done by entrypoint)

source /opt/ros/humble/setup.bash

Build the workspace

cd /LearnFlake
colcon build --symlink-install
source install/setup.bash

Run RoboSuite demo

cd /LearnFlake/src/external_pkgs/RoboSuite
python -m robosuite.demos.demo_random_action

Run RL training

cd /LearnFlake/src/rl_autonomy
python scripts/run_ppo.py

GUI Applications (WSLg)

Windows 11 and Windows 10 21H2+ include WSLg (Windows Subsystem for Linux GUI) which allows running Linux GUI apps seamlessly.

Test GUI is working

# Inside the container
apt-get update && apt-get install -y x11-apps
xeyes  # Should show animated eyes

MuJoCo rendering modes

  • MUJOCO_GL=egl - Headless rendering (default, fastest)
  • MUJOCO_GL=glfw - Window-based rendering (requires GUI)
  • MUJOCO_GL=osmesa - Software rendering (slowest, no GPU)

To change rendering mode:

export MUJOCO_GL=glfw
python your_script.py

Troubleshooting

"Cannot connect to X server"

Ensure WSLg is working:

# In WSL2 (not Docker)
echo $DISPLAY  # Should show something like :0

If empty, try:

export DISPLAY=:0

"nvidia-smi not found" in GPU container

  1. Verify NVIDIA drivers are installed on Windows
  2. Check Docker Desktop > Settings > Resources > GPU
  3. Ensure NVIDIA Container Toolkit is installed in WSL2

Container can't see GPU

# Verify GPU is visible
docker run --rm --gpus all nvidia/cuda:11.8.0-base-ubuntu22.04 nvidia-smi

Build fails at RoboSuite installation

The Dockerfile continues even if RoboSuite fails to install (using || true). Install manually inside the container:

pip install -e /LearnFlake/src/external_pkgs/RoboSuite

Slow performance

  • Use MUJOCO_GL=egl for headless rendering
  • For GPU containers, ensure NVIDIA runtime is active
  • Close unnecessary applications on the host

File Structure

LearnFlake/
├── Dockerfile.ubuntu           # Ubuntu 22.04 LTS Dockerfile
├── docker-compose.ubuntu.yml   # Windows-optimized compose file
├── docker/
│   └── entrypoint.sh          # Container startup script
├── src/
│   ├── external_pkgs/
│   │   └── RoboSuite/         # Physics simulation
│   └── rl_autonomy/           # RL algorithms
└── DOCKER_WINDOWS.md          # This file

Tips

  1. Live code editing: The ./:/LearnFlake volume mount means changes on Windows are immediately reflected in the container.

  2. Persist installed packages: If you install additional packages, commit the container:

    docker commit learnflake_cpu myusername/learnflake:custom

  3. Use VS Code Remote: Install the "Dev Containers" extension in VS Code to develop directly inside the container.

  4. TensorBoard: The rover_dev service exposes port 6006 for TensorBoard:

    # Inside container
tensorboard --logdir=/LearnFlake/logs --bind_all
# Access at http://localhost:6006 on Windows