Skip to main content
Stage 2 · Chapter 5 · Practice

5. Software Environment Setup

Chapter 5 of the SO-ARM Physical AI Beginner's Course — set up the LeRobot software environment on Ubuntu x86 or Jetson: Miniforge, the verified Seeed LeRobot repository, ffmpeg, PyTorch, and the feetech motor dependencies.

RequirementsInstall LeRobotVerify

5.1 System Requirements​

Requirements

5.1 System Requirements

For Ubuntu x86:

  • Ubuntu 22.04
  • CUDA 12+
  • Python 3.12
  • Torch 2.6+

For Jetson Orin:

  • Jetson JetPack 6.0 and 6.1 (JetPack 6.2 is not supported yet)
  • Python 3.12
  • Torch 2.3+

Remember from Chapter 3: avoid WSL or virtual-machine Ubuntu on Windows — native Ubuntu or Jetson gives you the fewest surprises.

5.2 Install LeRobot​

Step by step

5.2 Install LeRobot

Environments such as PyTorch and torchvision need to be installed based on your CUDA version. Follow the tab for your device.

1. Install Miniforge:

wget https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-x86_64.sh
chmod +x Miniforge3-Linux-x86_64.sh
./Miniforge3-Linux-x86_64.sh
# Once the installation is complete:
source ~/.bashrc
# Initialize all shells
conda init --all

2. Create and activate a fresh conda environment for LeRobot:

conda create -y -n lerobot python=3.12 && conda activate lerobot

3. Clone the verified LeRobot repository:

git clone https://github.com/Seeed-Projects/lerobot.git ~/lerobot

4. Install ffmpeg in your environment:

conda install ffmpeg -c conda-forge
tip

This usually installs ffmpeg 7.X compiled with the libsvtav1 encoder. If libsvtav1 is not supported (check with ffmpeg -encoders), explicitly install ffmpeg 7.1.1:

conda install ffmpeg=7.1.1 -c conda-forge

This also fixes the No valid stream found in input file error that appears later when encoding dataset videos.

5. Install LeRobot with the feetech motor dependencies:

tip

On a newly configured Ubuntu 22.04 machine — especially a minimal install — gcc and other build tools may be missing, and the evdev Python package used by the feetech extra will fail to build. Install them first:

sudo apt update
sudo apt install build-essential
cd ~/lerobot && pip install -e ".[feetech]"

5.3 Verify the Environment​

Verify

5.3 Verify the Environment

Installing LeRobot via pip can replace the original PyTorch and torchvision with CPU versions, so always check before training. In the lerobot conda environment, run:

python   # start Python in the terminal
import torch
print(torch.cuda.is_available())
exit()   # exit Python
  • False → the current environment uses the CPU version of PyTorch. For GPU training or inference, reinstall the PyTorch and torchvision versions that match your CUDA version following the official PyTorch guide.
  • True → GPU acceleration is available.
tip

NVIDIA RTX 50-series GPUs require the preview version of CUDA 12.8 or higher:

pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128

Finally, check that the LeRobot command-line tools are on your PATH:

lerobot-find-port --help

If the command prints help text, the environment is ready for the next chapter, where we configure the motors and assemble the arms.

Loading Comments...