Skip to content

About

Workshop to teaching learning from demonstration

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

SO-101 robot learning workshop

Four notebooks take you from a fresh environment to a fine-tuned SmolVLA policy controlling an SO-101 follower arm:

  1. Environment setup
  2. Leader/follower data collection
  3. SmolVLA training
  4. Deployment

Open the notebooks in order. Commands that require an interactive terminal, GPU, or physical robot are presented as copyable terminal commands. Use a terminal on the machine connected to the hardware, not a remote notebook kernel, for calibration, teleoperation, recording and deployment.

The examples use a local dataset and do not upload it or the model. They pin a specific LeRobot source revision; GPU drivers, system libraries and downloaded model weights are not locked. The default commands assume macOS or Linux with a POSIX shell, two OpenCV cameras, and an assembled SO-101 leader and follower. Training expects a CUDA GPU; deployment can run on a different machine.

Safety: Robots can move unexpectedly. Secure the arm, clear its workspace, keep people and loose objects away, and have a person ready to cut motor power. Only operate under supervision and within the robot's mechanical limits.

License

This workshop is licensed under the MIT License. Third-party dependencies and model weights are subject to their own licenses.

About

Workshop to teaching learning from demonstration

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages