LeRobot

An open-source robot learning toolkit from Hugging Face, covering the whole pipeline from data collection to training to deployment.

LeRobot is a PyTorch robot-learning library open-sourced by Hugging Face in 2024, aimed at lowering the barrier to learning on real robots. It provides a unified dataset format (LeRobotDataset, hosted on the Hugging Face Hub), ships built-in implementations of policies such as ACT, Diffusion Policy, π0, and SmolVLA, and includes scripts for teleoperated data collection, training, evaluation, and running on a real robot. It pairs with low-cost open-source arms such as the SO-100/SO-101, letting a student complete the whole pipeline — collect demonstration data, train with imitation learning, deploy on the real robot — with only a few hundred to a couple thousand dollars of hardware, which has made it one of the most common starting points for getting into embodied AI.

ExampleTwo SO-101 arms are used in a leader-follower teleoperation setup to record a few dozen grasping demonstrations, which are then used to train an ACT policy in LeRobot and run it on the real arm.

Also called

Hugging Face LeRobot, lerobot

Related

Hugging Face · LeRobotDataset · SO-100 / SO-101 Arm · SmolVLA · Action Chunking with Transformers · Imitation Learning

Sources

huggingface/lerobot (GitHub)
LeRobot documentation