LeRobot
End-to-end robot learning. Datasets, ACT, SmolVLA, training, deployment.
Overview
LeRobot is a PyTorch framework by Hugging Face for imitation and reinforcement learning on real robots. Includes datasets (DROID-100, ALOHA, SO-100), policy architectures (ACT, SmolVLA), training scripts, and deployment tools.
Official Links
- huggingface.co/docs/lerobot — Docs
- github.com/huggingface/lerobot — Code
Continue the loop
A dataset format is step one; the loop trains on it, then scores the policy on real hardware.