◐ModelRobotics & ControlFree
LeRobot
State-of-the-art robot learning library from Hugging Face with pretrained policies for real and simulated manipulation tasks.
LeRobot
LeRobot is Hugging Face's end-to-end robot learning library designed to bridge the gap between cutting-edge research and real-world robot deployment. It provides pretrained policies (ACT, Diffusion Policy, TDMPC), datasets, and simulation environments so you can train and evaluate robot manipulation agents with minimal friction.
Key Features
- Pretrained policies ready for fine-tuning on your hardware (ACT, Diffusion Policy, TD-MPC2)
- Unified dataset format with Hugging Face Hub integration for sharing robot trajectories
- Simulation environments (Gym-PushT, Gym-Aloha, Gym-xArm) for fast iteration
- Real-robot support for low-cost arms (Koch v1.1, SO-100, Moss v1)
- Weights & Biases training visualisation out of the box
Quick Start
pip install lerobot
# Run a pretrained ACT policy in simulation
python lerobot/scripts/eval.py \
-p lerobot/act_aloha_sim_transfer_cube_human \
eval.n_episodes=10
npx ai-supply add lerobot-robot-learning
Curated mirror of the open-source LeRobot (Apache-2.0). Get it from the source.