Repolode
Submit
huggingface

huggingface/lerobot

PyTorch library of models, datasets and tools for real-world robotics, covering hardware control, data collection, policy training and evaluation.

View on GitHubWebsite
Stars
28,733
Forks
6,032
Open issues
883
Last commit
18 hours ago

What is it?

LeRobot is a Hugging Face library for end-to-end robot learning in PyTorch. It aims to lower the barrier to entry for robotics by providing a shared dataset format, pretrained policies, and tools for controlling physical hardware, from low-cost arms such as the SO-100 and SO-101 to humanoids.

The project offers a unified Robot class that separates control logic from hardware specifics, so the same data collection, training and visualization tools work across supported arms, mobile bases, humanoids, gamepads and keyboards. Custom robots can be added by implementing the same interface. Third-party plugins for other robots, teleoperators and cameras are discovered automatically when installed.

Datasets use the LeRobotDataset format, which stores synchronized video or images with Parquet state and action data and can be hosted on the Hugging Face Hub. Policies span imitation learning, reinforcement learning, vision-language-action models and world models, and can be evaluated in simulation with lerobot-eval or deployed on real robots with lerobot-rollout.

Key features

  • 01Unified Robot interface for arms, mobile robots, humanoids, gamepads and keyboards
  • 02LeRobotDataset format with Parquet and MP4 storage hosted on the Hugging Face Hub
  • 03Training scripts for many policies, including ACT, Diffusion and SmolVLA
  • 04Simulation evaluation on benchmarks such as LIBERO and MetaWorld
  • 05Plugin system for third-party robots, teleoperators and cameras

Use cases

  • Recording demonstrations with a low-cost arm and training an imitation learning policy
  • Sharing and reusing robotics datasets through the Hugging Face Hub
  • Benchmarking pretrained VLA or imitation policies in simulation
  • Adding support for a custom robot by implementing the Robot interface

Quick start

  1. 1

    Install LeRobot from PyPI

    pip install lerobot
  2. 2

    Check the installed version and environment info

    lerobot-info
  3. 3

    Add extras for the features you need, e.g. training

    pip install 'lerobot[training]'

Things to know

  • Base install is lightweight, but many features need extra dependencies installed separately.
  • Hardware features depend on specific robots, and some policies need GPUs for training.

In collections

Reinforcement learning frameworks & training environmentsRobotics & embodied AI projects

Topics

#robotics#pytorch#imitation-learning#reinforcement-learning#dataset#vision-language-action#hugging-face

FAQ

What is lerobot?
PyTorch library of models, datasets and tools for real-world robotics, covering hardware control, data collection, policy training and evaluation. LeRobot is a Hugging Face library for end-to-end robot learning in PyTorch. It aims to lower the barrier to entry for robotics by providing a shared dataset format, pretrained policies, and tools for controlling physical hardware, from low-cost arms such as the SO-100 and SO-101 to humanoids.
Is lerobot free to use? What license does it have?
lerobot is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is lerobot still maintained?
Actively maintained. The last commit to lerobot was on Nov 7, 2026.
What can I use lerobot for?
Recording demonstrations with a low-cost arm and training an imitation learning policy Sharing and reusing robotics datasets through the Hugging Face Hub Benchmarking pretrained VLA or imitation policies in simulation Adding support for a custom robot by implementing the Robot interface
How do I install lerobot?
Install LeRobot from PyPI: pip install lerobot Check the installed version and environment info: lerobot-info Add extras for the features you need, e.g. training: pip install 'lerobot[training]'
Where can I get lerobot?
The source code and installation instructions are on GitHub: https://github.com/huggingface/lerobot

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects