Repository files navigation

Meet HopeJR – A humanoid robot arm and hand for dexterous manipulation!

Control it with exoskeletons and gloves for precise hand movements.

Perfect for advanced manipulation tasks! 🤖

See the full HopeJR tutorial here.

Meet the updated SO100, the SO-101 – Just €114 per arm!

Train it in minutes with a few simple moves on your laptop.

Then sit back and watch your creation act autonomously! 🤯

See the full SO-101 tutorial here.

Want to take it to the next level? Make your SO-101 mobile by building LeKiwi!

Check out theLeKiwi tutorial and bring your robot to life on wheels.

🤗 LeRobot aims to provide models, datasets, and tools for real-world robotics in PyTorch. The goal is to lower the barrier to entry to robotics so that everyone can contribute and benefit from sharing datasets and pretrained models.

🤗 LeRobot contains state-of-the-art approaches that have been shown to transfer to the real-world with a focus on imitation learning and reinforcement learning.

🤗 LeRobot already provides a set of pretrained models, datasets with human collected demonstrations, and simulation environments to get started without assembling a robot. In the coming weeks, the plan is to add more and more support for real-world robotics on the most affordable and capable robots out there.

🤗 LeRobot hosts pretrained models and datasets on this Hugging Face community page:huggingface.co/lerobot

ACT policy on ALOHA env	TDMPC policy on SimXArm env	Diffusion policy on PushT env

• The LeRobot team 🤗 for building SmolVLAPaper,Blog.
• Thanks to Tony Zhao, Zipeng Fu and colleagues for open sourcing ACT policy, ALOHA environments and datasets. Ours are adapted fromALOHA andMobile ALOHA.
• Thanks to Cheng Chi, Zhenjia Xu and colleagues for open sourcing Diffusion policy, Pusht environment and datasets, as well as UMI datasets. Ours are adapted fromDiffusion Policy andUMI Gripper.
• Thanks to Nicklas Hansen, Yunhai Feng and colleagues for open sourcing TDMPC policy, Simxarm environments and datasets. Ours are adapted fromTDMPC andFOWM.
• Thanks to Antonio Loquercio and Ashish Kumar for their early support.
• Thanks toSeungjae (Jay) Lee,Mahi Shafiullah and colleagues for open sourcingVQ-BeT policy and helping us adapt the codebase to our repository. The policy is adapted fromVQ-BeT repo.

Download our source code:

git clone https://github.com/huggingface/lerobot.gitcd lerobot

Create a virtual environment with Python 3.10 and activate it, e.g. withminiconda:

conda create -y -n lerobot python=3.10conda activate lerobot

When usingminiconda, installffmpeg in your environment:

conda install ffmpeg -c conda-forge

NOTE: This usually installsffmpeg 7.X for your platform compiled with thelibsvtav1 encoder. Iflibsvtav1 is not supported (check supported encoders withffmpeg -encoders), you can:

• [On any platform] Explicitly installffmpeg 7.X using:

conda install ffmpeg=7.1.1 -c conda-forge

• [On Linux only] Installffmpeg build dependencies andcompile ffmpeg from source with libsvtav1, and make sure you use the corresponding ffmpeg binary to your install withwhich ffmpeg.

Install 🤗 LeRobot:

pip install -e.

NOTE: If you encounter build errors, you may need to install additional dependencies (cmake,build-essential, andffmpeg libs). On Linux, run:sudo apt-get install cmake build-essential python3-dev pkg-config libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev libswscale-dev libswresample-dev libavfilter-dev. For other systems, see:Compiling PyAV

For simulations, 🤗 LeRobot comes with gymnasium environments that can be installed as extras:

• aloha
• xarm
• pusht

For instance, to install 🤗 LeRobot with aloha and pusht, use:

pip install -e".[aloha, pusht]"

To useWeights and Biases for experiment tracking, log in with

wandb login

(note: you will also need to enable WandB in the configuration. See below.)

Check outexample 1 that illustrates how to use our dataset class which automatically downloads data from the Hugging Face hub.

You can also locally visualize episodes from a dataset on the hub by executing our script from the command line:

python -m lerobot.scripts.visualize_dataset \    --repo-id lerobot/pusht \    --episode-index 0

or from a dataset in a local folder with theroot option and the--local-files-only (in the following case the dataset will be searched for in./my_local_data_dir/lerobot/pusht)

python -m lerobot.scripts.visualize_dataset \    --repo-id lerobot/pusht \    --root ./my_local_data_dir \    --local-files-only 1 \    --episode-index 0

It will openrerun.io and display the camera streams, robot states and actions, like this:

Our script can also visualize datasets stored on a distant server. Seepython -m lerobot.scripts.visualize_dataset --help for more instructions.

A dataset inLeRobotDataset format is very simple to use. It can be loaded from a repository on the Hugging Face hub or a local folder simply with e.g.dataset = LeRobotDataset("lerobot/aloha_static_coffee") and can be indexed into like any Hugging Face and PyTorch dataset. For instancedataset[0] will retrieve a single temporal frame from the dataset containing observation(s) and an action as PyTorch tensors ready to be fed to a model.

A specificity ofLeRobotDataset is that, rather than retrieving a single frame by its index, we can retrieve several frames based on their temporal relationship with the indexed frame, by settingdelta_timestamps to a list of relative times with respect to the indexed frame. For example, withdelta_timestamps = {"observation.image": [-1, -0.5, -0.2, 0]} one can retrieve, for a given index, 4 frames: 3 "previous" frames 1 second, 0.5 seconds, and 0.2 seconds before the indexed frame, and the indexed frame itself (corresponding to the 0 entry). See example1_load_lerobot_dataset.py for more details ondelta_timestamps.

Under the hood, theLeRobotDataset format makes use of several ways to serialize data which can be useful to understand if you plan to work more closely with this format. We tried to make a flexible yet simple dataset format that would cover most type of features and specificities present in reinforcement learning and robotics, in simulation and in real-world, with a focus on cameras and robot states but easily extended to other types of sensory inputs as long as they can be represented by a tensor.

Here are the important details and internal structure organization of a typicalLeRobotDataset instantiated withdataset = LeRobotDataset("lerobot/aloha_static_coffee"). The exact features will change from dataset to dataset but not the main aspects:

dataset attributes:  ├ hf_dataset: a Hugging Face dataset (backed by Arrow/parquet). Typical features example:  │  ├ observation.images.cam_high (VideoFrame):  │  │   VideoFrame = {'path': path to a mp4 video, 'timestamp' (float32): timestamp in the video}  │  ├ observation.state (list of float32): position of an arm joints (for instance)  │  ... (more observations)  │  ├ action (list of float32): goal position of an arm joints (for instance)  │  ├ episode_index (int64): index of the episode for this sample  │  ├ frame_index (int64): index of the frame for this sample in the episode ; starts at 0 for each episode  │  ├ timestamp (float32): timestamp in the episode  │  ├ next.done (bool): indicates the end of an episode ; True for the last frame in each episode  │  └ index (int64): general index in the whole dataset  ├ episode_data_index: contains 2 tensors with the start and end indices of each episode  │  ├ from (1D int64 tensor): first frame index for each episode — shape (num episodes,) starts with 0  │  └ to: (1D int64 tensor): last frame index for each episode — shape (num episodes,)  ├ stats: a dictionary of statistics (max, mean, min, std) for each feature in the dataset, for instance  │  ├ observation.images.cam_high: {'max': tensor with same number of dimensions (e.g. `(c, 1, 1)` for images, `(c,)` for states), etc.}  │  ...  ├ info: a dictionary of metadata on the dataset  │  ├ codebase_version (str): this is to keep track of the codebase version the dataset was created with  │  ├ fps (float): frame per second the dataset is recorded/synchronized to  │  ├ video (bool): indicates if frames are encoded in mp4 video files to save space or stored as png files  │  └ encoding (dict): if video, this documents the main options that were used with ffmpeg to encode the videos  ├ videos_dir (Path): where the mp4 videos or png images are stored/accessed  └ camera_keys (list of string): the keys to access camera features in the item returned by the dataset (e.g. `["observation.images.cam_high", ...]`)

ALeRobotDataset is serialised using several widespread file formats for each of its parts, namely:

• hf_dataset stored using Hugging Face datasets library serialization to parquet
• videos are stored in mp4 format to save space
• metadata are stored in plain json/jsonl files

Dataset can be uploaded/downloaded from the HuggingFace hub seamlessly. To work on a local dataset, you can specify its location with theroot argument if it's not in the default~/.cache/huggingface/lerobot location.

Check outexample 2 that illustrates how to download a pretrained policy from Hugging Face hub, and run an evaluation on its corresponding environment.

We also provide a more capable script to parallelize the evaluation over multiple environments during the same rollout. Here is an example with a pretrained model hosted onlerobot/diffusion_pusht:

python -m lerobot.scripts.eval \    --policy.path=lerobot/diffusion_pusht \    --env.type=pusht \    --eval.batch_size=10 \    --eval.n_episodes=10 \    --policy.use_amp=false \    --policy.device=cuda

Note: After training your own policy, you can re-evaluate the checkpoints with:

python -m lerobot.scripts.eval --policy.path={OUTPUT_DIR}/checkpoints/last/pretrained_model

Seepython -m lerobot.scripts.eval --help for more instructions.

Check outexample 3 that illustrates how to train a model using our core library in python, andexample 4 that shows how to use our training script from command line.

To use wandb for logging training and evaluation curves, make sure you've runwandb login as a one-time setup step. Then, when running the training command above, enable WandB in the configuration by adding--wandb.enable=true.

A link to the wandb logs for the run will also show up in yellow in your terminal. Here is an example of what they look like in your browser. Please also checkhere for the explanation of some commonly used metrics in logs.

Note: For efficiency, during training every checkpoint is evaluated on a low number of episodes. You may use--eval.n_episodes=500 to evaluate on more episodes than the default. Or, after training, you may want to re-evaluate your best checkpoints on more episodes or change the evaluation settings. Seepython -m lerobot.scripts.eval --help for more instructions.

We provide some pretrained policies on ourhub page that can achieve state-of-the-art performances.You can reproduce their training by loading the config from their run. Simply running:

python -m lerobot.scripts.train --config_path=lerobot/diffusion_pusht

reproduces SOTA results for Diffusion Policy on the PushT task.

If you would like to contribute to 🤗 LeRobot, please check out ourcontribution guide.

Once you have trained a policy you may upload it to the Hugging Face hub using a hub id that looks like${hf_user}/${repo_name} (e.g.lerobot/diffusion_pusht).

You first need to find the checkpoint folder located inside your experiment directory (e.g.outputs/train/2024-05-05/20-21-12_aloha_act_default/checkpoints/002500). Within that there is apretrained_model directory which should contain:

• config.json: A serialized version of the policy configuration (following the policy's dataclass config).
• model.safetensors: A set oftorch.nn.Module parameters, saved inHugging Face Safetensors format.
• train_config.json: A consolidated configuration containing all parameters used for training. The policy configuration should matchconfig.json exactly. This is useful for anyone who wants to evaluate your policy or for reproducibility.

To upload these to the hub, run the following:

huggingface-cli upload${hf_user}/${repo_name} path/to/pretrained_model

Seeeval.py for an example of how other people may use your policy.

An example of a code snippet to profile the evaluation of a policy:

fromtorch.profilerimportprofile,record_function,ProfilerActivitydeftrace_handler(prof):prof.export_chrome_trace(f"tmp/trace_schedule_{prof.step_num}.json")withprofile(activities=[ProfilerActivity.CPU,ProfilerActivity.CUDA],schedule=torch.profiler.schedule(wait=2,warmup=2,active=3,    ),on_trace_ready=trace_handler)asprof:withrecord_function("eval_policy"):foriinrange(num_episodes):prof.step()# insert code to profile, potentially whole body of eval_policy function

If you want, you can cite this work with:

@misc{cadene2024lerobot,author ={Cadene, Remi and Alibert, Simon and Soare, Alexander and Gallouedec, Quentin and Zouitine, Adil and Palma, Steven and Kooijmans, Pepijn and Aractingi, Michel and Shukor, Mustafa and Aubakirova, Dana and Russi, Martino and Capuano, Francesco and Pascale, Caroline and Choghari, Jade and Moss, Jess and Wolf, Thomas},title ={LeRobot: State-of-the-art Machine Learning for Real-World Robotics in Pytorch},howpublished ="\url{https://github.com/huggingface/lerobot}",year ={2024}}

Additionally, if you are using any of the particular policy architecture, pretrained models, or datasets, it is recommended to cite the original authors of the work as they appear below:

• SmolVLA

@article{shukor2025smolvla,title={SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics},author={Shukor, Mustafa and Aubakirova, Dana and Capuano, Francesco and Kooijmans, Pepijn and Palma, Steven and Zouitine, Adil and Aractingi, Michel and Pascal, Caroline and Russi, Martino and Marafioti, Andres and Alibert, Simon and Cord, Matthieu and Wolf, Thomas and Cadene, Remi},journal={arXiv preprint arXiv:2506.01844},year={2025}}

• Diffusion Policy

@article{chi2024diffusionpolicy,author ={Cheng Chi and Zhenjia Xu and Siyuan Feng and Eric Cousineau and Yilun Du and Benjamin Burchfiel and Russ Tedrake and Shuran Song},title ={Diffusion Policy: Visuomotor Policy Learning via Action Diffusion},journal ={The International Journal of Robotics Research},year ={2024},}

• ACT or ALOHA

@article{zhao2023learning,title={Learning fine-grained bimanual manipulation with low-cost hardware},author={Zhao, Tony Z and Kumar, Vikash and Levine, Sergey and Finn, Chelsea},journal={arXiv preprint arXiv:2304.13705},year={2023}}

• TDMPC

@inproceedings{Hansen2022tdmpc,title={Temporal Difference Learning for Model Predictive Control},author={Nicklas Hansen and Xiaolong Wang and Hao Su},booktitle={ICML},year={2022}}

• VQ-BeT

@article{lee2024behavior,title={Behavior generation with latent actions},author={Lee, Seungjae and Wang, Yibin and Etukuru, Haritheja and Kim, H Jin and Shafiullah, Nur Muhammad Mahi and Pinto, Lerrel},journal={arXiv preprint arXiv:2403.03181},year={2024}}

• HIL-SERL

@Article{luo2024hilserl,title={Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learning},author={Jianlan Luo and Charles Xu and Jeffrey Wu and Sergey Levine},year={2024},eprint={2410.21845},archivePrefix={arXiv},primaryClass={cs.RO}}