A web-based GUI for editing and merging LeRobot datasets.
Note: This is an unofficial tool and is not affiliated with Huggingface, LeRobot or the LeRobot team.
- Merge multiple datasets into a single dataset
- Easily remove episodes and create new clean datasets
- Speed control and keyboard shortcuts to streamline dataset cleaning
- UV Python package and project manager:
curl -LsSf https://astral.sh/uv/install.sh | sh
- Python 3.10+ (You can use uv to install and manage python versions e.g.
uv python install 3.12
) - Node.js 24+ (Install using nvm - see instructions below)
- A Huggingface account (free)
- Huggingface CLI
# Install nvm (Node Version Manager)
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.3/install.sh | bash
# Reload your shell configuration
source ~/.bashrc # or ~/.zshrc if using zsh
# Install and use Node.js 24
nvm install 24
nvm use 24
# Verify installation
node --version
git clone https://github.com/jackvial/lerobot-data-studio
cd lerobot-data-studio
# Create a virtual environment with UV
uv venv
# Activate the virtual environment
source .venv/bin/activate
# Install all packages using UV
uv sync
After completing the python installation, install frontend dependencies:
cd packages/lerobot-data-studio/frontend
npm install
Use the provided script to start both frontend and backend servers:
./run_dev.sh
Dataset creation for merged and filtered (AKA edited) datasets is always none destructive and will always create a new dataset and upload it to the Huggingface Hub.
Merges multiple datasets into a single new dataset and uploads it to the Huggingface Hub.
Editing/filtering a dataset creates a new dataset that only excludes the episodes that were selected in the UI.
uv run pytest
cd src/lerobot_data_studio/frontend
npm run test
Contributions are welcome!
This project is licensed under the MIT License - see the LICENSE file for details.