Episodes Preview SO-100 Visualizer
50 episodes · 30 fps · 2 cameras · 640×480 av1

SO-101 pick and place

50 teleoperated episodes of an SO-101 arm putting a pink LEGO brick into a transparent box (about 7 minutes at 30 fps, two cameras), in LeRobot v3.0 format. It's the SO-101 sample behind Try a sample when you train a model on Dreamscale.

Source and credit

This is an unchanged copy of lerobot/svla_so101_pickplace (revision f641879e22172be7e8161d5e6c1503c2d2feb657), recorded and published by Hugging Face's LeRobot team among the datasets released with SmolVLA. All credit for the data goes to them. The episodes, videos and metadata are theirs, byte for byte, and the commit history is kept; only this card is ours. It is shared under the original Apache-2.0 license.

If you use it, please cite the original work:

@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 Pascal, Caroline and Choghari, Jade and Meftah, Khalil and Ellerbach, Maxime 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}
}

@misc{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},
    year = {2025},
    eprint = {2506.01844},
    archivePrefix = {arXiv},
    primaryClass = {cs.RO},
    url = {https://arxiv.org/abs/2506.01844}
}

Use it

dreamscale train --dataset hf://dreamscale-labs/so101-pick-place --base pi0.5

Or open Training → Train a model in the dashboard and choose Try a sample.

Downloads last month
65

Paper for dreamscale-labs/so101-pick-place