Fifty real pick-and-place episodes from a low-cost SO-100/SO-101-class arm, recorded at 30 fps with a side camera and a top-down camera, and used as the SmolVLA fine-tuning example.
lerobot/pusht
Indexed from Hugging FaceLinked to its home on Hugging Face. FrontierOS indexes it; nothing is hosted or mirrored here.About
About 200 human demonstrations in a 2D simulation where a round pusher slides a T-shaped block onto a target outline, stored in LeRobot v3 format.
Why it matters for physical AI
It is the usual first dataset for checking that a diffusion-policy training loop works end to end before anyone touches a real robot.
Summary written by FrontierOS in our own words. The upstream card and license govern use. Metadata checked 27 Sep 2026. Read the full card on Hugging Face
Files, weights and version history live on Hugging Face. FrontierOS indexes this repo; it does not host or mirror it.
hf download lerobot/pusht --repo-type datasetUses the Hugging Face CLI (pip install -U huggingface_hub). This downloads the whole repo; add --include "<pattern>" to fetch only the files you need.
Check your teleop data before you train
Step through a few episodes in the LeRobot dataset visualizer before spending GPU hours.
More robot demonstrations
Browse allA filtered collection of 128 SO-100 datasets from 55 community contributors (about 11k episodes and 5.1M frames), with hand-cleaned task descriptions, released as the pretraining subset for SmolVLA.
Fifty human-teleoperated simulated episodes of a two-arm ALOHA robot handing a cube from one gripper to the other, recorded at 50 fps.