LeRobot xvla-base (robot policy)
LeRobot is Hugging Face’s growing toolbox for “robot learning” workflows: datasets of trajectories, standardized policy training, and repeatable evaluation runs. lerobot/xvla-base is a lightweight baseline repo that uses the xvla policy type and is mainly useful as a starting point and reference for how to wire LeRobot’s CLI.
The model card is short, but the value is in the “getting started” commands: lerobot-train to train a policy from a Hub dataset and push checkpoints back to the Hub, and lerobot-record to run inference/evaluation episodes (with an example robot type). If you’re trying LeRobot for the first time, this kind of minimal “known-good” repo is handy because it gives you a concrete target to diff against when your own policy configs or dataset wiring don’t behave the way you expect.
Quick stats from the listing feed: pipeline: robotics · 16 likes · 2110 downloads.
Source listing: https://huggingface.co/models?sort=modified