Import a dataset from Hugging Face

Bring a LeRobot dataset from the Hugging Face Hub onto this hub in one command, and read its episodes.

Prompt for your AI

Read https://mousemouse.ai/docs/import-from-hugging-face.md first. Import a LeRobot dataset from Hugging Face onto the hub for me, and tell me what the plan says before you confirm it. Ask me before anything that costs money or moves a robot.

A LeRobot dataset on the Hugging Face Hub comes onto this hub in one command. The command runs on your computer: the files go from Hugging Face to your machine, then to the hub. Once it lands, the dataset's Episodes tab plays each episode against its per-joint charts.

You need the lucen CLI and a key with the write scope. Train on your own machine has the install and sign-in lines.

1 · Get the command

Open Upload a dataset. Under Or import from Hugging Face, type the dataset's id as it appears on Hugging Face, such as lerobot/pusht. Add @ and a branch, tag or commit to import a revision. The card writes the command for you to copy.

2 · Run it

lucen datasets import hf://lerobot/pusht

It reads the dataset's description on Hugging Face first. Then it prints a plan: the revision, the episodes, the cameras, the size and the licence. Answer y to import it.

You'll see:

Import hf://lerobot/pusht @ 7628202a2180 (main) → you/pusht (private)
  LeRobotDataset v3.0 · 206 episodes · 25,650 frames · 10 fps · 1 camera (observation.image) · robot_type unknown
  8 file(s), 7.7 MB on the Hub · licence mit · README.md: the dataset's own
  Files go from Hugging Face to /Users/you/.cache/lucen/hf/lerobot/pusht/7628202a2180972f291ba1bc6723834921e72c19, then to the hub; bytes the hub already holds are not uploaded again.
Import it? [Y/n]: y
Created and imported hf://lerobot/pusht → you/pusht (private): 8 file(s), 8 uploaded (7.7 MB)
  LeRobotDataset v3.0 · 206 episodes · 25,650 frames · 10 fps · 1 camera (observation.image) · robot_type unknown; the hub reads its episodes.
  description set to the provenance.
https://mousemouse.ai/you/pusht?tab=episodes

The files reach your machine through Hugging Face's own client, so an interrupted download resumes where it stopped. Bytes the hub already holds are not uploaded again: run it twice and the second run uploads nothing. The new repo is private, and its description records where it came from.

Options, a private dataset, a script

--to you/other-name picks the repo, and --revision a branch or commit.
--public publishes it once the hub has read its episodes back.
--yes skips the question in a script, and --json prints the plan and the result as one object.
A dataset that needs access on Hugging Face opens with your HF_TOKEN.
Nothing runs on the hub itself. The import is your computer's job.

3 · Check it landed

lucen datasets ls --owner you

repo       kind     visibility  updated     description
you/pusht  dataset  private     2026-10-09  Imported from https://huggingface.co/datasets/ler…

Open the link the import printed. The Episodes tab lists the dataset's episodes, each with its video and joint charts.

What it refuses

The hub reads LeRobotDataset v2.0, v2.1 and v3.0. Anything else is refused before a byte moves, with the reason. Dataset format lists what the hub reads and what it leaves out.