Skip to content

Dataset · nvidia

Robot trajectories for GR00T N1

nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim

Sets of trajectories across various robots and tasks for post-training GR00T N1.

What's inside

The dataset contains trajectories from various robot embodiments and tasks, gathered for post-training GR00T N1. The card highlights the cross-embodied bimanual manipulation section: around 9,000 trajectories, including bimanual_panda_gripper.Threading, bimanual_panda_hand.LiftTray, and others. The data is presented in sets with names and trajectory counts.

What you can do with it

  • Fine-tune GR00T N1 or another robotic VLA model
  • Evaluate cross-embodied manipulation
  • Build a post-training pipeline for bimanual robots
Tasks
Robotics
Downloads this month
1.3M
Likes
295

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim")