typoverflow/libero_plus_goal

libero_plus_goal: detailed LeRobot v3.0 This dataset was converted from the LIBERO Plus LeRobot v2.1 libero_plus_goal partition. The original 8D state and 7D action vectors are preserved exactly as raw_state.ref_state and raw_action.ref_action. Canonical low-dimensional fields follow failure_rollout_data/dataset.md; debug.gripper_eef_* contains the ground-truth next-step relative EEF motion for inspection. Required camera transform for canonical training The… See the full description on the dataset page: https://huggingface.co/datasets/typoverflow/libero_plus_goal.