LeRobot v2.1 dataset for the Unitree G1 right-hand bottle-grab task. Merges two
teleoperation datasets and applies Dynamic Programming (DP) resampling on the
right-wrist XYZ trajectory to normalise the action density across all episodes. It is
the training data for the
v7 GR00T N1.7 fine-tune
a speedup-resampling reference baseline (trained 2026-06-25, checkpoint-20000
only).
