granola-bar-case-pack-test-2

LeRobotDataset v3.0 — robot_type: anvil_yam (OpenYAM bimanual, Quest teleop), 60 fps, 18 episodes / 53,830 frames. Task: "pack the granola bars into the box". Captured 2026-10-09.

• Cameras: chest, wrist_l, wrist_r (480×270, h264)
• State/action: 14-D (left gripper + 6 joints, right gripper + 6 joints); also observation.velocity, observation.effort
• Converted raw episodes 0008–0019, 0021–0026. Excluded: 0001, 0003–0007 (wrist cams out of focus), 0020 (4 s, no motion).
• Converter: anvil-embodied-ai @ 5d1c8eb (conversion_config.yaml included)

Train: uv run anvil-trainer --dataset.root=<path> --policy.type=act