slobot/aic-act-nahid-v2

ACT policy trained with LeRobot on the slobot/aic dataset.

Training setup

• Dataset: slobot/aic
• Dataset revision: main
• Policy: act
• Cameras: observation.images.center_camera, observation.images.left_camera, observation.images.right_camera
• Observation state: concatenation of unnormalized observation.task_id.* and normalized observation.tcp_offset.*
• Action target: normalized action.tcp_offset
• Training steps: 100000

Files

This repo contains the final exported LeRobot policy bundle from checkpoint 100000.