I have successfully trained and tested a pointpillars network on my own point cloud dataset using https://github.com/shangjie-li/pointpillars/tree/master. I modified demo_in_ros.py from that repository to apply the trained model for 3D object detection using ROS2 (Humble). I used https://answers.ros.org/question/367793/including-a-python-module-in-a-ros2-package/ and was able to install all python codes (located in different folders) as modules in ROS2 workspace (ros2_pointpillars) /ros2_pointpillars/install/lidar_object_detection/lib/python3.10/site-packages/lidar_object_detection. I need to use iou3d_nms_cuda.cpython-37m-x86_64-linux-gnu.so which is located in /ops/iou3d_nms folder. How can I import/install it as a module in ROS2?

I have successfully trained and tested a pointpillars network on my own point cloud dataset using https://github.com/shangjie-li/pointpillars/tree/master. I modified demo_in_ros.py from that repository to apply the trained model for 3D object detection using ROS2 (Humble). I used https://answers.ros.org/question/367793/including-a-python-module-in-a-ros2-package/ and was able to install all python codes (located in different folders) as modules in ROS2 workspace (ros2_pointpillars) /ros2_pointpillars/install/lidar_object_detection/lib/python3.10/site-packages/lidar_object_detection. I need to use iou3d_nms_cuda which is located in /ops/iou3d_nms folder. How can I import/install it as a module in ROS2?