Failed to launch tf_openpose when running Docker quick test on AGX Orin

[kevin-chau]

Hi, I'm trying to get the Docker quick test to work on an NVIDIA Jetson AGX Orin. I'm running into this error trying to open libcudart.so.10.2:

moca@ubuntu:~/Desktop/ext_storage$ sudo docker run -it --rm                 --net host                      koide3/monocular_person_following:jp45                 roslaunch monocular_person_following jetson_person_following.launch camera_name:=/top_front_camera/qhd
sourcing   /opt/ros/noetic/setup.bash
ROS_ROOT   /opt/ros/noetic/share/ros
ROS_DISTRO noetic
... logging to /root/.ros/log/c4b629fe-fb5c-11ee-9a78-48b02de4deef/roslaunch-ubuntu-1.log
Checking log directory for disk usage. This may take a while.
Press Ctrl-C to interrupt
Done checking log file disk usage. Usage is <1GB.

started roslaunch server http://ubuntu:37301/

SUMMARY
========

PARAMETERS
 * /monocular_people_tracking/association_maha_sq_thresh: 9.0
 * /monocular_people_tracking/association_neck_ankle_max_dist: 200
 * /monocular_people_tracking/association_neck_max_dist: 150
 * /monocular_people_tracking/detection_border_thresh_h: 25
 * /monocular_people_tracking/detection_border_thresh_w: 100
 * /monocular_people_tracking/detection_confidence_thresh: 0.1
 * /monocular_people_tracking/init_cov_scale: 0.25
 * /monocular_people_tracking/measurement_noise_pix_cov: 100
 * /monocular_people_tracking/process_noise_height_cov: 1e-10
 * /monocular_people_tracking/process_noise_pos_cov: 0.03
 * /monocular_people_tracking/process_noise_vel_cov: 0.01
 * /monocular_people_tracking/tracking_newtrack_dist2exists_thersh: 100
 * /monocular_people_tracking/tracking_remove_trace_thresh: 2.0
 * /monocular_person_following/id_switch_detection_thresh: -0.1
 * /monocular_person_following/imprinting_max_dist: 4.0
 * /monocular_person_following/initial_training_num_samples: 10
 * /monocular_person_following/min_target_confidence: 0.1
 * /monocular_person_following/reid_confidence_thresh: 0.1
 * /monocular_person_following/reid_positive_count: 5
 * /monocular_person_following/use_body: True
 * /monocular_person_following/use_face: False
 * /pose_estimator/allow_growth: False
 * /pose_estimator/camera: /top_front_camera...
 * /pose_estimator/model: mobilenet_thin
 * /pose_estimator/resolution: 656x368
 * /rosdistro: noetic
 * /rosversion: 1.15.11
 * /visualization_node/show: False
 * /visualization_node/use_face: False

NODES
  /
    compress_visualize (image_transport/republish)
    monocular_people_tracking (monocular_people_tracking/monocular_people_tracking_node)
    monocular_person_following (monocular_person_following/monocular_person_following_node)
    pose_estimator (tfpose_ros/broadcaster_ros.py)
    simple_gesture_recognition (monocular_person_following/simple_gesture_recognition.py)
    throttle_visualize (topic_tools/throttle)
    visualization_node (monocular_person_following/visualization.py)

ROS_MASTER_URI=http://localhost:11311

process[pose_estimator-1]: started with pid [75]
process[monocular_people_tracking-2]: started with pid [76]
process[monocular_person_following-3]: started with pid [77]
process[visualization_node-4]: started with pid [78]
process[compress_visualize-5]: started with pid [79]
process[throttle_visualize-6]: started with pid [80]
process[simple_gesture_recognition-7]: started with pid [81]
[ INFO] [1713209153.852244116]: construct body classifier
[ INFO] [1713209153.854402206]: add cnn10 to channel bank
--- simple_gesture_recognition ---
wait for service
done
2024-04-15 19:25:55.038606: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'libcudart.so.10.2'; dlerror: libcudart.so.10.2: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /root/catkin_ws/devel/lib:/opt/ros/noetic/lib:/usr/local/cuda-10.2/targets/aarch64-linux/lib
2024-04-15 19:25:55.038613: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'libcudart.so.10.2'; dlerror: libcudart.so.10.2: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /root/catkin_ws/devel/lib:/opt/ros/noetic/lib:/usr/local/cuda-10.2/targets/aarch64-linux/lib
2024-04-15 19:25:55.038693: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2024-04-15 19:25:55.038695: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2024-04-15 19:25:55.038903: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'libcudart.so.10.2'; dlerror: libcudart.so.10.2: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /root/catkin_ws/devel/lib:/opt/ros/noetic/lib:/usr/local/cuda-10.2/targets/aarch64-linux/lib
2024-04-15 19:25:55.038934: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2024-04-15 19:25:55.038943: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'libcudart.so.10.2'; dlerror: libcudart.so.10.2: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /root/catkin_ws/devel/lib:/opt/ros/noetic/lib:/usr/local/cuda-10.2/targets/aarch64-linux/lib
2024-04-15 19:25:55.038977: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
================================================================================REQUIRED process [pose_estimator-1] has died!
process has died [pid 75, exit code -11, cmd /root/catkin_ws/src/tf-pose-estimation/scripts/broadcaster_ros.py __name:=pose_estimator __log:=/root/.ros/log/c4b629fe-fb5c-11ee-9a78-48b02de4deef/pose_estimator-1.log].
log file: /root/.ros/log/c4b629fe-fb5c-11ee-9a78-48b02de4deef/pose_estimator-1*.log
Initiating shutdown!
================================================================================
[simple_gesture_recognition-7] killing on exit
[throttle_visualize-6] killing on exit
[compress_visualize-5] killing on exit
[visualization_node-4] killing on exit
[monocular_person_following-3] killing on exit
[monocular_people_tracking-2] killing on exit
[pose_estimator-1] killing on exit
shutting down processing monitor...
... shutting down processing monitor complete
done

[kevin-chau]

When trying to run with GPU I get this error:

moca@ubuntu:~/Desktop/ext_storage$ sudo docker run -it --rm                 --net host                 --gpus all                 koide3/monocular_person_following:jp45                 roslaunch monocular_person_following jetson_person_following.launch camera_name:=/top_front_camera/qhd
docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'csv'
invoking the NVIDIA Container Runtime Hook directly (e.g. specifying the docker --gpus flag) is not supported. Please use the NVIDIA Container Runtime instead.: unknown.