Problems with the instalaltion of the trt_pose and torch2trt in the dcoker container

[A98]

Hello.I couldnt install trt_pose and torch2trt in the docker container and i need help.
Sequence of my actions:

Attempt1:

1.sudo docker run -it 98aa2f10ae96
2.git clone GitHub - NVIDIA-AI-IOT/torch2trt: An easy to use PyTorch to TensorRT converter
3.cd torch2trt
4.python3 setup.py install

Error:
Traceback (most recent call last):
File “setup.py”, line 3, in
import torch
File “/usr/local/lib/python3.6/dist-packages/torch/init.py”, line 196, in
_load_global_deps()
File “/usr/local/lib/python3.6/dist-packages/torch/init.py”, line 149, in _load_global_deps
ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
File “/usr/lib/python3.6/ctypes/init.py”, line 348, in init
self._handle = _dlopen(self._name, mode)
OSError: libcurand.so.10: cannot open shared object file: No such file or directory
Attempt1.txt (3.4 KB)

Attempt2:

1.cd /home/atlas/jetson-inference
2.docker/run.sh -c face_recognition_only
3.git clone GitHub - NVIDIA-AI-IOT/torch2trt: An easy to use PyTorch to TensorRT converter
4.cd torch2trt
5.python3 setup.py install
6.exit
7.exit
8.docker/run.sh -c face_recognition_only
9.git clone GitHub - NVIDIA-AI-IOT/trt_pose: Real-time pose estimation accelerated with NVIDIA TensorRT
10. cd trt_pose
11.python3 setup.py install
12.exit
13.exit
14.sudo docker ps -a
15.sudo docker images
16.sudo docker commit 185bb86ae284 pose_detection_and_face_detection
17. docker/run.sh -c pose_detection_and_face_detection --volume /home/atlas/Desktop/Jupiter1:/home/atlas/jetson-inference
18.cd /home/atlas/jetson-inference
19.python3 pose_need.py

Error:

ModuleNotFoundError: No module named ‘trt_pose’
ModuleNotFoundError: No module named ‘torch2trt’

Attempt2.txt (60.7 KB)

Full log is here:
log (2).txt (64.1 KB)

What did i do wrong and how to install torch2trt and trt_pose in the docker container ?

[dusty_nv]

@A98

1.sudo docker run -it 98aa2f10ae96
OSError: libcurand.so.10: cannot open shared object file: No such file or directory

Hi @A98, can you try starting the container with --runtime nvidia ? That is needed on JetPack 4 so that the CUDA/cuDNN/TensorRT libraries automatically get mounted into the container from the device. You can also make your own Dockerfile with your steps so that it is built in an automated way.

[A98]

Thanks for your reply. --runtime nvidia solved the problem with the torch2trt installation.