VPI installation error In docker container

Autonomous Machines Jetson & Embedded Systems Jetson Nano
#1

I want to install VPI pack in the docker container.Before this,I installed
opencv:
OpenCV-4.1.1-2-gd5a58aa75-aarch64-libs.deb
OpenCV-4.1.1-2-gd5a58aa75-aarch64-dev.deb
cuda:
cuda-repo-l4t-10-2-local-10.2.89_1.0-1_arm64.deb
cudnn:
libcudnn8_8.0.0.180-1+cuda10.2_arm64.deb
finally,install the vpi:
vpi-lib-0.4.4-cuda10-aarch64-l4t.deb

Error occur:

Blockquote
root@6c0cbe58163b:~/host_sdk# dpkg -i vpi-lib-0.4.4-cuda10-aarch64-l4t.deb
(Reading database … 35607 files and directories currently installed.)
Preparing to unpack vpi-lib-0.4.4-cuda10-aarch64-l4t.deb …
Unpacking vpi (0.4.4) over (0.4.4) …
dpkg: dependency problems prevent configuration of vpi:
vpi depends on cuda-cufft-10-2 | libcufft-10-2 | cuda-cufft-10-5 | libcufft-10-5 | cuda-cufft-10-4 | libcufft-10-4 | cuda-cufft-10-3 | libcufft-10-3 | cuda-cufft-10-1 | libcufft-10-1 | cuda-cufft-10-0 | libcufft-10-0; however:
Package cuda-cufft-10-2 is not installed.
Package libcufft-10-2 is not installed.
Package cuda-cufft-10-5 is not installed.
Package libcufft-10-5 is not installed.
Package cuda-cufft-10-4 is not installed.
Package libcufft-10-4 is not installed.
Package cuda-cufft-10-3 is not installed.
Package libcufft-10-3 is not installed.
Package cuda-cufft-10-1 is not installed.
Package libcufft-10-1 is not installed.
Package cuda-cufft-10-0 is not installed.
Package libcufft-10-0 is not installed.
vpi depends on cuda-cudart-10-2 | cuda-cudart-10-3 | cuda-cudart-10-1 | cuda-cudart-10-0; however:
Package cuda-cudart-10-2 is not installed.
Package cuda-cudart-10-3 is not installed.
Package cuda-cudart-10-1 is not installed.
Package cuda-cudart-10-0 is not installed.
dpkg: error processing package vpi (–install):
dependency problems - leaving unconfigured
Errors were encountered while processing:
vpi

#3

Hi,

Please noted that our L4T base image doesn’t contain CUDA toolkit.
It mounts the library from the Jetson folder to save the image size.

You will need to copy the CUDA library into the container first if a package requires it in the building time.

Thanks.

#4

Yes,The installation process in the container running status,and I install the CUDA from shell,

#6

Hi,

An alternative is to add the VPI csv into the /etc/nvidia-container-runtime/host-files-for-container.d.
Currently, we don’t have a .csv file for VPI library.

You can create one so you don’t need to install it manually at the running time.

Thanks.