Error building Jetson Utils as standalone using the new Tensorrt container for cloud native.

[SamORichards]

I have been trying to use the new tensorrt container from Nvidia as the base image for building jetson utils as a standalone but keep running into this error:

CMake Error: The following variables are used in this project, but they are set to NOTFOUND. Please set them or make sure they are set and tested correctly in the CMake files: CUDA_nppicc_LIBRARY (ADVANCED)

Everything I have found online about this points to updating the FindCuda module but in the repo you seem to have already done this but this issue still persists.

Any suggestions?

`ARG BASE_IMAGE=nvcr.io/nvidia/l4t-tensorrt:r8.0.1-runtime FROM ${BASE_IMAGE}

ENV DEBIAN_FRONTEND=noninteractive ENV SHELL /bin/bash

WORKDIR jetson-utils

COPY f f WORKDIR f

install development packages

RUN apt update &&
apt install -y --no-install-recommends
build-essential cmake
nano
&& rm -rf /var/lib/apt/lists/*

# pip dependencies for pytorch-ssd

RUN pip3 install --verbose --upgrade Cython &&
pip3 install --verbose boto3 pandas

# alias python3 -> python

RUN rm /usr/bin/python &&
ln -s /usr/bin/python3 /usr/bin/python &&
ln -s /usr/bin/pip3 /usr/bin/pip

RUN mkdir build && cd build && cmake .. && make

LABEL Name=jetsonutils Version=0.0.1`

[dusty-nv]

Hi @SamORichards, the container would also need NVIDIA Performance Primitives (NPP) libraries/headers installed into it. Normally these come along with Jetpack/CUDA and are installed on the device and available to use via l4t-base container.

I think you would run into further issues trying to build this project in l4t-tensorrt, because l4t-tensorrt is a runtime container and doesn't have the devel packages (i.e. headers) which this project needs in order to compile.