Thank you for getting these containers to build. I’ll give the ROS Noetic + DeepStream container a try with my code…
Would you send / upload the Dockerfiles also? I have some additional dependencies (DeepStream python bindings) that I need to install for my purposes. Specifically, I had to add (to the dustynv/ros:humble-ros-base-deepstream-l4t-r35.1.0 container):
mkdir -p ~/ros2_ws/srcGitHub - NVIDIA-AI-IOT/ros2_deepstream: ROS 2 package for NVIDIA DeepStream applications on Jetson Platforms GitHub - NVIDIA-AI-IOT/deepstream_python_apps: DeepStream SDK Python bindings and sample applications
pip3 install pyds-1.1.4-py3-none-linux_aarch64.whl
Then, to get the ros2_deepstream example to work:
I ran into an additional snag when I try to run my own code in your dustynv/ros:noetic-ros-base-deepstream-l4t-r35.1.0 container…
I also need to install the custom bounding box parser from here:
I’ve successfully installed and used this parser in another container (built by a coworker - the Dockerfile is attached), which uses nvcr.io/nvidia/deepstream-l4t:6.1.1-triton as its base.
However, when I try to install into dustynv/ros:noetic-ros-base-deepstream-l4t-r35.1.0 I get this error:
root@ubuntu:/opt/DeepStream-Yolo# CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yolo#include “NvInfer.h”
I believe I’ve seen this error before, but am not sure what the solution is. Any idea on this?Dockerfile (3.0 KB)
It looks like the deepstream-l4t container doesn’t have the TensorRT development headers installed - can you install libnvinfer-dev in the container from apt?
BTW this is my modified Dockerfile.ros.noetic (with only that one -DSETUPTOOLS_DEB_LAYOUT=OFF change and the BASE_IMAGE change)
#
# this dockerfile roughly follows the 'Installing from source' from:
# http://wiki.ros.org/noetic/Installation/Source
#
ARG BASE_IMAGE=nvcr.io/nvidia/deepstream-l4t:6.1.1-samples
FROM ${BASE_IMAGE}
ARG ROS_PKG=ros_base
ENV ROS_DISTRO=noetic
ENV ROS_ROOT=/opt/ros/${ROS_DISTRO}
ENV ROS_PYTHON_VERSION=3
ENV DEBIAN_FRONTEND=noninteractive
WORKDIR /workspace
#
# add the ROS deb repo to the apt sources list
#
RUN apt-get update && \
apt-get install -y --no-install-recommends \
git \
cmake \
build-essential \
curl \
wget \
gnupg2 \
lsb-release \
ca-certificates \
&& rm -rf /var/lib/apt/lists/*
RUN sh -c 'echo "deb http://packages.ros.org/ros/ubuntu $(lsb_release -sc) main" > /etc/apt/sources.list.d/ros-latest.list'
RUN curl -s https://raw.githubusercontent.com/ros/rosdistro/master/ros.asc | apt-key add -
#
# install bootstrap dependencies
#
RUN apt-get update && \
apt-get install -y --no-install-recommends \
libpython3-dev \
python3-rosdep \
python3-rosinstall-generator \
python3-vcstool \
build-essential && \
rosdep init && \
rosdep update && \
rm -rf /var/lib/apt/lists/*
#
# download/build the ROS source
#
RUN mkdir ros_catkin_ws && \
cd ros_catkin_ws && \
rosinstall_generator ${ROS_PKG} vision_msgs --rosdistro ${ROS_DISTRO} --deps --tar > ${ROS_DISTRO}-${ROS_PKG}.rosinstall && \
mkdir src && \
vcs import --input ${ROS_DISTRO}-${ROS_PKG}.rosinstall ./src && \
apt-get update && \
rosdep install --from-paths ./src --ignore-packages-from-source --rosdistro ${ROS_DISTRO} --skip-keys python3-pykdl -y && \
python3 ./src/catkin/bin/catkin_make_isolated --install --install-space ${ROS_ROOT} -DCMAKE_BUILD_TYPE=Release -DSETUPTOOLS_DEB_LAYOUT=OFF && \
rm -rf /var/lib/apt/lists/*
#
# setup entrypoint
#
COPY ./packages/ros_entrypoint.sh /ros_entrypoint.sh
RUN echo 'source /opt/ros/${ROS_DISTRO}/setup.bash' >> /root/.bashrc
ENTRYPOINT ["/ros_entrypoint.sh"]
CMD ["bash"]
WORKDIR /
OK, did apt install libnvinfer-dev, which looked to run through fine.
However, now when I run CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yolo I get this error:
root@ubuntu:/opt/DeepStream-Yolo# CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yoloinclude
I found your thread here, which mentions this issue:
With your help, I went through my source code carefully. I found that part of the TensorRT sample common file was used in my source code, which led me to mistakenly think that the container did not contain CUDA and TensorRT packages. I did the following:
docker run -it -v MY_WORKSPACE:/home dustynv/ros:galactic-ros-base-l4t-r34.1.1 /bin/bash
docker cp /usr/src/tensorrt/samples/common MY_CONTAINER_ID:/usr/src/tensorrt/samples/
When I compiled using colcon toolkit, several outputs like the foll…
But when I search for “cuda_runtime_api.h” it’s not present in the container:
root@ubuntu:/usr/local# ls -la
So I tried adding the command from further up in the current thread, which installs the full CUDA Toolkit:
RUN apt-get update &&
This runs through fine, but now the CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yolo command causes this error:
Step 42/47 : RUN cd DeepStream-Yolo && CUDA_VER=11.4 make -C nvdsinfer_custom_impl_Yoloinclude
Attempting to use nvcr.io/nvidia/deepstream-l4t:6.1.1-triton as the base image to solve these issues…
Cool, thanks for posting the Dockerfile.
Hi @coreyslick , try installing the libnvinfer-plugin-dev package as well.
Dockerfile.ros.noetic (4.3 KB)
It ended up working using nvcr.io/nvidia/deepstream-l4t:6.1.1-triton as the base image. The Dockerfile is attached.
I really appreciate all of your work getting these containers to build - thank you!
No problem at all @coreyslick , glad that you were able to get it working!
system
October 19, 2022, 3:17pm
32
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.