Deepstream-l4t devel docker image

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide development (devel) docker images on NGC for deepstream-l4t[1], analogous to those available on x86-64’s deepstream[2].

[1] DeepStream-l4t | NVIDIA NGC
[2] DeepStream | NVIDIA NGC

3

Why did you need the development docker for Jetson? What is the reason that you can not install DeepStream directly in the board?

4

I was wanting to deploy pycuda to the image, which needed additional development dependencies for pip install to succeed. I was able to just do a multi-stage build using the triton image which had the needed depencencies to get the wheel built and then copy / install it in a base image:

ARG DS_DIST=deepstream-l4t
ARG DS_DEVEL=triton
ARG DS_VER=6.2
FROM nvcr.io/nvidia/${DS_DIST}:${DS_VER}-${DS_DEVEL} as devel
ARG PYCUDA_VER=2022.2.2
RUN apt-get update && \
    DEBIAN_FRONTEND=noninteractive apt-get install -y python3-dev && \
    apt-get clean
RUN install -d /root/pycuda && cd /root/pycuda && python -m pip wheel "pycuda~=${PYCUDA_VER}"
FROM nvcr.io/nvidia/${DS_DIST}:${DS_VER}-base
COPY --from=devel /root/pycuda/ /tmp/pycuda/
RUN pip install --user --no-index --find-links /tmp/pycuda pycuda && \
    rm -rf /tmp/pycuda

Worked like a charm and had a smaller image at the end.

5

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

The nvcr.io/nvidia/deepstream-l4t:6.2-triton docker contains all DeepStream development dependencies.

7

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.