Need Deepstream-Python docker containers for dGPU and Jetson

Hi,

Does Nvidia provide any docker containers for dGPUs and Jetson devices for Deepstream with Python?

Thanks,
Archana

You can get python package repo in docker container. They are not conflict.

Hi @ChrisDing,

Thanks for the reply. I have installed the DeepStream container on Nano and will try to setup Python-app repo there. I was curious to know, if Nvidia NGC have such ready DeepStream containers for Python.

@ChrisDing, I have tried to run Jetson l4t deepstream docker image in x86_64 architecture with referring this link. But when I downloaded docker image, cuda samples were missing. Also there were few more libs and files were missing inside /usr/lib/aarch64-linux-gpu, like tegra. I have a Jetson Nano, where all these files are present.

Can you please help me to setup https://ngc.nvidia.com/catalog/containers/nvidia:deepstream-l4t docker image in my Ubuntu desktop with x86_64?

Hi Archana,
It’s not feasiable to run L4T docker on x86 platform since L4T docker is for Jetson/aarch64 platform.

You could refer to Running the Deepstream Python apps in nvcr.io/nvidia/deepstream:4.0.2-19.12-devel container, headles to create your own x86 docker based on NGC x86 DeepStream docker.

Hi @mchi,

I tried to create the docker as mentioned in the post, it failed at step 2:

Step 2/14 : ADD python_release_bind /root/deepstream_sdk_v4.0.2_x86_64/sources/python
ADD failed: stat /var/lib/docker/tmp/docker-builder396232019/python_release_bind: no such file or directory

Any idea, why this has failed?

In the same folder of the dockerfile, you need to download the DeepStream python - https://github.com/NVIDIA-AI-IOT/deepstream_python_apps and name it as python_release_bind

Hi @mchi,

The custom docker with nvcr.io/nvidia/deepstream:4.0.2-19.12-devel works fine and I’m able to test at least first 3 examples. Thank you for the reference post link.
But somehow the other x86 based docker like nvcr.io/nvidia/deepstream:4.0.2-19.12-samples is not working for me. I have opened other issue and posting my comments there since last 15 days - Deepstream-app: error while loading shared libraries: libnvinfer.so.6: cannot open shared object file + ubuntu 18.04

Anyways, thanks for the solution from Running the Deepstream Python apps in nvcr.io/nvidia/deepstream:4.0.2-19.12-devel container, headles

Any plans for L4T docker on x86 platform?

Thanks,
Archana

Hi Archana,
Glad to know you get it running.

Deepstream-app: error while loading shared libraries: libnvinfer.so.6: cannot open shared object file + ubuntu 18.04
Will you still want to track the issue?

Any plans for L4T docker on x86 platform?
We will still provides multiple dockers with DS5.0 release for x86 and Jetson.

Thanks!

1 Like

@mchi,

Will you still want to track the issue?

I faced issue for DeepStream installation with separate prerequisites installations and versions. Somehow, I can’t switch to Cuda 10.1 from Cuda 10.2 due to other project dependencies. Then I found DeepStream docker and I just wonder why it’s is not working in docker provided for X86 platform, which has all the required versions. I spent lot of time on it.

We will still provides multiple dockers with DS5.0 release for x86 and Jetson.
Any road-map for future releases?

Thanks,
Archana

No doubt, DeepStream keeps adopting more use cases and requirements from users and improving itself with more releases.
Very soon, we will release DS5.0. For future release after DS5.0, its POR is still under discussion.

Any more issues about this topic?

Thanks!

1 Like

@mchi, Thanks for the details. No issues.