Tensorflow container for Jetson Nano

Hi, I’ve done some searching, seems that there is yet to have a tenforflow container for Jetson platform. Please advice:

  1. would there be one? I am not prying for secret here, only ask because my project depends on it.

  2. if there would not be an official tensorflow for jetson container, can we have some instruction on how to build one from scratch for Jetson platform?

  3. I understand that the CUDA container is on the pipeline, but it is unlikely that my project can depend on the availability of CUDA container. According to stackoverflow, the way to access CUDA from a tensorflow container, other than nvidia-docker is the following:

$ sudo docker run -ti --device /dev/nvidia0:/dev/nvidia0 --device /dev/nvidiactl:/dev/nvidiactl --device /dev/nvidia-uvm:/dev/nvidia-uvm tensorflow/tensorflow:latest-gpu

I can’t varify this since I only run tensorflow docker on nvidia-docker, and there is no CUDA on my computer. Please advice if that is a valid way to run tensorflow-gpu without nvidia-docker.

resources:

https://devtalk.nvidia.com/default/topic/1043462/jetson-agx-xavier/nvidia-docker-for-jetson-xavier-/

Hi,

nvidia-docker is going to support Jetson in the future JetPack release.

Currently, you can install the TensorFlow on the Jetson directly.
[url]https://devtalk.nvidia.com/default/topic/1048776/jetson-nano/official-tensorflow-for-jetson-nano-/[/url]

And try to access the libraries inside the official docker.
Here are some sample for your reference: [url]https://github.com/Technica-Corporation/Tegra-Docker[/url]

Thanks…

Hello,
when the next JetPack release will be?
Regards

Any update on this? Can we run TensorFlow in a docker container?
I have an application which runs on a container and uses tensor flow functions.

Hi,

We have released an L4T docker container based on the JetPack4.2.1:
[url]https://github.com/NVIDIA/nvidia-docker/wiki/NVIDIA-Container-Runtime-on-Jetson[/url]

We only preinstalled the CUDA toolkit.
You can install all the requirements and packages from the JetPack download folder.

Thanks.

See this instruction for current (very easy) steps to get Docker and Kubernetes running on Jetson Nano with GPU enabled:

(I’m one of the authors of this article.)

Great! Thanks for sharing!

As you said, only CUDA toolkit is installed in the container. I need Cudnn to build my app inside the container and the Cudnn librairies are not in the container.

Can you provide a link or some explanations about where is this Jetpack download folder ?

Hi julien.jehl,

Please help to open a new topic for the issue you met.

Thanks

Hi! I’m create same topic here. Can you answer please!