Machine Learning Containers for Jetson

Hi All,

We’ve released the following framework containers for Jetson and JetPack 4.4 Developer Preview on NVIDIA GPU Cloud (NGC):

  • TensorFlow Container (l4t-tensorflow) - contains TensorFlow pre-installed in a Python 3.6 environment to get up & running quickly with TensorFlow on Jetson.

  • PyTorch Container (l4t-pytorch) - contains PyTorch and torchvision pre-installed in a Python 3.6 environment to get up & running quickly with PyTorch on Jetson.

  • Machine Learning Container (l4t-ml) - contains TensorFlow, PyTorch, JupyterLab, and other popular ML and data science frameworks such as scikit-learn, scipy, and Pandas pre-installed in a Python 3.6 environment.

You can also find the Dockerfiles and build scripts on GitHub - have fun!


the TensorFlow container contains TensorFlow 1.15. How about TF 2.x?

Do you… Waiting tensorflower for releasing TF 2.2 stable?

Hi @smankusors, for now it contains TF 1.15, at a later time we will push another tag for TF 2.1 after some additional testing. If you want TF 2.1, for now you can find the TF 2.1 wheels for JetPack 4.4 here and the Dockerfiles that you can modify on GitHub.

Thanks for releasing those containers! Appreciated!

A question though (my inexperience/unfamiliarity is speaking now):

  1. Do those containers run on containerd that is used by kubernetes and Docker or is it specifically Docker (API/CLI) only?

I just realized (or never understood) the exact layering of the various underlying components (source: versus the exact ins/out of the Jetpack (Jetson Nano/NX/AGX) “container” support.

Hi @remko.lems, my testing/development of these containers was using the Docker daemon (i.e. docker build ..., docker run ... commands). I’m not personally familiar with containerd to know if it’s used/compatible or not, although I do know that several folks have built Kubernetes pods with Jetson.