TensorRT L4T docker image Python version Issue

AI & Data Science Deep Learning (Training & Inference) TensorRT
23

Thanks for your reply @ework, I had 2 further questions:

Q1: The outdated Dockerfile’s provided on nvidia/container-images/l4t-base are quite simple, I genuinely wonder if there’s more to it than just that Dockerfile- why does it take so long - planning since early 2021?

Q2: May I also ask why this repo is not up to date. According to the L4T base NGC page, this seems like a commitment to open source the Dockerfiles - see below quote. Could you please pass feedback to the team, that this repo hasn’t been updated.

Starting with the r32.4.3 release, the Dockerfile for the l4t-base docker image is also being provided. This can be accessed at this link. Users can use this to modify the contents to suit their need.

Thank you :)

24

This might be an answer to both of your questions. To be honest I don’t know why that repo is so outdated or why what should be as simple as updating a Dockerfile is not being worked on sooner. There are QA testing requirements and it’s possible some of the dependent Python modules don’t work as expected. I know we’re planning to move away from the approach where libraries from the host are overlaid into the container to instead using standalone containers. People working in this areas are also quite busy right now with Orin related work. I’m more familiar with the packaging and distribution of our deep learning libraries than the Jetpack distribution.

1 Like