Container for Pytorch >= 2.5.0

[esteban.gallardo]

I’m developing a project that was running Flux for image generation. So far I was working fine with the current Jetson-Container Pytorch 2.3.0, but now the image generation model requires from Pytorch 2.5.0, so I cannot longer build their repository with the previous container.

I’ve tried to build it by myself unsuccessfully with:
https://developer.download.nvidia.com/compute/redist/jp/v61/pytorch/torch-2.5.0a0+872d972e41.nv24.08.17622132-cp310-cp310-linux_aarch64.whl

Are there any plans to upgrade the Jetson-Containers to offer docker with Pytorch >= 2.5.0 version support?

[carolyuu]

Hi,
Here are some suggestions for the common issues:

1. Performance

Please run the below command before benchmarking deep learning use case:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

2. Installation

Installation guide of deep learning frameworks on Jetson:

• TensorFlow: Installing TensorFlow for Jetson Platform - NVIDIA Docs
• PyTorch: Installing PyTorch for Jetson Platform - NVIDIA Docs
  We also have containers that have frameworks preinstalled:
  Data Science, Machine Learning, AI, HPC Containers | NVIDIA NGC

3. Tutorial

Startup deep learning tutorial:

• Jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson
• TensorRT sample: Jetson/L4T/TRT Customized Example - eLinux.org

4. Report issue

If these suggestions don’t help and you want to report an issue to us, please attach the model, command/step, and the customized app (if any) with us to reproduce locally.

Thanks!

[AastaLLL]

Hi

Have you tried the PyTorch container from the NGC website?

For example: nvcr.io/nvidia/pytorch:24.11-py3-igpu
PyTorch | NVIDIA NGC

Thanks.

[esteban.gallardo]

Thanks for your support.

My device is a a Jetson Orin AGX.

I’ve already tried to install Pytorch and even if it’s possible to install it there seem to be no way to get CUDA support enabled. Nor compatibility with torchvision and torchaudio.

I’m looking to know if there is going to be a jetson-container update that support Pytorch 2.5.0 the same way it’s now supporting Pytorch 2.3.0.

I just need that this command creates a docker container that support Pytorch 2.5.0:

jetson-containers run $(autotag torchvision)

Dusty already told me several months ago that I should give up on installing things by myself and work on jetson-containers. I’m just doing that.

If there are no plans to create that container it would be nice to know so I could move on.

[AastaLLL]

Hi,

The container with the iGPU tag should work on Orin AGX.
Or you can build it with jetson-containers build pytorch.

The compatible package for JetPack 6.1 can also be found in the below link:
http://jetson.webredirect.org/jp6/cu126

Thanks.

[dusty_nv]

@esteban.gallardo

I’m looking to know if there is going to be a jetson-container update that support Pytorch 2.5.0 the same way it’s now supporting Pytorch 2.3.0.

Hi @esteban.gallardo, sorry just seeing this, and just to round out on it - yes, on JetPack 6.1, pytorch 2.5 is currently the default in jetson-containers. I have also tried 2.6.0-rc1. It would seem because of FlexAttention people are upgrading more often.

Anyways, I tend to keep the containers for the latest JetPack on the latest stable PyTorch. A lot of other packages get rebuilt off of those, so I tend to upgrade it every couple months.