JetPack 4.5 Production Release with L4T 32.5

We are pleased to announce JetPack 4.5, a production release supporting Jetson AGX Xavier series, Jetson Xavier NX, Jetson TX2 series, Jetson TX1, and Jetson Nano.

JetPack 4.5 includes VPI 1.0, Security features including enhanced secure boot and support for full disk encryption, enhanced bootloader functionality, and a new way of flashing Jetson devices using NFS.

We will be hosting a webinar on Feb 9th for an in-depth overview of JetPack 4.5 features where we will also demo select features and answer any questions you may have. Please register here.

Learn how to implement computer vision and image processing pipelines using VPI by registering to the webinar we will be hosting on Feb 11th

Highlights of JetPack 4.5 are:

  • First Production release of Vision Programming Interface (VPI).
  • Support for Docker 19.03.
  • Support for installing JetPack components on Linux host computer using Debian package management tool.
  • Secure boot enhanced to extend encryption support to kernel, kernel-dtb and initrd.
  • Disk encryption supported to protect data at rest.
  • New way of flashing eMMC, SDCard, USB or NVME connected to Jetson using Network File System.

Visit JetPack 4.5 and L4T 32.5 page for more details.

JetPack 4.5 components:

  • L4T R32.5
  • CUDA 10.2
  • cuDNN 8.0
  • TensorRT 7.1.3
  • VisionWorks 1.6
  • OpenCV 4.1.1
  • Vulkan 1.2
  • VPI 1.0 (First Production Release)
  • Nsight Systems 2020.4
  • Nsight Graphics 2020.5
  • Nsight Compute 2019.3

Install JetPack 4.5 using SDK Manager or SD card image (for Jetson Xavier NX Developer Kit) or upgrade via Debian package management tool (refer instructions from here)

Notes: (We will update this announcement accordingly when we take care of below notes)

2 Likes

Thanks! I was eagerly waiting for this. One question. The webinar is 9 AM PT which translates to 2 AM local time for me. I know it will be recorded, but is it possible to have another session at an Asia friendly time?

Good suggestion @jaiyamsharma. We will surely consider it.

1 Like

Is the OpenCV 4.1.1 here built with CUDNN and CUDA enabled? I usually have to build the source package with these enabled and my understanding is OpenCV prior to 4.3 didn’t support past CUDNN v7

Hi
From the links in the current docs https://docs.nvidia.com/jetson/archives/l4t-archived/l4t-325/index.html#page/Tegra%20Linux%20Driver%20Package%20Development%20Guide/flashing.html#wwpID0E0YK0HA

it appears that native boot support from NVME is still not supported.

Can you share if/when such support will appear?

1 Like

According to this page Jetpack 4.5 Archive | NVIDIA Developer 4.5 supports Cudnn 10.2. (above it says 8) Is possible to update 4.4.1 with just cudnn 10.2 while I wait for the 4.5 board support package to be completed for my carrier board?

Hello Suhash,

This is great news! Regarding the notes section, do you mean to say that Jetpack 4.5 does not support the current containers listed on NGC? If so, what’s the dependency between the container and the Jetpack update? Will the containers go through a similar update each time Jetpack version gets updated?

Can you give a timeline when the containers will be updated?

Regards,

uersoy

@uersoy , The current containers listed on NGC is for JetPack 4.4.x. We will be releasing NGC containers in one or two business days for JetPack 4.5.
The present mechanism for containers we have has a depedency that it brings CUDA, cuDNN, TensorRT and othe components form host. We are working on changing this in future. So the depedency will be removed in future.

2 Likes

@alex15 , the cuDNN versoin in the JetPack 4.5 Archive page is wrong and we will correct it. The JetPack 4.5 page (JetPack SDK | NVIDIA Developer) mentions cuDNN correct version as cuDNN 8.0.

Thanks Suhash, this make sense.

1 Like

@luisvale , native boot from NVMe is not yet supported on Jetson AGX Xavier and Jetson Xavier NX. We are looking into it for JetPack 4.6.

1 Like

Is the VPI 1.0 docs available somewhere? It says 22nd in the announcement but I can’t find them.

Also here: Release Notes :: NVIDIA JetPack Documentation it says VPI 1.0 has support for Optical Flow on the cpu and gpu. Does that mean it is not supported on the PVA?
On this thread (Any details about Stereo & Optical Flow Engine? - #2 by dusty_nv) there is talk of a hardware optical flow engine. Is that what is being used?

What I’m trying to determine is if there is any performance improvement in running VPI Optical Flow compared to just running OpenCV Optical Flow?

@suhash Hi, Suhash, can I pull and run the JetPack 4.5 base container on my current JetPack 4.4 board? I have tried to run nvcr.io/nvidia/l4t-pytorch:r32.5.0-pth1.7-py3 but got this:

import torch
torch.version
‘1.7.0’

torch.cuda.is_available()
/usr/local/lib/python3.6/dist-packages/torch/cuda/init.py:52: UserWarning: CUDA initialization: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx (Triggered internally at /media/nvidia/WD_NVME/PyTorch/JetPack_4.4.1/pytorch-v1.7.0/c10/cuda/CUDAFunctions.cpp:100.)
return torch._C._cuda_getDeviceCount() > 0
False

Hi @joey.su , no currently the containers have dependency on the root file system (which we are planning to remove going forward) and hence JP 4.5 base container need to be pulled on JetPack 4.5 only

@oscar.thorn , VPI 1.0 docs are available now VPI - Vision Programming Interface: Main Page

Thanks, Suhash.

Thanks!

Hi Suhash (@suhash )

I was just testing the jetson gaze and pose containers and they don’t seem to work with Jetpack 4.5. When will they be updated to run on Jetpack 4.5?

Thanks!

@uersoy , those pose and gaze containers were meant mostly for a demo purpose and we do not plan to update that for JetPack 4.5. The pose container pulled trt_pose project from GitHub - NVIDIA-AI-IOT/trt_pose: Real-time pose estimation accelerated with NVIDIA TensorRT