Somehow I downloaded the May release, I’ll try again with the correct July release. That said, I let it run overnight and in the morning woke up to see that it had finally completed.
Hi Dustin, After I install the Updated SD card image. I could recognize it is a new OS. But I could not do the nvidia-docker command to verify the nvidia container runtime is installed. Could you let me know if there is an issue with the updated SD card image?
I believe the nvidia-docker command was deprecated when nvidia-docker2 came out, but they kept a wrapper for backwards compatability. Instead you can specify the runtime with “–runtime=nvidia” when you do “docker run”, so instead of:
nvidia-docker run somecudathing:latest
it would be
docker run --runtime=nvidia somecudathing:latest
@nvidia employees, I don’t know if I say this enough, but thank you again. This is great work.
Question about using this new build L4T Ver 32.2 on the Jetbot SD where the current construction Wiki download image is L4t 32.1. Is it easy to add the packages and modules to run the Jetbot, or better to wait until NVidia releases an updated version?
With the release of nvidia-docker for the jetson line up I was wondering if you guys would consider adding the appropriate patches to the next kernel release to add support for all of apparmor’s features.
conical provides the patches and they apply without issue to the current release 4.9.140.
currently the kernel has basic apparmor support however almost all of the extra features are missing for example dbus integration.
the patches i’m talking about add all the missing features.
Also it would be great if you guys could patch in support for fuse mounts in user namespace.
this would be a bit more of a task but the patches are available and it could be done.
So if flashed this, I would need to install python3-dev etc, pip3, Jupyter Notebook, Jetbot package from the github, configure stats & jupyter to autostsrt on boot, Adafruit motor diver & PiOled package? Any other essential packages?
Dusty,
by chance does the new Jetpack version come with OpenCV “prebuilt with the GStreamer flag set as true”?
I wasn’t able to work with the MIPI-lane connected raspberry pi camera for the Nano I bought, originally. At least given the OpenCV version that came with Jetpack 4.2 because it was not prebuilt with the GStreamer flag set as true. The process of re-building OpenCV on the board was fairly time consuming. I’m sure a lot of developers would appreciate the the newest Jetpack release to include a version of OpenCV prebuilt with the GStreamer flag set already as true, to avoid this cumbersome task.
Example issue:
I was not able to access the camera data via GStreamer “through OpenCV” until I performed the aforementioned task.
dusty_nv, regarding your Posted 07/20/2019 06:29 PM, am I correct that this r32.2 Image does not contain the jetpack 4.2.1 and that the “Create image from Sctratch” process must include installing Jetpack as linked from that “Create” instructions? The Jetpack install It appears to create another flash? Regards.
Thanks, but following the “Create Image from Scratch” Step 5. Install the pre-built TensorFlow pip wheel by following these instructions:, goes to the Installing TensorFlow for Jetson Platform :: NVIDIA Deep Learning Frameworks Documentation which in “2. Prerequisites and Dependencies” requires the step “Install JetPack on your Jetson device.” which I could not due limited space on my Ubuntu PC, so I did ignore that and continued with the other dependencies to install Tensorflow and 3. Install TensorFlow, which all completed OK. However, the “4. Verifying The Installation” step $ python3, $ import tensor flow, failed with an error.
So is TensorFlow not required for Jetbot and should I have not installed that long list of Tensorflow depandnces, like numpy, continue wi the remaining Jetbot packages anyway? And what is the command or method to list the Jetpack and L4T versions? I tried the jetsonhacks/jetsonUtilities method “$ python jetsonInfo.py” which failed as there was no /etc/nvidia-tegra folder.
Regards
If you already re-flashed your SD card with the JetPack 4.2.1 image, then you already have JetPack on your SD card, and you don’t need to install it on your PC.
I believe TensorFlow is only used in the Object Detection portion of JetBot, so you should be able to proceed for now. Then once the refreshed JetBot image is available (that has been upgraded by us to JetPack 4.2.1), you can switch over to using the new JetBot image.