Login succeeded.
Retrieving data…
Data retrieved successfully.
Installation of this software is under the terms and conditions of the license agreements located in /opt/nvidia/sdkmanager/Eula/
Installing target to folder: /home//nvidia/nvidia_sdk
Download:
File System and OS [▇▇▇▇————————————————————————————————] 6% 16.08MB/s
Drivers for Jetson TX1/Nano [▇▇▇▇————————————————————————————————] 6% 13.36MB/s
In prior versions you would get:
Logging in…
Login succeeded.
Retrieving data…
Data retrieved successfully.
Installation of this software is under the terms and conditions of the license agreements located in /opt/nvidia/sdkmanager/Eula/
Download:
File System and OS ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
Drivers for Jetson TX1/Nano ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
CUDA Toolkit for L4T ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
cuDNN on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
TensorRT on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
OpenCV on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
VisionWorks on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
NVIDIA Container Runtime with Docke… ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
Multimedia API ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100
Install log file is available at: /home//.nvsdkm/sdkm.log
All done!
If I log into the UI and select download only, it will download all components. This issue is separate from the missing deepstream download via the CLI.
And this effects multiple devices and multiple JetPack versions. I would roll back to the older SDK Manager, but all of its downloads broke when the new SDK manager was released. The command line is completely unusable at this point.
I am also having similar issues downloading JetPack 4.2 using the CLI. I Get two download bars and then an “All Done”, but the rest of the packages don’t download. What has changed on the backend. Can someone from Nvidia please explain why the new update bricked the CLI?
@Nvidia: any update on this? Would be nice to have the CLI working again as before as currently there is no way to fully automate the initial setup headless as part of a CI/CD pipeline.
@Nvidia is there any updates on this or some sort of timeline as to when you guys are expecting the CLI to get fixed as it bricks alot of our internal processes since we’ve established methods to flash our devices using this method from a backend and deployment perspective. A prompt response would be beneficial.
Thanks Dusty. I’ll give this a try and report back. Are there plans to update the documentation for this? It is very outdated and missing a lot of this nuance now: Command-Line Install :: NVIDIA SDK Manager Documentation - I believe the command line app help itself also needs to be updated.
The application just hangs and never does anything. If I then open the SDK Manager, log in saving credentials, then close the SDK Manager, I can now run the command line.
It also tells me it is saving to /home//nvidia/nvidia_sdk which is doesn’t actually install anything to:
Login succeeded.
Retrieving data...
Data retrieved successfully.
Installation of this software is under the terms and conditions of the license agreements located in /opt/nvidia/sdkmanager/Eula/
Installing target to folder: /home/<user>/nvidia/nvidia_sdk
Download:
File System and OS ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
Drivers for Jetson TX2 & TX2i ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
CUDA Toolkit for L4T ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
cuDNN on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
TensorRT on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
OpenCV on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
VisionWorks on Target ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
NVIDIA Container Runtime with Docke... ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
Multimedia API ✔ [▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇] 100%
Install log file is available at: /home/idavis/.nvsdkm/sdkm.log
So we can no longer log in via the command line when running the sdkmanager. We must log in first, the close it, then run the cli.
There is no way to download the DeepStream or TensorFlow components without going through the UI
The log messaging is incorrect with regard to the installation folder.
–installontarget isn’t found in the command line help
The CLI options we’re using here are marked as only for install/uninstall in the command line help.
Thanks for your feedback. Please see a few answers below.
Can you let me know your sdkmanager client version? I tested with 0.9.14-4964, but couldn’t reproduce the issue. If you are using the same version, can you please upload the ~/.nvsdkm/sdkm.log file for more investigation?
Yes. We’ll fix the issue with additional sdk in next release.
sdkmanager --help command also lists downloadonly option, like the following:
sdkmanager --help
Options
-h, --help Displays this usage guide.
--query Optional. Prints all options available for the user. Must be executed with
the --user or --offline settings
--cli install|uninstall|update|downloadonly CLI mode option. This allows SDK Manager to run install, uninstall, download
only or update without displaying a user interface.
--view none|live|log Optional. This option allows you to view the cli install while it is in
process, output to a log file, or turn off viewing the install completely.
Default value is live.
--logs Optional. Exports the log files when the install process is complete.
--logintype devzone|nvonline Optional. Login with devzone or nvonline, default is devzone.
--archivedversions Optional. Display only archived versions.