GPU acceleration for PixInsight with Linux (Kubuntu or Ubuntu) using RC Astro Tools (eg StarXterminator) or Starnet ++

I have a very cook book approach to the command line in Ubuntu but you obviously understand it. thank you very much for pointing out what to do if it all goes pear-shaped.
I started with
sudo apt update
sudo apt upgrade

but I did not reboot the computer. After reading further, that might have been all I had to do.
Once I got the mismatch I googled and found this info which I posted above
I admit I did "cook book" it on this fix. It did work, though.
 
Thought I'd let you all know that...... this works with Kubuntu 24.04 ! Just ran through it step by step, after failing the first time I tried using a different guide, and I just removed the stars in a 4 panel mosaic of the heart and soul nebula in 2.5 minutes. It took 30 mins to do a 4500x3500 (ish) before following this guide. mosaic is 14050x9935. I'm ridiculously happy !

just tested Noise XT and Blur XT. same result. now to get INDI sorted....

Thankyou lblock. Looks like this knackered old fx8350 cpu system has a bit more life in it thanks to the gpu now doing the work. Cheers (y)
 
side note.... I did have to add "--override" when installing cuda as the gcc compiler is newer version. worked after that though. :)
 
What nvidia drivers version are you on?
I'm on the proprietary tested driver. 535.171.04.
when I tried setting up cuda using the nvidia instructions it ended up installing driver 550, which isn't available through ubuntu-drivers. probably why it didn't work the first time
 
Thank you very much for the guide @lblock I am trying the trail version out on my older desktop to see if it will work. I have a GeForce GTX 1060 3GB and Ubuntu 22.04 / KDE Plasma. I wasn't sure if these newer drivers would work with my 2017 hardware, but they did, and dropped BlurX from 5-10 minutes down to around a minute or so! Huge difference.
 
Many thanks for your step-by-step guide @lblock !! I have installed a dual boot WIN/UBUNTU (latest version 24.04) on a fresh new custom built laptop (Nvidia RTX3050) and with this guide I finally got PI to use the CUDA power ;)

I have installed some different versions compared to the one you pointed in this guide but everything worked perfect. Here are what I have installed:

- Nvidia drivers 535.171.04
- Nvidia CUDA Toolkit 12.2
- cuDNN v8.9.7 for CUDA 12.x (Local Installer for Ubuntu 22.04 x86_64 Deb)
- libtensorflow-gpu-linux-x86_64-2.15.0

I have tested the GPU usage with nvtop and launching starnet. Worked perfectly.
 
Install Cuda11.8 I get this error

Install CUDA 11.8
sudo wget https://developer.download.nvidia.c...al_installers/cuda_11.8.0_520.61.05_linux.run

~$ sudo wget https://developer.download.nvidia.c...al_installers/cuda_11.8.0_5
20.61.05_linux.run
--2024-05-05 17:21:36-- https://developer.download.nvidia.c...al_installers/cuda_11.8.0_520.61.05_linux.run
Resolving developer.download.nvidia.c...al_installers (developer.download.nvidia.c...al_installers)... failed:
Name or service not known.
wget: unable to resolve host address ‘developer.download.nvidia.c...al_installers’
 
Install Cuda11.8 I get this error

Install CUDA 11.8
sudo wget https://developer.download.nvidia.c...al_installers/cuda_11.8.0_520.61.05_linux.run

~$ sudo wget https://developer.download.nvidia.c...al_installers/cuda_11.8.0_5
20.61.05_linux.run
--2024-05-05 17:21:36-- https://developer.download.nvidia.c...al_installers/cuda_11.8.0_520.61.05_linux.run
Resolving developer.download.nvidia.c...al_installers (developer.download.nvidia.c...al_installers)... failed:
Name or service not known.
wget: unable to resolve host address ‘developer.download.nvidia.c...al_installers’
The ... in the URL isn't really there, it's just shorthand for a truncated link. The full text of the URL is
 
The URL's in blue are links and may appear trucated. In this instance hover over the blue URL, copy it and then in the terminal type out sudo wget and then paste the link you copied after that. I have added instructions for this into the guide.
 
Last edited:
Hi there,
I am working my way thru the instructions (great instructions btw) but am stuck when I try and move the libtensorflow files (near the end of the instructions). I just can't seem to get past this stage. Any help would be much appreciated. The file has been downloaded fine as described, I can see it in my Download directory. This is what I get from my terminal when I try and carry on (per the instructions):

jim@jim-HP-Z6-G4-Workstation:~$ sudo tar -C/usr/local -xzf libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz
[sudo] password for jim:
tar (child): libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz: Cannot open: No such file or directory
tar (child): Error is not recoverable: exiting now
tar: Child returned status 2
tar: Error is not recoverable: exiting now

Again, any help with this would be so grateful. I can almost see the end in sight. Cheers JIM
 
Last edited:
Hi there,
I am working my way thru the instructions (great instructions btw) but am stuck when I try and move the libtensorflow files (near the end of the instructions). I just can't seem to get past this stage. Any help would be much appreciated. The file has been downloaded fine as described, I can see it in my Download directory. This is what I get from my terminal when I try and carry on (per the instructions):

jim@jim-HP-Z6-G4-Workstation:~$ sudo tar -C/usr/local -xzf libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz
[sudo] password for jim:
tar (child): libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz: Cannot open: No such file or directory
tar (child): Error is not recoverable: exiting now
tar: Child returned status 2
tar: Error is not recoverable: exiting now

Again, any help with this would be so grateful. I can almost see the end in sight. Cheers JIM
Make sure that the command is spelled correctly. For example in the command line there should be a space between -C and /usr/local.
It should look exactly like this:

sudo tar -C /usr/local -xzf libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz

Otherwise the system will interpret -C/usr/local as an incorrect location.
 
Make sure that the command is spelled correctly. For example in the command line there should be a space between -C and /usr/local.
It should look exactly like this:

sudo tar -C /usr/local -xzf libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz

Otherwise the system will interpret -C/usr/local as an incorrect location.
Thanks for the monos, I did put that extra space in but I still get the same Terminal response:

jim@jim-HP-Z6-G4-Workstation:~$ sudo tar -C /usr/local -xzf libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz
tar (child): libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz: Cannot open: No such file or directory
tar (child): Error is not recoverable: exiting now
tar: Child returned status 2
tar: Error is not recoverable: exiting now

I checked the name of the file in my Downloads and it is exactly as entered above. I presume that file being in the Downloads is where it is meant to be accessed? is it meant to be anywhere else to be accessed? Any ideas from anyone? Frustrating to say the least. cheers
 
Thanks for the monos, I did put that extra space in but I still get the same Terminal response:

jim@jim-HP-Z6-G4-Workstation:~$ sudo tar -C /usr/local -xzf libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz
tar (child): libtensorflow-gpu-linux-x86_64-2.14.0.tar.gz: Cannot open: No such file or directory
tar (child): Error is not recoverable: exiting now
tar: Child returned status 2
tar: Error is not recoverable: exiting now

I checked the name of the file in my Downloads and it is exactly as entered above. I presume that file being in the Downloads is where it is meant to be accessed? is it meant to be anywhere else to be accessed? Any ideas from anyone? Frustrating to say the least. cheers
I assume that the file to be unzipped has been previously downloaded to the download folder. To be able to apply the command you would have to place the console before in that folder so that the downloaded file is accessible. This is not obvious in the instructions I have reviewed but it would be enough to do:

cd ~/Downloads

Before executing the command to make it work.... or so I guess.
 
echo the above...looks like you're not in the same directory as the downloaded file so the command won't do anything as there's no file to unzip.

doing as Monos says should sort it for you. (y)
 
I have now a problem using BlurXterminator plugin into PI. Does anyone else got this error while using it?

Screenshot from 2024-05-21 18-58-00.png

Can it be related to a wrong CUDA and Tensorflow setup? The strange thing is that starnet is working correctly and it can use the GPU.
 
Thanks sharpie and monos, what you said made sense (in hindsight) and of course that was the problem. Cheers for you help. It all works fine now and my processing time has decreased heaps!!
 
Last edited:
I have now a problem using BlurXterminator plugin into PI. Does anyone else got this error while using it?

View attachment 23177

Can it be related to a wrong CUDA and Tensorflow setup? The strange thing is that starnet is working correctly and it can use the GPU.
Not sure on that one. It seems like reinstalling BlurXterminator might be the answer since it works on Starnet.
 
Back
Top