Experimental TensorFlow GPU acceleration repository

munger1 said:
Good for you, I'm not
Click to expand...
Well, I don't know what you expect PI to do. Its developers can hardly support libraries and operating system extensions that it does not use. PI works fully as advertised and promised, and the RC tools do, as well.
 
cloudbait said:
Well, I don't know what you expect PI to do. Its developers can hardly support libraries and operating system extensions that it does not use. PI works fully as advertised and promised, and the RC tools do, as well.
Click to expand...
And why does it work in trial? You expect it working after buying it too. That's what I expect, should be an easy task.
 
munger1 said:
And why does it work in trial? You expect it working after buying it too. That's what I expect, should be an easy task.
Click to expand...
It does work after buying. You're having problems with something that isn't part of PI. And those problems are almost certainly operator error. What feature of PI or any third-party plugin is failing to meet its operational guarantee in your purchased version?
 
munger1 said:
And why does it work in trial?
Click to expand...
GPU acceleration doesn't work in the trial version and in the consumer version of Pixinsight for any module/script shipped with Pixinsight.
Only the RC-Tools have GPU acceleration support, and it works in both, the trial and the consumer version.
If the RC-Tools are not having the GPU acceleration, then the problem is roughly 60cm in front of the screen.
After uninstalling the trial version and install the consumer version, you have to enable the GPU acceleration again for the RC-Tools.

Cheers
Tom
 
if it works in trial after instlaling RC astros tensorflow GPU repo and it doesnt work in commercial and if its not pixinsights thing, than the problem isnt roughly 60cm in front of the screen, it is roughly about 10 000km away from the screen. dont worry, i solved it, from RC i didnt even get an answer. there is a solution on the rc site, but its outdated and you can not find what you need. someone sent me a very good instruction
 
munger1 said:
if it works in trial after instlaling RC astros tensorflow GPU repo and it doesnt work in commercial and if its not pixinsights thing, than the problem isnt roughly 60cm in front of the screen, it is roughly about 10 000km away from the screen. dont worry, i solved it, from RC i didnt even get an answer. there is a solution on the rc site, but its outdated and you can not find what you need. someone sent me a very good instruction
Click to expand...
And I'm sure the solution had nothing to do with PI, and everything to do with how you installed and configured CUDA on your system.
 
munger1 said:
if it works in trial after instlaling RC astros tensorflow GPU repo and it doesnt work in commercial and if its not pixinsights thing, than the problem isnt roughly 60cm in front of the screen, it is roughly about 10 000km away from the screen. dont worry, i solved it, from RC i didnt even get an answer. there is a solution on the rc site, but its outdated and you can not find what you need. someone sent me a very good instruction
Click to expand...
Can you share where to find that instruction?
Larry
 
for ppl who have nvidia 50xx (in my case nVidia 5070 Ti) on windows - the point is, that CUDA supports this GPU from version 12.x, so i searched many forums and came with this setup :)

cudnn-windows-x86_64-9.10.1.4_cuda12-archive.zip
cuda_12.9.0_576.02_windows.exe
libtensorflow-gpu-windows-x86_64-2.10.0.zip (windows latest)
zlib123dllx64.zip

install it just like its in the link from munger1, but, on top of new versions, it needed few tweaks:
- path variable - add also C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.9\libnvvp, not only to /bin folder
- system variables have to be CUDA_PATH and CUDA_PATH_v12_9 both pointing to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.9

Screen Shot 2025-05-24 at 03.19.18.762 PM.png

Screen Shot 2025-05-24 at 03.19.31.848 PM.png



Then, it should work fine. For first time i tryed it, it seemed that it froze again, but according to system manager the GPU memory was full, so i waited and after few minutes it started working like a charm.
Screen Shot 2025-05-24 at 03.20.02.544 PM.png


Hope this post will save someone few hours of tinkering :)
 
Does anyone know why TensorFlow would stop working all of a sudden? And know how to get it working again? I've had this repository downloaded for a while now and haven't had any issues. I'm currently on Pix ver 1.9.3 Lockhart.

CS, Tommy
 
darko7411 said:
Does anyone know why TensorFlow would stop working all of a sudden? And know how to get it working again? I've had this repository downloaded for a while now and haven't had any issues. I'm currently on Pix ver 1.9.3 Lockhart.

CS, Tommy
Click to expand...
Disregard... I reset updates, applied all the updates and it started working again.
 
You must log in or register to reply here.
Back
Top