Error when I try to use ai v4 on blurxterminator

[tantalusthief]

I updated the blurxterminator for pixinsight. I'm able to select ai v4, but when I try to use it I get this error

The ai file is there and it works fine for v2.

 

[Marcelofig]

Go to the RC Astro FAQs:

BlurXTerminator 2.0/AI4 error in PixInsight: MLLoad() could not load tensorflow graph – RC Astro

www.rc-astro.com

 

[cloudbait]

I updated the blurxterminator for pixinsight. I'm able to select ai v4, but when I try to use it I get this error
View attachment 21316
The ai file is there and it works fine for v2.

No problem for me, PI build 1593 on Windows 11. CUDA accelerated. Might want to check with RC-Astro tech support.

 

[tantalusthief]

Go to the RC Astro FAQs:

BlurXTerminator 2.0/AI4 error in PixInsight: MLLoad() could not load tensorflow graph – RC Astro

www.rc-astro.com

t

that fixed it thank you

 

[Warhen]

Hi tantalustheif, may I ask what exactly fixed it? Did you find the original PI version of TensorFlow, or a version, which allowed GPU acceleration? I’d like to try version 4 AI of BXT but as I don’t have the original TF file, I’d like to avoid having to reinstall the program. Thanks all!

 

[cloudbait]

Hi tantalustheif, may I ask what exactly fixed it? Did you find the original PI version of TensorFlow, or a version, which allowed GPU acceleration? I’d like to try version 4 AI of BXT but as I don’t have the original TF file, I’d like to avoid having to reinstall the program. Thanks all!

Why would you need to reinstall anything? If you're currently CUDA accelerated, you should be fine as long as you are using the latest tensorflow library, which you almost certainly are. And if you get the error, you just need to download that library from the link on Russ's site and stick it in the PI /bin folder.

 

[Neven Krcmarek]

Hi tantalustheif, may I ask what exactly fixed it? Did you find the original PI version of TensorFlow, or a version, which allowed GPU acceleration? I’d like to try version 4 AI of BXT but as I don’t have the original TF file, I’d like to avoid having to reinstall the program. Thanks all!

Had this error, this fixed it: https://www.williamliphotos.com/starnet-cuda
You have to update your tensorflow.dll to a newer version from the libtensorflow-gpu-windows-x86_64-2.7.0
Also you have to update CUDA (NVIDIA GPU Computing Toolkit) from 10.1 to 11.8

 

[tantalusthief]

Hi tantalustheif, may I ask what exactly fixed it? Did you find the original PI version of TensorFlow, or a version, which allowed GPU acceleration? I’d like to try version 4 AI of BXT but as I don’t have the original TF file, I’d like to avoid having to reinstall the program. Thanks all!

I downloaded the tensor dll from the rcastro faq link that Marcelofig posted.I downloaded the one for no gpu acceleration and replaced the tensor dll that was already in my computer.

 

[Warhen]

Thanks guys, I’m actually golden with the TF file that I already have. Not only does it work with version 4 AI, but it was 10 seconds faster.