Access cuda using tensorflow crash touchdesigner b24410.

Access Cuda 9.0 using TensorFlow crash touchdesigner.

Details :

Added in preferences, Python module path:

Which use external python.
of my machine , python 3.5

I posted as bug cause it crashed but,
what rules i should follow ?
Any other way to use Anaconda ?

thank you ,
Keep up the great work

The 2018.20000 Official is built against CUDA 8.0. That is likely the issue. We’ll be posting a branch build with ZED support soon that is built against CUDA 9.2, if you can find a version of that library built against that.

I been told that the 9.2 raising other issues (I cant run the tests with it).
Maybe there is option to be flexible about it ?

anyway, thanks for the answer - I will update.

There is chance its all because of using external python ?
Im not use cuda anymore ( could be im wrong, but this time I installed the cpu version of tensor)
and its still kill Touch.

I think Tensorflow uses Cuda internally. Possibly if you spawn the process as a seperate thread it might remove the cuda conflict.

hello! maybe this will help you? viewtopic.php?f=4&t=12505#p48408

so now I used 2020.20000series in TouchDesigner, which is built against CUDA 10.0,Cuda right? so now my configuration is tensorflow == 2.0.0 python =3.7 CUDA==10.0 cuDNN==7.4

i thought I totally follow version, but TD still cashed