Tensorflow 2.0 python segfault on Jetson AGX Xavier

I’m managed to successfully install tensorflow 2.0.0 following this guide:

Now I’m trying to run the object detection tutorial from tensorflow:

and it seem to always crash on a segmentation fault:

Thread 17 "python3.6" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7f3b7fe1f0 (LWP 20006)]
0x0000007fa829d744 in tensorflow::NonMaxSuppressionV2GPUOp::Compute(tensorflow::OpKernelContext*) ()
   from /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/_pywrap_tensorflow_internal.so

There is no memory problems, at the point of the crash I’m only using 9/16GB, and none on the swap.
I found some people had that issue before, but it should be already fixed on my version of tensorflow it seems, yet it’s still happening on the jetson:


A common issue is that TensorFlow is built with the incorrect GPU architecture.

Xavier is using GPU capacity sm=72.
Did you add sm=72 when compiling TensorFlow?


Hi vladpuuz4w,

We haven’t heard back from you in a couple weeks, so marking this topic closed.
Please open a new forum issue when you are ready and we’ll pick it up there.