I built Pytorch from source on Jetson TX2 with CUDA 10. I think it builds with CUDA integration but I am not able to use CUDA.
>>> torch.cuda.is_available()
True
>>> torch.rand(1).cuda()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
RuntimeError: CUDA error: unknown error
I build with following version.
Python = 3.7.3
Pytorch = 1.0.0
CUDA = 10.0
Any idea about how should I solve this issue?