when running test_nn.py test_Softmax2d fails

test_Softmax2d (test_nn.TestNN) ... pure virtual method called
terminate called without an active exception
Traceback (most recent call last):
File "run_test.py", line 342, in
main()
File "run_test.py", line 334, in main
raise RuntimeError(message)
RuntimeError: test_nn failed! Received signal: SIGIOT

When running it multiple times using
$ python -m unittest test_nn.TestNN.test_Softmax2d -v

sometimes it succeeds, but most of the time it fails with the above failure
When I compile with DEBUG=1 , it seems to succeed more than it fails

PyTorch version: 0.5.0a0+6eec411
Is debug build: No
CUDA used to build PyTorch: 9.0.176

OS: Red Hat Enterprise Linux Server 7.4 (Maipo) ppc64le Power8
GCC version: (GCC) 4.8.5 20150623 (Red Hat 4.8.5-16)
CMake version: 3.6.3

Python version: 3.6
Is CUDA available: Yes
CUDA runtime version: 9.0.176
GPU models and configuration:
GPU 0: Tesla P100-SXM2-16GB
GPU 1: Tesla P100-SXM2-16GB

Nvidia driver version: 384.81
cuDNN version: Probably one of the following:
/usr/local/cuda-9.0/targets/ppc64le-linux/lib/libcudnn.so
/usr/local/cuda-9.0/targets/ppc64le-linux/lib/libcudnn.so.7
/usr/local/cuda-9.0/targets/ppc64le-linux/lib/libcudnn.so.7.0.4
/usr/local/cuda-9.0/targets/ppc64le-linux/lib/libcudnn_static.a

Versions of relevant libraries:
[pip3] numpy (1.14.0)
[conda] torch 0.5.0a0+6eec411