As reported in TensorFlow issue https://github.com/tensorflow/tensorflow/issues/5777, a number of users have problem using the library after their PC has woken up from suspended state.
My system details:
TensorFlow: tensorflow-gpu 1.0.0 installed via pip in Anaconda sandbox
GPU: GeForce GTX 680
NVIDIA driver: 375.39
OS: Ubuntu 16.10 64 bit
The details of steps to reproduce the crash in my system, and the failure messages, are in this comment: https://github.com/tensorflow/tensorflow/issues/5777#issuecomment-292814193. The attached nvidia-bug-report is from a system after reboot – please let me know if I should produce one straight after the TF/vectorAdd crash.
TensorFlow contributor zheng-xq suspects this is a problem with cuda itself, since the crash can be reproduced using vectorAdd example, so I am reporting it here.
nvidia-bug-report.log.gz (224 KB)