FS#56766 - [python-tensorflow-opt-cuda] crash on cuda 9.1

Attached to Project: Community Packages
Opened by John Tukey (John_Tukey) - Monday, 18 December 2017, 17:56 GMT
Last edited by Sven-Hendrik Haase (Svenstaro) - Monday, 25 December 2017, 12:07 GMT
Task Type Bug Report
Category Packages
Status Closed
Assigned To Sven-Hendrik Haase (Svenstaro)
Architecture All
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:
With the latest packages of python-tensorflow-opt-cuda, cuda and cudnn, the TensorFlow scripts crash when launched. Error message below:

`Failed to get the number of CUDA devices: CUDA driver version is insufficient for CUDA runtime version
Floating point exception (core dumped)`

To make it work again, python-tensorflow-opt-cuda has to be downgraded to 1.4.0 and cuda to 9.0

Related issues on the github page:
https://github.com/tensorflow/tensorflow/issues/15140
https://github.com/tensorflow/tensorflow/issues/12052

Additional info:
* python-tensorflow-opt-cuda 1.4.1-1
cuda 9.1.85-1
cudnn 7.0.5-1
This task depends upon

Closed by  Sven-Hendrik Haase (Svenstaro)
Monday, 25 December 2017, 12:07 GMT
Reason for closing:  No response
Comment by John Tukey (John_Tukey) - Monday, 18 December 2017, 18:04 GMT
Sorry, wrong Severity, it should be High
Comment by Sven-Hendrik Haase (Svenstaro) - Wednesday, 20 December 2017, 07:02 GMT
I'm pretty sure your system is either running outdated nvidia drivers or simply not upgraded. It does work fine with nvidia 387. Upgrade and reboot if unsure.

Loading...